Douglas 2011

Ground-motion prediction equations
19642010
Final Report
BRGM/RP-59356-FR
February, 2011
Ground-motion prediction equations

19642010
Final Report
BRGM/RP-59356-FR
February, 2011
Study carried out as part of research activities
BRGM 2010 - PDR10RNS41
John Douglas
Checked by:
Original signed by
Julien Rey
Approved by:
Original signed by
Hormoz Modaressi
If the present report has not been signed in its digital form, a signed original
of this document will be available at the Information and Documentation Unit (STI).
BRGMs quality management system is certied ISO 9001:2008 by AFAQ.
Keywords: earthquakes, seismic hazard, ground-motion prediction equations (GMPEs),

ground-motion models, peak ground acceleration, elastic response spectrum
In bibliography, this report should be cited as follows:

Douglas, J. (2011) Ground-motion prediction equations 19642010, BRGM/RP-59356-FR,
444 pages, 9 illustrations.
BRGM, 2011. No part of this document may be reproduced without the prior permission of BRGM.
Synopsis
This report summarizes all empirical ground-motion prediction equations (GMPEs), to estimate earthquake peak ground acceleration (PGA) and elastic response spectral ordinates,
published between 1964 and 2010 (inclusive). This report replaces: the Imperial College London report of Douglas (2004a), which provided a summary of all GMPEs from 1964 until the
end of 2003; the BRGM report of Douglas (2006), which summarized all GMPEs from 2004
to 2006 (plus some earlier models); and the report of Douglas (2008), concerning GMPEs
published in 2007 and 2008 (plus some earlier models). In addition, this report lists published
GMPEs derived from simulations, although details are not given since the focus here is on
empirical models. Studies that only present graphs are only listed as are those nonparametric formulations that provide predictions for different combinations of distance and magnitude
because these are more difcult to use for seismic hazard analysis than those which give
a single formula. Equations for single earthquakes or for earthquakes of approximately the
same size are excluded due to their limited usefulness. Those relations based on conversions
from macroseismic intensity are only listed.
This report was compiled as part of Task 2 (Compilation of list of candidate GMPEs) of the
Global Component on GMPEs coordinated by the Pacic Earthquake Engineering Research
Center (PEER) for the Global Earthquake Model (GEM) and Workpackage 4 (Strong ground
motion modeling) of the Seismic Hazard Harmonization in Europe (SHARE) project of the
Seven Framework Programme of the European Commission (grant agreement no. 226769).
This report summarizes, in total, the characteristics of 289 empirical GMPEs for the prediction of PGA and 188 empirical models for the prediction of elastic response spectral ordinates.
In addition, many dozens of simulation-based models to estimate PGA and elastic response
spectral ordinates are listed but no details are given.
It should be noted that the size of this report means that it may contain some errors or
omissions. No discussion of the merits, ranges of applicability or limitations of any of the
relationships is included herein except those mentioned by the authors or inherent in the data
used. This report is not a critical review of the models. The GMPEs are generally reported in
the form used in the original references.
Ground-motion prediction equations 19642010
BRGM/RP-59356-FR Final Report
Contents
1 Introduction
17
1.1
Other summaries and reviews of GMPEs . . . . . . . . . . . . . . . . . . . . . . 18
1.2
GMPEs summarised here . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2 Summary of published GMPEs for PGA
2.1
Esteva & Rosenblueth (1964) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2
Kanai (1966) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3
Milne & Davenport (1969) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4
Esteva (1970) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5
Denham & Small (1971) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.6
Davenport (1972) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.7
Donovan (1973) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.8
Denham et al. (1973) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.9
Esteva & Villaverde (1973) & Esteva (1974) . . . . . . . . . . . . . . . . . . . .
2.10 McGuire (1974) & McGuire (1977) . . . . . . . . . . . . . . . . . . . . . . . . .
2.11 Orphal & Lahoud (1974) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.12 Ambraseys (1975b), Ambraseys (1975a) & Ambraseys (1978a) . . . . . . . . .
2.13 Trifunac & Brady (1975), Trifunac (1976) & Trifunac & Brady (1976) . . . . . . .
2.14 Blume (1977) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.15 Milne (1977) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.16 Ambraseys (1978b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.17 Donovan & Bornstein (1978) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.18 Faccioli (1978) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.19 McGuire (1978) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.20 A. Patwardhan, K. Sadigh, I.M. Idriss, R. Youngs (1978) reported in Idriss (1978)
2.21 Cornell et al. (1979) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.22 Faccioli (1979) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.23 Faccioli & Agalbato (1979) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.24 Aptikaev & Kopnichev (1980) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.25 Blume (1980) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.26 Iwasaki et al. (1980) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.27 Matuschka (1980) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.28 Ohsaki et al. (1980a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.29 Campbell (1981) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.30 Chiaruttini & Siro (1981) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.31 Joyner & Boore (1981) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.32 Bolt & Abrahamson (1982) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.33 Joyner & Boore (1982b) & Joyner & Boore (1988) . . . . . . . . . . . . . . . . .
2.34 PML (1982) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
23
23
23
23
24
24
24
24
25
25
25
26
26
26
27
28
28
29
29
30
30
31
31
31
33
33
34
34
35
35
37
38
39
40
40
2.35
2.36
2.37
2.38
2.39
2.40
2.41
2.42
2.43
2.44
2.45
2.46
2.47
2.48
2.49
2.50
2.51
2.52
2.53
2.54
2.55
2.56
2.57
2.58
2.59
2.60
2.61
2.62
2.63
2.64
2.65
2.66
2.67
2.68
2.69
2.70
2.71
2.72
2.73
2.74
2.75
2.76
2.77
2.78
2.79
2.80
2.81
2.82
2.83
6
Schenk (1982) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Brillinger & Preisler (1984) . . . . . . . . . . . . . . . . . . . . . . . . . .
Joyner & Fumal (1984), Joyner & Fumal (1985) & Joyner & Boore (1988) .
Kawashima et al. (1984) & Kawashima et al. (1986) . . . . . . . . . . . .
McCann Jr. & Echezwia (1984) . . . . . . . . . . . . . . . . . . . . . . . .
Schenk (1984) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Xu et al. (1984) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Brillinger & Preisler (1985) . . . . . . . . . . . . . . . . . . . . . . . . . .
Kawashima et al. (1985) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Peng et al. (1985b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Peng et al. (1985a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
PML (1985) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
McCue (1986) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
C.B. Crouse (1987) reported in Joyner & Boore (1988) . . . . . . . . . . .
Krinitzsky et al. (1987) & Krinitzsky et al. (1988) . . . . . . . . . . . . . .
Sabetta & Pugliese (1987) . . . . . . . . . . . . . . . . . . . . . . . . . .
K. Sadigh (1987) reported in Joyner & Boore (1988) . . . . . . . . . . . .
Singh et al. (1987) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Algermissen et al. (1988) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Annaka & Nozawa (1988) . . . . . . . . . . . . . . . . . . . . . . . . . . .
K.W. Campbell (1988) reported in Joyner & Boore (1988) . . . . . . . . .
Fukushima et al. (1988) & Fukushima & Tanaka (1990) . . . . . . . . . .
Gaull (1988) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
McCue et al. (1988) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Petrovski & Marcellini (1988) . . . . . . . . . . . . . . . . . . . . . . . . .
Tong & Katayama (1988) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Yamabe & Kanai (1988) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Youngs et al. (1988) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Abrahamson & Litehiser (1989) . . . . . . . . . . . . . . . . . . . . . . . .
Campbell (1989) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ordaz et al. (1989) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Alfaro et al. (1990) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ambraseys (1990) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Campbell (1990) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Dahle et al. (1990b) & Dahle et al. (1990a) . . . . . . . . . . . . . . . . .
Jacob et al. (1990) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sen (1990) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sigbjrnsson (1990) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Tsai et al. (1990) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ambraseys & Bommer (1991) & Ambraseys & Bommer (1992) . . . . . . .
Crouse (1991) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Garca-Fernndez & Canas (1991) & Garcia-Fernandez & Canas (1995) .
Geomatrix Consultants (1991), Sadigh et al. (1993) & Sadigh et al. (1997)
Huo & Hu (1991) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
I.M. Idriss (1991) reported in Idriss (1993) . . . . . . . . . . . . . . . . . .
Loh et al. (1991) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Matuschka & Davis (1991) . . . . . . . . . . . . . . . . . . . . . . . . . .
Niazi & Bozorgnia (1991) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Rogers et al. (1991) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
41
41
42
42
43
44
44
45
45
46
46
47
47
47
48
49
50
51
51
51
52
53
54
55
55
56
56
57
58
59
60
61
61
61
62
63
63
63
63
64
65
66
67
67
68
69
69
70
71
2.84
2.85
2.86
2.87
2.88
2.89
2.90
2.91
2.92
2.93
2.94
2.95
2.96
2.97
2.98
2.99
2.100
2.101
2.102
2.103
2.104
2.105
2.106
2.107
2.108
2.109
2.110
2.111
2.112
2.113
2.114
2.115
2.116
2.117
2.118
2.119
2.120
2.121
2.122
2.123
2.124
2.125
2.126
2.127
2.128
2.129
2.130
2.131
Stamatovska & Petrovski (1991) . . . . . . . . . . . . . . . . .

Abrahamson & Youngs (1992) . . . . . . . . . . . . . . . . . .
Ambraseys et al. (1992) . . . . . . . . . . . . . . . . . . . . .
Kamiyama et al. (1992) & Kamiyama (1995) . . . . . . . . . .
Sigbjrnsson & Baldvinsson (1992) . . . . . . . . . . . . . . .
Silva & Abrahamson (1992) . . . . . . . . . . . . . . . . . . . .
Taylor Castillo et al. (1992) . . . . . . . . . . . . . . . . . . . .
Tento et al. (1992) . . . . . . . . . . . . . . . . . . . . . . . .
Theodulidis & Papazachos (1992) . . . . . . . . . . . . . . . .
Abrahamson & Silva (1993) . . . . . . . . . . . . . . . . . . . .
Boore et al. (1993), Boore et al. (1997) & Boore (2005) . . . .
Campbell (1993) . . . . . . . . . . . . . . . . . . . . . . . . . .
Dowrick & Sritharan (1993) . . . . . . . . . . . . . . . . . . . .
Gitterman et al. (1993) . . . . . . . . . . . . . . . . . . . . . .
McVerry et al. (1993) & McVerry et al. (1995) . . . . . . . . .
Singh et al. (1993) . . . . . . . . . . . . . . . . . . . . . . . .
Steinberg et al. (1993) . . . . . . . . . . . . . . . . . . . . . .
Sun & Peng (1993) . . . . . . . . . . . . . . . . . . . . . . . .
Ambraseys & Srbulov (1994) . . . . . . . . . . . . . . . . . . .
Boore et al. (1994a) & Boore et al. (1997) . . . . . . . . . . .
El Hassan (1994) . . . . . . . . . . . . . . . . . . . . . . . . .
Fukushima et al. (1994) & Fukushima et al. (1995) . . . . . . .
Lawson & Krawinkler (1994) . . . . . . . . . . . . . . . . . . .
Lungu et al. (1994) . . . . . . . . . . . . . . . . . . . . . . . .
Musson et al. (1994) . . . . . . . . . . . . . . . . . . . . . . .
Radu et al. (1994), Lungu et al. (1995a) & Lungu et al. (1996)
Ramazi & Schenk (1994) . . . . . . . . . . . . . . . . . . . . .
Xiang & Gao (1994) . . . . . . . . . . . . . . . . . . . . . . . .
Aman et al. (1995) . . . . . . . . . . . . . . . . . . . . . . . .
Ambraseys (1995) . . . . . . . . . . . . . . . . . . . . . . . . .
Dahle et al. (1995) . . . . . . . . . . . . . . . . . . . . . . . .
Lee et al. (1995) . . . . . . . . . . . . . . . . . . . . . . . . .
Lungu et al. (1995b) . . . . . . . . . . . . . . . . . . . . . . .
Molas & Yamazaki (1995) . . . . . . . . . . . . . . . . . . . . .
Sarma & Free (1995) . . . . . . . . . . . . . . . . . . . . . . .
Ambraseys et al. (1996) & Simpson (1996) . . . . . . . . . . .
Ambraseys & Simpson (1996) & Simpson (1996) . . . . . . . .
Aydan et al. (1996) . . . . . . . . . . . . . . . . . . . . . . . .
Bommer et al. (1996) . . . . . . . . . . . . . . . . . . . . . . .
Crouse & McGuire (1996) . . . . . . . . . . . . . . . . . . . . .
Free (1996) & Free et al. (1998) . . . . . . . . . . . . . . . . .
Inan et al. (1996) . . . . . . . . . . . . . . . . . . . . . . . . .
Ohno et al. (1996) . . . . . . . . . . . . . . . . . . . . . . . .
Romeo et al. (1996) . . . . . . . . . . . . . . . . . . . . . . .
Sarma & Srbulov (1996) . . . . . . . . . . . . . . . . . . . . .
Singh et al. (1996) . . . . . . . . . . . . . . . . . . . . . . . .
Spudich et al. (1996) & Spudich et al. (1997) . . . . . . . . . .
Stamatovska & Petrovski (1996) . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
72
72
73
73
74
74
75
76
76
77
78
79
79
80
80
81
81
81
82
82
83
84
84
85
85
86
86
87
88
88
89
90
92
92
93
94
95
95
95
96
97
98
98
99
99
100
100
101
7
2.132 Campbell (1997), Campbell (2000), Campbell (2001) & Campbell & Bozorgnia
(1994) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.133 Munson & Thurber (1997) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.134 Pancha & Taber (1997) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.135 Rhoades (1997) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.136 Schmidt et al. (1997) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.137 Youngs et al. (1997) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.138 Zhao et al. (1997) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.139 Baag et al. (1998) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.140 Bouhadad et al. (1998) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.141 Costa et al. (1998) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.142 Manic (1998) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.143 Reyes (1998) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.144 Rinaldis et al. (1998) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.145 Sadigh & Egan (1998) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.146 Sarma & Srbulov (1998) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.147 Sharma (1998) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.148 Smit (1998) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.149 Cabaas et al. (1999), Cabaas et al. (2000) & Benito et al. (2000) . . . . . .
2.150 Chapman (1999) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.151 Cousins et al. (1999) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.152 lafsson & Sigbjrnsson (1999) . . . . . . . . . . . . . . . . . . . . . . . . . .
2.153 Si & Midorikawa (1999, 2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.154 Spudich et al. (1999) & Spudich & Boore (2005) . . . . . . . . . . . . . . . . .
2.155 Wang et al. (1999) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.156 Zar et al. (1999) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.157 Ambraseys & Douglas (2000), Douglas (2001b) & Ambraseys & Douglas (2003)
2.158 Bozorgnia et al. (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.159 Campbell & Bozorgnia (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.160 Field (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.161 Jain et al. (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.162 Kobayashi et al. (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.163 Monguilner et al. (2000a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.164 Sharma (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.165 Smit et al. (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.166 Takahashi et al. (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.167 Wang & Tao (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.168 Chang et al. (2001) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.169 Herak et al. (2001) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.170 Lussou et al. (2001) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.171 Sanchez & Jara (2001) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.172 Wu et al. (2001) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.173 Chen & Tsai (2002) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.174 Gregor et al. (2002a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.175 Glkan & Kalkan (2002) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.176 Khademi (2002) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.177 Margaris et al. (2002a) & Margaris et al. (2002b) . . . . . . . . . . . . . . . . .
2.178 Saini et al. (2002) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.179 Schwarz et al. (2002) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
102
103
104
104
105
106
107
108
108
108
109
109
109
110
111
112
112
113
114
114
115
116
117
118
118
119
121
121
122
124
125
126
127
127
128
129
129
130
131
132
132
133
133
134
135
136
137
137
2.180
2.181
2.182
2.183
2.184
2.185
2.186
2.187
2.188
2.189
2.190
2.191
2.192
2.193
2.194
2.195
2.196
2.197
2.198
2.199
2.200
2.201
2.202
2.203
2.204
2.205
2.206
2.207
2.208
2.209
2.210
2.211
2.212
2.213
2.214
2.215
2.216
2.217
2.218
2.219
2.220
2.221
2.222
2.223
2.224
2.225
2.226
2.227
2.228
Stamatovska (2002) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Tromans & Bommer (2002) . . . . . . . . . . . . . . . . . . . . . . . . .
Zonno & Montaldo (2002) . . . . . . . . . . . . . . . . . . . . . . . . . .
Alarcn (2003) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Alchalbi et al. (2003) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Atkinson & Boore (2003) . . . . . . . . . . . . . . . . . . . . . . . . . .
Boatwright et al. (2003) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bommer et al. (2003) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Campbell & Bozorgnia (2003d,a,b,c) & Bozorgnia & Campbell (2004b) .
Halldrsson & Sveinsson (2003) . . . . . . . . . . . . . . . . . . . . . .
Shi & Shen (2003) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sigbjrnsson & Ambraseys (2003) . . . . . . . . . . . . . . . . . . . . .
Skarlatoudis et al. (2003) . . . . . . . . . . . . . . . . . . . . . . . . . .
Beauducel et al. (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Beyaz (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bragato (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Gupta & Gupta (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Kalkan & Glkan (2004a) . . . . . . . . . . . . . . . . . . . . . . . . . .
Kalkan & Glkan (2004b) and Kalkan & Glkan (2005) . . . . . . . . . .
Lubkowski et al. (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Marin et al. (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Midorikawa & Ohtake (2004) . . . . . . . . . . . . . . . . . . . . . . . .
zbey et al. (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Pankow & Pechmann (2004) and Pankow & Pechmann (2006) . . . . . .
Sunuwar et al. (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Skarlatoudis et al. (2004) . . . . . . . . . . . . . . . . . . . . . . . . . .
Ulusay et al. (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ambraseys et al. (2005a) . . . . . . . . . . . . . . . . . . . . . . . . . .
Ambraseys et al. (2005b) . . . . . . . . . . . . . . . . . . . . . . . . . .
Bragato (2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bragato & Slejko (2005) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Frisenda et al. (2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Garca et al. (2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Liu & Tsai (2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
McGarr & Fletcher (2005) . . . . . . . . . . . . . . . . . . . . . . . . . .
Nowroozi (2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ruiz & Saragoni (2005) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Takahashi et al. (2005), Zhao et al. (2006) and Fukushima et al. (2006)
Wald et al. (2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Atkinson (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Beyer & Bommer (2006) . . . . . . . . . . . . . . . . . . . . . . . . . .
Bindi et al. (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Campbell & Bozorgnia (2006a) and Campbell & Bozorgnia (2006b) . . .
Costa et al. (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Gmez-Sobern et al. (2006) . . . . . . . . . . . . . . . . . . . . . . .
Hernandez et al. (2006) . . . . . . . . . . . . . . . . . . . . . . . . . .
Kanno et al. (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Laouami et al. (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Luzi et al. (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
138
138
139
140
140
141
144
145
147
150
150
151
151
153
153
154
155
155
156
157
158
159
160
161
162
162
163
164
167
168
169
171
172
173
174
175
176
176
179
179
181
182
184
186
186
187
187
189
189
9
2.229
2.230
2.231
2.232
2.233
2.234
2.235
2.236
2.237
2.238
2.239
2.240
2.241
2.242
2.243
2.244
2.245
2.246
2.247
2.248
2.249
2.250
2.251
2.252
2.253
2.254
2.255
2.256
2.257
2.258
2.259
2.260
2.261
2.262
2.263
2.264
2.265
2.266
2.267
2.268
2.269
2.270
2.271
2.272
2.273
2.274
2.275
2.276
10
Mahdavian (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
McVerry et al. (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Moss & Der Kiureghian (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Pousse et al. (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Souriau (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Zare & Sabzali (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Akkar & Bommer (2007b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ghodrati Amiri et al. (2007a) & Ghodrati Amiri et al. (2007b) . . . . . . . . . .
Aydan (2007) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bindi et al. (2007) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bommer et al. (2007) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Boore & Atkinson (2007) & Boore & Atkinson (2008) . . . . . . . . . . . . . . .
Campbell & Bozorgnia (2007), Campbell & Bozorgnia (2008b) & Campbell & Bozorgnia (2008a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Danciu & Tselentis (2007a) & Danciu & Tselentis (2007b) . . . . . . . . . . . .
Douglas (2007) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hong & Goda (2007) & Goda & Hong (2008) . . . . . . . . . . . . . . . . . . .
Graizer & Kalkan (2007) & Graizer & Kalkan (2008) . . . . . . . . . . . . . . . .
Massa et al. (2007) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Popescu et al. (2007) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sobhaninejad et al. (2007) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Tavakoli & Pezeshk (2007) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Tejeda-Jcome & Chvez-Garca (2007) . . . . . . . . . . . . . . . . . . . . . .
Abrahamson & Silva (2008) & Abrahamson & Silva (2009) . . . . . . . . . . . .
gstsson et al. (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Aghabarati & Tehranizadeh (2008) . . . . . . . . . . . . . . . . . . . . . . . . .
Cauzzi & Faccioli (2008), Cauzzi (2008) & Cauzzi et al. (2008) . . . . . . . . .
Chiou & Youngs (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Cotton et al. (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Humbert & Viallet (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Idriss (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Lin & Lee (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Massa et al. (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Mezcua et al. (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Morasca et al. (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Slejko et al. (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Srinivasan et al. (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Akyol & Karagz (2009) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bindi et al. (2009a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bindi et al. (2009b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bragato (2009) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hong et al. (2009b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hong et al. (2009a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Kuehn et al. (2009) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Mandal et al. (2009) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Moss (2009) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ptursson & Vogfjrd (2009) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Rupakhety & Sigbjrnsson (2009) . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
190
191
196
196
197
198
198
199
200
200
201
203
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
207
210
211
212
213
214
215
216
217
218
219
222
224
226
228
231
232
233
234
235
237
238
239
239
240
241
242
244
245
247
247
248
249
250
251
253
2.277
2.278
2.279
2.280
2.281
2.282
2.283
2.284
2.285
2.286
2.287
2.288
Akkar & Bommer (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Akkar & agnan

(2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Arroyo et al. (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bindi et al. (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Cua & Heaton (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Douglas & Halldrsson (2010) . . . . . . . . . . . . . . . . . . . . . . . .
Faccioli et al. (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Graizer et al. (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hong & Goda (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Jayaram & Baker (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Montalva (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ornthammarath et al. (2010b), Ornthammarath (2010) & Ornthammarath
(2010a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.289 Ulutas & Ozer (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
et al.
. . .
. . .
.
.
.
.
.
.
.
.
.
.
.
. 265
. 266
3 General characteristics of GMPEs for PGA

4 Summary of published GMPEs for spectral ordinates
4.1
Johnson (1973) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2
McGuire (1974) & McGuire (1977) . . . . . . . . . . . . . . . . . . . . .
4.3
Kobayashi & Nagahashi (1977) . . . . . . . . . . . . . . . . . . . . . . .
4.4
Trifunac (1977) & Trifunac & Anderson (1977) . . . . . . . . . . . . . . .
4.5
Faccioli (1978) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.6
McGuire (1978) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.7
Trifunac (1978) & Trifunac & Anderson (1978a) . . . . . . . . . . . . . .
4.8
Trifunac & Anderson (1978b) . . . . . . . . . . . . . . . . . . . . . . . .
4.9
Cornell et al. (1979) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.10 Faccioli & Agalbato (1979) . . . . . . . . . . . . . . . . . . . . . . . . .
4.11 Trifunac & Lee (1979) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.12 Ohsaki et al. (1980b) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.13 Ohsaki et al. (1980a) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.14 Trifunac (1980) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.15 Devillers & Mohammadioun (1981) . . . . . . . . . . . . . . . . . . . . .
4.16 Joyner & Boore (1982a) . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.17 Joyner & Boore (1982b) . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.18 Kobayashi & Midorikawa (1982) . . . . . . . . . . . . . . . . . . . . . .
4.19 Joyner & Fumal (1984), Joyner & Fumal (1985) & Joyner & Boore (1988)
4.20 Kawashima et al. (1984) . . . . . . . . . . . . . . . . . . . . . . . . . .
4.21 Kawashima et al. (1985) . . . . . . . . . . . . . . . . . . . . . . . . . .
4.22 Trifunac & Lee (1985) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.23 Kamiyama & Yanagisawa (1986) . . . . . . . . . . . . . . . . . . . . . .
4.24 C.B. Crouse (1987) reported in Joyner & Boore (1988) . . . . . . . . . .
4.25 Lee (1987) & Lee (1993) . . . . . . . . . . . . . . . . . . . . . . . . . .
4.26 K. Sadigh (1987) reported in Joyner & Boore (1988) . . . . . . . . . . .
4.27 Annaka & Nozawa (1988) . . . . . . . . . . . . . . . . . . . . . . . . . .
4.28 Crouse et al. (1988) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.29 Petrovski & Marcellini (1988) . . . . . . . . . . . . . . . . . . . . . . . .
4.30 Yokota et al. (1988) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.31 Youngs et al. (1988) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.32 Kamiyama (1989) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
254
255
256
258
259
260
260
262
262
263
264
269
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
297
297
297
297
298
300
300
300
301
302
302
302
303
303
303
304
305
305
305
306
306
306
306
307
308
308
309
309
309
310
310
311
311
11
4.33
4.34
4.35
4.36
4.37
4.38
4.39
4.40
4.41
4.42
4.43
4.44
4.45
4.46
4.47
4.48
4.49
4.50
4.51
4.52
4.53
4.54
4.55
4.56
4.57
4.58
4.59
4.60
4.61
4.62
4.63
4.64
4.65
4.66
4.67
4.68
4.69
4.70
4.71
4.72
4.73
4.74
4.75
4.76
4.77
4.78
4.79
4.80
4.81
12
Trifunac & Lee (1989) . . . . . . . . . . . . . . . . . . . . . . . . .

Atkinson (1990) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Campbell (1990) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Dahle et al. (1990b) & Dahle et al. (1990a) . . . . . . . . . . . . .
Tamura et al. (1990) . . . . . . . . . . . . . . . . . . . . . . . . .
Tsai et al. (1990) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Crouse (1991) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Dahle et al. (1991) . . . . . . . . . . . . . . . . . . . . . . . . . .
Geomatrix Consultants (1991), Sadigh et al. (1993) & Sadigh et al.
I.M. Idriss (1991) reported in Idriss (1993) . . . . . . . . . . . . . .
Loh et al. (1991) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Matuschka & Davis (1991) . . . . . . . . . . . . . . . . . . . . . .
Mohammadioun (1991) . . . . . . . . . . . . . . . . . . . . . . . .
Stamatovska & Petrovski (1991) . . . . . . . . . . . . . . . . . . .
Benito et al. (1992) . . . . . . . . . . . . . . . . . . . . . . . . . .
Niazi & Bozorgnia (1992) . . . . . . . . . . . . . . . . . . . . . . .
Silva & Abrahamson (1992) . . . . . . . . . . . . . . . . . . . . . .
Tento et al. (1992) . . . . . . . . . . . . . . . . . . . . . . . . . .
Abrahamson & Silva (1993) . . . . . . . . . . . . . . . . . . . . . .
Boore et al. (1993) & Boore et al. (1997) . . . . . . . . . . . . . .
Caillot & Bard (1993) . . . . . . . . . . . . . . . . . . . . . . . . .
Campbell (1993) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Electric Power Research Institute (1993a) . . . . . . . . . . . . . .
Sun & Peng (1993) . . . . . . . . . . . . . . . . . . . . . . . . . .
Boore et al. (1994a), Boore et al. (1997) & Boore (2005) . . . . .
Climent et al. (1994) . . . . . . . . . . . . . . . . . . . . . . . . .
Fukushima et al. (1994) & Fukushima et al. (1995) . . . . . . . . .
Lawson & Krawinkler (1994) . . . . . . . . . . . . . . . . . . . . .
Lee & Manic (1994) & Lee (1995) . . . . . . . . . . . . . . . . . .
Mohammadioun (1994a) . . . . . . . . . . . . . . . . . . . . . . .
Mohammadioun (1994b) . . . . . . . . . . . . . . . . . . . . . . .
Musson et al. (1994) . . . . . . . . . . . . . . . . . . . . . . . . .
Theodulidis & Papazachos (1994) . . . . . . . . . . . . . . . . . .
Dahle et al. (1995) . . . . . . . . . . . . . . . . . . . . . . . . . .
Lee & Trifunac (1995) . . . . . . . . . . . . . . . . . . . . . . . . .
Ambraseys et al. (1996) & Simpson (1996) . . . . . . . . . . . . .
Ambraseys & Simpson (1996) & Simpson (1996) . . . . . . . . . .
Bommer et al. (1996) . . . . . . . . . . . . . . . . . . . . . . . . .
Crouse & McGuire (1996) . . . . . . . . . . . . . . . . . . . . . . .
Free (1996) & Free et al. (1998) . . . . . . . . . . . . . . . . . . .
Molas & Yamazaki (1996) . . . . . . . . . . . . . . . . . . . . . . .
Ohno et al. (1996) . . . . . . . . . . . . . . . . . . . . . . . . . .
Sabetta & Pugliese (1996) . . . . . . . . . . . . . . . . . . . . . .
Spudich et al. (1996) & Spudich et al. (1997) . . . . . . . . . . . .
Abrahamson & Silva (1997) . . . . . . . . . . . . . . . . . . . . . .
Atkinson (1997) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Campbell (1997), Campbell (2000) & Campbell (2001) . . . . . . .
Schmidt et al. (1997) . . . . . . . . . . . . . . . . . . . . . . . . .
Youngs et al. (1997) . . . . . . . . . . . . . . . . . . . . . . . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
(1997)
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
312
312
313
313
313
314
314
315
316
317
317
317
317
317
318
318
319
319
320
321
321
322
322
323
323
323
324
324
324
325
325
326
326
327
327
327
327
328
328
328
328
328
329
329
330
331
332
333
333
4.82
4.83
4.84
4.85
4.86
4.87
4.88
4.89
4.90
4.91
4.92
4.93
4.94
4.95
4.96
4.97
4.98
4.99
4.100
4.101
4.102
4.103
4.104
4.105
4.106
4.107
4.108
4.109
4.110
4.111
4.112
4.113
4.114
4.115
4.116
4.117
4.118
4.119
4.120
4.121
4.122
4.123
4.124
4.125
4.126
4.127
4.128
4.129
4.130
Bommer et al. (1998) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Perea & Sordo (1998) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Reyes (1998) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Shabestari & Yamazaki (1998) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapman (1999) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Spudich et al. (1999) & Spudich & Boore (2005) . . . . . . . . . . . . . . . . .
Ambraseys & Douglas (2000), Douglas (2001b) & Ambraseys & Douglas (2003)
Bozorgnia et al. (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Campbell & Bozorgnia (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chou & Uang (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Field (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Kawano et al. (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Kobayashi et al. (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
McVerry et al. (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Monguilner et al. (2000b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Shabestari & Yamazaki (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Smit et al. (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Takahashi et al. (2000) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Lussou et al. (2001) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Das et al. (2002) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Glkan & Kalkan (2002) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Khademi (2002) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Manic (2002) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Schwarz et al. (2002) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Zonno & Montaldo (2002) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Alarcn (2003) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Atkinson & Boore (2003) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Berge-Thierry et al. (2003) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bommer et al. (2003) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Campbell & Bozorgnia (2003d,a,b,c) & Bozorgnia & Campbell (2004b) . . . . .
Fukushima et al. (2003) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Kalkan & Glkan (2004a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Kalkan & Glkan (2004b) and Kalkan & Glkan (2005) . . . . . . . . . . . . . .
Matsumoto et al. (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
zbey et al. (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Pankow & Pechmann (2004) and Pankow & Pechmann (2006) . . . . . . . . . .
Sunuwar et al. (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Takahashi et al. (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Yu & Hu (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ambraseys et al. (2005a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ambraseys et al. (2005b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bragato & Slejko (2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Garca et al. (2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
McGarr & Fletcher (2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Pousse et al. (2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Takahashi et al. (2005), Zhao et al. (2006) and Fukushima et al. (2006) . . . .
Wald et al. (2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Atkinson (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Beyer & Bommer (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
333
334
335
335
335
335
336
336
336
336
337
337
338
338
339
340
341
341
341
341
342
342
342
342
342
343
343
343
345
345
345
346
346
346
347
347
347
348
349
350
351
351
351
351
352
353
353
353
354
13
4.131
4.132
4.133
4.134
4.135
4.136
4.137
4.138
4.139
4.140
4.141
4.142
4.143
4.144
4.145
4.146
4.147
4.148
4.149
4.150
4.151
4.152
4.153
4.154
4.155
4.156
4.157
4.158
4.159
4.160
4.161
4.162
4.163
4.164
4.165
4.166
4.167
4.168
4.169
4.170
4.171
4.172
4.173
4.174
4.175
4.176
4.177
4.178
14
Bindi et al. (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Campbell & Bozorgnia (2006a) and Campbell & Bozorgnia (2006b) . . . . . . .
Hernandez et al. (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Kanno et al. (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
McVerry et al. (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Pousse et al. (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sakamoto et al. (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sharma & Bungum (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Zare & Sabzali (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Akkar & Bommer (2007b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bindi et al. (2007) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bommer et al. (2007) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Boore & Atkinson (2007) & Boore & Atkinson (2008) . . . . . . . . . . . . . . .
Campbell & Bozorgnia (2007), Campbell & Bozorgnia (2008b) & Campbell & Bozorgnia (2008a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Danciu & Tselentis (2007a) & Danciu & Tselentis (2007b) . . . . . . . . . . . .
Fukushima et al. (2007b) & Fukushima et al. (2007a) . . . . . . . . . . . . . .
Hong & Goda (2007) & Goda & Hong (2008) . . . . . . . . . . . . . . . . . . .
Massa et al. (2007) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Tejeda-Jcome & Chvez-Garca (2007) . . . . . . . . . . . . . . . . . . . . . .
Abrahamson & Silva (2008) & Abrahamson & Silva (2009) . . . . . . . . . . . .
Cauzzi & Faccioli (2008), Cauzzi (2008) & Cauzzi et al. (2008) . . . . . . . . .
Chen & Yu (2008b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chen & Yu (2008a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chiou & Youngs (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Cotton et al. (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Dhakal et al. (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Idriss (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Lin & Lee (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Massa et al. (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Morasca et al. (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Yuzawa & Kudo (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Akyol & Karagz (2009) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bindi et al. (2009a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bindi et al. (2009b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bragato (2009) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ghasemi et al. (2009) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hong et al. (2009b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hong et al. (2009a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Kuehn et al. (2009) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Moss (2009) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Rodriguez-Marek & Montalva (2010) . . . . . . . . . . . . . . . . . . . . . . . .
Rupakhety & Sigbjrnsson (2009) . . . . . . . . . . . . . . . . . . . . . . . . .
Sharma et al. (2009) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Akkar & Bommer (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Akkar & agnan

(2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ghodrati Amiri et al. (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
354
354
354
354
355
355
355
356
357
357
358
358
358
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
359
359
359
360
361
361
361
361
361
362
362
362
362
363
363
364
364
364
364
364
365
365
365
365
365
366
366
367
367
367
368
368
369
370
370
4.179
4.180
4.181
4.182
4.183
4.184
4.185
4.186
Arroyo et al. (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Bindi et al. (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Douglas & Halldrsson (2010) . . . . . . . . . . . . . . . . . . . . . . . .
Faccioli et al. (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hong & Goda (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Jayaram & Baker (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . .
Montalva (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ornthammarath et al. (2010b), Ornthammarath (2010) & Ornthammarath
(2010a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.187 Sadeghi et al. (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.188 Saffari et al. (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
et al.
. . .
. . .
. . .
.
.
.
.
.
.
.
371
371
371
371
371
372
372
. 372
. 372
. 373
5 General characteristics of GMPEs for spectral ordinates
375
6 List of other ground-motion models
397
15
16
Chapter 1
Introduction
ESEE Report 01-1 A comprehensive worldwide summary of strong-motion attenuation relationships for peak ground acceleration and spectral ordinates (1969 to 2000) (Douglas,
2001a) was completed and released in January 2001. A report detailing errata of this report
and additional studies was released in October 2002 (Douglas, 2002). These two reports were
used by Douglas (2003) as a basis for a review of previous ground-motion prediction equations (GMPEs). Following the release of these two reports, some further minor errors were
found in the text and tables of the original two reports, and additional studies were found in the
literature that were not included in ESEE 01-1 or the follow-on report. Also some new studies
were published. Rather than produce another report listing errata and additions it was decided
to produce a new report that included details on all the studies listed in the rst two reports
(with the corrections made) and also information on the additional studies. This report was
published as a research report of Imperial College London at the beginning of 2004 (Douglas,
2004a). At the end of 2006 a BRGM report was published (Douglas, 2006) detailing studies
published in 20042006 plus a few earlier models that had been missed in previous reports.
Finally, at the end of 2008 another BRGM report was published (Douglas, 2008) containing
summaries of GMPEs from 2007 and 2008 and some additional earlier models that had been
recently uncovered.
Because of the large number of new GMPEs published in the past couple of years and
the discovery of some additional earlier studies and various errors in the previous reports, it
was decided to publish a new comprehensive report to replace the previous reports (Douglas,
2001a, 2002, 2004a, 2006, 2008) containing all previous reports plus additional material rather
than publish yet another addendum to the 2004 report. It was also decided that, for completeness and due to the lack of another comprehensive and public source for this information, to
include a list of GMPEs developed using other methods than regression of strong-motion data,
e.g. simulation-based models (e.g. Douglas & Aochi, 2008). However, due to the complexity
of briey summarizing these models it was decided not to provide details here but only references a public report on these models may be published later. Douglas (2007) compares
predicted response spectra from many of the stochastic models listed.
This report summarizes, in total, the characteristics of 289 empirical GMPEs for the prediction of peak ground acceleration (PGA) and 188 models for the prediction of elastic response
spectral ordinates. With this many ground-motion estimation equations available it is important
to have criteria available for the selection of appropriate models for seismic hazard assessment in a given region Cotton et al. (2006) and, more recently, Bommer et al. (2010)
suggest selection requirements for the choice of models. For the selection of GMPEs routinely applicable to state-of-the-art hazard analyses of ground motions from shallow crustal
earthquakes Bommer et al. (2010) summarize their criteria thus.
17
1. Model is derived for an inappropriate tectonic environment (such as subduction-zone

earthquakes or volcanic regions).
2. Model not published in a Thomson Reuters ISI-listed peer-reviewed journal (although an
exception can be made for an update to a model that did meet this criterion).
3. The dataset used to derive the model is not presented in an accessible format; the
minimum requirement would be a table listing the earthquakes and their characteristics,
together with the number of records from each event.
4. The model has been superseded by a more recent publication.
5. The model does not provide spectral predictions for an adequate range of response
periods, chosen here to be from 0 to 2 s.
6. The functional form lacks either non-linear magnitude dependence or magnitude-dependent
decay with distance.
7. The coefcients of the model were not determined with a method that accounts for interevent and intra-event components of variability; in other words, models must be derived using one- or two-stage maximum likelihood approaches or the random effects
approach.
8. Model uses inappropriate denitions for explanatory variables, such as ML or repi , or
models site effects without consideration of Vs,30 .
9. The range of applicability of the model is too small to be useful for the extrapolations
generally required in PSHA: Mmin > 5, Mmax < 7, Rmax < 80 km.
10. Model constrained with insufciently large dataset: fewer than 10 earthquakes per unit
of magnitude or fewer than 100 records per 100 km of distance.
Similar criteria could be developed for other types of earthquakes (e.g. subduction). Application of these criteria would lead to a much reduced set of models. The aim of this report,
however, is not to apply these, or any other, criteria but simply to summarize all models that
have been published. Bommer et al. (2010) also note that: [i]f one accepts the general approach presented in this paper, then it becomes inappropriate to develop and publish GMPEs
that would subsequently be excluded from use in PSHA [probabilistic seismic hazard analysis]
on the basis of not satisfying one or more of the requirements embodied in the criteria.
Predictions of median ground motions from GMPEs show great dispersion (Douglas, 2010a,b)
demonstrating the large epistemic uncertainties involved in the estimation of earthquake shaking. This uncertainty should be accounted for within seismic hazard assessments by, for example, logic trees (e.g. Bommer & Scherbaum, 2008).
1.1 Other summaries and reviews of GMPEs

A number of reviews of GMPEs have been made in the past that provide a good summary
of the methods used, the results obtained and the problems associated with such relations.
Trifunac & Brady (1975, 1976) provide a brief summary and comparison of published relations. McGuire (1976) lists numerous early relations. Idriss (1978) presents a comprehensive
review of published attenuation relations up until 1978, including a number which are not easily available elsewhere. Hays (1980) presents a good summary of ground-motion estimation
18
procedures up to 1980. Boore & Joyner (1982) provide a review of attenuation studies published in 1981 and they comment on empirical prediction of strong ground motion in general.
Campbell (1985) contains a full survey of attenuation equations up until 1985. Joyner & Boore
(1988) give an excellent analysis of ground motion prediction methodology in general, and
attenuation relations in particular; Joyner & Boore (1996) update this by including more recent studies. Ambraseys & Bommer (1995) provide an overview of relations that are used for
seismic design in Europe although they do not provide details about methods used. Recent reviews include those by Campbell (2003c,a) and Bozorgnia & Campbell (2004a), which provide
the coefcients for a number of commonly-used equations for peak ground acceleration and
spectral ordinates, and Douglas (2003). Bommer (2006) discusses some pressing problems
in the eld of empirical ground-motion estimation.
Summaries and reviews of published ground-motion models for the estimation of strongmotion parameters other than PGA and elastic response spectral ordinates are available1 .
For example: Bommer & Martnez-Pereira (1999), Alarcn (2007) and Bommer et al. (2009)
review predictive equations for strong-motion duration; Tromans (2004) summarizes equations
for the prediction of PGV and displacement (PGD); Bommer & Alarcn (2006) provide a more
recent review of GMPEs for PGV; Hancock & Bommer (2005) discuss available equations for
estimating number of effective cycles; Stafford et al. (2009) briey review GMPEs for Arias
intensity; and Rathje et al. (2004) summarize the few equations published for the prediction
of frequency-content parameters (e.g. predominant frequency).
1.2
GMPEs summarised here
Equations for single earthquakes (e.g. Bozorgnia et al. , 1995) or for earthquakes of approximately the same size (e.g. Seed et al. , 1976; Sadigh et al. , 1978) are excluded because
they lack a magnitude-scaling term and, hence, are of limited use. Also excluded are those
originally developed to yield the magnitude of an earthquake (e.g. Espinosa, 1980), i.e. the
regression is performed the other way round, which should not be used for the prediction of
ground motion at a site. Models such as that by Olszewska (2006), who uses source energy logarithms to characterize mining-induced events, have been excluded because such
a characterization of event size is rare in standard seismic hazard assessments. Similarly,
equations derived using data from nuclear tests, such as those reported by Hays (1980), are
not included. Those based on simulated ground motions from stochastic source models (e.g
Atkinson & Boore, 1990) and other types of simulations (e.g. Megawati et al. , 2005), those
derived using the hybrid empirical technique (e.g Campbell, 2003b; Douglas et al. , 2006) and
those relations based on intensity measurements (e.g. Battis, 1981) are listed in Chapter 6
but no details are given because the focus here is on empirical models derived from groundmotion data. Studies which provide graphs to give predictions (e.g. Schnabel & Seed, 1973)
are only listed and not summarized as are those nonparametric formulations that give predictions for different combinations of distance and magnitude (e.g. Anderson, 1997), both of
which are more difcult to use for seismic hazard analysis than those which report a single
formula. For similar reasons, models derived using neural networks (e.g. Gll & Erelebi,
2007) are only listed.
GMPEs for the prediction of PGA are summarized in Chapters 2 and 3 and those for
spectral ordinates are summarized in Chapters 4 and 5. The nal chapter (Chapter 6) lists
other ground-motion models that are not detailed in the previous chapters. All the studies that
1
Note that a number of the models summarized in this report also provide coefcients for peak ground velocity
(PGV).
19
present the same GMPE are mentioned at the top of the section and in the tables of general
characteristics (Illustrations 1 & 2). The information contained within each section, and within
tables, is the sum of information contained within each of the publications, i.e. not all the information may be from a single source. Note that GMPEs are ordered in chronological order both
in the section titles and the order of the sections. Therefore, a well-known model presented in
a journal article may not be listed where expected since it had previously been published in a
conference proceedings or technical report. To nd a given model it is recommended to examine the table of content carefully or apply a keyword search to the PDF. Some models (e.g.
Abrahamson & Silva, 1997) provide GMPEs for spectral accelerations up to high frequencies
(e.g. 100 Hz) but do not explicitly state that these equations can be used for the prediction
of PGA. Therefore, they are only listed in the chapters dealing with GMPEs for the prediction
of spectral ordinates (Chapters 4 and 5). This should be considered when searching for a
particular model.
To make it easier to understand the functional form of each GMPE the equations are given
with variable names replacing actual coefcients and the derived coefcients and the standard
deviation, , are given separately (for PGA equations). These coefcients are given only for
completeness and if an equation is to be used then the original reference should be consulted.
If a coefcient is assumed before the analysis is performed then the number is included directly
in the formula.
Obviously all the details from each publication cannot be included in this report because
of lack of space but the most important details of the methods and data used are retained.
The number of records within each site and source mechanism category are given if this
information was reported by the authors of the study. Sometimes these totals were found by
counting the numbers in each category using the tables listing the data used and, therefore,
they may be inaccurate.
This report contains details of all studies for PGA and response spectra that could be found
in the literature (journals, conference proceedings, technical reports and some Ph.D. theses)
although some may have been inadvertently missed2 . Some of the studies included here have
not been seen but are reported in other publications and hence the information given here may
not be complete or correct. Since this report has been written in many distinct periods over
a decade (20002010), the amount of information given for each model varies, as does the
style.
In the equations unless otherwise stated, D, d, R, r, or similar are distance and M
or similar is magnitude and all other independent variables are stated. PGA is peak ground
acceleration, PGV is peak ground velocity and PSV is relative pseudo-velocity.
In Illustrations 1 & 2 the gross characteristics of the data used and equation obtained are
only given for the main equation in each study. The reader should refer to the section on a
particular publication for information on other equations derived in the study.
In earlier reports the name attenuation relation(ships) is used for the models reported.
The current de facto standard is to refer to such models as ground-motion prediction equations (GMPEs) and, therefore, this terminology is adopted here. However, as discussed by
Boore & Atkinson (2007, Appendix A) there is some debate over the best name for these
models (e.g. ground-motion model or ground-motion estimation equations) and some people disagree with the use of the word prediction in this context.
No discussion of the merits, ranges of applicability or limitations of any of the relationships
is included herein except those mentioned by the authors or inherent in the data used. This
report is not a critical review of the models. The ground-motion models are reported in the
2
Generally GMPEs from technical reports and Ph.D. theses are only summarized if they have been cited in
journal or conference articles.
20
form given in the original references except sometimes the equation is simplied if this can
be easily done. Note that the size of this report means that it may contain some errors or
omissions the reader is encouraged to consult the original reference if a model is to be
used.
21
22
Chapter 2
Summary of published GMPEs for

PGA
2.1
Esteva & Rosenblueth (1964)
Ground-motion model is:
a = c exp(M )R
where a is in cm/s2 , c = 2000, = 0.8 and = 2 ( is not given).
2.2 Kanai (1966)

a
1 10a2 M P log10 R+Q
TG
= a3 + a4 /R
a =
P
Q = a5 + a6 /R
where a is in cm/s2 , a1 = 5, a2 = 0.61, a3 = 1.66, a4 = 3.60, a5 = 0.167 and
a6 = 1.83 ( is not given).
TG is the fundamental period of the site.
2.3
Milne & Davenport (1969)
A=
a1 ea2 M
a3 ea4 M + 2
where A is in percentage of g, a1 = 0.69, a2 = 1.64, a3 = 1.1 and a4 = 1.10 ( not

given).
Use data from Esteva & Rosenblueth (1964).
23
2.4 Esteva (1970)

a = c1 ec2 M (R + c3 )c4
where a is in cm/s2 , c1 = 1230, c2 = 0.8, c3 = 25, c4 = 2 and = 1.02 (in terms of
natural logarithm).
Records from soils comparable to stiff clay or compact conglomerate.
Records from earthquakes of moderate duration.
2.5 Denham & Small (1971)

log Y = b1 + b2 M + b3 log R
where Y is in g, b1 = 0.2, b2 = 0.2 and b3 = 1.1 ( not given).
Records from near dam on recent unconsolidated lake sediments which are 50 m
thick.
Note need for more points and large uncertainty in b1 , b2 and b3 .
2.6 Davenport (1972)

A = em R
where A is in g, = 0.279, = 0.80, = 1.64 and = 0.74 (in terms of natural
logarithms).
2.7 Donovan (1973)

y = b1 eb2 M (R + 25)b3
where y is in gal, b1 = 1080, b2 = 0.5, b3 = 1.32 and = 0.71. 25 adopted from Esteva
(1970).
214 (32%) records from San Fernando (9/2/1971) earthquake and 53% of records with
PGA less than 0.5 m/s2 .
Considers portions of data and nds magnitude dependence increases with increasing
distance from source and more small accelerations increase magnitude dependence.
Thus magnitude and distance cannot be considered independent variables.
24
2.8
Denham et al. (1973)
log Ya = a1 + a2 ML + b3 log R
where Ya is in cm/s2 , a1 = 2.91, a2 = 0.32 and a3 = 1.45 ( is not given).
Use records from Yonki station (20 records) which is on 50 m of recent alluvium and
from Paguna station (5 records) which is on unconsolidated volcanic rock.
Question validity of combining data at the two sites because of differences in geological
foundations.
Note large standard errors associated with coefcients preclude accurate predictions of
ground motions.
Also derive equation for Yonki station separately.
2.9
Esteva & Villaverde (1973) & Esteva (1974)
Yc = b1 eb2 M (R + b4 )b3
where Yc is in cm/s2 , b1 = 5600, b2 = 0.8, b3 = 2, b4 = 40 and = 0.64 (in terms of
natural logarithm).
2.10
McGuire (1974) & McGuire (1977)
E[v] = a10bM (R + 25)c

where E indicates expectation, v is in gal, a = 472, b = 0.278, c = 1.301.
Excludes records for which signicant soil amplication established but makes no distinction between rock and soil sites.
Focal depths between 9 and 70 km with most about 10 km. Most records from earthquakes with magnitudes about 6.5 and most distances less than 50 km. Uses records
from 21 different sites.
Notes that physical laws governing ground motion near the source are different than
those governing motion at greater distances therefore excludes records with epicentral
distance or distance to fault rupture smaller than one-half of estimated length of rupture.
Examines correlation among the records but nd negligible effect.
25
2.11 Orphal & Lahoud (1974)

A = 10M R
where A is in g, = 6.6102 , = 0.40, = 1.39 and = 1.99 (this is multiplication
factor).
Use 113 records with distances between 15 to 350 km from San Fernando earthquake
to nd distance dependence, .
Use 27 records of Wiggins Jr. (1964) from El Centro and Ferndale areas, with magnitudes between 4.1 and 7.0 and distances between 17 and 94 km (assuming focal depth
of 15 km), to compute magnitude dependent terms assuming distance dependence is
same as for San Fernando.
2.12 Ambraseys (1975b), Ambraseys (1975a) & Ambraseys (1978a)

log Y = b1 + b2 ML + b3 log R
where Y is in cm/s2 , b1 = 0.46, b2 = 0.63, b3 = 1.10 and = 0.321
Ambraseys & Bommer (1995) state that uses earthquakes with maximum focal depth of
15 km.
2.13 Trifunac & Brady (1975), Trifunac (1976) & Trifunac & Brady
(1976)
log10 amax = M + log10 A0 (R) log10 a0 (M, p, s, v)
ap + bM + c + ds + ev + f M 2 f (M Mmax )2
for M Mmax
ap + bM + c + ds + ev + f M 2
log10 a0 (M, p, s, v) =
for Mmax M Mmin
ap
+
bM
min + c + ds + ev + f Mmin
for M Mmin
where amax is in cm/s2 , log10 A0 (R) is an empirically determined attenuation function
from Richter (1958) used for calculation of ML , p is condence level and v is component
direction (v = 0 for horizontal and 1 for vertical). Coefcients are: a = 0.898, b =
1.789, c = 6.217, d = 0.060, e = 0.331, f = 0.186, Mmin = 4.80 and Mmax = 7.50
(log10 A0 (R) not given here due to lack of space).
Use three site categories:
s = 0 Alluvium or other low velocity soft deposits: 63% of records.

1
26
From Ambraseys & Bommer (1995).
s = 1 Intermediate type rock: 23% of records.

s = 2 Solid hard basement rock: 8% of records.
Exclude records from tall buildings.
Do not use data from other regions because attenuation varies with geological province
and magnitude determination is different in other countries.
Records baseline and instrument corrected. Accelerations thought to be accurate between 0.07 and 25 Hz or between 0.125 and 25 Hz for San Fernando records.
Most records (71%) from earthquakes with magnitudes between 6.06.9, 22% are from
5.05.9, 3% are from 4.04.9 and 3% are from 7.07.7 (note barely adequate data from
these two magnitude ranges). 63% of data from San Fernando earthquake.
Note that for large earthquakes, i.e. long faults, log10 A0 (R) would have a tendency
to atten out for small epicentral distances and for low magnitude shocks curve would
probably have a large negative slope. Due to lack of data . 20 km this is impossible to
check.
Note difculty in incorporating anelastic attenuation because representative frequency
content of peak amplitudes change with distance and because relative contribution of
digitization noise varies with frequency and distance.
Note that log10 A0 (R) may be unreliable for epicentral distances less than 10 km because of lack of data.
Change of slope in log10 A0 (R) at R = 75 km because for greater distances main
contribution to strong shaking from surface waves, which are attenuated less rapidly
( 1/R1/2 ) than near-eld and intermediate-eld ( 1/R24 ), or far-eld body waves
( 1/R).
Note lack of data to reliably characterise log10 a0 (M, p, s, v) over a sufciently broad
range of their arguments. Also note high proportion of San Fernando data may bias
results.
Firstly partition data into four magnitude dependent groups: 4.04.9, 5.05.9, 6.06.9
and 7.07.9. Subdivide each group into three site condition subgroups (for s = 0, 1 and
2). Divide each subgroup into two component categories (for v = 0 and 1). Calculate
log10 a0 (M, p, s, v) = M +log10 A0 (R)log10 amax within each of the 24 parts. Arrange
each set of n log10 a0 values into decreasing order with increasing n. Then mth data
point (where m equals integer part of pn) is estimate for upper bound of log10 a0 for p%
condence level. Then t results using least squares to nd a, . . . f .
Check number of PGA values less than condence level for p = 0.1, . . . , 0.9 to verify adequacy of bound. Find simplifying assumptions are acceptable for derivation of
approximate bounds.
2.14
Blume (1977)
a = b1 eb2 ML (R + 25)b3
27
where a is in gal, for ML 6 12 b1 = 0.318291.14b , b2 = 1.03, b3 = 1.14b and = 0.930
(in terms of natural logarithm) and for ML > 6 12 b1 = 26.0 291.22b , b2 = 0.432,
b3 = 1.22b and = 0.592 (in terms of natural logarithm).
Assumes all earthquakes have focal depth of 8 km.
Makes no distinction for site conditions in rst stage where uses only earthquake records.
Studies effects of PGA cutoff (no cutoff, 0.01, 0.02 and 0.05 m/s2 ), distance cutoff (no
cutoff and < 150 km) and magnitude cutoff (all, 5 21 , 6, 6 12 , 6 34 and 6 12 ).
Selects 6 12 as optimum magnitude cutoff but uses all data to derive equation for ML
6 12 because not much difference and dispersion is slightly lower (in terms of 1 standard
deviation have 2.53 and 2.61).
In second stage uses only records from underground nuclear explosions, consistent with
natural earthquake records, to derive site factor.
Uses 1911 alluvium and 802 rock records and derive PGA ratio of alluvium to rock
assuming their PGAs equal at 4 km.
Finds site impedance Vs , where is density and Vs is shear-wave velocity under site,
is best measure of site condition. Use 2000 fps (600 m/s) as shear-wave velocity of
alluvium stations.
Multiplies equation (after taking logarithms) by b =
1
2
log10 (Vs ) and normalise to 4 km.
Notes may not be a good model for other regions.
2.15 Milne (1977)

ACC = a1 ea2 M Ra3

where ACC is in g, a1 = 0.04, a2 = 1.00 and a3 = 1.4.
2.16 Ambraseys (1978b)

a2 exp(a3 M
)
a
= a1 R
where a
is in cm/s2 , a1 = 1.31, a2 = 0.92 and a3 = 1.455 ( is not given).
Uses data from former USSR, former Yugoslavia, Portugal, Italy, Iran, Greece and Pakistan.
Peak ground accelerations have either been taken from true-to-scale accelerograms or
have been supplied by local networks. Records have not been high- or low-pass ltered
because it was found not to work with short records.
Believes body-wave or local magnitude are the appropriate magnitude scales because
interested in the high-frequency range of spectra, which are seen and sampled by
strong-motion instruments, and most engineering structures have high natural frequencies.
28
Most of the magnitudes were recalculated using P-waves of periods of not more than
1.2 s because it was found that the magnitude was dependent on the period of the Pwaves used for its determination.
Groups data into intervals of 0.5 magnitude units by 10 km in which the mean and
standard deviations of the PGAs is calculated. This grouping minimises distance and
magnitude-dependent effects. Notes that the number of observations is barely sufcient
to allow a statistical treatment of the data and hence only test general trend. Notes that
scatter is signicant and decreases with increasing magnitude.
2.17
Donovan & Bornstein (1978)
y = b1 eb2 M (R + 25)b3
where
b1 = c1 Rc2
b2 = d1 + d2 log R
b3 = e1 + e2 log R
where y is in gal, c1 = 2, 154, 000, c2 = 2.10, d1 = 0.046, d2 = 0.445, e1 = 2.515,

e2 = 0.486, for y = 0.01 g = 0.5, for y = 0.05 g = 0.48, for y = 0.10 g = 0.46
and for y = 0.15 g = 0.41 (in terms of natural logarithm).
Use 25 because assume energy centre of Californian earthquakes to be at depth 5 km.
Consider two site conditions but do not model:
1. Rock: (21 records)
2. Stiff soil: (38 records)
32% of records from San Fernando (9/2/1971) but veries that relationship is not signicantly biased by this data.
Most records within 50 km and most from earthquakes with magnitudes of about 6.5.
Recognises that magnitude and distance are not independent variables.
Find b1 , b2 and b3 by dividing data according to distance and computing b parameters
for each set using least squares. Find a distinct trend with little scatter.
2.18
Faccioli (1978)
y = a10bM (R + 25)c
where y is in gal, a = 108.60, b = 0.265, c = 0.808 and = 0.236 (in terms of logarithm
to base 10).
Records from sites underlain by cohesive or cohesionless soils with shear-wave velocities less than about 100 m/s and/or standard penetration resistance N 10 in uppermost 10 m with layers of considerably stiffer materials either immediately below or at
depths not exceeding a few tens of metres.
29
Focal depths between 9 and 100 km.

Free-eld accelerograms, to minimize soil-structure interaction.
Excludes records with PGA < 0.4 m/s2 .
21 Japanese records processed with frequency cutoffs of bandpass lter, for baseline
correction, adjusted so as to account for length and mean sampling rate of records and
response characteristics of SMAC-2. 4 of remaining 7 records processed in same way.
2.19 McGuire (1978)

ln x = b1 + b2 M + b3 ln R + b4 Ys
where x is in cm/s2 , b1 = 3.40, b2 = 0.89, b3 = 1.17, b4 = 0.20 and = 0.62.
Uses two site categories:
Ys = 0 Rock: sedimentary or basement rock or soil less than 10 m thick, 11 records.

Ys = 1 Soil: alluvium or other soft material greater than 10 m thick, 59 records.
Uses records from basement of buildings or from free-eld. Uses no more than seven
records from same earthquake and no more than nine from a single site to minimize
underestimation of calculated variance. Retains records which give a large distance
and magnitude range.
Notes that near-eld ground motion governed by different physical laws than intermediate and far eld so excludes near-eld data, for example El Centro (19/5/1940) and
Cholame-2, from Parkeld earthquake (28/6/1966)
Considers a distance dependent site term but not statistically signicant. Also uses
a magnitude dependent site term and although it was statistically signicant it did not
reduce the scatter and also since largest magnitude for a rock site is 6.5, result may be
biased.
2.20 A. Patwardhan, K. Sadigh, I.M. Idriss, R. Youngs (1978) reported in Idriss (1978)
ln y = ln A + BMs + E ln[R + d exp(f Ms )]

where y is in cm/s2 , d = 0.864 and f = 0.463 and for path A (rock): A = 157 (for
median), A = 186 (for mean), B = 1.04 and E = 1.90, for path A (stiff soil): A = 191
(for median), A = 224 (for mean), B = 0.823 and E = 1.56 and for path B (stiff soil):
A = 284 (for median), A = 363 (for mean), B = 0.587 and E = 1.05 ( not given).
Separate equations for two types of path:
A Shallow focus earthquakes (California, Japan, Nicaragua and India), 63 records.
30
B Subduction (Benioff) zone earthquakes (Japan and South America), 23 earthquakes, 5.3 Ms 7.8, 32 records.
Use two site categories for path A earthquakes for which derive separate equations:
1. Rock: 21 records.
2. Stiff soil: 42 records.
Use only stiff soil records for deriving subduction zone equation.
Most earthquakes for path A have 5 Ms 6.7.
All data corrected. PGA for corrected Japanese and South American records much
higher than uncorrected PGA.
2.21
Cornell et al. (1979)
ln Ap = a + bML + c ln(R + 25)

where Ap is in cm/s2 , a = 6.74, b = 0.859, c = 1.80 and = 0.57.
No more than 7 records from one earthquake to avoid biasing results.
Records from basements of buildings or free-eld.
2.22
Faccioli (1979)
log y = b1 + b2 M + b3 log(R + 25)

where y is in cm/s2 , b1 = 0.44, b2 = 0.33, b3 = 2.66 and = 0.12.
Uses data from three sedimentary rock sites (Somplago, San Rocco and Robic) because aim of study to provide zoning criteria as free as possible from inuence of local
conditions.
Compares predictions and observations and nd close t, possibly because of restricted
distance range.
Note that use of simple functional form and rhypo acceptable approximation because of
short rupture lengths.
2.23
Faccioli & Agalbato (1979)
log y = b1 + b2 M + b3 log(R + )
where y is in cm/s2 , b1 = 1.59 0.69, b2 = 0.25 0.03, b3 = 0.79 0.12, = 0 and
= 0.25 for horizontal PGA and b1 = 1.38 1.89, b2 = 0.24 0.09, b3 = 0.78 0.25
and = 0.25 for vertical PGA.
31
Use two site classes:

Soil Includes alluvium and moraine deposits of varying thicknesses and characteristics.
Rock-like Includes limestone, dolomite, ysch and cemented conglomerates, even if heavily
fractured, overlain by not more than 45 m of alluvium.
Use published and unpublished material for classication.
Use data from Friuli 1976 mainshock and subsequent earthquakes from four networks
including temporary stations (ENEL, CNEN, IZIIS and CEA/DSN). Data from ENEL,
CNEN and IZIIS from RFT-250 and SMA-1 instruments and data from CEA/DSN from
short-period seismographs. Some records not available in digital form so used reported
PGAs.
Almost all records from free-eld stations.
58 PGAs from rhypo 20 km.
13 cm/s2 P GA 515 cm/s with 93% above 30 cm/s2 .
Best-recorded earthquake (mainshock) contributed 24 PGAs.
One station contributed 17 PGAs.
Also regresses just using data from mainshock.
is either 0 or 25 in regression. Prefer results with = 0 because smaller standard
errors in b3 .
Statistical tests show b2 and b3 are signicantly different than 0.
Also present coefcients for rock-like stations only and soil stations only. Find that effect
of selection by site class does not greatly affect coefcients.
Process a smaller set of records available in digitized form (76 horizontal components)
using high-pass lter (cut-off and roll-off of 0.40.8 Hz) based on digitization noise. Note
difculty in standard processing due to high-frequency content and short durations. Use
sampling rate of 100 Hz. Find that corrected horizontal PGAs are on average 6% lower
than uncorrected PGAs and 15% show difference larger than 10%. For vertical PGAs
average difference is 12%. Develop equations based on this subset (for horizontal PGA
b1 = 1.51 0.77, b2 = 0.24 0.04, b3 = 0.70 0.21 and = 0.24). Note similarity to
results for uncorrected PGAs.
Also derive equation using only 39 PGAs from rhypo 20 km and note weak magnitude
and distance dependence. Compare to data from shallow soil sites at Forgaria-Cornino
and Breginj and note that local site conditions can signicantly modify bedrock motions
even at close distances.
32
2.24
Aptikaev & Kopnichev (1980)
log Ae = a1 M + a2 log R + a3
where Ae is in cm/s2 , for Ae 160 cm/s2 a1 = 0.28, a2 = 0.8 and a3 = 1.70 and for
Ae < 160 cm/s2 a1 = 0.80, a2 = 2.3 and a3 = 0.80 ( not given).
As a rule, PGA corresponds to S-wave.
Use ve source mechanism categories (about 70 records, 59 earthquakes from W. N.
America including Hawaii, Guatemala, Nicaragua, Chile, Peru, Argentina, Italy, Greece,
Romania, central Asia, India and Japan):
1. Contraction faulting (uplift and thrust), about 16 earthquakes.
2. Contraction faulting with strike-slip component, about 6 earthquakes.
3. Strike-slip, about 17 earthquakes.
4. Strike-slip with dip-slip component, about 6 earthquakes.
5. Dip-slip, about 9 earthquakes.
Use these approximately 70 records to derive ratios of mean measured, A0 , to predicted
PGA, Ae , log(A0 /Ae ), and for ratios of mean horizontal to vertical PGA, log Ah /Av , for
each type of faulting. Use every earthquake with equal weight independent of number
of records for each earthquake.
Results are:
log A0 /Ae
log Ah /Av
Category 1
Category 2
Category 3
Category 4
Category 5
0.35 0.13 (16)

0.32 0.13 (12)
0.11 0.17 (5)

0.32 0.08 (5)
0.22 0.08 (17)

0.27 0.07 (12)
0.06 0.13 (6)

0.18 0.10 (5)
0.06 0.20 (9)

0.17 0.11 (5)
where gives 0.7 condence intervals and number in brackets is number of earthquakes used.
Also calculate mean envelope increasing speed for P-wave amplitudes, A, obtained at
teleseismic distances: n = d ln A/dt, where t is time for P-wave arrival and try to relate
to ratios for each type of faulting.
2.25
Blume (1980)
a = b1 eb2 M (R + k)b3
where a is in gal, for method using distance partitioning b1 = 18.4, b2 = 0.941, b3 = 1.27
and k = 25 and for ordinary one-stage method b1 = 102, b2 = 0.970, b3 = 1.68 and
k = 25 ( not given).
Does not use PGA cutoff because PGA is, by itself, a poor index of damage in most
cases.
Mean magnitude is 5.4 and mean distance is 84.4 km.
33
Notes problem of regression leverage for some attenuation studies. Lots of data in
fairly narrow distance band, e.g. records from San Fernando earthquake, can dominate
regression and lead to biased coefcients.
Divides data into ten distance bands (A-J) which are 10 km wide up to 60 km and then
60-99.9 km, 100139.9 km, 140199.9 km and 200 km. Fits log10 a = bM c to data
in each band and ts Ground-motion model to selected point set in M , R and a.
Also ts equation using all data using normal least squares.
Adds 52 records (3.2 M 6.5, 5 R 15 km) and repeats; nds little change.
2.26 Iwasaki et al. (1980)

PGA = a1 10a2 M ( + 10)a3

where PGA is in gal, for type I sites a1 = 46.0, a2 = 0.208 and a3 = 0.686 , for type
II sites a1 = 24.5, a2 = 0.333 and a3 = 0.924, for type III sites a1 = 59.0, a2 = 0.261
and a3 = 0.886, for type IV sites a1 = 12.8, a2 = 0.432, a3 = 1.125 and for all sites
a1 = 34.1, a2 = 0.308 and a3 = 0.925 ( not given).
Use four site categories:
Type I Tertiary or older rock (dened as bedrock) or diluvium with depth to bedrock, H <
10 m, 29 records.
Type II Diluvium with H 10 m or alluvium with H < 10 m, 74 records.
Type III Alluvium with H < 25 m including soft layer (sand layer vulnerable to liquefaction
or extremely soft cohesive soil layer) with thickness < 5 m, 130 records.
Type IV Other than above, usually soft alluvium or reclaimed land, 68 records.
Select earthquakes with Richter magnitude 5.0, hypocentral depth 60 km and which
include at least one record with PGA 50 gals (0.5 m/s2 ). Exclude records with PGA
< 10 gals (0.1 m/s2 ).
All records for M 7.0 are from distance > 60 km.
Do regression separately for each soil category and also for combined data.
2.27 Matuschka (1980)

Yc = b1 eb2 M (R + b4 )b3
Coefcients unknown.
34
2.28
Ohsaki et al. (1980a)
A = 10a1 M a2 log x+a3

where A is in cm/s2 , for horizontal PGA a1 = 0.440, a2 = 1.381 and a3 = 1.04 and for
vertical PGA a1 = 0.485, a2 = 1.85 and a3 = 1.38 ( not given).
All records from free-eld bedrock sites.
2.29
Campbell (1981)
PGA = a exp(bM )[R + c1 exp(c2 M )]d

where PGA is in g, for unconstrained model a = 0.0159, b = 0.868, c1 = 0.0606,
c2 = 0.700, d = 1.09 and = 0.372 (on natural logarithm) and for constrained model
a = 0.0185, b = 1.28, c1 = 0.147, c2 = 0.732, d = 1.75 and = 0.384 (in terms of
natural logarithm).
Uses this functional form because capable of modelling possible nonlinear distance scaling in near eld and because distance at which transition from near eld to far eld
occurs probably proportional to fault rupture zone size.
Considers six site classications but does not model:
A Recent alluvium: Holocene Age soil with rock 10 m deep, 71 records.
B Pleistocene deposits: Pleistocene Age soil with rock 10 m deep, 22 records.
C Soft rock: Sedimentary rock, soft volcanics, and soft metasedimentary rock, 14
records.
D Hard rock: Crystalline rock, hard volcanics, and hard metasedimentary rock, 9
records.
E Shallow soil deposits: Holocene or Pleistocene Age soil < 10 m deep overlying
soft or hard rock, 17 records. Not used in analysis.
F Soft soil deposits: extremely soft or loose Holocene Age soils, e.g. beach sand or
recent oodplain, lake, swamp, estuarine, and delta deposits, 1 record. Not used
in analysis.
Notes that data from areas outside western USA may be substantially different than
those from western USA due to tectonics and recording practices but far outweighed
by important contribution these data can make to understanding of near-source ground
motion.
Notes use of only near-source data has made differences in anelastic attenuation negligible to inherent scatter from other factors.
Selects data from shallow tectonic plate boundaries generally similar to western N.
America, deep subduction events excluded because of differences in travel paths and
stress conditions.
35
Selects data from instruments with similar dynamic characteristics as those used in USA
to avoid bias, therefore excludes data from SMAC accelerographs in Japan.
Selects data which meet these criteria:
1. Epicentres known with an accuracy of 5 km or less, or accurate estimate of closest
distance to fault rupture surface known.
2. Magnitudes accurate to within 0.3 units.
3. Distances were within 20, 30, and 50 km for magnitudes less than 4.75 between
4.75 and 6.25 and greater than 6.25 respectively. Only uses data from earthquakes
with magnitude 5.0 because of greatest concern for most design applications.
4. Hypocentres or rupture zones within 25 km of ground surface.
5. PGA 0.2 m/s2 for one component, accelerographs triggered early enough to capture strong phase of shaking.
6. Accelerograms either free-eld, on abutments of dams or bridges, in lowest basement of buildings, or on ground level of structures without basements. Excluded
Pacoima Dam record, from San Fernando (9/2/1971) earthquake due to topographic, high-frequency resonance due to large gradation in wave propagation
velocities and amplication due to E-W response of dam.
Well distributed data, correlation between magnitude and distance only 6%.
Uses PGA from digitised, unprocessed accelerograms or from original accelerograms
because fully processed PGAs are generally smaller due to the 0.02 s decimation and
frequency band-limited ltering of records.
Uses mean of two horizontal components because more stable peak acceleration parameter than either single components taken separately or both components taken together.
Magnitude scale chosen to be generally consistent with Mw . Division point between
using ML and Ms varied between 5.5 and 6.5; nds magnitudes quite insensitive to
choice.
Notes rrup is a statistically superior distance measure than epicentral or hypocentral
and is physically consistent and meaningful denition of distance for earthquakes having
extensive rupture zones.
Does not use all data from San Fernando earthquake to minimize bias due to large
number of records.
Uses seven different weighting schemes, to control inuence of well-recorded earthquakes (e.g. San Fernando and Imperial Valley earthquakes). Giving each record or
each earthquake equal weight not reasonable representation of data. Uses nine distance dependent bins and weights each record by a relative weighting factor 1/ni,j ,
where ni,j is total number of recordings from ith earthquake in j th interval.
Finds unconstrained coefcients and all coefcients statistically signicant at 99%.
Finds coefcients with d constrained to 1.75 (representative of far-eld attenuation of
PGA) and c2 = b/d, which means PGA is independent of magnitude at the fault rupture
surface. All coefcients statistically signicant at 99%. Notes similarity between two
models.
36
Plots normalised weighted residuals against distance, magnitude2 and predicted acceleration2 . Finds that residuals uncorrelated, at 99%, with these variables.
Normal probability plots, observed distribution of normalised weighted residuals and
Kolmogorov-Smirnov test, at 90%, conrms that PGA can be accepted as being lognormally distributed.
Finds effects of site geology, building size, instrument location and mechanism to be
extensively interrelated so selects only records from free-eld or small structures.
Analyses all selected data, nd sites of classes E and F signicantly higher PGA , at
90% level, so removes records from E and F.
Finds differences in PGA from other site categories to be negligible but notes that it
cannot be extended to PGV, PGD, spectral ordinates or smaller magnitudes or further
distances.
Distribution with mechanism is: 69 from strike-slip, 40 from reverse, 5 from normal and 2
records from oblique. Finds that reverse fault PGAs are systematically higher, signicant
at 90%, than those from other fault types although size of bias is due to presence of data
from outside N. America.
Considers soil (A and B) records from small buildings (115 components) and in free-eld
and those obtained in lowest basement of large buildings (40 components). Finds PGA
signicantly lower, at 90% level, in large buildings.
Finds topographic effects for 13 components used in nal analysis (and for 11 components from shallow soil stations) to be signicantly higher, at 90%, although states size
of it may not be reliable due to small number of records.
Removes Imperial Valley records and repeats analysis. Finds that saturation of PGA
with distance is not strongly dependent on this single set of records. Also repeats analysis constraining c2 = 0, i.e. magnitude independent saturation, and also constraining
c1 = c2 = 0, i.e. no distance saturation, nds variance when no distance saturation is
signicantly higher, at 95%, than when there is saturation modelled.
Finds that magnitude saturation effects in modelling near-source behaviour of PGA is
important and c2 is signicantly greater than zero at levels of condence exceeding
99%. Also variance is reduced when c2 6= 0 although not at 90% or above.
Repeats analysis using distance to surface projection of fault, nds reduced magnitude
saturation but similar magnitude scaling of PGA for larger events.
2.30
Chiaruttini & Siro (1981)
log a = b0 + bAN XAN + bAB XAB + bM ML + bd log d

where a is in g/100, b0 = 0.04, bAN = 0.24, bAB = 0.23, bM = 0.41 and bd = 0.99 (
not given).
2
Not shown in paper.
37
Use three site categories for Friuli records, although note that information is rather supercial:
ThA Alluvium with depth > 20 m, 36 records.
RI Rock-like: hard rock or stiff soil, 243 records.
thA Alluvium-like with depth 20 m: includes sites for which thickness of deposit is reported to be very small which accounts for a few metres of weathering of underlying
bedrock, 60 records.
Alpide belt records divided into two categories: rock-like (25 records) and alluvium-like
(40 records).
Use data from free-eld instruments or from instruments in basements of small structures and divide data into three regions: those from 1976 Friuli shocks (120 records)
XAN = XAB = 0, those from 1972 Ancona swarm (40 records) XAN = 1 &
XAB = 0 and those from Alpide Belt (Azores to Pakistan excluding those from Friuli and
Ancona) (64 records) XAN = 0 & XAB = 1. Exclude records with PGA < 0.15 m/s2
to avoid possible bias at low acceleration values.
Assume average focal depth of 6 km.
Note some PGA values derived from velocity records which are retained because compatible with other data. No instrument corrections applied to Friuli records because
correction does not substantially alter PGA.
Use ML because determined at short distances and allows homogenous determination
from lowest values up to saturation at ML = 7.0 and it is determined at frequencies of
nearly 1 Hz, close to accelerographic band.
Perform regression on PGAs from each of the three regions and each soil types considered within that region.
Group rock-like (R) and thick alluvium (ThA) records together for Friuli. Find bd for Friuli
equations derived for thin alluvium-like and rock and thick alluvium not signicantly different but bM is signicantly different, at 95% level. Repeat analysis using only Tolmezzo
records because of large scatter in residuals but decide it is in thA category.
For Alpide belt equations nd bM is almost the same for Rl and Al records and the
difference in bd is less than standard error, thus repeat analysis using a dummy variable
XAl which equals 0 for Rl and 1 for Al records.
2.31 Joyner & Boore (1981)

log y = + M log r + br
where r = (d2 + h2 )1/2
where y is in g, = 1.02, = 0.249, b = 0.00255, h = 7.3 and = 0.26.
Use two site categories (not all records have category):
3
38
Typographic error in their Table 1 because only 14 records are listed for rock-like sites
S = 0 Rock: sites described as granite, diorite, gneiss, chert, greywacke, limestone,

sandstone or siltstone and sites with soil material less than 4 to 5 m thick overlying rock, 29 records. Indicate caution in applying equations for M > 6.0 due to
limited records.
S = 1 Soil: sites described as alluvium, sand, gravel, clay, silt, mud, ll or glacial outwash
except where soil less than 4 to 5 m thick, 96 records.
Restrict data to western North American shallow earthquakes, depth less than 20 km,
with M > 5.0. Most records from earthquakes with magnitudes less than 6.6.
Exclude records from base of buildings three or more storeys high and from abutments
of dams.
Exclude records associated with distances which had an uncertainty greater than 5 km.
Exclude records from distances greater than or equal to the shortest distance to an
instrument which did not trigger.
Six earthquakes recorded at only one station so not included in second stage regression.
Include quadratic dependence term, M2 , but not signicant at 90% level so omitted.
Include site term, cS , but not signicant so omitted.
Examine residuals against distance for difference magnitude ranges, no obvious differences in trends are apparent among the different magnitude classes.
Consider a magnitude dependent h = h1 exp(h2 [M 6.0]) but reduction in variance not
signicant. Also prefer magnitude independent h because requires fewer parameters.
Examine effect of removing records from different earthquakes from data.
Examine effect of different h on residuals and b. Note coupling between h and b.
Note coincidence of anelastic coefcient, b, and measured Q values. Also note similarity
between h and proportions of depth of seismogenic zone in California.
2.32
Bolt & Abrahamson (1982)
y = a{(x + d)2 + 1}c eb(x+d)

where y is in g, for 5 M < 6 a = 1.2, b = 0.066, c = 0.033, d = 23 and standard
error for one observation of 0.06 g, for 6 M < 7 a = 1.2, b = 0.044, c = 0.042, d = 25
and standard error for one observation of 0.10 g, for 7 M 7.7 a = 0.24 b = 0.022,
c = 0.10, d = 15 and standard error for one observation of 0.05 g and for 6 M 7.7
a = 1.6, b = 0.026, c = 0.19, d = 8.5 and standard error for one observation of 0.09 g.
Use data of Joyner & Boore (1981).
Form of equation chosen to satisfy plausible physical assumptions but near-eld behaviour is not determined from overwhelming contributions of far-eld data.
39
Apply nonlinear regression on y not on log y to give more weight to near-eld values.
Split data into four magnitude dependent groups: 5 M < 6, 6 M < 7, 7 M 7.7
and 6 M 7.7.
Use form of equation and regression technique of Joyner & Boore (1981), after removing 25 points from closer than 8 km and nd very similar coefcients to Joyner & Boore
(1981). Conclude from this experiment and their derived coefcients for the four magnitude groups that using their form of equation predicted near-eld accelerations are not
governed by far-eld data.
Find no evidence of systematic increase in PGA near the source as a function of magnitude and that the large scatter prevents attaching signicance to differences in near-eld
PGA which are predicted using their attenuation relations for different magnitude ranges.
2.33 Joyner & Boore (1982b) & Joyner & Boore (1988)
log y = + (M 6) + (M 6)2 p log r + br + cS

r = (d2 + h2 )1/2
where y is in g, = 0.23, = 0, p = 1, b = 0.0027, c = 0, h = 8.0 and = 0.28 and
for randomly oriented component = 0.43 and for larger component = 0.49.
Use same data and method as Joyner & Boore (1981), see Section 2.31, for PGA.
Use data from shallow earthquakes, dened as those for which fault rupture lies mainly
above a depth of 20 km.
2.34 PML (1982)

ln(a) = C1 + C2 M + C3 ln[R + C4 exp(C5 M )]

where a is in g, C1 = 1.17, C2 = 0.587, C3 = 1.26, C4 = 2.13, C5 = 0.25 and
= 0.543.
Use data from Italy (6 records, 6 earthquakes), USA (18 records, 8 earthquakes),
Greece (13 records, 9 earthquakes), Iran (3 records, 3 earthquakes), Pakistan (3 records,
1 earthquake), Yugoslavia (3 records, 1 earthquake), USSR (1 record, 1 earthquake),
Nicaragua (1 record, 1 earthquake), India (1 record, 1 earthquake) and Atlantic Ocean
(1 record, 1 earthquake).
Develop for use in UK.
40
2.35
Schenk (1982)
log Amean = aM b log R + c

where Amean is in cm/s2 , a = 1.1143, b = 1.576 and c = 2.371 ( not given).
Fits equation by eye because least squares method is often strictly dependent on marginal
observations, particularly for little pronounced dependence.
2.36
Brillinger & Preisler (1984)
A1/3 = a1 + a2 M + a3 ln(d2 + a24 )

where A is in g, a1 = 0.432(0.072), a2 = 0.110(0.012), a3 = 0.0947(0.0101), a4 =
6.35(3.24), 1 = 0.0351(0.0096) (inter-event) and 2 = 0.0759(0.0042) (intra-event),
where numbers in brackets are the standard errors of the coefcients.
Use exploratory data analysis (EDA) and alternating conditional expectations (ACE)
techniques.
Firstly sought to determine functions (A), (M ) and (d) so that (A) = (M )+(d),
i.e. an approximately additive relationship. Prefer additivity because of linearity, ease of
interpolation and interpretation and departures from t are more easily detected.
Use ACE procedure to nd model. For set of data, with response yi and predictors wi
and xi nd functions to minimize: ni=1 [(yi ) (wi ) (xi )]2 subject to

(wi ) = 0,
(xi ) = 0,
(yi ) = 0 and
(yi )2 = n. Search amongst unrestricted curves
or unrestricted monotonic curves. Use EDA to select specic functional forms from the
estimates of , and at each data point.
Do not use weighting because does not seem reasonable from statistical or seismological points of view.
Do not want any individual earthquake, e.g. one with many records, overly inuencing
results.
Note that because each earthquake has its own source characteristics its records are
intercorrelated. Therefore use random effects model which accounts for perculiarities
of individual earthquakes and correlation between records from same event.
On physical grounds, restrict , and to be monotonic and nd optimal transformation
of magnitude is approximately linear, optimal transformation of distance is logarithmic
and cube root is optimal for acceleration transformation.
Note that need correlations between coefcients, which are provided, to attach uncertainties to estimated PGAs.
Provide method of linearization to give 95% condence interval for acceleration estimates.
41
Also provide a graphical procedure for estimating accelerations that does not rely on an
assumed functional form.
Examine residual plots (not shown) and found a candidate for an outlying observation
(the record from the Hollister 1974 earthquake of 0.011 g at 17.0 km).
Find that assumption of normality after transformation seems reasonable.
2.37 Joyner & Fumal (1984), Joyner & Fumal (1985) & Joyner &
Boore (1988)
log y = c0 + c1 (M 6) + c2 (M 6)2 + c3 log r + c4 r + S

1
where r = (d2 + h2 ) 2
and: S
0 for rock site

c6 log VV0 for soil site
where y is in g, coefcients c0 to c4 , h and are from Joyner & Boore (1981) and c6
and V0 are not signicant at 90% level so do not report them.
Use data of Joyner & Boore (1981).
Continuous site classication for soil sites in terms of shear-wave velocity, V , to depth
of one quarter wavelength of waves of period of concern. V measured down to depths
of at least 30 m and then extrapolated using geological data. V known for 33 stations.
Soil amplication factor based on energy conservation along ray tubes, which is a body
wave argument and may not hold for long periods for which surface waves could be
important. Does not predict resonance effects.
Regress residuals, Rij , w.r.t. motion predicted for rock sites on log Rij = Pi + c6 Vj ,
where j corresponds to j th station and i to ith earthquake. Decouples site effects variation from earthquake-to-earthquake variation. Find unique intercept by requiring average
site effect term calculated using shear-wave velocity to be same as that calculated using
rock/soil classication.
No signicant, at 90%, correlation between residuals and V for PGA.
Repeat regression on residuals using V and depth to underlying rock (dened as either
shear-wave velocity > 750 m/s or > 1500 m/s). Find no correlation.
2.38 Kawashima et al. (1984) & Kawashima et al. (1986)

X(M, , GCi ) = a(GCi )10b(GCi )M ( + 30)c

where X(M, , GCi ) is in gal, c = 1.218, for group 1 sites a(GC1 ) = 987.4, b(GC1 ) =
0.216 and = 0.216, for group 2 sites a(GC2 ) = 232.5, b(GC2 ) = 0.313 and = 0.224
and for group 3 sites a(GC3 ) = 403.8, b(GC3 ) = 0.265 and = 0.197.
42

Group 1 Tertiary or older rock (dened as bedrock) or diluvium with H < 10 m or fundamental period TG < 0.2 s.
Group 2 Diluvium with H 10 m, alluvium with H < 10 m or alluvium with H < 25 m
including soft layer with thickness < 5 m or fundamental period 0.2 < TG < 0.6 s.
Group 3 Other than above, normally soft alluvium or reclaimed land.
Only includes free-eld records with MJMA 5.0 and focal depths Dp < 60 km. Excludes records from structures with rst oor or basement.
Records instrument corrected, because Japanese instruments substantially suppress
high frequencies, considering accuracy of digitization for frequencies between 13 and
12 Hz.
Note that MJMA and not necessarily most suitable parameters to represent magnitude
and distance but only ones for all records in set.
Note lack of near-eld data for large magnitude earthquakes, approximately
from MJMA < 7.0.
3
4
of records
Use 30 km in distance dependence term because focal depth of earthquakes with magnitudes between 7.5 and 8.0 are between 30 and 100 km so 30 is approximately half the
fault length.
Try equation: log X = f1 +f2 M +f3 log(+30)+f4 Dp +f5 M log(+30)+f6 M Dp +
f7 Dp log( + 30) + f8 M 2 + f9 {log( + 30)}2 + f10 Dp2 where fi are coefcients to be
found considering each soil category separately. Apply multiple regression analysis to
36 combinations of retained coefcients, fi , and compute multiple correlation coefcient,
R, and adjusted multiple correlation coefcient, R . Find that inclusion of more than
three coefcients does not give signicant increase in R , and can lead to unrealistic
results. Conclude due to insufcient data.
Consider a, b and c dependent and independent of soil type and examine correlation
coefcient, R, and adjusted correlation coefcient, R . Find that c is not strongly dependent on soil type.
Find match between normal distribution and histograms of residuals.
2.39
McCann Jr. & Echezwia (1984)
Four Ground-motion models:
log10 Y
= a + bM + d log10 [(R2 + h2 )1/2 ]
log10 Y
= a + bM + d log10 [R + c1 exp(c2 M )]
Model II
[c
c2 ]
1
+
+ eR
Model III
= a + bM + d log10
R2
R
= a + bM + d log10 [R + 25]
Model IV
log10 Y
log10 Y
Model I
where Y is in g, for model I a = 1.320, b = 0.262, d = 0.913, h = 3.852 and

= 0.158, for model II a = 1.115, b = 0.341, c1 = 1.000, c2 = 0.333, d = 1.270
and = 0.154, for model III a = 2.000, b = 0.270, c1 = 0.968, c2 = 0.312, d = 0.160,
43
e = 0.0105 and = 0.175 and for model IV a = 1.009, b = 0.222, d = 1.915 and
= 0.174.
Note 25 in Model IV should not be assumed but should be found by regression.
Note tectonics and travel paths may be different between N. American and foreign
records but consider additional information in near eld more relevant.
Selection procedure composite of Campbell (1981) and Joyner & Boore (1981). Exclude
data from buildings with more than two storeys.
Weighted least squares, based on distance, applied to control inuence of well recorded
events (such as San Fernando and Imperial Valley). Similar to Campbell (1981)
Test assumption that logarithm of residuals are normally distributed. Cannot disprove
assumption.
Variability between models not more than 20% at distances > 10 km but for distances
< 1 km up to 50%.
2.40 Schenk (1984)

log Amean = aM b log R + c

where Amean is in cm/s2 , a = 0.37, b = 1.58 and c = 2.35 ( not given).
Considers two site conditions but does not model:
1. Solid
2. Soft
Fits equation by eye.
States applicable approximately for: Rlower R Rupper where log Rlower = 0.1M +
.
0.5 and log Rupper = 0.35M + 0.4, due to distribution of data.
Notes great variability in recorded ground motions up to R = 30 km due to great inuence of different site conditions.
Notes for M 4 source can be assumed spherical but for M > 4 elongated (extended)
shape of focus should be taken into account.
2.41 Xu et al. (1984)

PGA = a1 exp(a2 M )(R + a3 )a4

where PGA is in g, a1 = 0.1548, a2 = 0.5442, a3 = 8 and a4 = 1.002 ( not given).
All records from aftershocks of 1975 Haicheng earthquake and from 1976 Tangshan
earthquake and aftershocks.
44
Most records from earthquakes with magnitude less than 5.8 and from distances <
30 km.
Exclude records with PGA < 0.5 m/s2 to avoid too much contribution from far eld.
Due to small number of records simple regression technique justied.
States valid for 4 M 6.5 and R 100 km.
Also use 158 records from western N. America to see whether signicantly different than
N. Chinese data. Derive equations using both western N. American and N. Chinese
data and just western N. American data and nd that predicted PGAs are similar, within
uncertainty.
Insufcient data to nd physically realistic anelastic term.
2.42
Brillinger & Preisler (1985)
log A = a1 + a2 M log r + a3 r
where r2
= d2 + a24
where A is in g, a1 = 1.229(0.196), a2 = 0.277(0.034), a3 = 0.00231(0.00062),

a4 = 6.650(2.612), 1 = 0.1223(0.0305) (inter-event) and = 0.2284(0.0127) (intraevent), where numbers in brackets are the standard errors of the coefcients.
Provide algorithm for random effects regression.
Note that the functional form adopted in Brillinger & Preisler (1984) is strictly empirical
and hence repeat analysis using functional form of Joyner & Boore (1981), which is
based on physical reasoning.
Note that need correlations between coefcients, which are provided, to attach uncertainties to estimated PGAs.
2.43
Kawashima et al. (1985)
Use very similar data to Kawashima et al. (1984); do not use some records because
missing due to recording and digitizing processes. Use equation and method (although
do not check all 36 combinations of forms of equation) used by Kawashima et al. (1984),
see section 2.38.
X(M, , GCi ) is in gal. Coefcients are: c = 1.190 and for ground group 1 a = 117.0
and b = 0.268 and for ground group 2 a = 88.19 and b = 0.297 and for group ground 3
a = 13.49 and b = 0.402 with = 0.253.
45
2.44 Peng et al. (1985b)

log10 a = A + BM + C log10 R + DR
where a is in cm/s2 , for N.E. China A = 0.474, B = 0.613, C = 0.873 and D =
0.00206 ( not given) and for S.W. China A = 0.437, B = 0.454, C = 0.739 and
D = 0.00279 ( not given).
Consider two site conditions for NE records but do not model:
1. Rock: 28 records.
2. Soil: 45 records.
Consider all records to be free-eld.
Note that Chinese surface-wave magnitude, M , is different than Ms and may differ by
0.5 or more. Use mb or Ms and nd larger residuals.
Most records from M 5.8.
Note isoseismals are not elongated for these earthquakes so use of another distance
measure will not change results by much.
Also derives equation for SW China (3.7 M 7.2, 6.0 R 428.0 km all but
one record 106.0 km , 36 records from 23 earthquakes) and note difference between
results from NE China although use less data.
Note that some scatter may be due to radiation pattern.
Note that data is from limited distance range so need more data to conrm results.
2.45 Peng et al. (1985a)

log Am = a1 + a2 M log R a3 R
R =
d2 + h2
where Am is g, a1 = 1.49, a2 = 0.31, a3 = 0.0248, h = 9.4 km and = 0.32 (for
horizontal components) and a1 = 1.92, a2 = 0.29, a3 = 0.0146, h = 6.7 km and
= 0.36 (for vertical components).
Data from experimental strong-motion array consisting of 12 Kinemetrics PDR-1 instruments deployed in the epicentral area of the Ms = 7.8 Tangshan earthquake of 28th
July 1976. Provide details of site geology at each station; most stations are on soil.
Records from earthquakes recorded by only one station were excluded from analysis.
Note that equations are preliminary and more rened equations await further studies of
magnitudes and distances used in analysis.
Note that high anelastic attenuation coefcient may be due to biases introduced by the
distribution in magnitude-distance space and also because of errors in magnitude and
distances used.
46
2.46
PML (1985)
ln(a) = C1 + C2 M + C3 ln[R + C4 exp(C5 M )] + C6 F

where a is in g, C1 = 0.855, C2 = 0.46, C3 = 1.27, C4 = 0.73, C5 = 0.35,
C6 = 0.22 and = 0.49.
Use data from Italy (47 records, 9 earthquakes), USA (128 records, 18 earthquakes),
Greece (11 records, 8 earthquakes), Iran (2 records, 2 earthquakes), Yugoslavia (7
records, 2 earthquake), Nicaragua (1 record, 1 earthquake), New Zealand (3 records, 3
earthquakes), China (2 records, 2 earthquakes) and Canada (2 records, 1 earthquake).
Develop for use in UK.
Select earthquakes with Ms < 7 and R 40 km.
Focal depths < 40 km.
Use two source mechanism categories (40 records have no source mechanism given):
F = 0 Strike-slip and normal, 85 records.

F = 1 Thrust, 78 records.
Also derive equation not considering source mechanism, i.e. C6 = 0.
2.47
McCue (1986)
A = a1 (ea2 ML )(dh )a3

where A is in g, a1 = 0.00205, a2 = 1.72 and a3 = 1.58 ( not given).
2.48
C.B. Crouse (1987) reported in Joyner & Boore (1988)
ln y = a + bMs + cMs2 + d ln(r + 1) + kr

where y is in gal, a = 2.48456, b = 0.73377, c = 0.01509, d = 0.50558, k =
0.00935 and = 0.58082.
Records from deep soil sites (generally greater than 60 m in thickness).
Data from shallow crustal earthquakes.
47
2.49 Krinitzsky et al. (1987) & Krinitzsky et al. (1988)

Ground-motion model is (for shallow earthquakes):
log A = a1 + a2 M log r + a3 r
where A is in cm/s2 , a1 = 1.23 (for hard sites), a1 = 1.41 (for soft sites), a2 = 0.385
and a3 = 0.00255 ( is not given).
Ground-motion model is (for subduction zone earthquakes):
log A = b1 + b2 M log
r2 + 1002 + b3 r
where A is in cm/s2 , b1 = 2.08 (for hard sites), b1 = 2.32 (for soft sites), b2 = 0.35 and
b3 = 0.0025 ( is not given).
1 Rock
2 Stiff soil
3 Deep cohesionless soil ( 16 m)
4 Soft to medium stiff clay ( 16 m)
Categories 1 and 2 are combined into a hard (H) class and 3 and 4 are combined into a
soft (S) class. This boundary established using eld evidence at a shear-wave velocity
of 400 m/s and at an SPT N count of 60.
Use data from ground oors and basements of small or low structures (under 3 stories)
because believe that small structures have little effect on recorded ground motions.
Separate earthquakes into shallow (h 19 km) and subduction (h 20 km) because
noted that ground motions have different characteristics.
Use epicentral distance for Japanese data because practical means of representing
deep subduction earthquakes with distant and imprecise fault locations.
Do not use rupture distance or distance to surface projection of rupture because believe
unlikely that stress drop and peak motions will occur with equal strength along the fault
length and also because for most records fault locations are not reliably determinable.
Note that there is a paucity of data but believe that the few high peak values observed
(e.g. Pacoima Dam and Morgan Hill) cannot be dismissed without the possibility that
interpretations will be affected dangerously.
For subduction equations, use records from Japanese SMAC instruments that have not
been instrument corrected, even though SMAC instruments show reduced sensitivity
above 10 Hz, because ground motions > 10 Hz are not signicant in subduction earthquakes. Do not use records from SMAC instruments for shallow earthquakes because
high frequency motions may be signicant.
Examine differences between ground motions in extensional (strike-slip and normal
faulting) and compressional (reverse) regimes for shallow earthquakes but do not model.
Find that the extensional ground motions seem to be higher than compressional motions, which suggest is because rupture propagation comes closer to ground surface in
extensional faults than in compressional faults.
48
Group records into 1 M unit intervals and plot ground motions against distance. When
data is numerous enough the data points are encompassed in boxes (either one, two or
three) that have a range equal to the distribution of data. The positions of the calculated
values within the boxes were used as guides for shaping appropriate curves. Initially
curves developed for M = 6.5 were there is most data and then these were extended
to smaller and larger magnitudes.
2.50
Sabetta & Pugliese (1987)
log y = a + bM log(R2 + h2 )1/2 + eS

where y is in g and for distance to surface projection of fault a = 1.562, b = 0.306,
e = 0.169, h = 5.8 and = 0.173.
Use two site categories:
S = 0 Stiff and deep soil: limestone, sandstone, siltstone, marl, shale and conglomerates
(Vs > 800 m/s) or depth of soil, H , > 20 m, 74 records.
S = 1 Shallow soil: depth of soil, H , 5 H 20 m, 21 records.
Select records which satisfy these criteria:
1. Reliable identication of the triggering earthquake.
2. Magnitude greater than 4.5 recorded by at least two stations.
3. Epicentres determined with accuracy of 5 km or less.
4. Magnitudes accurate to within 0.3 units.
5. Accelerograms from free-eld. Most are from small electric transformer cabins, 4
from one- or two-storey buildings with basements and 5 from near abutments of
dams.
Depths between 5.0 and 16.0 km with mean 8.5 km.
Focal mechanisms are: normal and oblique (7 earthquakes, 48 records), thrust (9 earthquakes, 43 records) and strike-slip (1 earthquake, 4 records).
Notes lack of records at short distances from large earthquakes.
Records baseline-, instrument-corrected and ltered with cutoff frequencies determined
by visual inspection in order to maximise signal to noise ratio within band. Cutoff frequencies ranged from 0.2 to 0.4 Hz and from 25 to 35 Hz. This correction routine thought
to provide reliable estimates of PGA so uncorrected PGA do not need to be used.
For well separated multiple shocks, to which magnitude and focal parameters refer, use
only rst shock.
Magnitude scale assures a linear relationship between logarithm of PGA and magnitude
and avoids saturation effects of ML .
49
Distance to surface projection of fault rupture thought to be a more physically consistent

denition of distance for earthquakes having extensive rupture zones and is easier to
predict for future earthquakes. Also reduces correlation between magnitude and distance.
Use Exploratory Data Analysis using the ACE procedure to nd transformation functions
of distance, magnitude and PGA.
Include anelastic attenuation term but it is positive and not signicant.
Include magnitude dependent h equal to h1 exp(h2 M ) but nd h2 not signicantly different than zero. Note distribution of data makes test not denitive.
Find geometric attenuation coefcient, c, is close to 1 and highly correlated with h so
constrain to 1 so less coefcients to estimate.
Consider deep soil sites as separate category but nd difference between them and stiff
sites is not signicant.
Also use two-stage method but coefcients and variance did not change signicantly
with respect to those obtained using one-stage method, due to uniform distribution of
recordings among earthquakes.
Find no signicant trends in residuals, at 99% level and also no support for magnitude
dependent shape for attenuation curves.
Exclude records from different seismotectonic and geological regions and repeat analysis. Find that predicted PGA are similar.
Plot residuals from records at distances 15 km or less against magnitude; nd no support
for magnitude dependence of residuals.
Note some records are affected by strong azimuthal effects, but do not model them
because they require more coefcients to be estimated, direction of azimuthal effect
different from region to region and azimuthal effects have not been used in other relationships.
2.51 K. Sadigh (1987) reported in Joyner & Boore (1988)

ln y = a + bM + c1 (8.5 M)c2 + d ln[r + h1 exp(h2 M)]

where y is in g. For strike-slip earthquakes: b = 1.1, c1 = 0, c2 = 2.5, for PGA at
soil sites a = 2.611 and d = 1.75, for M < 6.5 h1 = 0.8217, h2 = 0.4814 and
for M 6.5 h1 = 0.3157 and h2 = 0.6286, for PGA at rock sites a = 1.406 and
d = 2.05, for M < 6.5 h1 = 1.353 and h2 = 0.406 and for M 6.5 h1 = 0.579
and h2 = 0.537. For reverse-slip increase predicted values by 20%. For M < 6.5
= 1.26 0.14M and for M 6.5 = 0.35.
1. Soil
50
2. Rock
Use two source mechanism categories:
1. Strike-slip
2. Reverse-slip
Supplement data with signicant recordings of earthquakes with focal depths < 20 km
from other parts of world.
Different equations for M < 6.5 and M 6.5.
2.52
Singh et al. (1987)
log ymax = Ms c log R +

where ymax is in cm/s2 , = 0.429, c = 2.976, = 5.396 and = 0.15.
More complicated functional form unwarranted due to limited distance range.
Depths between 15 and 20 km.
Only use data from a single rm site (Ciudad Universitaria), on a surface layer of lava
ow or volcanic tuff.
Only records from coastal earthquakes.
Residuals plotted against distance, no trends seen.
Give amplication factor for lake bed sites (25 to 80 m deposit of highly compressible,
high water content clay underlain by resistant sands), but note based on only a few sites
so not likely to be representative of entire lake bed.
2.53
Algermissen et al. (1988)
ln(A) = a1 + a2 Ms + a3 ln(R) + a4 R
where A is in g, a1 = 1.987, a2 = 0.604, a3 = 0.9082, a4 = 0.00385 and = 0.68.
2.54
Annaka & Nozawa (1988)
log A = Cm M + Ch H Cd log(R + A exp BM ) + Co

where A is in cm/s2 , A and B so PGA becomes independent of magnitude at fault
rupture, H is depth of point on fault plane when R becomes closest distance to fault
plane, Cm = 0.627, Ch = 0.00671, Cd = 2.212, Co = 1.711 and = 0.211.
51

Convert records from sites with Vs < 300 m/s into records from sites with Vs > 300 m/s
using 1-D wave propagation theory.
Introduce term Ch H because it raises multiple correlation coefcient for PGA.
Note equations apply for site where 300 Vs 600 m/s.
2.55 K.W. Campbell (1988) reported in Joyner & Boore (1988)

ln y = a + bM + d ln[r + h1 exp(h2 M )] + s
where s = e1 K1 + e2 K2 + e3 K3 + e4 K4 + e5 K5 + e6 (K4 + K5 ) tanh(e7 r)
where y is in g, a = 2.817, b = 0.702, d = 1.20, h1 = 0.0921, h2 = 0.584, e1 = 0.32,
e2 = 0.52, e3 = 0.41, e4 = 0.85, e5 = 1.14, e6 = 0.87, e7 = 0.068 and = 0.30.
K3 = 1 Soils 10 m deep.
K3 = 0 Other.
Uses three embedment categories:
K4 = 1, K5 = 0 Basements of buildings 39 storeys.

K5 = 1, K4 = 0 Basements of buildings 10 storeys.
K4 = 0, K5 = 0 Other.
Selects data using these criteria:
1. Largest horizontal component of peak acceleration was 0.02 g [ 0.2 m/s2 ].
2. Accelerograph triggered early enough to record strongest phase of shaking.
3. Magnitude of earthquake was 5.0.
4. Closest distance to seismogenic rupture was < 30 or < 50 km, depending on
whether magnitude of earthquake was < 6.25 or > 6.25.
5. Shallowest extent of seismogenic rupture was 25 km.
6. Recording site located on unconsolidated deposits.
Excludes records from abutments or toes of dams.
Derives two equations: unconstrained (coefcients given above) and constrained which
includes a anelastic decay term kr which allows equation to be used for predictions
outside near-source zone (assumes k = 0.0059 for regression, a value appropriate for
region of interest should be chosen).
Uses two source mechanism categories:
K1 = 0 Strike-slip.
K1 = 1 Reverse.
52
Uses two directivity categories:
K2 = 1 Rupture toward site.

K2 = 0 Other.
2.56
Fukushima et al. (1988) & Fukushima & Tanaka (1990)
log A = aM log(R + c10aM ) bR + d

where A is in cm/s2 , a = 0.41, b = 0.0034, c = 0.032, d = 1.30 and = 0.21.
Use four site categories for some Japanese stations (302 Japanese records not classied):
1. Rock: 41 records
2. Hard: ground above Tertiary period or thickness of diluvial deposit above bedrock
< 10 m, 44 records.
3. Medium: thickness of diluvial deposit above bedrock > 10 m, or thickness of alluvial deposit above bedrock < 10 m, or thickness of alluvial deposit < 25 m and
thickness of soft deposit is < 5 m, 66 records.
4. Soft soil: other soft ground such as reclaimed land, 33 records.
Use 1100 mean PGA values from 43 Japanese earthquakes (6.0 MJMA 7.9, focal
depths 30 km) recorded at many stations to investigate one and two-stage methods. Fits log A = c b log X (where X is hypocentral distance) for each earthquake
and computes mean of b, b. Also ts log A = aM b log X + c using one-stage
method. Find that b > b and shows that this is because magnitude and distance
are strongly correlated (0.53) in data set. Find two-stage method of Joyner & Boore
(1981) very effective to overcome this correlation and use it to nd similar distance coefcient to b. Find similar effect of correlation on distance coefcient for two other models:
log A = aM b log(+30)+c and log A = aM log X bX +c, where is epicentral
distance.
Japanese data selection criteria: focal depth < 30 km, MJMA > 5.0 and predicted PGA
0.1 m/s2 . US data selection criteria: dr 50 km, use data from Campbell (1981).
Because a affects distance and magnitude dependence, which are calculated during
rst and second steps respectively use an iterative technique to nd coefcients. Allow
different magnitude scaling for US and Japanese data.
For Japanese data apply station corrections before last step in iteration to convert PGAs
from different soil conditions to standard soil condition using residuals from analysis.
Two simple numerical experiments performed. Firstly a two sets of articial acceleration data was generated using random numbers based on attenuation relations, one
with high distance decay and which contains data for short distance and one with lower
distance decay, higher constant and no short distance data. Find that the overall equation from regression analysis has a smaller distance decay coefcient than individual
coefcients for each line. Secondly nd the same result for the magnitude dependent
coefcient based on similar articial data.
53
Exclude Japanese data observed at long distances where average acceleration level
was predicted (by using an attenuation relation derived for the Japanese data) to be
less than the trigger level (assume to be about 0.05 m/s2 ) plus one standard deviation
(assume to be 0.3), i.e. 0.1 m/s2 , to avoid biasing results and giving a lower attenuation
rate.
Use the Japanese data and same functional form and method of Joyner & Boore (1981)
to nd an attenuation relation; nd the anelastic coefcient is similar so conclude attenuation rate for Japan is almost equal to W. USA.
Find difference in constant, d, between Japanese and W. USA PGA values.
Plot residuals against distance and magnitude and nd no bias or singularity.
2.57 Gaull (1988)

log PGA = [(a1 log R + a2 )/a3 ](ML a4 ) a5 log R a6 R + a7

where PGA is in m/s2 , a1 = 5, a2 = 3, a3 = 20, a4 = 6, a5 = 0.77, a6 = 0.0045 and
a7 = 1.2 ( not given).
Considers three site categories but does not model:
1. Rock: 6 records
2. Alluvium: 5 records
3. Average site: 10 records
Most records from earthquakes with magnitudes about 3 and most from distances below
about 20 km.
Band pass lter records to get PGA associated with waves with periods between 0.1
and 0.5 s because high frequency PGA from uncorrected records not of engineering
signicance.
Adds 4 near source (5 R 10 km) records from US, Indian and New Zealand earthquakes with magnitudes between 6.3 and 6.7 to supplement high magnitude range.
Add some PGA points estimated from intensities associated with 14/10/1968 ML = 6.9
Meckering earthquake in Western Australia.
Plot 6 records from one well recorded event with ML = 4.5 and t an attenuation curve of
form log PGA = b1 b2 log R b3 R by eye. Plot PGA of all records with 2 R 20 km
against magnitude, t an equation by eye. Use these two curves to normalise all PGA
values to ML = 4.5 and R = 5 km from which estimates attenuation relation.
54
2.58
McCue et al. (1988)
(
A = a(exp(bM ))
)d
R
+c
R0
where A is in g, ln a = 5.75, b = 1.72, c = 0, d = 1.69 and R0 = 1 ( not given).

Few records from free-eld, most are in dams or special structures.,
Because only 62 records, set R0 = 1 and c = 0.
Most records from earthquakes with ML between 1.5 and 2.0.
Maximum PGA in set 3.05 m/s2 .
Nonuniform distribution of focal distances. One quarter of records from same hypocentral distance. Therefore plot PGA of these records against magnitude (1.2 . ML . 4.3
most less than 2.1) to nd b. Then plot bM ln A against ln(R/R0 ) for all records to
nd a and d.
Notes limited data.
2.59
Petrovski & Marcellini (1988)
ln(a) = b01 + b2 M + b3 ln(R + c)

where a is in cm/s2 , b01 = 6.4830, b2 = 0.5438, b3 = 1.3330, c = 20 km and =
0.6718 (for horizontal PGA) and b1 = 5.6440, b2 = 0.5889, b3 = 1.3290, c = 20 km
and = 0.6690 (for vertical PGA) (also give coefcients for other choices of c).
Data from moderate soil conditions.
Data mainly from SMA-1s but 17 from RFT-250s.
Data from northern Greece (5 records, 4 stations, 3 earthquakes), northern Italy (45
records, 18 stations, 20 earthquakes) and former Yugoslavia (70 records, 42 stations,
23 earthquakes).
Data from free-eld or in basements of structures.
Select records from earthquakes with 3 M 7. Most earthquakes with M 5.5.
4 earthquakes (4 records) with M 3.5, 20 (27 records) with 3.5 < M 4.5, 13 (25
records) with 4.5 < M 5.5, 8 (50 records) with 5.5 < M 6.5 and 1 (14 records)
with M > 6.5.
Select records from earthquakes with h 40 km. Most earthquakes with h 10 km.
6 earthquakes with h 5 km, 30 with 5 < h 10 km, 5 with 10 < h 20 km, 4 with
20 < h 30 km and 1 with h > 30.
Select records that satised predetermined processing criteria so that their amplitude
would be such as to give negligible errors after processing.
55
Select records to avoid concentration of records w.r.t. certain sites, magnitudes, hypocentral distances or earthquakes. Most well-recorded earthquakes is 15/4/1979 Montenegro
earthquake with 14 records.
Try values of c between 0 and 40 km. Find standard deviation does not vary much for
different choices.
Test assumption of the log-normal probability distribution of data using graph in a coordinate system for log-normal distribution of probability, by 2 test and by the KolmogorovSmirnov test (not shown). Find assumption is acceptable.
2.60 Tong & Katayama (1988)

log A = M log( + 10) + T +

is in gal, T is predominant period of site, = 0.509, = 2.32, = 0.039 and
where A
= 2.33 ( not given).
Correlation coefcient between magnitude and distance is 0.84, so magnitude and distance cannot be considered independent, so attenuation rate, , is difcult to nd.
= i log( + 10) + i to each earthquake. Dene reliability param First step t log A
2
eter, i = Ni Ri , where Ni is degrees of freedom for i earthquake and Ri is correlation
coefcient. Plot i against i and nd attenuation rate scattered, between 6 and 9, for
i < 1 (Group B) and for 1 > 1 attenuation rate converges (Group U).
Group B includes earthquakes with focal depths > 388 km, earthquakes with small
magnitudes and records from distances 100 km, earthquakes with records from great
distances where spread of distances is small, earthquakes recorded by only 3 stations
and earthquakes with abnormal records. Exclude these records.
Apply multiple regression on Group U to nd , , and simultaneously. Also x
= i i / i and nd , and . Find different coefcients but similar correlation

coefcient. Conclude due to strong correlation between M and so many regression
planes exist with same correlation coefcient.
Perform Principal Component Analysis (PCA) on log A, M , log( + 10), T and log A/A
and nd that equation found by xing is not affected by ill-effect of correlation between
M and .
Omit T from regression and nd little effect in estimation.
2.61 Yamabe & Kanai (1988)

log10 a = log10 x
where
= b1 + b2 M
and:
= c1 + c2 M
where a is in gal, b1 = 3.64, b2 = 1.29, c1 = 0.99 and c2 = 0.38 ( not given).

56

Regress recorded PGA of each earthquake, i, on log10 a = i i log10 x, to nd i and
i . Then nd b1 and b2 from = b1 + b2 M and c1 and c2 from = c1 + c2 M .
Also consider = d1 .
Find and from 6 earthquakes (magnitudes between 5.4 and 6.1) from Tokyo-Yokohama
area are much higher than for other earthquakes, so ignore them. Conclude that this is
due to effect of buildings on ground motion.
2.62
Youngs et al. (1988)
ln(amax ) = C1 + C2 Mw C3 ln[R + C4 exp(C5 Mw )] + BZt

where amax is in g, C1 = 19.16, C2 = 1.045, C3 = 4.738, C4 = 205.5, C5 = 0.0968,
B = 0.54 and = 1.55 0.125Mw .
Use only rock records to derive equation but use some (389 records) for other parts of
study. Classication using published shear-wave velocities for some sites.
Exclude data from very soft lake deposits such as those in Mexico City because may
represent site with special amplication characteristics.
Data from subduction zones of Alaska, Chile, Peru, Japan, Mexico and Solomon Islands.
Use two basic types of earthquake:
Zt = 0 Interface earthquakes: low angle, thrust faulting shocks occurring on plate interfaces.
Zt = 1 Intraslab earthquakes: high angle, predominately normal faulting shocks occurring
within down going plate.
Classication by focal mechanisms or focal depths (consider earthquakes with depths
> 50 km to be intraslab). Note that possible misclassication of some intraslab shocks
as interface events because intraslab earthquakes do occur at depths < 50 km.
Plots PGA from different magnitude earthquakes against distance; nd near-eld distance saturation.
Originally include anelastic decay term C6 R but C6 was negative (and hence nonphysical) so remove.
Plot residuals from original PGA equation (using rock and soil data) against Mw and
R; nd no trend with distance but reduction in variance with increasing Mw . Assume
standard deviation is a linear function of Mw and nd coefcients using combined rock
and soil data (because differences in variance estimation from rock and soil are not
signicant).
Use derived equation connecting standard deviation and Mw for weighted (weights inversely proportional to variance dened by equation) nonlinear regression in all analyses.
57
Plot residuals from original PGA equation; nd that hypothesis that coefcients of equations for interface and intraslab earthquakes are the same can be rejected (using likelihood ratio test for nonlinear regression models) at 0.05 percentile level for both soil
and rock. Try including a term proportional to depth of rupture into equation (because
intraslab deeper than interface events) but nd no signicant reduction in standard error.
Introduce BZt term into equation; nd B is signicant at 0.05 percentile level. Try including rupture type dependence into other coefcients but produces no further decrease in
variance so reject.
Use only data from sites with multiple recordings of both interface and intraslab earthquakes and include dummy variables, one for each site, to remove differences due to
systematic site effects. Fix C1 to C5 to values from entire set and nd individual site
terms and B ; nd B is very similar to that from unconstrained regression.
Examine residuals for evidence of systematic differences between ground motion from
different subduction zones; nd no statistically signicant differences in PGA among
different subduction zones.
Use geometric mean of two horizontal components to remove effect of component-tocomponent correlations that affect validity of statistical tests assuming individual components of motion represent independent measurements of ground motion. Results
indicate no signicant difference between estimates of variance about median relationships obtained using geometric mean and using both components as independent data
points.
Extend to Mw > 8 using nite difference simulations of faulting and wave propagation
modelled using ray theory. Method and results not reported here.
2.63 Abrahamson & Litehiser (1989)

log10 a = + M c log10 [r + exp(h2 M )] + F + Ebr

where F = 1 for reverse or reverse oblique events and 0 otherwise and E = 1 for
interplate events and 0 otherwise, a is in g, for horizontal PGA = 0.62, = 0.177,
c = 0.982, h2 = 0.284, = 0.132, b = 0.0008 and = 0.277 and for vertical PGA
= 1.15, = 0.245, c = 1.096, h2 = 0.256, = 0.096, b = 0.0011 and = 0.296.
Consider three site classications, based on Joyner & Boore (1981):
1. Rock: corresponds to C, D & E categories of Campbell (1981), 159 records.
2. Soil: corresponds to A,B & F categories of Campbell (1981), 324 records.
3. Unclassied: 102 records.
Use to examine possible dependence in residuals not in regression because of many
unclassied stations.
Data based on Campbell (1981).
58
Fault mechanisms are: strike-slip (256 records from 28 earthquakes), normal (14 records
from 7 earthquakes), normal oblique (42 records from 12 earthquakes), reverse (224
records from 21 earthquakes) and reverse oblique (49 records from 8 earthquakes).
Grouped into normal-strike-slip and reverse events. Weakly correlated with magnitude
(0.23), distance (0.18) and tectonic environment (0.03).
Tectonic environments are: interplate (555 records from 66 earthquakes) and intraplate
(30 records from 10 earthquakes) measurements. Weakly correlated with magnitude
(0.26), distance (0.17) and fault mechanism (0.03).
Depths less than 25 km.
Use array average (37 instruments are in array) from 10 earthquakes recorded at SMART 1
array in Taiwan.
Most records from distances less than 100 km and magnitude distribution is reasonably
uniform but correlation between magnitude and distance of 0.52.
Try two-stage technique and model (modied to include fault mechanism and tectonic
environment parameters) of Joyner & Boore (1981), nd inadmissable positive anelastic
coefcient, so do not use it.
Use a hybrid regression technique based on Joyner & Boore (1981) and Campbell
(1981). A method to cope with highly correlated magnitude and distance is required.
First step: t data to f2 (r) = c log10 (r + h) and have separate constants for each earthquake (like in two-stage method of Joyner & Boore (1981)). Next holding c constant nd
, , b and h2 from tting h = exp(h2 M ). Weighting based on Campbell (1981) is used.
Form of h chosen using nonparametric function, H(M ), which partitions earthquakes
into 0.5 unit bins. Plot H(M ) against magnitude. Find that H(M ) = h1 exp(h2 M ) is
controlled by Mexico (19/9/1985) earthquake and h1 and h2 are highly correlated, 0.99,
although does given lower total variance. Choose H(M ) = exp(h2 M ) because Mexico earthquake does not control t and all parameters are well-determined, magnitude
dependent h signicant at 90%.
Try removing records from single-recorded earthquakes and from shallow or soft soil but
effect on predictions and variance small (< 10%).
Plot weighted residuals within 10 km no signicant, at 90%, trends are present.
Find no signicant effects on vertical PGA due to site classication.
2.64
Campbell (1989)
ln PHA = a + bML 1.0 ln[R + c1 ]

where PHA is in g, a = 2.501, b = 0.623, c1 = 7.28 and = 0.506.
Selects records from deep soil (> 10 m). Excludes data from shallow soil ( 10 m) and
rock sites and those in basements of buildings or associated with large structures, such
as dams and buildings taller than two storeys. Selects records with epicentral distances
59
20 km for ML < 4.75 and distances 30 km for ML 4.75 to minimize regional

differences in anelastic attenuation and potential biases associated with nontriggering
instruments and unreported PGAs.
Focal depths, H , between 1.8 and 24.3 km with mean of 8.5 km.
PGAs scaled from either actual or uncorrected accelerograms in order to avoid potential
bias due to correction.
Uses weighted nonlinear least squares technique of Campbell (1981).
Tries two other forms of equation: ln PHA = a + bML 1.0 ln[R + c1 ] + e1 H and
ln PHA = a + bML 1.0 ln[R + c1 ] + e2 ln H for epicentral and hypocentral distance.
Allows saturation of PGA for short distances but nds nonsignicant coefcients, at 90%.
Also tries distance decay coefcient other than 1.0 but nds instability in analysis.
Examines normalised weighted residuals against focal depth, ML and distance. Finds
that although residuals seem to be dependent on focal depth there are probably errors
in focal depth estimation for deep earthquakes in the study so the dependence may not
be real. Finds residuals not dependent on magnitude or distance.
Uses 171 records (0.9 R 28.1 km) from 75 earthquakes (2.5 ML 5.0,
0.7 H 24.3 km) excluded from original analysis because they were on shallow
soil, rock and/or not free-eld, to examine importance of site geology and building size.
Considers difference between PGA from records grouped according to instrument location, building size, embedment, and site geology and the predicted PGA using the
attenuation equation to nd site factors, S . Groups with nonsignicant, at 90%, values
of S are grouped together. Finds two categories: embedded alluvial sites from all building sizes (38 records) and shallow-soil (depth of soil 10 m) sites (35 records) to have
statistically signicant site factors.
Performs regression analysis on all records (irrespective of site geology or building size)
from Oroville (172 records from 32 earthquakes) and Imperial Valley (71 records from 42
earthquakes) to nd individual sites that have signicant inuence on prediction of PGA
(by using individual site coefcients for each station). Finds equations predict similar
PGA to those predicted by original equation. Finds signicant differences between PGA
recorded at different stations in the two regions some related to surface geology but for
some nds no reason.
Uses 27 records (0.2 R 25.0 km) from 19 earthquakes (2.5 MbLG 4.8,
0.1 H 9 km) from E. N. America to examine whether they are signicantly different
than those from W. N. America. Finds residuals signicantly, at 99% level, higher than
zero and concludes that it is mainly due to site effects because most are on shallow soils
or other site factors inuence ground motion. Correcting the recorded PGAs using site
factors the difference in PGA between E. N. America and W. N. America is no longer
signicant although notes may not hold for all of E. N. America.
2.65 Ordaz et al. (1989)

Ground-motion model is unknown.
60
2.66
Alfaro et al. (1990)
Ground-motion model for near eld is:

1
log(A) = a1 + a2 Ms log(r2 + a23 ) 2
where A is in g, a1 = 1.116, a2 = 0.312, a3 = 7.9 and = 0.21.

Ground-motion model for far eld is:
1
log(A) = b1 + b2 Ms + b3 log(r2 + b24 ) 2
where A is in g, b1 = 1.638, b2 = 0.438, b3 = 1.181, b4 = 70.0 and = 0.21.

Separate crustal and subduction data because of differences in travel path and stress
conditions:
1. Near eld
2. Far eld, 20 records from San Salvador, 20 earthquakes, 4.2 Ms 7.2, depths
between 36 and 94 km, 31 r 298 km.
2.67
Ambraseys (1990)
log y = + Mw log r + br
where r = (d2 + h2 )1/2
where y is in g, = 1.101, = 0.2615, b = 0.00255, h = 7.2 and = 0.25.
Uses data and method of Joyner & Boore (1981) but re-evaluates Mw for all earthquakes. Finds some large changes, e.g. Santa Barbara changes from Mw = 5.1 to
Mw = 5.85. Uses ML for 2 earthquakes (ML = 5.2, 6.2).
Find effect of uncertainty in Mw causes less than 10% change in .
Also calculates equation using Ms instead of Mw .
Finds assumption Ms = Mw introduces bias, particularly for small magnitude shocks,
on unsafe side, and this can be signicant in cases where there is a preponderance of
small earthquakes in set.
2.68
Campbell (1990)
ln(Y ) = a + bM + d ln[R + c1 exp(c2 M )] + eF + f1 tanh[f2 (M + f3 )]

+ g1 tanh(g2 D) + h1 K1 + h2 K2 + h3 K3
where Y is in g, a = 2.245, b = 1.09, c1 = 0.361, c2 = 0.576, d = 1.89, e = 0.218,
f1 = 0, f2 = 0, f3 = 0, g1 = 0, g2 = 0, h1 = 0.137, h2 = 0.403 and h3 = 0.
= 0.517 for M 6.1 and = 0.387 for M 6.2. Also given is = 0.450 for
M 4.7.
61
Records from rm soil and soft rock sites. Characterises site conditions by depth to
basement rock (sediment depth) in km, D.
Records from different size buildings. K1 = 1 for embedded buildings 311 storeys,
K2 = 1 for embedded buildings with >11 storeys and K3 = 1 for non-embedded buildings >2 storeys in height. K1 = K2 = K3 = 0 otherwise.
Uses two fault mechanisms:
F = 0 Strike-slip
F = 1 Reverse
2.69 Dahle et al. (1990b) & Dahle et al. (1990a)

ln A = c1 + c2 M + c4 R + ln G(R, R0 )
where G(R, R0 ) = R1
and: G(R, R0 ) =
R01
for R R0
R0
R
)5/6
for R > R0
where A is in m/s2 , c1 = 1.471, c2 = 0.849, c4 = 0.00418 and = 0.83.

Use records from rock sites (presumably with hard rock or rm ground conditions).
Assume intraplate refers to area that are tectonically stable and geologically more uniform than plate boundary areas. Select records from several reasonably intraplate
areas (eastern N. America, China, Australia, and some parts of Europe), due to lack of
data.
Select records which are available unprocessed and with sufcient information on natural frequency and damping of instrument.
Use Ms , when available, because reasonably unbiased with respect to source dimensions and there is globally consistent calculation method.
Most (72%) records from earthquakes with M 5.5. Tangshan and Friuli sequence
comprise a large subset. Correlation coefcient between magnitude and distance is
0.31.
Instrument correct records and elliptical lter with pass band 0.25 to 25.0 Hz.
If depth unknown assume 15 km.
Choose R0 = 100 km although depends on crustal structure and focal depth. It is
distance at which spherical spreading for S waves overtaken by cylindrical spreading for
Lg waves.
PGA attenuation relation is pseudo-acceleration equation for 0.025 s period and 5%
damping.
Plot residuals against magnitude and distance.
Note rst order results, because data from several geological regions and use limited
data base.
62
2.70
Jacob et al. (1990)
A = 10(a1 +a2 M +a3 log d+a4 d)

where A is in g, a1 = 1.43, a2 = 0.31, a3 = 0.62 and a4 = 0.0026 ( not given).
Note equation only for hard rock sites.
Equation from a composite of two separate regressions: one using data from 6 earthquakes, 4.7 M 6.4 and d primarily between 40 and 820 km and one using the
same data supplemented with data from 2 earthquakes with M = 1.8 and M = 3.2
and d 20 km to extend results to smaller M and d. Give no details of this composite
regression.
Note regressions are preliminary and should be tested against more data.
Note careful assessment of uncertainties is required.
2.71
Sen (1990)
ln PGA = a + bM + c ln(r + h) + F
where PGA is in cm/s2 , a = 1.375, b = 1.672, c = 1.928 and = 0.213 (h not
given). Standard deviation is composed of two parts, inter-site = 0.261 and intra-site
= 0.653. F = 1 for thrust mechanism and 0 otherwise.
Computes theoretical radiation pattern and nds a linear trend between residuals and
radiation pattern but does not model.
2.72
Sigbjrnsson (1990)
apeak = 0 exp(1 M ) exp(2 R)R P

where P = 1.
Notes that data are very limited and any denite conclusions should, therefore, be
avoided.
Does not give coefcients, only predictions.
2.73
Tsai et al. (1990)
ln y = C0 + C1 M + C2 (8.5 M )2.5 + C3 ln[D + C4 exp(C5 M )]

where y is in g, C3 = 2.1, C4 = 0.616, C5 = 0.524 and for M 6.5 C0 = 1.092,
C1 = 1.10, C2 = 0 and = 0.36 and for M < 6.5 C0 = 0.442, C1 = 1.0, C2 = 0 and
= 1.27 0.14M .
63
All records from rock or rock-like sites.

Separate equation for M < 6.5 and M 6.5.
Use only shallow crustal thrust earthquakes.
Use another database of rock and soil site records and simulated acceleration time
histories to nd conversion factors to predict strike-slip and oblique ground motions from
the thrust equation given above. For strike-slip conversion factor is 0.83 and for oblique
conversion factor is 0.91.
Standard deviation, , for M 6.5 from regression whereas for M < 6.5 from previous results. Conrm magnitude dependence of standard deviation using 803 recordings
from 124 earthquakes, 3.8 Mw 7.4, D < 100 km.
2.74 Ambraseys & Bommer (1991) & Ambraseys & Bommer (1992)
log a = + M log r + br
where r = (d2 + h20 )1/2
or: r = (d2 + h2 )1/2
where a is in g, for horizontal PGA = 1.09, = 0.238, b = 0.00050, h = 6.0
and = 0.28 and for vertical PGA = 1.34, = 0.230, b = 0, h = 6.0 and
= 0.27. When use focal depth explicitly: for horizontal PGA = 0.87, = 0.217,
b = 0.00117 and = 0.26 and for vertical PGA = 1.10, = 0.200, b = 0.00015
and = 0.26.
Consider two site classications (without regard to depths of deposits) but do not model:
1. Rock
2. Alluvium
Select records which have: Ms 4.0 and standard deviation of Ms known and reliable
estimates of source-site distance and focal depth, h 25 km, regardless of local soil
conditions from free-eld and bases of small buildings. No reliable data or outliers excluded. Records from instruments at further distances from the source than the closest
non-triggered instrument were non-excluded because of non-homogeneous and irregularly spaced networks and different and unknown trigger levels.
Most data, about 70%, with distances less than 40 km. Note strong bias towards smaller
values of magnitude and PGA.
PGA read from analogue and digitised data, with different levels of processing. Differences due to different processing usually below 5%, but some may be larger.
Errors in distances for small shocks may be large.
Prefer one-stage technique because second step of two-stage method would ignore
records from singly-recorded earthquakes which compose over half the events, also
nd more realistic, b, and h0 using one-stage method. Do not use weighting because
involves assumptions which are difcult to verify.
64
Find inadmissable and positive b for vertical PGA so remove and repeat.
Remove records from distances less than or equal to half their focal depth and also less
than or equal to their focal depth, nd that h0 is governed by near-eld data.
Use focal depth explicitly, by replacing r = (d2 + h20 )1/2 by r = (d2 + h2 )1/2 . Find lower
standard deviation and that it is very signicant.
Repeat analysis on subsets of records grouped by focal depth. Find no correlation
between h0 and focal depth of subset. Use h0 equal to mean focal depth in each subset
and nd similar results to when focal depth used explicitly.
Repeat analysis with geometric attenuation coefcient equal to 0.83, corresponding to
the Airy phase, as opposed to 1.0.
Find small dependence of horizontal PGA on site classication, note due to level of
information available.
2.75
Crouse (1991)
ln PGA = p1 + p2 M + p4 ln[R + p5 exp(p6 M )] + p7 h

where PGA is in gal, using all PGA values p1 = 6.36, p2 = 1.76, p4 = 2.73, p5 = 1.58,
p6 = 0.608, p7 = 0.00916 and = 0.773.
Use data from stiff soil sites (depth of soil < 25 m).
Include data from any zones with strong seismic coupling, such as the younger subduction zones (S.W. Japan, Alaska, C. America (Mexico), C. Chile, Peru and northern
Honshu and Kuril subduction zones in Japan) unless compelling reasons to exclude
data. Do this because lack of data from Cascadia. Most (> 70%) are from Japan.
Focal depths, h, between 0 and 238 km.
Compare Japanese and Cascadia PGA values for earthquakes with similar magnitude
and depths and nd similar.
Do not exclude data from buildings or which triggered on S-wave. Note could mean
some PGAs are underestimated.
Plot ground motion amplitude (PGA and also some maximum displacements from seismograms) against distance for a number of large magnitude shocks (including some
data from rock sites which not included in set for regression). Find that rate of attenuation becomes smaller for shorter distances and process is magnitude dependent. Also
plot Japanese PGA data, from earthquakes with h 50 km, split into three distance
groups (between 50 and 75 km, between 100 and 150 km and between 250 and 300 km)
nd as distance increases magnitude scaling becomes larger and possible saturation in
PGA for large magnitudes. Fit ln PGA = p1 + p2 ln(R + C) to some PGA values from
large magnitude shocks for C = 0 and C > 0, nd lower standard deviation for C > 0.
65
Fit ln PGA = a + bM and ln PGA = a + bM + cM 2 to Japanese data split into the three
distance groups (mentioned above); nd b increases with increasing distance range but
both equations t data equally well.
Constrain p4 to negative value and p5 and p6 to positive values.
Include quadratic magnitude term, p3 M 2 , but nd equal to zero.
Plot residuals against M ; nd uniformly distributed and evidence for smaller residuals
for larger M .
Plot residuals against R4 and nd decreasing residuals for increasing R.
Give equation using only those records available in digital form (235 records).
2.76 Garca-Fernndez & Canas (1991) & Garcia-Fernandez & Canas

(1995)
ln PGA = ln C0 + C1 M 0.5 ln r r
where PGA is in cm/s2 , for Iberian Peninsula ln C0 = 5.13, C1 = 2.12 and =
0.0039, for NE region ln C0 = 4.74, C1 = 2.07 and = 0.0110 and for SSE region
ln C0 = 5.30, C1 = 2.21 and = 0.0175 ( is not given).
Derive equations for two regions:
SSE South south-east part of the Iberian peninsula, from the Guadalquivir basin to the
Mediterranean Sea, including the Betic Cordillera, 140 records from 5 stations.
NE North-east part of the Iberian peninsula, including the Pyrenees, the Catalan Coastal
Ranges, the Celtiberian chain and the Ebro basin, 107 records from 3 stations.
Use vertical-component short-period analogue records of Lg-waves (which are believed
to have the largest amplitudes for the period range 0.1 to 1s) from regional earthquakes
in Iberian Peninsula.
Processing procedure is: digitise seismogram using irregular sampling rate to get better sampling at peaks and kinks, select baseline, apply cubic spline interpolation and
compare original and digitised seismograms. Next the Fourier amplitude spectrum is
computed and the instrument amplitude response is removed.
Estimate PGA using the maximum value of pseudo-absolute acceleration obtained from
Fourier amplitude spectra. Derived equations are for characteristic frequency of 5 Hz.
Compare estimated PGAs with observed PGAs from ve earthquakes and nd good
agreement.
Use 5 Hz values from Garcia-Fernandez & Canas (1992) and Vives & Canas (1992).
4
66
Not shown in paper.
2.77
Geomatrix Consultants (1991), Sadigh et al. (1993) & Sadigh

et al. (1997)
(
)
ln PGA = C1 + C2 M + C3 ln rrup + C4 eC5 M + C6 ZT
where PGA is in g, for horizontal PGA, rock sites and strike-slip faulting C3 = 0 and
C4 = 2.100, for M 6.5 C1 = 0.624, C2 = 1.0, C5 = 1.29649 and C6 = 0.250
and for M > 6.5, C1 = 1.274, C2 = 1.1, C5 = 0.48451 and C6 = 0.524. For
reverse and thrust earthquakes multiply strike-slip prediction by 1.2. = 1.39 0.14M
for M < 7.21 and = 0.38 for M 7.21. For horizontal PGA and deep soil C2 = 1.0,
C3 = 1.70 and C6 = 0, for strike-slip faulting C1 = 2.17 and for reverse or thrust
faulting C1 = 1.92, for M 6.5 C4 = 2.1863 and C5 = 0.32 and for M > 6.5
C4 = 0.3825 and C5 = 0.5882. = 1.52 0.16M for M 7 and = 0.40 for M = 7.
For vertical PGA, rock sites and strike-slip faulting C3 = 0 and C4 = 2.300, for M
6.5 C1 = 0.430, C2 = 1.0, C5 = 1.2726 and C6 = 0.228 and for M > 6.5, C1 =
1.080, C2 = 1.1, C5 = 0.3524 and C6 = 0.478. For reverse and thrust earthquakes
multiply strike-slip prediction by 1.1 and for oblique faulting multiply by 1.048. = 0.48
for M 6.5, = 3.08 0.40M for 6 < M < 6.5 and = 0.68 for M 6.
Use two site categories (for horizontal motion):
1. Rock: bedrock within about a metre of surface. Note that many such sites are soft
rock with Vs 750 m/s and a strong velocity gradient because of near-surface
weathering and fracturing, 274 records.
2. Deep soil: greater than 20 m of soil over bedrock. Exclude data from very soft soil
sites such as those from San Francisco bay mud, 690 records.
Vertical equations only for rock sites.
Crustal earthquakes dened as those that occur on faults within upper 20 to 25 km of
continental crust.
Use source mechanism: RV=reverse (26+2) ZT = 1 and SS=strike-slip (and some
normal) (89+0) ZT = 0. Classied as RV if rake> 45 and SS if rake< 45 . Find
peak motions from small number of normal faulting earthquakes not to be signicantly
different than peak motions from strike-slip events so were including in SS category.
Records from instruments in instrument shelters near ground surface or in ground oor
of small, light structures.
4 foreign records (1 from Gazli and 3 from Tabas) supplement Californian records.
Separate equations for Mw < 6.5 and Mw 6.5 to account for near-eld saturation
effects and for rock and deep soil sites.
2.78
Huo & Hu (1991)
Ground-motion model is (case II):
log y = C1 + C2 M C4 log[R + C5 exp(C6 M )]

67
where y is in gal, C5 = 0.231 and C6 = 0.626, for rock C1 = 0.894, C2 = 0.563,

C4 = 1.523 and = 0.220 and for soil C1 = 1.135, C2 = 0.462, C4 = 1.322 and
= 0.243 (these coefcients are from regression assuming M and R are without error).
1. Rock
2. Soil
Supplement western USA data in large magnitude range with 25 records from 2 foreign
earthquakes with magnitudes 7.2 and 7.3.
Note that there are uncertainties associated with magnitude and distance and these
should be considered in derivation of attenuation relations.
Develop method, based on weighted consistent least-square regression, which minimizes residual error of all random variables not just residuals between predicted and
measured ground motion. Method considers ground motion, magnitude and distance
to be random variables and also enables inverse of attenuation equation to be used
directly.
Note prediction for R > 100 km may be incorrect due to lack of anelastic attenuation
term.
Use both horizontal components to maintain their actual randomness.
Note most data from moderate magnitude earthquakes and from intermediate distances
therefore result possibly unreliable outside this range.
Use weighted analysis so region of data space with many records are not overemphasized. Use M -R subdivisions of data space: for magnitude M < 5.5, 5.5 M 5.9,
6.0 M 6.4, 6.5 M 6.9, 7.0 M 7.5 and M > 7.5 and for distance R < 3,
3 R 9.9, 10 R 29.9, 30 R 59.9, 60 R 99.9, 100 R 300 and
R > 300 km. Assign equal weight to each subdivision, and any data point in subdivision
i containing ni data has weight 1/ni and then normalise.
To nd C5 and C6 use 316 records from 7 earthquakes (5.6 M 7.2) to t log Y =
m
m
i=1 C2,i Ei C4 log[r+ i=1 R0,i Ei ], where Ei = 1 for ith earthquake and 0 otherwise.
Then t R0 = C5 exp(C6 M ) to results.
Also try equations: log y = C1 +C2 M C4 log[R+C5 ] (case I) and log y = C1 +C2 M
C3 M 2 C4 log[R + C5 exp(C6 M )] (case III) for M Mc , where impose condition
C3 = (C2 C4 C6 / ln 10)/(2Mc ) so ground motion is completely saturated at M = Mc
(assume Mc = 8.0).
Find equations for rock and soil separately and for both combined.
2.79 I.M. Idriss (1991) reported in Idriss (1993)

ln(Y ) = [0 + exp(1 + 2 M )] + [0 exp(1 + 2 M )] ln(R + 20) + aF

68
where Y is in g, a = 0.2, for M 6 0 = 0.150, 1 = 2.261, 2 = 0.083, 0 = 0,

1 = 1.602, 2 = 0.142 and = 1.39 0.14M and for M > 6 0 = 0.050,
1 = 3.477, 2 = 0.284, 0 = 0, 1 = 2.475, 2 = 0.286 and for M < 7 14
= 1.39 0.14M and for M 7 14 = 0.38.
Records from rock sites.
Uses three fault mechanisms:
F=0 Strike slip
F=0.5 Oblique
F=1 Reverse
Separate equations for M 6 and M > 6.
Examines residuals for PGA. Finds average residual almost zero over entire distance
range; trend reasonable up to about 60 km but beyond 60 km relationship would underestimate recorded PGA.
Finds standard deviation to be linear function of magnitude.
2.80
Loh et al. (1991)
a = b1 eb2 M (R + b4 )b3
where a is in g, b1 = 1.128, b2 = 0.728, b3 = 1.743, b4 = 32 km and = 0.563 (in terms

of ln).
Use only data from rock sites.
Focal depths, h, between 0.2 and 97.4 km. Most records from h < 30 km.
Also derive equations for PGA using log10 (a) = b1 + b2 M + b3 log R2 + b25 and a =
b1 eb2 M (R+b4 eb5 M )b3 in order to have diversity in the characterisation of ground motion.
Use rhypo because no clear fault ruptures identied for Taiwanese earthquakes.
All data from SMA-1s.
PGAs between 7.3 and 360.2 cm/s2 .
2.81
Matuschka & Davis (1991)
Exact functional form unknown but based on those of Campbell (1981), Fukushima &
Tanaka (1990) and Abrahamson & Litehiser (1989).
Use three site classes.
Develop separate equations for each site class. Only possible for two classes. Therefore, modify equation derived for site class C to obtain coefcients for other two classes.
Digitization sampling rate of records used is 50 Hz. Most data low-pass ltered at
24.5 Hz.
69
Most data high-pass ltered with cut-offs above 0.25 Hz.

Due to limited data, advise caution when using model.
2.82 Niazi & Bozorgnia (1991)

ln Y = a + bM + d ln[R + c1 ec2 M ]
where Y is in g, for horizontal PGA a = 5.503, b = 0.936, c1 = 0.407, c2 = 0.455,
d = 0.816 and = 0.461 and for vertical PGA a = 5.960, b = 0.989, c1 = 0.013,
c2 = 0.741, d = 1.005 and = 0.551.
All records from SMART-1 array so essentially identical site conditions and travel paths.
All records from free-eld instruments mounted on 4inch (10 cm) thick concrete base
mats, approximately 2 by 3 feet (60 by 90 cm) across.
Select earthquakes to cover a broad range of magnitude, distance and azimuth and
ensuring thorough coverage of the array. Criteria for selection is: at least 25 stations
recorded shock, focal depth < 30 km, hypocentral distance < 50 km except for two
large earthquakes from beyond 50 km to constrain distance dependence.
Focal depths between 0.2 and 27.2 km with all but one 13.9 km.
Azimuths between 60 and 230 .
Most records (78%) have magnitudes between 5.9 and 6.5. Note magnitude and distance are not independent (correlation coefcient is 0.6).
Records have sampling interval of 0.01 s. Processed using trapezoidal band passed
lter with corner frequencies 0.07, 0.10, 25.0 and 30.6 Hz.
Not enough information to use distance to rupture zone.
Source mechanisms of earthquakes are: 4 normal, 2 reverse, 1 reverse oblique and 1
normal oblique with 4 unknown. Do not model source mechanism dependence because
of 4 unknown mechanisms.
Use weighted regression, give equal weight to recordings from each earthquake within
each of 10 distance bins (< 2.5, 2.55.0, 5.07.5, 7.510.0, 10.014.1, 14.120.0, 20
28.3, 28.340.0, 40.056.6 and 56.6130 km). Do this so earthquakes with smaller number of recordings are not overwhelmed by those with a larger coverage and also to
give additional weight to shocks recorded over multiple distance bins. Apply two-stage
regression, because of high correlation between magnitude and distance, excluding 3
earthquakes (M = 3.6, 5.0, 7.8) with 162 records from rst stage to reduce correlation
between M and R to 0.1. Also do one-stage regression although do not give coefcients.
Use mean horizontal component because reduces uncertainty in prediction.
70
Examine coefcient of variation for each earthquake using median and normalized standard deviation of recordings in inner ring of array. Find evidence for magnitude dependent uncertainty (large magnitude shocks show less uncertainty). Find that main contribution to scatter is inter-event variations again by examining coefcient of variation;
although note may be because using dense array data.
Examine mean residuals of observations from each earthquake. Find evidence for
higher than predicted vertical PGA from reverse faulting earthquakes and lower than
predicted vertical PGA from normal faulting earthquakes, although due to lack of information for 4 earthquakes note that difcult to draw any conclusions.
Examine mean residuals of observations from each station in inner ring. Find mean
residuals are relatively small compared with standard deviation of regression so variation
between stations is less than variation between earthquakes. Find for some stations
some large residuals.
2.83
Rogers et al. (1991)
log ap = a1 +0.36M 0.002R+a2 log R+a3 S1 +a4 S1 log R+a5 S5 +a6 S5 log R+a7 S6 log R
where ap is in g, a1 = 1.62, a2 = 1.01, a3 = 0.246, a4 = 0.212, a5 = 0.59,
a6 = 0.29, a7 = 0.21 and = 0.29.
Use six local site classications:
S1 Holocene
S2 Pleistocene soil
S3 Soft rock
S4 Hard rock
S5 Shallow (< 10 m depth) soil
S6 Soft soil (e.g. bay mud)
Data from about 800 different stations.
Note that inclusion of subduction-zone events in analysis may affect results with unmodelled behaviour, particularly with regard to distance scaling although believe use of rrup
partially mitigates this problem.
Firstly compute an equation does not include site coefcients. Conduct regression analysis on site-condition subsets of the residuals using M or log R as dependent variable.
Find several regressions are not statistically signicant at the 5% level and/or the predicted effects are small at the independent variable extremes. Find strongest effects and
most signicant results are for shallow soil sites and soft soil sites although because of
the high correlation between M and log R in the set used it is difcult to construct unbiased models.
Use a stochastic random-vibration approach to nd theoretical equations for estimating
PGA that include the effect of local site conditions as distance-dependent terms. Using
the results from this analysis construct equation based on the observed PGAs. Try
71
including terms for S1 , S2 , S5 , S6 and corresponding log R terms for each site type but
iterate to retain only the signicant terms.
Fix magnitude scaling (0.36M ) and anelastic attenuation (0.002R). Do not try to optimise
the t other than using xed values similar to those given by the stochastic analysis.
Note that anelastic coefcient may be too low but it produces an acceptable geometric
spreading term.
Note that because Moho critical reections can increase amplitudes beyond about 50 km
the effects of anelastic or geometric attenuation may be masked.
Allowing all the coefcients in the equation to be free produces a smaller magnitude scaling coefcient, a smaller geometric spreading coefcient, and a non-signicant anelastic
attenuation term.
Note that data from S5 and S6 are sparse.
Compare estimated PGAs with data from within small magnitude ranges. Find that PGAs
from Morgan Hill earthquake are overestimated, which believe is due to the unilateral
rupture of this earthquake masking the effect of the local site conditions.
2.84 Stamatovska & Petrovski (1991)

Acc = b1 exp(b2 M )(Rh + c)b3

Acc is in cm/s2 , b1 = 534.355, b2 = 0.46087, b3 = 1.14459, c = 25 and ln Acc =
0.72936.
Data from 141 different sites, which are considered to have average soil conditions.
Data from Yugoslavia (23 earthquakes), Italy (45 earthquakes), northern Greece (3
earthquakes), Romania (1 earthquake), Mexico (1 earthquake) and the USA (5 earthquakes). Select earthquakes to have range of magnitudes and focal depths.
Data processed using standard procedure.
Conduct Pearson 2 and Kolmogorov-Smirnov tests to test acceptability of log-normal
assumption using a 5% signicance level. Conclude that assumption is justied.
Note the strong inuence of the data used on results and the need to improve it.
2.85 Abrahamson & Youngs (1992)

ln y = a + bM + d ln(r + c) + eF
where a = 0.0586, b = 0.696, c = 12.0, d = 1.858, e = 0.205, = 0.399 (intra-event)
and = 0.201 (inter-event) (units of y are not given but probably g).
F is fault type (details not given).
Develop new algorithm for one-stage maximum-likelihood regression, which is more
robust than previous algorithms.
72
2.86
Ambraseys et al. (1992)
log(a) = c1 + c2 M + c3 r + c4 log r
1
r = (d2 + h20 ) 2
where a is in g, c1 = 1.038, c2 = 0.220, c3 = 0.00149, c4 = 0.895, h0 = 5.7 and

= 0.260.
Investigate equations of PML (1982) and PML (1985) using criteria:
1. Is the chosen data set of earthquake strong-motion records suitable to represent
the UK seismic environment?
2. Are the associated seismological and geophysical parameters used in these reports reliable and consistent?
3. Is the methodology used to derive attenuation laws and design spectra from the
data set reliable?
Investigate effect of different Ground-motion model, one and two-stage regression technique, record selection technique and recalculation of associated parameters. Find
these choice cause large differences in predictions.
Coefcients given above are for PML (1985) data with recalculated magnitudes and
distances and addition of extra records from some earthquakes.
2.87
Kamiyama et al. (1992) & Kamiyama (1995)
Ground-motion model is (note that there is a typographical error in Kamiyama et al.

(1992); Kamiyama (1995) because rt has been replaced by rc in equations):
log10 amax = 1.64R0 + b1 R1 + b2 R2 + ca +

{
R0 =
{
R1 =
{
R2 =
N
1
Ai Si
i=1
0 for r rt
log10 r log10 rc for r > rt
0 for r rt
1 for r > rt
0 for r rt
M for r > rt
where Si = 1 for i station, S0 = 0 otherwise, amax is in cm/s2 , b1 = 1.164, b2 = 0.358,

ca = 2.91, rc = 5.3 km and = 0.247 (Ai given in publications but not reported here
due to lack of space).
Instrument correct records and lter with pass band between 0.24 and 11 Hz.
Model individual soil conditions at each site as amplication factors, AMPi , as described
by Kamiyama & Yanagisawa (1986).
Most records are from hypocentral distances between 30 and 200 km.
73

Models peak ground accelerations independent of magnitude and distance in a fault
zone, rt , where rt = rc 10(b1 +b2 M )/1.64 .
Constrain decay with distance in far eld to 1.64 using results from other studies to
avoid problems due to correlation between M and log10 r.
Use trial and error method to nd rc so that resulting values of rt are consistent with
empirical estimates of fault length from past studies.
Also give expression using shortest distance to fault plane (rupture distance), R, by
replacing the expression for r rc and r > rc by one expression given by replacing r,
hypocentral distance, by R + rc in expression for r > rc . This gives PGA independent
of magnitude at distance R = 0 km.
Note that use of rhypo is not necessarily best choice but use it due to simplicity.
Check residual plots; nd no trends so conclude adequate from statistical point of view.
2.88 Sigbjrnsson & Baldvinsson (1992)

log A = + M log R + bR
d2 + h2
with: R =
where A is in g, for average horizontal PGA and 4 < M < 6 = 1.98, = 0.365,
b = 0.0039 and = 0.30, for larger horizontal PGA and 4 < M < 6 = 1.72,
= 0.327, b = 0.0043 and = 0.30 and for both horizontal PGAs and 2 < M < 6
= 2.28, = 0.386, b = 0 and = 0.29.
Find that Icelandic data does not t other published relations.
Find equation using only records with M 4.0, h equal to focal depth and both the
horizontal components.
Find equation using only records with M 4.0, h equal to focal depth and larger horizontal component.
Also repeated with all data. Anelastic coefcient constrained to zero because otherwise
positive.
Also done with h free.
Note that large earthquakes have h 10 km while small events have h 5 km.
2.89 Silva & Abrahamson (1992)

ln pga = c1 + 1.2M + c3 ln(r + 20) + 0.25F

where pga is in g, c1 = 3.27, c3 = 1.79 and total = 0.46 for deep soil and c1 =
3.56, c3 = 1.67 and total = 0.46 for rock/shallow soil.
74
Originally use ve site classes (chosen based on site response analyses using broad
categories and generic site proles):
1. Rock. 66 records
2. Shallow soil (< 250 ft. 6 records.)
3. Intermediate depth soil (2501000 ft). 2 records.
4. Deep soil (> 1000 ft). 51 records.
5. Alluvium of unknown depth. 10 records.
but insufcient records in shallow and intermediate classes to evaluate separately so
combine rock and shallow classes and intermediate, deep and unknown depth categories to leave two classes: < 250 ft and > 250 ft.
Use two faulting mechanisms:
F = 0 Strike-slip
F = 1 Reverse or oblique
Process data by: 1) interpolation of uncorrected unevenly sampled records to 400 samples per second; 2) frequency domain low-pass ltering using a causal ve-pole Butterworth lter with corner frequencies selected based on visual examination of Fourier
amplitude spectrum; 3) removal of instrument response; 4) decimation to 100 or 200
samples per second depending on low-pass lter corner frequencies; and 5) application
of time-domain baseline correction, using polynomials of degrees zero to ten depending on integrated displacements, and nal high-pass lter chosen based on integrated
displacements that is at at corner frequency and falls off proportional to frequency on
either side, which is applied in the time domain twice (forward and backwards) to result
in zero phase shift.
Note that due to limited magnitude range of data, magnitude dependence is not well
constrained nor is dependency on mechanism. Hence these coefcients are xed based
on previous studies.
Plot residuals w.r.t. distance. Find slight increase at 70100 km. To test if due to Moho
bounce repeat regression assuming functional form that is at between 70 and 90 km
but this produced a smaller likelihood. Conclude that data does not support signicant
attening at < 100 km.
Note that model is preliminary.
2.90
Taylor Castillo et al. (1992)
ln(A) = a1 + a2 Ms + a3 ln(R) + a4 R
where A is in m/s2 , a1 = 0.339, a2 = 0.455, a3 = 0.67, a4 = 0.00207 and = 0.61.
75
2.91 Tento et al. (1992)

ln PGA = b1 + b2 M + b3 R ln R
where R = (d2 + h2 )1/2
where PGA is in gal, b1 = 4.73, b2 = 0.52, b3 = 0.00216, h is mean focal depth of
group into which each earthquake is classied and = 0.67.
Most records from distances between 10 km and 40 km.
Correction technique based on uniform Caltech correction procedure. Most (125) were
automatically digitised, rest were manually digitised. Roll-on and cutoff frequencies of
Ormsby lter were selected by adopting a record dependent criteria. Cutoff frequencies
range between 0.13 Hz and 1.18 Hz with a median of 0.38 Hz.
Records included from analysis were from free-eld stations. Excluded those not complete (e.g. started during strong-motion phase). Excluded those with epicentral distances greater than that of rst nontriggered station.
Note relatively small inuence of form of equation adopted although two step method
seems preferable.
Note correction procedure plays a relevant role in analysis.
Note using d instead of R causes greater scatter in data.
Note moderate underestimation for low magnitude in near eld and for high magnitude
in far eld.
2.92 Theodulidis & Papazachos (1992)

ln Y = C1 + C2 M + C3 ln(R + R0 ) + C4 S
where Y is in cm/s2 , C1 = 3.88, C2 = 1.12, C3 = 1.65, R0 = 15, C4 = 0.41 and
= 0.71.
Use two site categories (mean opinion of seven specialists who classied sites into three
categories: soft alluvium, crystalline rock and intermediate):
S=1 Rock: 34+4 records. Japanese sites have diluvium with depth to bedrock H <
10 m. Alaskan sites have PGV/PGA 66 7 cms1 g1 .
S=0 Alluvium: 71+12 records. Japanese sites have diluvium H > 10 m or alluvium
H < 10 m, and alluvium with H < 25 m as well as soft layers with thickness
< 5 m. Alaskan sites have PGV/PGA > 66 7 cms1 g1 .
70% of records from ground level or basement of buildings with two storeys or less. Rest
from buildings with up to eight storeys.
76
Some (16) Greek records manually digitized and baseline corrected, some (22) Greek
records manually digitized and ltered and rest of the Greek records automatically digitized and ltered.
Due to lack of data for 7.0 < Ms < 7.5 include shallow subduction data from other
regions with similar seismotectonic environments (Japan and Alaska) using criteria i)
depth < 35 km, ii) Mw or MJMA between 7.0 and 7.5, iii) instruments triggered before
S-wave, iv) free-eld recording, v) surface geology known at station. Note Ms , Mw and
MJMA are equivalent between 6.0 and 8.0.
Focal depths between 0 km (13 km) and 18 km (31 km).
Most data from Ms < 5.5 and from R < 50 km.
Use four step regression procedure. First step use only Greek data from Ms > 6.0 (9
R 128 km, 14 records) for which distances are more reliable (use both hypocentral
and epicentral distance nd epicentral distance gives smaller standard deviation) to nd
geometrical coefcient C31 and R0 ignoring soil conditions. Next nd constant (C12 ),
magnitude (C22 ) and soil (C42 ) coefcients using all data. Next recalculate geometrical
(C33 ) coefcient using only Greek data with Ms > 6.0. Finally nd constant (C14 ),
magnitude (C24 ) and soil (C44 ) coefcients using all the data; nal coefcients are C14 ,
C24 , C33 and C44 .
Plot residuals against Ms and R and nd no apparent trends. Find residuals (binned
into 0.2 intervals) t normal distribution.
2.93
Abrahamson & Silva (1993)
ln pgarock = 1 + 2 M + 3 ln[r + exp(4 + 5 M )] + 11 F1

ln pgasoil = 6 + 7 M + 8 ln[r + exp(9 + 10 )] + 11 F1
where pga is in g, 1 = 4.364, 2 = 1.016, 3 = 1.285, 4 = 3.34, 5 = 0.79,
6 = 8.698, 7 = 1.654, 8 = 1.166, 9 = 6.80, 10 = 1.40, 11 = 0.17, = 0.44,
= 0.00 (sic) and total = 0.44.
Originally use ve site classes (chosen based on site response analyses using broad
categories and generic site proles):
1. Rock. 78 records
2. Shallow soil (< 250 ft. 25 records.)
3. Intermediate depth soil (2501000 ft). 5 records.
4. Deep soil (> 1000 ft). 62 records.
5. Alluvium of unknown depth. 31 records.
but insufcient records in shallow and intermediate classes to evaluate separately so
combine rock and shallow classes and intermediate, deep and unknown depth categories to leave two classes: < 250 ft and > 250 ft.
77
F1 = 0 Strike-slip or normal
F1 = 1 Reverse
Based on Silva & Abrahamson (1992) (see Section 2.89.
Only use Nahanni records for spectral ordinates and not PGA because more representative of eastern US rock than western US rock.
2.94 Boore et al. (1993), Boore et al. (1997) & Boore (2005)
log Y
= b1 + b2 (M 6) + b3 (M 6)2 + b4 r + b5 log r + b6 GB + b7 GC
where r = (d2 + h2 )1/2

where Y is in g, for randomly-oriented horizontal component (or geometrical mean)
b1 = 0.105, b2 = 0.229, b3 = 0, b4 = 0, b5 = 0.778, b6 = 0.162, b7 = 0.251, h = 5.57
and = 0.230 (for geometrical mean = 0.208) and for larger horizontal component
b1 = 0.038, b2 = 0.216, b3 = 0, b4 = 0, b5 = 0.777, b6 = 0.158, b7 = 0.254, h = 5.48
and = 0.205.
Due to an error in Equation (3) of Boore et al. (1994a) and Equation (6) of Boore
et
al. (1997) c reported in Boore et al. (1994a, 1997) are too large by a factor of
2. Therefore correct values of standard deviations are: f = 0.431, c = 0.160,
r = 0.460, s = 0.184 and ln Y = 0.495.
Class A Vs,30 > 750 m/s, some categorised using measured shear-wave velocity, most
estimated GB = 0, GC = 0, 48 records
Class B 360 < Vs,30 750 m/s, some categorised using measured shear-wave velocity,
most estimated GB = 1, GC = 0, 118 records.
Class C 180 < Vs,30 360 m/s,some categorised using measured shear-wave velocity,
most estimated GB = 0, GC = 1, 105 records.
where Vs,30 is average shear-wave velocity to 30 m.
Dene shallow earthquakes as those for which fault rupture lies mainly above a depth
of 20 km.
Peak acceleration scaled directly from accelerograms, in order to avoid bias from sparsely
sampled older data.
Do not use data from structures three storeys or higher, from dam abutments or from
base of bridge columns. Do not use data from more than one station with the same site
condition within a circle of radius 1 km (note that this is a somewhat arbitrary choice).
Exclude records triggered by S wave.
Do not use data beyond cutoff distance which is dened as equal to lesser of distance to
the rst record triggered by S wave and closest distance to an operational nontriggered
instrument.
78
Note that little data beyond 80 km.

Due to positive values of b4 when b5 = 1, set b4 to zero and let b5 vary.
2.95
Campbell (1993)
ln(Y ) = 0 + a1 M + 1 tanh[a2 (M 4.7)] ln(R2 + [a3 exp(a1 M )]2 )1/2

(4 + 5 M )R + a4 F + [2 + a5 ln(R)]S + 3 tanh(a6 D)
where Y is in g, 0 = 3.15, 1 = 0, 2 = 0, 3 = 0, 4 = 0.0150, 5 = 0.000995,
a1 = 0.683, a2 = 0.647, a3 = 0.0586, a4 = 0.27, a5 = 0.105, a6 = 0.620 and
= 0.50.
S=0 Quaternary deposits (soil).
S=1 Tertiary or older sedimentary, metamorphic, and igneous deposits (rock).
Also includes depth to basement rock ( km), D.
Uses two fault mechanisms:
F=0 Strike-slip.
F=1 Reverse, reverse-oblique, thrust, and thrust-oblique.
Recommends use F = 0.5 for normal or unknown mechanisms.
Gives estimates of average minimum depths to top of seismogenic rupture zone.
Uses stochastic simulation model to nd anelastic coefcients 4 and 5 because uses
only near-source records.
Uses weighted nonlinear regression method based on Campbell (1981) to control dominance of well-recorded earthquakes.
2.96
Dowrick & Sritharan (1993)
where r = (d2 + h2 )1/2
Coefcients are unknown.
Data from earthquakes occurring between 1987 and 1991.
79
2.97 Gitterman et al. (1993)

log Y = a + bM log
r2 + h2 cr
where Y is in g, a = 5.026, b = 0.989, h = 2.7 and c = 0.00443 ( not reported).

Some data from velocity sensors have been used, after differentiation, to increase
amount of data at moderate and long distances.
2.98 McVerry et al. (1993) & McVerry et al. (1995)

Ground-motion model is (Type A):
log10 PGA = a + bMw cr d log10 r

where P GA is in g, a = 1.434 0.339, b = 0.209 0.036, c = 0.00297 0.00093,
d = 0.449 0.186 and = 0.276.
Find that ground motions in previous earthquakes were signicantly higher than the
motions predicted by equations derived from W. N. America data.
Only include records from earthquakes for which Mw is known because of poor correlation between ML and Mw in New Zealand.
Focal depths, he 122 km.
140 records from reverse faulting earthquakes.
Divide records into crustal and deep earthquakes.
Only use records for which reliable event information is available, regardless of their
distances with respect to untriggered instruments.
Only use records which triggered on the P-wave.
Also derive separate equations for shallow, upper crustal earthquakes (he 20 km, 102
records, 5.1 Mw 7.3, 13 r 274 km) and crustal earthquakes (he 50 km, 169
records, 5.1 Mw 7.3, 13 r 274 km).
Also try equations of form: log10 PGA = a + bMw d log10 r (Type B) and log10 PGA =
a + bMw cr log10 r (Type C) because of large standard errors and highly correlated
estimates for some of the coefcients (particularly c and d). Find Type B usually gives
much reduced standard errors for d than Type A model and have lowest correlation
between coefcients, but are sceptical of extrapolating to distance ranges shorter and
longer than the range of data. Type C usually has similar standard deviations to Type A.
Find that usually all three models give similar predictions over distance range of most of
the data, but sometimes considerably different values at other distances.
Derive separate equations for reverse faulting earthquakes only and usually nd similar
results to the combined equations.
Find deep earthquakes produce signicantly higher PGAs than shallow earthquakes for
similar r.
80
2.99
Singh et al. (1993)
log(A) = a1 + a2 M + a3 log[G(R0 )] + a4 R0
where R02
= R2 + (ea5 M )2
G(R0 ) = R0 for: R0 100 km
and: G(R0 ) =
(100R0 ) for: R0 > 100 km
where A is in cm/s2 , a1 = 2.74, a2 = 0.212, a3 = 0.99, a4 = 0.000943, a5 = 0.47
and = 0.26.
Use same data as Taylor Castillo et al. (1992).
Employ several different regression techniques.
Select equation found by Bayesian method (given above) for hazard study.
2.100
Steinberg et al. (1993)
log(Amax ) = a1 M + a2 log(D + a3 ) + a4
where Amax is in cm/s2 , a1 = 0.54, a2 = 1.5, a3 = 10 and a4 = 1.25 ( not reported).
2.101
Sun & Peng (1993)
ln A = a + bM c ln(R + h) + dTs
where A is in cm/s2 , a = 7.7, b = 0.49, c = 1.45, d = 0.19, h = 25.0 and = 0.46.
Model soil using its fundamental period of the overburden soil, Ts . Thickness of deposit dened as depth to rock base, dened either as Vs > 800 m/s or when ratio of
shear-wave velocity in ith layer to shear-wave velocity in i 1th layer is greater than 2
(only calculate period to 100 m because only have important effect on structure). For
outcropping rock, Ts = 0.05 s.
Eight distance intervals used for weighting, ve 10 km wide up to 50 km, 5069.9 km, 70
99.9 km and 100200 km. Within each interval each earthquake received equal weight,
inversely proportional to number of records from that earthquake in interval.
Use resolve accelerations in direction, , which gives largest value. Find scatter is lower
than for larger horizontal component.
Many (27) earthquakes only have one record associated with them and 60 records are
from San Fernando.
81
2.102 Ambraseys & Srbulov (1994)

log a = b1 + b2 Ms + b3 r + b4 log r
where r = (d2 + h20 )0.5
where a is in g, b1 = 1.58, b2 = 0.260, b3 = 0.00346, b4 = 0.625, h0 = 4 and
= 0.26.
Do not consider effect of site geology but expect it to be statistically insignicant for
PGA.
Focal depths, h < 25 km. Mean focal depth is 10 4 km.
Mean magnitude of earthquakes considered is 6.0 0.7.
Most records from d < 100 km.
Only use records with PGA > 0.01 g.
Records mainly from SMA-1s located at ground oor or in basements of buildings and
structures and free-eld sites regardless of topography.
Records from thrust earthquakes (46% of total), normal earthquakes (26%) and strikeslip earthquakes (28%).
Baseline correct and low-pass lter records. Select cut-offs from visual examination
of Fourier amplitude spectrum of uncorrected time-histories and choose cut-off below
which the Fourier amplitude spectrum showed an unrealistic energy increase due to
digitization noise and instrument distortions.
Find (from reprocessing about 300 records) that with very few exceptions differences in
PGAs arising from different methods of processing are not signicant, remaining below
3%.
Also derive equation which includes focal depth explicitly.
2.103 Boore et al. (1994a) & Boore et al. (1997)

Based on Boore et al. (1993) see Section 2.94
log Y
= b1 + b2 (M 6) + b3 (M 6)2 + b4 r + b5 log r + bV (log VS log VA )
where r = (d2 + h2 )1/2

where Y is in g, b1 to b5 , h and are same as for Boore et al. (1993) (see Section 2.94)
and for randomly oriented component bV = 0.371 and VA = 1400 and for larger
horizontal component bV = 0.364 and VA = 1390.
Model site effect as a continuous function of average shear-wave velocity to 30 m deep,
VS .
82
Coefcients b1 , b2 , b3 ,b4 and b5 from Boore et al. (1993).

Find no basis for different magnitude scaling at different distances.
Find evidence for magnitude dependent uncertainty.
Find evidence for amplitude dependent uncertainty.
Find marginal statistical signicance for a difference between strike-slip (dened as
those with a rake angle within 30 of horizontal) and reverse-slip motions but do not
model it. Modelled in Boore et al. (1994b) (by replacing b1 by bSS GSS + bRS GRS where
GSS = 1 for strike-slip shocks and 0 otherwise and GRS = 1 for reverse-slip shocks
and 0 otherwise) and reported in Boore et al. (1997). Coefcients for randomly oriented
horizontal component are: bSS = 0.136 and bRS = 0.0515 .
Analysis done using one and two-stage maximum likelihood methods; note that results
are very similar.
Earthquakes with magnitudes below 6.0 are poorly represented.
Note that few Class A records.
Note that VS does not model all the effects of site because it does not model effect of
the thickness of attenuating material on motion.
Note that ideally would like to model site in terms of average shear-wave velocity to
one-quarter wavelength.
Note lack measurements from distances greater than 100 km so that weak-motion data
from seismographic stations maybe should be used.
Note that use of cutoff distances independent of geology or azimuth may be over strict
but it is simple and objective. Note that methods based on data from nontriggered
stations or using seismogram data may be better.
2.104
El Hassan (1994)
log a = C1 + C2 M + C3 log(R + C4 )
where a is in cm/s2 , C1 = 8.65, C2 = 0.71, C3 = 1.6, C4 = 40 and = 0.6.
May not be an empirical GMPE but derived through a intensity-PGA relations.
5
These are taken from Table 8 of Boore et al. (1997) which uses natural logarithms so they were converted
into terms of logarithms to base 10.
83
2.105 Fukushima et al. (1994) & Fukushima et al. (1995)

log Y = aM + bX log X +
i ci
where Y is in cm/s2 , i = 1 at ith receiver and 0 otherwise, for horizontal PGA a = 0.918
and b = 0.00846 ( not given) and for vertical PGA a = 0.865 and b = 0.00741 (
not given). ci given in paper but are not reported here due to lack of space.
Data from three vertical arrays in Japan so predictions at surface and at different depths
down to 950 m.
Different denition of MJMA for focal depths > 60 km so exclude such data. Focal
depths between 2 and 60 km.
Exclude data from earthquakes M < 5.0 because errors are larger for smaller events.
Exclude data for which predicted, using a previous attenuation relation, PGV < 0.1 cm/s
in order to nd precise attenuation rate.
Most data from earthquakes with M 6.0 and most from X 100 km.
Records low-pass ltered with cutoff frequency 25 Hz for records from 2 sites and 30 Hz
for records from 1 site.
Use two-stage method because positive correlation between M and X . Also apply one
step; nd it is biased and two-stage method is most effective method to correct bias.
Check residuals (not shown) against M and X nd no remarkable bias.
2.106 Lawson & Krawinkler (1994)

log Y = a + b(M 6) + c(M 6)2 + d
R2 + h2 + e log
R2 + h2 + f SB + gSC

A Firm to hard rock: granite, igneous rocks, sandstones and shales with close to
widely spaced fractures, 750 Vs,30 1400 m/s SB = 0, SC = 0.
B Gravelly soils and soft to rm rocks: soft igneous rocks, sandstones and shales,
gravels and soils with > 20% gravel, 360 Vs,30 750 m/s SB = 1 , SC = 0.
C Stiff clays and sandy soils: loose to very dense sands, silt loams and sandy clays,
and medium stiff to hard clay and silty clays (N > 5 blows/ft), 180 Vs,30
360 m/s SB = 0, SC = 1.
For shallow (fault rupture within 20 km of earth surface) crustal earthquakes.
Use free-eld records. Records not signicantly contaminated by structural feedback,
excludes records from structures with >2 stories.
Chooses Ground-motion model because of simplicity. Note that other possible forms of
equation may have signicant effect on results, but including more terms complicates
relationships without reducing variability.
Do not give coefcients only predictions.
84
2.107
Lungu et al. (1994)
ln PGA = c1 + c2 Mw + c3 ln R + c4 h
where PGA is in g, c1 = 2.122, c2 = 1.885, c3 = 1.011, c4 = 0.012 and = 0.502.
Focal depth, h, between 79 and 131 km.
Consider to separate areas of 90 to investigate variation with respect to azimuth; nd
azimuthal dependence.
Find individual attenuation equations for three earthquakes. Note faster attenuation for
smaller magnitude and faster attenuation for deeper events.
2.108
Musson et al. (1994)
Ground-motion model is (model 1):
ln A = a + bM ln(R) + dR
where A is in cm/s2 , a = 2.11, b = 1.23 and d = 0.014.
Ground-motion model is (model 2):
ln A = c1 + c2 M + c4 R + ln G(R, R0 )
and: G(R, R0 ) = R01
for R R0
R0 5/6
for R > R0
R
where A is in m/s2 , c1 and c2 are from Dahle et al. (1990b), c4 = 0.0148 and is
recommended as 0.65 (although this is from an earlier study and is not calculated in
regression).
Use data from Canada (Saguenay earthquake and Nahanni sequence) and Belgium
(Roermond earthquake).
Focal depths, h, between 1 and 30 km with average 14.4 km.
Assume peak ground acceleration equals pseudo-acceleration at 30 Hz due to few unclipped horizontal UK records and because instrument response of UK instruments
means records unreliable above 30 Hz. Use only digital VME records for 30 Hz model.
Note poorness of data due to UK data and other data being widely separated thus preventing a comparison between the two sets. Also means straightforward regression
methods would be inadequate as there would be little control on shape of curves derived.
Note earlier models over predict UK data.
Use two-stage least squares method to give model 1. First stage t only UK/Belgian
data to nd b, in second stage use this value of b and use all data to nd a and d.
85
Do not recommend model 1 for general use because too inuenced by limitations of
data to be considered reliable. Canadian data probably insufcient to anchor curves at
small R/large M and extremely high Saguenay earthquake records carry undue weight.
Use model of Dahle et al. (1990b) to get model 2. Fix c1 and c2 to those of Dahle et al.
(1990b) and nd c4 . Prefer this model.
2.109 Radu et al. (1994), Lungu et al. (1995a) & Lungu et al.
(1996)
ln PGA = c1 + c2 M + c3 ln R + c4 h
where PGA is in cm/s2 , c1 = 5.432, c2 = 1.035, c3 = 1.358, c4 = 0.0072 and
= 0.397.
Sites have different soil conditions, some medium and stiff sites and some very soft soil
sites.
Use some records from Moldova and Bulgaria.
Records from free-eld or from basements of buildings.
Originally include data from a shallower (focal depth 79 km), smaller magnitude (ML =
6.1, Mw = 6.3) earthquake with shorter return period than other three earthquakes, but
exclude in nal analysis.
Originally do attenuation analysis for two orthogonal directions N45E (which is in direction of fault plane) and N35E (which is normal to fault plane). From this dene 3 90
circular sectors based roughly on tectonic regions, and calculate attenuation relations
for each of these sectors as well as for all data. Find azimuthal dependence.
Remove 1 to 3 anomalous records per sector.
Remove the only record from the 4/3/1977 earthquake, because it has a strong inuence
on results, and repeat analysis using model ln PGA = b1 + b2 M + b3 ln R, nd lower
predicted PGA.
Find slower attenuation in direction of fault plane compared with normal to fault plane.
Find faster attenuation and larger standard deviation (by nding attenuation equations
for two different earthquakes) for deeper focus and larger magnitude shocks.
2.110 Ramazi & Schenk (1994)

ah = a1 (a2 + d + H)a5 exp(a6 Ms )

H = |d a3 |a4
86
where for horizontal peak acceleration ah is in cm/s2 , a1 = 4000, a2 = 20, a3 = 16 and

a4 = 0.63 for soil sites a5 = 2.02 and a6 = 0.8 and for rock sites a5 = 2.11 and
a6 = 0.79 ( not given). For vertical peak acceleration on soil sites av is in cm/s2 a1 to
a3 are same as horizontal and a4 = 0.48, a5 = 1.75 and a6 = 0.53 ( not given).
Use two site categories (from original of four) for which derive two separate equations:
1. Rock: mainly category (2) a) loose igneous rocks (tuffs), friable sedimentary rocks,
foliated metamorphic rock and rocks which have been loosened by weathering, b)
conglomerate beds, compacted sand and gravel and stiff clay (argillite) beds where
soil thickness > 60 m from bed rock. 29 records.
2. Soil: mainly category (4) a) soft and wet deposits resulting from high level of water
table, b) gravel and sand beds with weak cementation and/or uncementated unindurated clay (clay stone) where soil thickness > 10 m from bed rock. 54 records.
Find equations using hypocentral distance but nd that poor t for Rudbar (Manjil) earthquake (Ms = 7.7) which conclude due to use of hypocentral rather than rupture distance.
Find equations using rupture distance6 for Rudbar (Manjil) earthquake and hypocentral distances for other earthquakes. Coefcients given above. They conclude that it
is important that equations are derived using rupture distance rather than hypocentral
distance because most destructive earthquakes rupture surface in Iran.
Do not know physical meaning of H term but nd that it causes curves to t data better.
2.111
Xiang & Gao (1994)
Ap = aebMs (R + )c
where Ap is in cm/s2 and for combined Yunnan and W. N. American data a = 1291.07,
b = 0.5275, c = 1.5785, = 15 and = 0.5203 (in terms of natural logarithm).
All records from basement rock.
Most Yunnan data from main and aftershocks of Luquan and Luncang-Gengma earthquakes.
Records from Lancang-Gengma sequence corrected.
Most Yunnan records with 3 Ms 5 and 10 R 40 km.
To overcome difculty due to shortage of large magnitude records and sample heterogeneous distribution in near and far elds use W. N. America data, because intensity
attenuation is similar.
Fit curves to Yunnan and Yunnan with W. N. American data. Find curve for combined
data has lower variance and t to observation data for large magnitudes is better (by
plotting predicted and observed PGA).
6
They state it is . . . closest distance from the exposure of ruptured part of the fault . . . so may not be rupture
distance.
87
2.112 Aman et al. (1995)

log(a1/M ) = b1 b3 log(R)
where a is in cm/s2 , b1 = 0.433, b3 = 0.073 and = 0.037.
Data from three earthquakes with MB of 5.7, one of MB of 5.8 and the other MB of 7.2.
Compare predicted and observed ground motions for 20/10/1991 Uttarkashi earthquake
(M 6.1) and nd good t.
2.113 Ambraseys (1995)

log a = A + BMs + Cr + D log r

where r2
= d2 + h20
where a is in g, for 4.0 M 7.4: for horizontal PGA not including focal depth
A = 1.43, B = 0.245, C = 0.0010, D = 0.786, h0 = 2.7 and = 0.24, for vertical
PGA not including focal depth A = 1.72, B = 0.243, C = 0.00174, D = 0.750,
h0 = 1.9 and = 0.24, for horizontal PGA including focal depth A = 1.06, B = 0.245,
C = 0.00045, D = 1.016, h0 = h and = 0.25 and for vertical PGA including focal
depth A = 1.33, B = 0.248, C = 0.00110, D = 1.000, h0 = h and = 0.25.
Reviews and re-evaluates distances, focal depths, magnitudes and PGAs because data
from variety of sources with different accuracy and reliability. For Ms > 6.0 distances
have acceptable accuracy but for Ms < 6.0 distance, depths and magnitudes are poorly
known. Errors in locations for Ms < 6.0 still large with no foreseeable means of improving them. Use of repi for Ms < 6.0 justied because difference between rjb and repi for
small earthquakes is not larger than uncertainty in epicentre. Check and redetermine
station locations; nd large differences in excess of 15 km for some stations.
Focal depths poorly determined. Revises 180 depths using S-start times (time between
P and S-wave arrival).
Focal depths h < 26 km; most (60%+) between 4 and 14 km.
Does not use ML because no ML values for Algeria, Iran, Pakistan, Turkey and former
USSR and unreliable for other regions. Does not use magnitude calculated from strongmotion records because magnitude calculation requires point source approximation to
be valid. Conversion from ML to Ms should not be done because of uncertainty in
conversion which should be retained.
Notes that Ms results in nonlinear scaling on PGA with Mw due to nonlinear relationship
between log M0 and Ms .
Uses PGAs in four forms: maximum values from accelerograms read by others (34%),
from corrected records (30%), scaled directly from accelerograms (13%) and from digitised plots (23%). Notes potential bias in using both corrected and uncorrected PGAs but
neglects it because small difference (. 4% for those checked). Excludes PGAs near
88
trigger level because processing errors can be large. Some unltered digital records
which require additional processing to simulate SMA-1 could be associated with larger
differences (. 10%).
Excludes records from basements and ground oors of structures with more than 3
levels. Retains the few records from dam abutments and tunnel portals.
Excludes records generated by close small magnitude earthquakes triggered by S-wave.
Does not exclude records obtained at distances greater than shortest distance to an
operational but not triggered instrument because of non-constant or unknown trigger
levels and possible malfunctions of instruments.
Uses weighted regression of Joyner & Boore (1988) for second stage.
Splits data into ve magnitude dependent subsets: 2.0 Ms 7.3 (1260 records
from 619 shocks), 3.0 Ms 7.3 (1189 records from 561 shocks), 4.0 Ms 7.3
(830 records from 334 shocks), , 5.0 Ms 7.3 (434 records from 107 shocks),
and 3.0 Ms 6.0 (976 records from 524 shocks). Calculates coefcients for each
subset. Finds only small differences 15% over distance range 1200 km between
predictions and uncertainties. Concludes results stable. Prefers results from subset
with 4.0 Ms 7.3.
Finds it difcult to obtain some vertical accelerations due to low ground motion so ignores data from > 100 km with PGA < 1%g (0.1 m/s2 ).
Repeats regression using r2 = d2 + h2 . Finds depth important.
Calculates using one-stage method; nds very similar results for 10 < d < 100 km.
Considers magnitude dependent function: log a = b1 +b2 Ms +b3 r+b4 [r+b5 exp(b6 Ms )].
Finds b5 is zero so drops b3 and repeats. Finds b5 close to zero so magnitude dependent
function not valid for this dataset.
Local shear-wave velocity, Vs , proles known for 44 stations (268 records from 132
earthquakes between 2.5 and 7.2) although only 14 from > 40 km so barely sufcient
to derive equation. Use 145 records from 50 earthquakes with Ms > 4.0 to t log a =
A + BMs + Cr + D log r + E log Vs30 , where Vs30 is average shear-wave velocity to
reference depth of 30 m. Finds C positive so constrain to zero. Find no reduction in
standard deviation.
Uses residuals from main equation to nd E . Notes that should not be used because
of small number of records. Considers different choices of reference depth; nds using
between 5 and 10 m leads to higher predicted amplications. Notes better to use Vs30
because no need for subjective selection of categories.
2.114
Dahle et al. (1995)
ln A = c1 + c2 Mw + c3 ln R + c4 R + c5 S
r2 + rh2
with: R =
89
where A is in m/s2 , c1 = 1.579, c2 = 0.554, c3 = 0.560, c4 = 0.0032, c5 = 0.326,

rh = 6 and = 0.3535
Use records from Costa Rica, Mexico, Nicaragua and El Salvador. Only Mexican earthquakes with Mw 6.5 were used.
S = 0 Rock: 92 records
S = 1 Soil: 88 records
Use a Bayesian one-stage regression method (Ordaz et al. , 1994) to yield physically
possible coefcients.
Consider tectonic type: subduction or shallow crustal but do not model.
Find no signicant difference between Guerrero (Mexico) and other data.
Find no signicant difference between subduction and shallow crustal data.
2.115 Lee et al. (1995)

Ground-motion models are (if dene site in terms of local geological site classication):
log amax = M +Att(/L, M, T )+b1 M +b2 s+b3 v+b4 +b5 M 2 +
bi6 SLi +b70 rR+b71 (1r)R
or (if dene site in terms of depth of sediment):
log amax = M +Att(/L, M, T )+b1 M +b2 h+b3 v+b4 +b5 M 2 +

bi6 SLi +b70 rR+b71 (1r)R
i
where:
b0 log10 for R Rmax

b0 log10 max (R Rmax )/200 for R > Rmax
)1/2
(
R2 + H 2 + S 2
= S ln 2
R + H 2 + S02
max = (Rmax , H, S)
1
Rmax =
( + 2 4H 2 )
2
Att(, M, T ) =
S0 is correlation radius of source function and can be approximated by S0 T /2 (for

PGA assume T 0.1 s so use S0 = 0.1 km), is shear-wave velocity in source region,
T is period, S is source dimension approximated by S = 0.2 for M < 3 and S =
25.34 + 8.51M for 3 M 7.25, L is rupture length of earthquake approximated by
L = 0.01 100.5M km and v is component direction (v = 0 for horizontal 1 for vertical).
Different b0 , b70 and b71 are calculated for ve different path categories. Coefcients are
not reported here due to lack of space.
Use four types of site parameter:
Local geological site classication (dened for all records):
90
s = 0 Sites on sediments.
s = 1 Intermediate sites.
s = 2 Sites on basement rock.
Depth of sediments from surface to geological basement rock beneath site, h (dened for 1675 records out of 1926).
Local soil type parameter describes average soil stiffness in top 100200 m (dened for 1456 records out of 1926):
sL = 0 Rock soil sites SL1 = 1, SL2 = 0 and SL3 = 0. Characterises soil up to depth
of less than 10 m.
sL = 1 Stiff soil sites SL1 = 1, SL2 = 0 and SL3 = 0 (shear-wave velocities < 800 m/s
up to depth of 75100 m).
sL = 2 Deep soil sites SL2 = 1, SL1 = 0 and SL3 = 0. (shear-wave velocities
< 800 m/s up to depth of 150200 m).
sL = 3 Deep cohesionless soil sites SL3 = 1, SL1 = 0 and SL2 = 0 (only use for one
site with 10 records).
Average soil velocity in top 30 m, vL (if unavailable then use soil velocity parameter,
sT ) (dened for 1572 records out of 1926):
Soil type A vL > 750 m/s.
Soil type B 360 m/s < vL 750 m/s.
Soil type C 180 m/s < vL 360 m/s.
Soil type D vL 180 m/s.
Only include records for which signicant subset of site parameters (s, h, sL , vL ) exist.
Almost all earthquakes have focal depths H 15 km; all focal depths H 43 km.
Use records from 138 aftershocks of Imperial Valley earthquake (15/10/1979), which
contribute most of M 3 records.
Use records from 109 earthquakes with M 3.
Use free-eld records.
Characterise path by two methods:
Fraction of wave path travelled through geological basement rock measured at
surface, from epicentre to station, 0 r 1.
Generalised path type classication:
1. Sediments to sediments.
2. Rock-to-sediments, vertically.
3. Rock-to-sediments, horizontally.
4. Rock-to-rock.
5. Rock-to-rock through sediments, vertically.
6. Rock-to-sediments through rock and sediments, vertically.
7. Rock-to-sediments though rock and sediments, horizontally.
8. Rock-to-rock through sediments, horizontally.
91
Due to lack of data combine path types 2 and 6 in new category 2, combine path
types 3 and 7 in new category 3, combine path types 4, 5 and 8 in new category
4 (when r 6= 1) and combine 4, 5 and 8 in new category 5 (when r = 1).
Plot PGA against magnitude and distance to get surface by interpolation. Plot without
smoothing and with light and intense smoothing. Find for small magnitude (M 34)
earthquakes attenuation is faster than for large magnitude (M 67) earthquakes.
Use a multi-step residue regression method. First t log amax = M + Att(, M, T ) +
b1 M + b2 s + b3 v + b4 + b5 M 2 (or log amax = M + Att(, M, T ) + b1 M + b2 h + b3 v + b4 +
b5 M 2 ) and calculate residuals = log amax log a
max where amax is estimated PGA
(1) (1)
(0) (0)
(1) (1)
(2) (2)
(3) (3)
and a
max is recorded PGA. Fit = b7 SL + b7 SL + b7 SL + b7 SL + b7 SL
(i)
(i)
where SL = 1 if sL = i and SL = 0 otherwise. Find signicant dependence.

Try including vL both as a continuous and discrete parameter in model but not signicant at 5% signicance level. Next calculate residuals from last stage and t =
b00 log10 (/L) + b04 + b60 rR + b61 (1 r)R for each of the ve path type groups (1 to 5).
Lastly combine all the individual results together into nal equation.
Note that b70 and b71 can only be applied for R . 100 km where data is currently available. For R & 100 km the predominant wave type changes to surface waves and so b70
and b71 do not apply.
2.116 Lungu et al. (1995b)

Study almost identical to Radu et al. (1994), see Section 2.109, but different coefcients
given: c1 = 3.672, c2 = 1.318, c3 = 1.349, c4 = 0.0093 and = 0.395.
2.117 Molas & Yamazaki (1995)

log y = b0 + b1 M + b2 r + b3 log r + b4 h + ci
where y is in cm/s2 , b0 = 0.206, b1 = 0.477, b2 = 0.00144, b3 = 1, b4 = 0.00311,
= 0.276 and ci is site coefcient for site i (use 76 sites), given in paper but are not
reported here due to lack of space.
Records from accelerometers on small foundations detached from structures; thus consider as free-eld.
Exclude records with one horizontal component with PGA < 1 cm/s2 [0.01 m/s2 ] because weaker records not reliable due to resolution (0.03 cm/s2 [0.0003 m/s2 ]) of instruments.
Exclude earthquakes with focal depths equal to 0 km or greater than 200 km, due to lack
of such data. Depths (depth of point on fault plane closest to site), h, between about
1 km to 200 km.
Apply a low-cut lter with cosine-shaped transition from 0.01 to 0.05 Hz.
Positive correlation between magnitude and distance so use two-stage method.
92
Note different denition for MJMA for focal depths > 60 km.
Firstly do preliminary analysis with b4 = 0 and no site coefcients; nd b2 is positive so
constrain to 0 but nd b3 < 1.0 so constrain b3 to 1.0 and unconstrain b2 . Find linear
dependence in residuals on h especially for h < 100 km. Find signicant improvement
in coefcient of determination, R2 , using terms b4 h and c.
Find singularity in matrices if apply two-stage method, due to number of coefcients, so
propose a iterative partial regression method.
Also separate data into ve depth ranges (A: h = 0.1 to 30 km, 553 records from 111
earthquakes; B: h = 30 to 60 km, 778 records from 136 earthquakes; C: h = 60 to
90 km, 526 records from 94 earthquakes; D: h = 90 to 120 km, 229 records from 31
earthquakes; E: h = 120 to 200 km, 112 records from 19 earthquakes) and nd attenuation equations for each range. Note results from D & E may not be reliable due to small
number of records. Find similar results from each group and all data together.
Find weak correlation in station coefcients with soil categories, as dened in Iwasaki
et al. (1980), but note large scatter.
2.118
Sarma & Free (1995)
log(ah ) = C1 + C2 M + C3 M 2 + C4 log(R) + C5 R + C6 S
d2 + h20
where R =
where ah is in g, C1 = 3.4360, C2 = 0.8532, C3 = 0.0192, C4 = 0.9011, C5 =
0.0020, C6 = 0.0316, h0 = 4.24 and = 0.424.
S = 0 Rock
S = 1 Soil
Use one-stage method because of the predominance of earthquakes with single recordings in the set.
Note that it is very important to choose a functional form based as much as possible
on physical grounds because the data is sparse or non-existent for important ranges of
distance and magnitude.
Carefully verify all the distances in set.
Use focal depths from (in order of preference): special reports (such as aftershock
monitoring), local agencies and ISC and NEIS determinations. Focal depths < 30 km.
Do not use ML or mb because of a variety of reasons. One of which is the saturation of
ML and mb at higher magnitudes (ML , mb > 6).
If more than one estimate of Mw made then use average of different estimates.
93
Use PGAs from: a) digital or digitised analogue records which have been baseline corrected and ltered, b) data listings of various agencies and c) other literature. Difference
between PGA from different sources is found to be small.
Also derive equations assuming C3 = 0 (using rock and soil records and only soil
records) and C3 = 0, C4 = 1 and C6 = 0 (using only rock records).
Include records from Nahanni region and nd similar results.
Also derive equations for Australia (115 records from 86 earthquakes, 2.4 Mw 6.1,
1 de 188 km) and N. E. China (Tangshan) (193 records from 64 earthquakes,
3.5 Mw 7.5, 2 de 199 km) . Find considerable difference in estimated PGAs
using the equations for the three different regions.
2.119 Ambraseys et al. (1996) & Simpson (1996)

log y = C10 + C2 M + C4 log r + CA SA + CS SS
where r =
d2 + h20
where y is in g, C10 = 1.48, C2 = 0.266, C4 = 0.922, CA = 0.117, CS = 0.124,
h0 = 3.5 and = 0.25.
Use four site conditions but retain three (because only three records from very soft (L)
soil which combine with soft (S) soil category):
R Rock: Vs > 750 m/s, SA = 0, SS = 0, 106 records.
A Stiff soil: 360 < Vs 750 m/s, SA = 1, SS = 0, 226 records.
S Soft soil: 180 < Vs 360 m/s, SA = 0, SS = 1, 81 records.
L Very soft soil: Vs 180 m/s, SA = 0, SS = 1, 3 records.
Lower limit of Ms = 4.0 because smaller earthquakes are generally not of engineering
signicance.
Focal depths less than 30 km, 81% between 5 and 15 km.
Note for some records distances have uncertainty of about 10 km.
Most records from distances less than about 40 km.
For some small events need to estimate Ms from other magnitude scales.
Most records from free-eld stations although some from basements or ground oors of
relatively small structures, and tunnel portals. Do not exclude records from instruments
beyond cutoff distance because of limited knowledge about triggered level.
All uncorrected records plotted, checked and corrected for spurious points and baseline
shifts.
94
Uniform correction procedure was applied for all records. For short records (< 5 s)
a parabolic adjustment was made, for long records (> 10 s) ltering was performed
with pass band 0.20 to 25 Hz and for intermediate records both parabolic and ltering
performed and the most realistic record was chosen. Instrument correction not applied
due to limited knowledge of instrument characteristics.
Also analyze using one-stage method, note results comparable.
2.120
Ambraseys & Simpson (1996) & Simpson (1996)
Based on Ambraseys et al. (1996), see Section 2.119.

Coefcients are: C10 = 1.74, C2 = 0.273, C4 = 0.954, CA = 0.076, CS = 0.058,
h0 = 4.7 and = 0.26.
2.121
Aydan et al. (1996)
amax = a1 [exp(a2 Ms ) exp(a3 R) a4 ]

where amax is in gal, a1 = 2.8, a2 = 0.9, a3 = 0.025 and a4 = 1 ( is not given).
Most records from rhypo > 20 km.
Note that data from Turkey is limited and hence equation may be rened as amount of
data increases.
Also give equation to estimate ratio of vertical PGA (av ) to horizontal PGA (ah ): av /ah =
0.217 + 0.046Ms ( is not given).
2.122
Bommer et al. (1996)
ln(A) = a + bM + d ln(R) + qh
where h is focal depth, A is in g, a = 1.47, b = 0.608, d = 1.181, q = 0.0089 and
= 0.54.
Only use subduction earthquakes.
Do not recommend equation used for hazard analysis, since derive it only for investigating equations of Climent et al. (1994).
95
2.123 Crouse & McGuire (1996)

ln Y = a + bM + d ln(R + c1 exp{c2 M }) + eF
where Y is in g, for site category B: a = 2.342699, b = 1.091713, c1 = 0.413033,
c2 = 0.623255, d = 1.751631, e = 0.087940 and = 0.427787 and for site category
C: a = 2.353903, b = 0.838847, c1 = 0.305134, c2 = 0.640249, d = 1.310188,
e = 0.051707 and = 0.416739.
Use four site categories, Vs is shear-wave velocity in upper 100 ft (30 m):
A Rock: Vs 2500 fps (Vs 750 m/s), 33 records
B Soft rock or stiff soil: 1200 Vs 2500 fps (360 Vs < 750 m/s), 88 records
C Medium stiff soil: 600 Vs < 1200 fps (180 Vs < 360 m/s), 101 records
D Soft clay: Vs < 600 fps (Vs < 180 m/s), 16 records
Use two source mechanisms: reverse (R): F = 1, 81 records and strike-slip (S)
F = 0, 157 records. Most (77) reverse records from Ms 6.7.
Most (231) records from small building (up to 3 storeys in height) or from instrument
shelters to reduce effect of soil-structure interaction. 6 records from 6 storey buildings
and 1 record from a 4 storey building, included because lack of data in site or distance
range of these records. Structures thought not to appreciably affect intermediate or long
period and at large distances short period ground motion more greatly diminished than
long period so less effect on predictions.
Exclude records from Eureka-Ferndale area in N. California because may be associated
with subduction source, which is a different tectonic regime than rest of data. Also
excluded Mammoth Lake records because active volcanic region, atypical of rest of
California.
Include one record from Tarzana Cedar Hills although exclude a different record from
this station due to possible topographic effects.
Most records between 6 M s 7.25 and 10 R 80 km.
Apply weighted regression separately for site category B and C. Data space split into 4
magnitude (6.06.25, 6.256.75, 6.757.25, 7.25+) and 5 distance intervals ( 10 km,
1020 km, 2040 km, 4080 km, 80 km+). Each recording within bin given same total
weight.
So that Y is increasing function of M and decreasing function of R for all positive M
and R apply constraints. Dene g = b/d and h = (g + c2 ), then rewrite equation
ln Y = a + d{gM + ln[R + c1 exp(c2 M )]} + eF and apply constraints g 0, d 0,
c 0, c2 0 and h 0.
Check plots of residuals (not shown in paper), nd uniform distribution.
Find e not signicantly different than 0 and inconsistency in results between different soil
classes make it difcult to attach any signicance to fault type.
96
Lack of records for A and D site categories. Find scale factors k1 = 0.998638 and
k2 = 1.200678 so that YA = k1 YB and YD = k2 YC , where YS is predicted ground
motion for site class S . Find no obvious dependence of k1 or k2 on acceleration from
examining residuals. Find k1 and k2 not signicantly different than 1.
Note limited data for R < 10 km, advise caution for this range.
Note equation developed to estimate site-amplication factors not for seismic hazard
analysis.
2.124
Free (1996) & Free et al. (1998)
log(Y ) = C1 + C2 M + C3 M2 + C4 log(R) + C5 (R) + C6 (S)
R =
d2 + h20
where Y is in g, for M > 1.5 using acceleration and velocity records, for horizontal PGA
C1 = 4.2318, C2 = 1.1962, C3 = 0.0651, C4 = 1, C5 = 0.0019, C6 = 0.261,
h0 = 2.9 and = 0.432 and for vertical PGA C1 = 4.1800, C2 = 1.0189, C3 =
0.0404, C4 = 1, C5 = 0.0019, C6 = 0.163, h0 = 2.7 and = 0.415.
S = 0 Rock, H: 470 records, V: 395 records.

S = 1 Soil, H: 88 records, V: 83 records.
Note that not most accurate approach but due to lack of site information consider this
technique makes most consistent use of available information.
Select data using these criteria:
1. Epicentre and recording station must be within the stable continental region boundaries dened by Johnston et al. (1994) because a) such regions form end of spectrum of regions described by intraplate and hence allows differences with interplate regions to be seen, b) they are clearly delineated regions and c) intraplate
oceanic crust is excluded.
2. Minimum magnitude level M = 1.5.
3. Use records from dam abutments and downstream free-eld sites but excludes
records from crests, slopes, toes, galleries, or basements.
4. Use records from acceleration and velocity instruments.
5. Specify no minimum PGA.
6. Specify no maximum source distance. Do not exclude records from distances
greater than shortest distance to a non-triggered station.
Data from Australia, N.W. Europe, Peninsular India and E. N. America.
Focal depths, 2 h 28 km.
Most records from M < 4.0.
97
Visually inspect all records including integrated velocities and displacements, identify
and remove traces dominated by noise, identify and correct transient errors (spikes,
ramps, linear sections, back time steps and clipped peaks), identify scaling errors, identify and remove multiple event records. Linear baseline correct and elliptically lter with
cut-off 0.25 to 0.5 Hz (determine frequency by visual inspection of adjusted record) and
33 to 100 Hz (generally pre-determined by Nyquist frequency).
Large proportion of records from velocity time histories which differentiate to acceleration. Test time domain method (central difference technique) and frequency domain
method; nd very similar results. Use time domain method.
Distribution with respect to magnitude did not allow two-stage regression technique.
In many analyses distribution of data with respect to distance did not allow simultaneous
determination of coefcients C4 and C5 , for these cases constrain C4 to 1.
Test effect of minimum magnitude cut-off for two cut-offs M = 1.5 and M = 3.5. Find
if include data from M < 3.5 then there is substantial over prediction of amplitudes for
d < 10 km for large magnitudes unless include C3 term. C3 effectively accounts for
large number of records from small magnitudes and so predictions using the different
magnitude cut-offs are very similar over broad range of M and d.
Try including focal depth, h, explicitly by replacing h0 with h because h0 determined for
whole set (which is dominated by small shocks at shallow depths) may not be appropriate for large earthquakes. Find improved t at small distances but it does not result in
overall improvement in t ( increases); this increase thought due to large errors in focal
depth determination.
Find larger standard deviations than those found in previous studies which note may be
due to intrinsic differences between regional subsets within whole set. Repeat analysis separately for Australia (for horizontal and vertical), N. America (for horizontal and
vertical) and N.W. Europe (horizontal); nd reduced standard deviations (although still
large), C5 varies signicantly between 3 regions.
Repeat analysis excluding velocity records.
Also repeat analysis using only rock records.
2.125 Inan et al. (1996)

log PGA = aM + b log R + c

where PGA is in an unknown unit but it is probably in gal, a = 0.65, b = 0.9 and
c = 0.44 ( not reported).
2.126 Ohno et al. (1996)

log S(T ) = a(T )M log Xeq b(T )Xeq + c(T ) + qs(T )

98
where S(0.02) is in gal, a(0.02) = 0.318, b(0.02) = 0.00164 and c(0.02) = 1.597
(s(0.02) and only given in graphs).
Use two site conditions:
q = 0 Pre-Quaternary: Rock (sandstone, siltstone, shale, granite, mudstone, etc.); thickness of surface soil overlying rock is less than 10 m; shallow soil or thin alluvium,
160 records. S-wave velocities > 600 m/s.
q = 1 Quaternary: Soil (alluvium, clay, sand, silt, loam, gravel, etc.), 336 records. S-wave
velocities 600 m/s.
Exclude records from very soft soil such as bay mud or articial ll because few such
records and ground motions may be strongly affected by soil nonlinearity.
Use equivalent hypocentral distance, Xeq , because strong motion in near-source region
affected from points other than nearest point on fault plane.
Use portion of record after initial S-wave arrival.
Approximates PGA by spectral acceleration for period of 0.02 s and 5% damping.
Plot the amplitude factors from rst stage against Mw ; nd well represented by linear
function.
2.127
Romeo et al. (1996)
log PHA = a1 + a2 Mw log(d2 + h2 )1/2 + a3 S

where PHA is in g, a1 = 1.870 0.182, a2 = 0.366 0.032, a3 = 0.168 0.045,
h = 6 km and = 0.173 for rjb and a1 = 2.238 0.200, a2 = 0.438 0.035,
a3 = 0.195 0.049, h = 5 km and = 0.190 for repi .
S = 0 Rock or stiff soils and deep alluvium.

S = 1 All other sites.
Use data and functional form of Sabetta & Pugliese (1987) but use Mw instead of magnitudes used by Sabetta & Pugliese (1987).
2.128
Sarma & Srbulov (1996)
log(Ap /g) = b1 + b2 Ms + b3 log r + b4 r

where r = (d2 + h20 )0.5
where Ap is in g, using both horizontal components b1 = 1.617, b2 = 0.248, b3 =
0.5402, b4 = 0.00392, h0 = 3.2 and = 0.26 and for larger horizontal component
b1 = 1.507, b2 = 0.240, b3 = 0.542, b4 = 0.00397, h0 = 3.0 and = 0.26.
99
Consider two soil categories but do not model:

1. Rock
2. Soil
Classify sites without regard to depth and shear-wave velocity of deposits.
Most records from W. USA but many from Europe and Middle East.
Records from instruments on ground oor or in basements of buildings and structures
up to 3 storeys and at free-eld sites, regardless of topography.
Records baseline corrected and low-pass ltered using elliptic lter.
2.129 Singh et al. (1996)

log10 AGM = b1 + 0.31M b3 log R

where AGM is in cm/s2 , b1 = 1.14 and b3 = 0.615 ( is not given). Note there are
typographical errors in the abstract.
Data from three earthquakes with mb = 5.7, one with mb = 5.8 and one with mb = 7.2.
Adopt magnitude scaling coefcient (0.31) from Boore (1983).
2.130 Spudich et al. (1996) & Spudich et al. (1997)

= b1 + b2 (M 6) + b3 (M 6)2 + b4 R + b5 log10 R + b6
2 + h2
rjb
where R =
log10 Y
where Y is in g,
b1 = 0.156, b2 = 0.229, b3 = 0, b4 = 0, b5 = 0.945, b6 = 0.077,
h = 5.57, = 12 + 22 + 32 where 1 = 0.216, 2 = 0, for randomly orientated
component 3 = 0.094 and for geometric mean 3 = 0.
Use two site categories (following classication of Joyner & Boore (1981)):
= 0 Rock: 35 records
= 1 Soil: 93 records
Applicable for extensional regimes, i.e. those regions where lithosphere is expanding
areally.
Reject records from structures of more than two storeys or from deeply embedded basements or those which triggered on S wave.
Include records from those instruments beyond cutoff distance, i.e. beyond rst instrument which did not trigger.
100
Correction technique based on uniform correction and processing. Determine passband for ltering based on visual inspection of Fourier amplitude spectra and doublyintegrated displacements. Apply instrument correction.
Not enough data to be able to nd all coefcients so use b2 and b3 from Boore et al.
(1994a)
Note that should only be used in distance range 0 to 70 km because further away ground
motions tend to be over predicted.
2.131
Stamatovska & Petrovski (1996)
Acc = exp(b) exp(bM )(Rh + C)bR

where Rh2
= (Re /)2 + h2
1 + tg 2
and =
2
a + tg 2
where Acc is in cm/s2 , is the azimuth of the site with respect to energy propagation
pattern, b = 3.49556, bM = 1.35431, C = 30, bR = 1.58527, a = 1.2 and = 0.48884
(denitions of t and g are not given).
Correct PGAs for local site effects so that PGAs used correspond to a site with a shearwave velocity of 700 m/s. Do not state how this is performed.
Most records from SMA-1s.
Not all records from free-eld.
Records from strong intermediate depth earthquakes in Vrancea region.
Focal depths, 89.1 h 131 km.
For each of the four earthquakes, calculate coefcents in equation ln Acc = b0 +
b1 ln(Re /), the main direction of energy propagation and the relation between the semiaxes of the ellipse in two orthogonal directions (a : b).
Also calculate coefcents in equation ln Acc = b+bM M +bR ln(Rh +C) for different azimuth by normalising the values of Re / by the azimuth. Give coefcients for Bucharest,
Valeni and Cerna Voda.
Note that uncertainty is high and suggest this is because of distribution of data with
respect to M , Re and h, the use of data processed in different ways, soil-structure
interaction and the use of an approximate correction method for local site effects.
101
2.132 Campbell (1997), Campbell (2000), Campbell (2001) & Campbell & Bozorgnia (1994)
Ground-motion model (horizontal component) is:
ln AH
= a1 + a2 M + a3 ln
2
+ [a4 exp(a5 M )]2
RSEIS
+ [a6 + a7 ln RSEIS + a8 M ]F + [a9 + a10 ln RSEIS ]SSR

+ [a11 + a12 ln RSEIS ]SHR + fA (D)
{
fA (D) =
0 for D 1 km
{[a11 + a12 ln(RSEIS )] [a9 + a10 ln(RSEIS )]SSR }(1 D)(1 SHR ) for D < 1 km
where AH is in g, a1 = 3.512, a2 = 0.904, a3 = 1.328, a4 = 0.149, a5 = 0.647,

a6 = 1.125, a7 = 0.112, a8 = 0.0957, a9 = 0.440, a10 = 0.171, a11 = 0.405,
a12 = 0.222, = 0.55 for AH < 0.068 g, = 0.173 0.140 ln(AH ) for 0.068 g
AH 0.21 g and = 0.39 for AH > 0.21 g (when expressed in terms of acceleration)
and = 0.889 0.0691M for M < 7.4 and = 0.38 for M 7.4 (when expressed in
terms of magnitude).
Ground-motion model (vertical component) is:
ln AV
= ln AH + b1 + b2 M + b3 ln[RSEIS + b4 exp(b5 M )]
+ b6 ln[RSEIS + b7 exp(b8 M )] + b9 F
where AV is in g, b1 = 1.58, b2 = 0.10, b3 = 1.5, b

4 = 0.079, b5 = 0.661,
b6 = 1.89, b7 = 0.361, b8 = 0.576, b9 = 0.11 and V = 2 + 0.362 (where is
standard deviation for horizontal PGA prediction).
Uses three site categories:
SSR = 0, SHR = 1 Hard rock: primarily Cretaceous and older sedimentary deposits, metamorphic rock, crystalline rock and hard volcanic deposits (e.g. basalt).
SSR = 1, SHR = 0 Soft rock: primarily Tertiary sedimentary deposits and soft volcanic deposits (e.g. ash deposits).
SSR = 0, SHR = 0 Alluvium or rm soil: rm or stiff Quaternary deposits with depths greater
than 10 m.
Also includes sediment depth (D) as a variable.
Restricts to near-source distances to minimize inuence of regional differences in crustal
attenuation and to avoid complex propagation effects that have been observed at longer
distances.
Excludes recordings from basement of buildings greater than two storeys on soil and
soft rock, greater than ve storeys on hard rock, toe and base of dams and base of
bridge columns. Excludes recordings from shallow and soft soil because previous analyses showed such sites have accelerations signicantly higher than those on deep, rm
alluvium. Include records from dam abutments because comprise a signicant number
of rock recordings and due to stiff foundations are expected to be only minimally affected
by dam. Some of these could be strongly affected by local topography.
102
Includes earthquakes only if they had seismogenic rupture within shallow crust (depths
less than about 25 km). Includes several large, shallow subduction interface earthquakes because previous studies found similar near-source ground motions to shallow
crustal earthquakes.
Includes only earthquakes with M about 5 or larger to emphasize those ground motions of greatest engineering interest and limit analysis to more reliable, well-studied
earthquakes.
Notes that distance to seismogenic rupture is a better measure than distance to rupture
or distance to surface projection because top layer of crust is non-seismogenic and will
not contribute to ground motion. Give estimates for average depth to top of seismogenic
rupture for hypothetical earthquakes.
Considers different focal mechanisms: reverse (H:6, V:5), thrust (H:9, V:6), reverseoblique (H:4, V:2) and thrust-oblique (0), total (H:19, V:13) F = 1 (H:278 records,
V:116 records) (reverse have a dip angle greater than or equal to 45 ), strike-slip (H:27,
V:13) F = 0 (H:367 records, V:109 records) (strike-slip have an absolute value
of rake less than or equal to 22.5 from the horizontal as measured along fault plane).
There is only one normal faulting earthquakes in set of records (contributing four horizontal records) so difference is not modelled although F = 0.5 given as rst approximation
(later revised to F = 0).
Mostly W. USA with 20 records from Nicaragua(1) Mexico (5), Iran (8), Uzbekistan (1),
Chile (3), Armenia (1) and Turkey (1).
Does regression rstly with all data. Selects distance threshold for each value of magnitude, style of faulting and local site condition such that the 16th percentile estimate of AH
was equal to 0.02 g (which corresponds to a vertical trigger of about 0.01 g). Repeats
regression repeated only with those records within these distance thresholds. Avoids
bias due to non-triggering instruments.
Finds dispersion (uncertainty) to be dependent on magnitude and PGA, models as linear
functions. Finds better t for PGA dependency.
2.133
Munson & Thurber (1997)
log10 PGA = b0 + b1 (M 6) + b2 r log10 r + b4 S
where r =
d2 + h2
PGA is in g, b0 = 0.518, b1 = 0.387, b2 = 0.00256, b4 = 0.335, h = 11.29 and
= 0.237.
S = 0 Lava: 38 records
S = 1 Ash: 60 . Vs . 200 m/s, 13 records
Depths between 4 and 14 km with average 9.6 km (standard deviation 2.3 km). Limit
of 15 km chosen to differentiate between large tectonic earthquakes and deeper mantle
events.
103
Attenuation greater than for western USA due to highly fractured volcanic pile.
Peak acceleration measured directly from accelerograms. Check against one from corrected records, small difference.
Excludes records triggered on S-wave and those beyond cutoff distance (the distance
to rst nontriggered instrument).
Does weighted and unweighted least squares analysis; nd some differences.
2.134 Pancha & Taber (1997)

where r = (d2 + h2 )1/2
Coefcients are unknown.
Also develop model using functional form of Molas & Yamazaki (1995).
All data from rock sites.
Data from seismographs of New Zealand National Seismograph Network and temporary
deployments on East Cape of the North Island, the Marlborough region of the South
Island and the central volcanic zone of the North Island.
Most data from more than 100 km from the source.
2.135 Rhoades (1997)

log10 a = + M log10 r + r
where r = (d2 + h2 )1/2
where a is in g, = 1.237 0.254, = 0.278 0.043, = 0.00220 0.00042,
h = 6.565 0.547, 2 = 0.00645 0.00382 and 2 = 0.0527 0.00525 (where 2 is the
inter-earthquake variance and 2 is the intra-earthquake variance and signies the
standard error of the estimate.
Notes that errors in magnitude determination are one element that contributes to the
between-earthquake component of variance and could thus cause apparent differences
between earthquakes, even if none existed.
Develops a method to explicitly include consideration of magnitude uncertainties in a
random earthquake effects model so that the between-earthquake component of variance can be split into the part that is due only to magnitude uncertainty (and is therefore of no physical consequence) and the part for which a physical explanation may be
sought.
104
Applies method to data of Joyner & Boore (1981). Assume two classes of magnitude
estimates: those with estimates of Mw , which assumes to be associated with a standard
error of 0.1, and those for which ML was used as a surrogate for Mw , which assumes
to be associated with a standard error of 0.3. Find that the inter-earthquake variance
is much lower than that computed assuming that the magnitudes are exact but that
other coefcients are similar. Believes that the high inter-earthquake variance derived
using the exact magnitudes model is largely explained by the large uncertainties in the
magnitude estimates using ML .
2.136
Schmidt et al. (1997)
ln A = c1 + c2 M + c3 ln r + c4 r + c5 S1 + c6 S2
where r =
R 2 + 62
where A is in m/s2 , c1 = 1.589, c2 = 0.561, c3 = 0.569, c4 = 0.003, c5 = 0.173,
c6 = 0.279 and = 0.80 (for all earthquakes), c1 = 1.725, c2 = 0.687, c3 = 0.742,
c4 = 0.003, c5 = 0.173, c6 = 0.279 and = 0.83 (for shallow crustal earthquakes)
and c1 = 0.915, c2 = 0.543, c3 = 0.692, c4 = 0.003, c5 = 0.173, c6 = 0.279 and
= 0.74 (for subduction zone earthquakes).
S1 = 0, S2 = 0 Rock, 54 records.
S1 = 1, S2 = 0 Hard soil, 63 records.
S1 = 0, S2 = 1 Soft soil, 83 records.
Most records from SMA-1s with 6 records from SSA-2.
Use PSA at 40 Hz (0.025 s) as peak ground acceleration.
Records instrument corrected and bandpass ltered with cut-offs of 0.2 and 20 Hz.
Use data from shallow crustal earthquakes (133 records) and subduction zone earthquakes (67 records).
Perform regression on combined shallow crustal and subduction zone records, on just
the shallow crustal records using rhypo and using repi and on just subduction zone
records.
Note that distribution w.r.t. distance improves in the near eld when epicentral distance
is used but only possible to use repi for shallow crustal earthquakes because for subduction zone earthquakes hypocentral distance is much greater than epicentral distance
so should use rhypo instead.
For 4 M 6 distribution w.r.t. epicentral distance is quite good but for M > 6 no
records from de < 40 km.
Use a two step procedure. Firstly use entire set and both horizontal components and
compute two soil terms (one for hard and one for soft soil). In second step use soil terms
to correct motions for rock conditions and then repeat regression.
105
Use Bayesian analysis (Ordaz et al. , 1994) so that derived coefcients comply with
physics of wave propagation because include a priori information on the coefcients to
avoid physically unrealistic values. Choose initial values of coefcients based on theory
and previous results
Cannot nd coefcient in r by regression so adopt 6 km from previous study.
Examine residuals w.r.t. distance and magnitude and nd no trends.
2.137 Youngs et al. (1997)

Ground-motion model for soil is:
ln PGA =
C1
+ C2 M +
with: C1
= C1 + C6 Zr
C3
C4
= C3 + C7 Zr
C3 ln
C4 C 2 M
C
rrup + e
]
+ C5 Zt + C9 H + C10 Zss
= C4 + C8 Zr
where PGA is in g, C1 = 0.6687, C2 = 1.438, C3 = 2.329, C4 = ln(1.097),

C5 = 0.3643, C9 = 0.00648 and = 1.45 0.1M (other coefcients in equation not
needed for prediction on deep soil and are not given in paper).
Ground-motion model for rock is:
]
[
C
C 2 M
+ C5 Zss + C8 Zt + C9 H
ln PGA = C1 + C2 M + C3 ln rrup + e 4 C3
with: C1
= C1 + C3 C4 C3 C4
C3 = C3 + C6 Zss
C4 = C4 + C7 Zss
where PGA is in g, C1 = 0.2418, C2 = 1.414, C3 = 2.552, C4 = ln(1.7818), C8 =
0.3846, C9 = 0.00607 and = 1.45 0.1M (other coefcients in equation not needed
for prediction on rock and are not given in paper).
Use different models to force rock and soil accelerations to same level in near eld.
Use three site categories to do regression but only report results for rock and deep soil:
Zr = 1, Zds = 0, Zss = 0 Rock: Consists of at most about a metre of soil over weathered
rock, 96 records.
Zds = 1, Zr = 0, Zss = 0 Deep soil: Depth to bedrock is greater than 20 m, 284 records.
Zss = 1, Zds = 0, Zr = 0 Shallow soil: Depth to bedrock is less than 20 m and a signicant
velocity contrast may exist within 30 m of surface, 96 records.
Use free-eld recordings, i.e. instruments in basement or ground-oor of buildings less
than four storeys in height. Data excluded if quality of time history poor or if portion of
main shaking not recorded.
Consider tectonic type: interface (assumed to be thrust) (98 records) Zt = 0, intraslab (assumed to be normal) (66 records) Zt = 1
106
Focal depths, H , between 10 and 229 km

Not enough data to perform individual regression on each subset so do joint regression
analysis.
Both effect of depth and tectonic type signicant.
Large differences between rock and deep soil.
Note differences between shallow crustal and interface earthquake primarily for very
large earthquakes.
Assume uncertainty to be linear function of magnitude.
2.138
Zhao et al. (1997)
Ground-motion model (Model 1) is:
log10 PGA = A1 Mw + A2 log10
r2 + d2 + A3 hc + A4 + A5 R + A6 A + A7 I
where PGA is in m/s2 , R = 1 for crustal reverse 0 otherwise, A = 1 for rock 0

otherwise, I = 1 for interface 0 otherwise, A1 = 0.298, A2 = 1.56, A3 = 0.00619,
A4 = 0.365, A5 = 0.107, A6 = 0.186, A7 = 0.124, d = 19 and = 0.230.
Models also given for soil sites only (Model 2), unspecied site (Model 3), focal mechanism and tectonic type unknown (Model 4) and only magnitude, depth and distance
known (Model 5)
Records from ground or base of buildings. 33 from buildings with more than 3 storeys;
nd no signicant differences.
Retain two site categories:
1. Rock: Topographic effects expected, very thin soil layer ( 3 m) overlying rock or
rock outcrop.
2. Soil: everything else
Use depth to centroid of rupture, hc , 4 hc 149. Only nine are deeper than 50 km.
Exclude records from deep events which travelled through mantle.
Consider tectonic type: C=crustal (24+17 records), I=interface (7+0 records) and S=slab
(20+0 records)
Consider source mechanism: N=normal (15+1 records), R=reverse (22+5 records) and
S=strike-slip (12+11 records). Classify mixed mechanisms by ratio of components
1.0.
For only ve records difference between the distance to rupture surface and the distance
to centroid could be more than 10%.
66 foreign near-source records (dr 10 km) from 17 crustal earthquakes supplement
NZ data. Mainly from western North America including 17 from Imperial Valley and 12
from Northridge.
107
Exclude one stations records (Atene A) due to possible topographical effects.

Exclude records which could have been affected by different attenuation properties in
the volcanic region.
Note regional difference between Fiordland and volcanic region and rest of country but
do model.
Retain coefcients if signicant at = 0.05.
Anelastic term not signicant.
2.139 Baag et al. (1998)

ln PGA = a1 + a2 M + a3 ln R + a4 R
2 + a2
where R =
Repi
5
where PGA is in cm/s2 , a1 = 0.4, a2 = 1.2, a3 = 0.76, a4 = 0.0094 and a5 = 10 (
not given).
This article has not been seen. The model presented may not be a fully empirical model.
2.140 Bouhadad et al. (1998)

A = c exp(M )[Rk + a]R

Coefcients not given, only predictions.
2.141 Costa et al. (1998)

log(A) = a + bM + c log(r)
where A is in g, a = 1.879, b = 0.431 and c = 1.908 (for vertical components) and
a = 2.114, b = 0.480 and c = 1.693 (for horizontal components).
All records from digital instruments.
Try including a term d log(M ) but tests show that d is negligible with respect to a, b and
c.
108
2.142
Manic (1998)
log(A) = c1 + c2 M + c3 log(D) + c4 D + c5 S
D = (R2 + d20 )1/2
where A is in g, c1 = 1.664, c2 = 0.333, c3 = 1.093, c4 = 0, c5 = 0.236, d0 = 6.6
and = 0.254.
Uses four site categories (following Ambraseys et al. (1996)) but only two have data
within them:
S = 0 Rock (R): vs > 750 m/s, 92 records.

S = 1 Stiff soil (A): 360 < vs 750 m/s, 184 records.
where vs is average shear-wave velocity in upper 30 m.
Uses both horizontal components to get a more reliable set of data.
Tries using ML rather than Ms , epicentral distance rather than hypocentral distance and
constraining anelastic decay coefcient, c4 , to zero. Chooses combination which gives
minimum .
2.143
Reyes (1998)
ln Sa = 1 + 2 (M 6) + 3 (M 6)2 + 4 ln R + 5 R
where Sa is in cm/s2 , 1 = 5.8929, 2 = 1.2457, 3 = 9.7565 102 , 4 = 0.50,
5 = 6.3159 103 and = 0.420.
Use data from one station, University City (CU) in Mexico City, a relatively rm site.
2.144
Rinaldis et al. (1998)
ln Y = C14 + C22 M + C31 ln(R + 15) + C43 S + C54 F

where Y is in cm/s2 , C14 = 5.57, C22 = 0.82, C31 = 1.59, C43 = 0.14, C54 = 0.18
and = 0.68. Assume 15 km inside ln(R + . . .) from Theodulidis & Papazachos (1992).
S = 0 Rock: includes stiff sites.

S = 1 Alluvium: includes both shallow and deep soil sites.
Use two source mechanism categories:
F = 0 Thrust and strike-slip earthquakes.

109
F = 1 Normal earthquakes.
Use epicentral distance because in Italy and Greece the surface geology does not show
any evident faulting, consequently it is impossible to use a fault distance denition.
Good distribution and coverage of data with respect to site category and source mechanism.
Consider six strong-motion records (three Italian and three Greek) with different associated distances, magnitudes and record length and apply the different processing techniques of ENEA-ENEL and ITSAK to check if data from two databanks can be merged.
Digitise six records using same equipment. ITSAK technique: subtract the reference
trace (either xed trace or trace from clock) from uncorrected accelerogram and select
band-pass lter based on either Fourier amplitude spectra of acceleration components
or selected using a different technique. ENEA-ENEL technique: subtract the reference
trace from uncorrected accelerogram and select band-pass lter by comparing Fourier
amplitude spectra of acceleration components with that of xed trace. Find small differences in PGA, PGV, PGD so can merge Italian and Greek data into one databank.
Use four step regression procedure, similar to that Theodulidis & Papazachos (1992)
use. First step use only data with M 6.0 (7 R 138 km) for which distances
are more accurate to nd geometrical coefcient C31 . Next nd constant (C12 ) and
magnitude (C22 ) coefcients using all data. Next nd constant (C13 ) and soil (C43 )
coefcients using all data. Finally nd constant (C14 ) and source mechanism (C54 )
coefcients using data with M 6.0 for which focal mechanism is better constrained;
nal coefcients are C14 , C22 , C31 , C43 and C54 . Investigate inuence of distance on
C54 by subdividing data in nal step into three categories with respect to distance (7
R 140 km, 7 R 100 km and 7 R 70 km).
Equation intended as rst attempt to obtain attenuation relations from combined databanks and site characteristics and fault rupture properties could and should be taken
into account.
2.145 Sadigh & Egan (1998)

Based on Sadigh et al. (1997), see Section 2.77.
ln PGA = C1 + C2 M + C3 ln[rrup + exp(C4 + C5 M )]

where PGA is in g, for M < 6.5 C4 = 1.29649 and C5 = 0.25 and for M 6.5
C4 = 0.48451 and C5 = 0.524. For rock sites: C3 = 2.100, for strike-slip mechanism
and M < 6.5 C1 = 0.949 and C2 = 1.05, for strike-slip mechanism and M 6.5 C1 =
1.274 and C2 = 1.10, for reverse-slip and M < 6.5 C1 = 0.276 and C2 = 0.90 and
for reverse-slip and M 6.5 C1 = 1.024 and C2 = 1.10. For soil sites: C3 = 1.75,
for strike-slip mechanism and M < 6.5 C1 = 1.1100 and C2 = 0.875, for strike-slip
mechanism and M 6.5 C1 = 1.3830 and C2 = 0.917, for reverse-slip mechanism
and M < 6.5 C1 = 0.0895 and C2 = 0.750 and for reverse-slip mechanism and
M 6.5 C1 = 1.175 and C2 = 0.917 ( not given).
110
1. Rock: bedrock within about a metre of surface. Note that many such sites are soft
rock with Vs 750 m/s and a strong velocity gradient because of near-surface
weathering and fracturing, 274 records.
2. Deep soil: greater than 20 m of soil over bedrock. Exclude data from very soft soil
sites such as those from San Francisco bay mud, 690 records.
Dene crustal earthquakes as those that occur on faults within upper 20 to 25 km of
continental crust.
Consider source mechanism: RV=reverse (26+2) and SS=strike-slip (and some normal)
(89+0). Classied as RV if rake> 45 and SS if rake< 45 . Find peak motions from
small number of normal faulting earthquakes not to be signicantly different than peak
motions from strike-slip events so include in SS category.
Separate equations for Mw < 6.5 and Mw 6.5 to account for near-eld saturation
effects, for rock and deep soil sites and reverse and strike-slip earthquakes.
Records from instruments in instrument shelters near ground surface or in ground oor
of small, light structures.
4 foreign records (1 from Gazli and 3 from Tabas) supplement Californian records.
2.146
Sarma & Srbulov (1998)
log(ap /g) = C1 + C2 Ms + C3 d + C4 log d

where ap is in g, for soil sites C1 = 1.86, C2 = 0.23, C3 = 0.0062, C4 = 0.230 and
= 0.28 and for rock sites C1 = 1.874, C2 = 0.299, C3 = 0.0029, C4 = 0.648
and = 0.33.
Use two site categories because of limited available information (based on nature of top
layer of site regardless of thickness) for which derive separate equations:
1. Soil
2. Rock
Use record from free-eld or in basements of buildings 3 storeys high.
Use Ms because better represents size of shallow earthquakes and is determined from
teleseismic readings with much smaller standard errors than other magnitude scales and
also saturates at higher magnitudes than all other magnitude scales except Mw which
is only available for relatively small portion of earthquakes. For some small earthquakes
convert to Ms from other magnitude scales.
For very short records, 5 s long, correct using parabolic baseline, for records > 10 s
long correct using elliptical lter and for records between 5 and 10 s long both parabolic
correction and ltering applied and select best one from appearance of adjusted time
histories.
Equations not any more precise than other attenuation relations but are simply included
for completeness and for a comparison of effects of dataset used with other dataset.
Data did not allow distinction between different source mechanisms.
111
2.147 Sharma (1998)

log A = c1 + c2 M b log(X + ec3 M )

where A is in g, c1 = 1.072, c2 = 0.3903, b = 1.21, c3 = 0.5873 and = 0.14.
Considers two site categories but does not model:
R Rock: generally granite/quartzite/sandstone, 41 records.
S Soil: exposed soil covers on basement, 25 records.
Focal depths between 7.0 and 50.0 km.
Most records from distances > 50 km. Correlation coefcient between M and X is 0.63.
Does not include source mechanism as parameter because not well dened and including many terms may lead to errors. Also neglects tectonic type because set is small and
small differences are expected.
Fit log A = b log X + c to data from each earthquake separately and nd average b
equal to 1.292. Then t log A = aM b log X + c to data from all earthquakes and nd
b = 0.6884. Fit log A = b log X + di li to all data, where li = 1 for ith earthquake
and 0 otherwise and nd b = 1.21, use this for rest of analysis.
Use weighted regression, due to nonuniform sampling over all M and X . Divide data
into distance bins 2.5 km wide up to 10 km and logarithmically dependent for larger
distances. Within each bin each earthquake is given equal weight by assigning a relative
weight of 1/nj,l , where nj,l is the number of recordings for j th earthquake in lth distance
bin, then normalise so that sum to total number of recordings.
Original data included two earthquakes with focal depths 91.0 km and 119.0 km and
M = 6.8 and 6.1 which caused large errors in regression parameters due to large
depths so excluded them.
Check capability of data to compute coefcients by deleting, in turn, c1 , c2 and c3 , nd
higher standard deviation.
Makes one coefcient at a time equal to values given in Abrahamson & Litehiser (1989),
nds sum of squares increases.
Notes lack of data could make relationship unreliable.
2.148 Smit (1998)

log Y = a + bM log R + dR
where Y is in nm/s2 , b = 0.868, d = 0.001059, = 0.35, for horizontal PGA a = 5.230
and for vertical PGA a = 5.054.
Most records from rock sites.
112
Focal depths between 0 and about 27 km (most less than 10 km).

Most records from ML < 3.5.
Most earthquakes have strike-slip mechanism.
Uses records from high gain short period seismographs and from strong-motion instruments.
Records are instrument corrected.
Eliminates some far-eld data from small magnitude earthquakes using signal to noise
ratio criterion.
Records cover entire azimuthal range.
Notes that need more data in near eld.
Notes that care must be taken when using equations for prediction of ground motion in
strong earthquakes (M 6) because of lack of data.
2.149
Cabaas et al. (1999), Cabaas et al. (2000) & Benito et al.

(2000)
ln A = C1 + C2 M + C3 (R + R0 ) + C4 ln(R + R0 ) + C5 S
where A is in cm/s2 , C1 = 0, C2 = 0.664, C3 = 0.009, C4 = 2.206, R0 = 20,
C5 = 8.365 (for S1), C5 = 8.644 (for S2), C5 = 8.470 (for S3) and C5 = 8.565 (for S4)
for horizontal PGA using repi and Ms and all Mediterranean data, C1 = 0, C2 = 0.658,
C3 = 0.008, C4 = 2.174, R0 = 20, C5 = 7.693 (for S1), C5 = 7.915 (for S2) and
C5 = 7.813 (for S4) (C5 not derived for S3) for vertical PGA using repi and Ms and all
Mediterranean data. is not given (R2 is reported).
S1 Hard basement rock.
S2 Sedimentary rock and conglomerates.
S3 Glacial deposits.
S4 Alluvium and consolidated sediments.
Derive separate equations using data from Mediterranean region and also just using
data from Spain.
Equations for Spain derived using mbLg .
Spanish data all from earthquakes with 2.5 mbLg 6.0 and 0 rhypo 300 km.
113
2.150 Chapman (1999)

log10 Y = a + b(M 6) + c(M 6)2 + d log(r2 + h2 )1/2 + eG1 + f G2

where Y is in cm/s2 , a = 3.098, b = 0.3065, c = 0.07570, d = 0.8795, h = 6.910,
e = 0.1452, f = 0.1893 and = 0.2124.
A & B Vs,30 > 760 m/s, 24 records G1 = 0, G2 = 0.
C 360 < Vs,30 760 m/s, 116 records G1 = 1, G2 = 0.
D 180 < Vs,30 360 m/s, 164 records G1 = 0, G2 = 1.
Uses records from ground level or in basements of structures of two stories or less, and
excludes records from dam or bridge abutments.
Selects records which include major motion portion of strong-motion episode, represented by S wavetrain. Excludes records triggered late on S wave or those of short
duration terminating early in coda.
Most records already corrected. Some records instrument corrected and 4-pole causal
Butterworth ltered (corner frequencies 0.1 and 25 Hz). Other records instrument corrected and 4-pole or 6-pole causal Butterworth bandpass ltered (corner frequencies 0.2
and 25 Hz). All data ltered using 6-pole causal high-pass Butterworth lter with corner
frequency 0.2 Hz and velocity and displacement curves examined.
Uses method of Campbell (1997) to reduce bias due to non-triggered instruments, for
some recent shocks. Firstly uses all data to determine minimum distances (which are
functions of magnitude and site condition) at which 16th percentile values of PGA are <
0.02 g[0.2 m/s] (corresponding to 0.01 g[0.1 m/s] vertical component trigger threshold).
Next delete records from larger distances and repeat regression.
Check residuals against distance and magnitude for each site class; nd no obvious
non-normal magnitude or distance dependent trends.
2.151 Cousins et al. (1999)

Based on Zhao et al. (1997) see Section 2.138
log10 PGA = A1 Mw + A2 log10 R + A3 hc + A4 + A5 + A6 + A7 R + A8 Mw + A9

+ A10 Rv
where PGA is in m/s2 , R = r2 + d2 and Rv is distance travelled by direct seismic

wave through volcanic region. A5 only for crustal reverse, A6 only for interface, A7 only
for strong and weak rock, A8 only for strong rock, A9 only for strong rock, A1 = 0.2955,
A2 = 1.603, A3 = 0.00737, A4 = 0.3004, A5 = 0.1074, A6 = 0.1468, A7 =
0.00150, A8 = 0.3815, A9 = 2.660, A10 = 0.0135, d = 19.0 and = 0.24.
114
Originally considers ve site categories but retain three:

1. Strong rock: Vs > 700 m/s
2. Weak rock: 375 Vs 700 m/s and category AV those sites with a very thin layer
( 3 m) overlying rock
3. Soil: everything else
Depth to centroid of rupture, hc , used, 4 hc 94 km.
60% on soil, 40% on rock
Consider tectonic type: C=Crustal (12+17), I=Interface (5+0) and S=Slab(8+0)
Consider source mechanism: N=normal (6+1), R=reverse (12+5) and S=strike-slip (7+11).
Mixed classied by ratio of components 1.0.
Mixture of analogue and digital accelerograms (72%) and seismograms (28%)
Accelerograms sampled at 100250 samples/sec. Bandpass frequencies chosen by
analysis of Fourier amplitude spectrum compared with noise spectrum. fmin between
0.15 and 0.5 Hz and fmax equal to 25 Hz. Instrument correction applied to analogue
records.
Seismograms sampled at 50100 samples/sec. Differentiated once. Instrument corrected and high pass ltered with fmin = 0.5 Hz. No low pass lter needed.
Clipped seismograms usually retained.
Directional effect noticed but not modelled.
Most records from more than 100 km away. Note lack of near-source data.
Records from accelerograms further away than rst operational non-triggering digital
accelerograph, which had a similar triggering level, were excluded.
Models difference between high attenuating volcanic and normal regions.
2.152
lafsson & Sigbjrnsson (1999)
log(amax ) = 1 + 2 log M0 3 log(R)

where amax is in cm/s2 , M0 is in dyn cm and R is in cm, 1 = 0.0451, 2 = 0.3089,
3 = 0.9642 and = 0.3148.
Instruments in basement of buildings located on rock or very stiff ground.
Records from 21 different stations.
Most records from digital instruments with 200 Hz sampling frequency and high dynamic
range.
Seismic moments calculated using the strong-motion data.
Most data from M0 5 1023 dyn cm and from de 40 km.
115
2.153 Si & Midorikawa (1999, 2000)

Ground-motion model for rupture distance is:
log A = aMw + hD +
di Si + e log(X + c1 10c2 Mw ) kX
where A is in cm/s2 , a = 0.50, h = 0.0036, d1 = 0, d2 = 0.09, d3 = 0.28, e = 0.60,

k = 0.003 and = 0.27 (c1 and c2 are not given).
Ground-motion model for equivalent hypocentral distance (EHD) is:
log A = aMw + hD +
di Si + e log Xeq kXeq
where A is in cm/s2 , a = 0.50, h = 0.0043, d1 = 0, d2 = 0.01, d3 = 0.22, e = 0.61,

k = 0.003 and = 0.28.
Use two site categories for most records following Joyner & Boore (1981):
1. Rock
2. Soil
Records from free-eld or small buildings where soil-structure interaction effects are
negligible.
Records from three different type of instrument so instrument correct. Filter with corner
frequencies, chosen according to noise level, a) 0.08 & 0.15 Hz, b) 0.10 & 0.20 Hz or c)
0.15 to 0.33 Hz.
Exclude records obviously affected by soil liquefaction.
Focal depth (dened as average depth of fault plane), D, between 6 and 120 km; most
less than 40 km.
Select records satisfying: distances < 300 km for Mw > 7, distances < 200 km for
6.6 Mw 7, distances < 150 km for 6.3 Mw 6.5 and distances < 100 km for
Mw < 6.3.
Fix k = 0.003.
Multiply rock PGAs by 1.4 to get soil PGA based on previous studies.
Use three fault types: crustal (<719 records from 9 earthquakes) S1 = 1, S2 =
0, S3 = 0, inter-plate (<291 records from 7 earthquakes) S2 = 1, S1 = 0, S3 = 0 and
intra-plate (<127 records from 5 earthquakes) S3 = 1, S1 = 0, S2 = 0.
Use weighted regression giving more weight to near-source records (weight factor of 8
for records < 25 km, 4 for records between 20 and 50 km, 2 for records between 50 and
100 km and 1 for records > 100 km). Use only three earthquakes with sufcient nearsource data to nd c1 and c2 then use all earthquakes to nd a, h, di , e in second stage
using weighted regression dependent on number of recordings for each earthquake
(weight factor of 3 for >83 records, 2 for between 19 and 83 records, 1 for <19 records.
Note that Mw and D are positively correlated so a and h may not be correctly determined
when using rupture distance. Constrain a for rupture distance model to that obtained for
EHD and constrain PGA to be independent of magnitude at 0 km and repeat regression.
Coefcients given above.
116
2.154
Spudich et al. (1999) & Spudich & Boore (2005)
Update of Spudich et al. (1997) see Section 2.130.

log10 Z = b1 + b2 (M 6) + b3 (M 6)2 + b5 log10 D + b6
2 + h2
with: D =
rjb
where
Z is in g, b1 = 0.299, b2 = 0.229, b3 = 0, b5 = 1.052, b6 = 0.112, h = 7.27 and
= 12 + 22 + 32 where 1 = 0.172, 2 = 0.108 and for randomly oriented horizontal
component 3 = 0.094 and for larger horizontal component 3 = 0.
Values of 3 (used to compute standard deviation for a randomly
orientated component)
reported in Spudich et al. (1999) are too large by a factor of 2.
Use two site categories (could not use more or Vs,30 because not enough data):
= 0 Rock: includes hard rock (12 records) (plutonic igneous rocks, lava ows, welded
tuffs and metamorphic rocks unless severely weathered when they are soft rock),
soft rock (16 records) (all sedimentary rocks unless there was some special characteristic noted in description, such as crystalline limestone or massive cliff-forming
sandstone when they are hard rock) and unknown rock (8 records). 36 records in
total.
= 1 Soil (alluvium, sand, gravel, clay, silt, mud, ll or glacial outwash of more than 5 m
deep): included shallow soil (8 records) (5 to 20 m deep), deep soil (77 records)
(> 20 m deep) and unknown soil (21 records). 106 records in total.
Applicable for extensional regimes, i.e. those regions where lithosphere is expanding
areally. Signicantly different ground motion than non-extensional areas.
Criteria for selection of records is: Mw 5.0, df 105 km. Reject records from
structures of more than two storeys or from deeply embedded basements or those which
triggered on S wave. Also reject those close to dams which may be affected by dam.
Also only use records already digitised.
Include records from those instrument beyond cutoff distance, i.e. beyond rst instrument which did not trigger, because of limited records and lack of data on non-triggering.
Not enough data to be able to nd all coefcients so use b2 and b3 from Boore et al.
(1993) and b6 from Boore et al. (1994a).
One-stage maximum likelihood method used because many events used which only
have one record associated with them and the two-stage method underestimates the
earthquake-to-earthquake component of variation in that case.
Correction technique based on uniform correction and processing using upper, fh , and
lower, fl , frequencies for passband based on a visual inspection of Fourier amplitude
spectrum and baseline tting with a polynomial of degree 5.
Check to see whether normal and strike-slip earthquakes give signicantly different
ground motions. No signicant difference.
117
2.155 Wang et al. (1999)

log A = a + bMs + c log R + dR

where A is in cm/s2 , using just soil records a = 0.430, b = 0.428, c = 0.764, d =
0.00480 and = 0.271.
Use records from aftershocks of Tangshan earthquake.
Note Ms values used may have some systematic deviation from other regions and errors, which decrease with increasing magnitude, can reach 0.5.
Errors in epicentral locations not less than 2 km. Reject 3 records because have R <
2 km, if include then nd standard deviation increases and c obtained is unreasonable.
Fit equation to all data (both rock and soil) but note that only for reference. Also t
equation to soil data only (2.1 R 41.3 km, 3.7 Ms 4.9, 33 records from 6
earthquakes).
Remove all four earthquakes with Ms < 4.0, for which error in magnitude determination
is large, and t equation to soil data only (2.8 R 41.1 km, 4.5 Ms 4.9, 13
records from 2 earthquakes). Find smaller uncertainties.
Also t data to log A = a + bMs c log(R + R0 ); nd similar results.
Also use resultant of both horizontal components; nd similar results to using larger
component.
Also t eastern North America data (3.9 R 61.6 km, 2.3 Ms 3.8, 7 records
from 3 earthquakes); nd similar attenuation characteristics.
All equations pass F-tests.
2.156 Zar et al. (1999)

log A = aM bX d log X + ci Si
where units of A not given (but probably m/s2 ), for vertical PGA a = 0.362, b = 0.0002,
c1 = 1.124, c2 = 1.150, c3 = 1.139, c4 = 1.064, d = 1 and = 0.336 and
for horizontal PGA a = 0.360, b = 0.0003, c1 = 0.916, c2 = 0.862, c3 = 0.900,
c4 = 0.859, d = 1 and = 0.333.
Use four site categories, which were based on H/V receiver function (RF) measurements (use geotechnical measurements at 50 sites and strong-motion accelerograms at
other sites):
Site class 1 RF does not exhibit any signicant amplication below 15 Hz. Corresponds to rock
and stiff sediment sites with average S-wave velocity in top 30 m (Vs,30 ) > 700 m/s.
Use c1 .
118
Site class 2 RF exhibits a fundamental peak exceeding 3 at a frequency between 5 and 15 Hz.
Corresponds to stiff sediments and/or soft rocks with 500 < Vs,30 700 m/s. Use
c2 .
Site class 3 RF exhibits peaks between 2 and 5 Hz. Corresponds to alluvial sites with 300 <
Vs,30 500 m/s. Use c3 .
Site class 4 RF exhibits peaks for frequencies < 2 Hz. Corresponds to thick soft alluvium. Use
c4 .
Only 100 records are associated with earthquakes with known focal mechanisms, 40
correspond to strike-slip/reverse, 31 to pure strike-slip, 24 to pure reverse and 4 to a
pure vertical plane. Note that use of equations should be limited to sources with such
mechanisms.
Use only records for which the signal to noise ratio was acceptable.
Source parameters from teleseismic studies available for 279 records.
Calculate source parameters directly from the strong-motion records for the remaining
189 digital records using a source model. Hypocentral distance from S-P time and
seismic moment from level of acceleration spectra plateau and corner frequency.
Focal depths from 9 to 133 km but focal depth determination is very imprecise and majority of earthquakes are shallow.
Suggest that whenever estimation of depth of earthquake is impossible use distance to
surface projection of fault rather than hypocentral distance because differences between
hypocentral and epicentral distances are not signicant for shallow earthquakes.
Also derive equations based only on data from the Zagros thrust fault zone (higher seismic activity rate with many earthquakes with 4 M 6) and based only on data from
the Alborz-Central Iran zone (lower seismic activity rate but higher magnitude earthquakes). Find some differences between regions.
Investigate xing d to 1 (corresponding to body waves) and to 0.5 (corresponding to
surface waves).
Note that there are very few (only two) near-eld (from less than 10 km from surface
fault rupture) records from earthquakes with Mw > 6.0 and so results are less certain
for such combinations of magnitude and distance.
2.157
Ambraseys & Douglas (2000), Douglas (2001b) & Ambraseys

& Douglas (2003)
log y = b1 + b2 Ms + b3 d + bA SA + bS SS
where y is in m/s2 , for horizontal PGA b1 = 0.659, b2 = 0.202, b3 = 0.0238,
bA = 0.020, bS = 0.029 and = 0.214 and for vertical PGA b1 = 0.959, b2 = 0.226,
b3 = 0.0312, bA = 0.024, bS = 0.075 and = 0.270.
Assume decay associated with anelastic effects due to large strains and cannot use
both log d and d because highly correlated in near eld.
119
Use four site categories (often use shear-wave velocity proles):

L Very soft soil: approximately Vs,30 < 180 m/s, (combine with category S) SA =
0, SS = 1, 4 records.
S Soft soil: approximately 180 Vs,30 < 360 m/s SA = 0, SS = 1, 87 records.
A Stiff soil: approximately 360 Vs,30 < 750 m/s SA = 1, SS = 0, 68 records.
R Rock: approximately Vs,30 > 750 m/s SA = 0, SS = 0, 23 records.
where Vs,30 is average shear-wave velocity to 30 m. Know no site category for 14
records.
Use only records from near eld where importance of vertical acceleration is greatest.
Select records with Ms 5.8, d 15 km and focal depth h 20 km. Do not use
magnitude dependent denition to avoid correlation between magnitude and distance
for the records.
Focal depths, 1 h 19 km.
Majority (133 records, 72%) of records from W. N. America, 40 records (22%) from
Europe and rest from Canada, Nicaragua, Japan and Taiwan.
Consider three source mechanisms but do not model:
1. Normal, 8 earthquakes, 16 records.
2. Strike-slip, 18 earthquakes, 72 records.
3. Thrust, 16 earthquakes, 98 records.
Use only free-eld records using denition of Joyner & Boore (1981), include a few
records from structures which violate this criterion but feel that structure did not affect
record in period range of interest.
Records well distributed in magnitude and distance so equations are well constrained
and representative of entire dataspace. Note lack of records from normal earthquakes.
Correlation coefcient between magnitude and distance is 0.10.
Use same correction procedure (elliptical lter with pass band 0.2 to 25 Hz, roll-off frequency 1.001 Hz, sampling interval 0.02 s, ripple in pass-band 0.005 and ripple in stopband 0.015 with instrument correction) for almost all records. Use 19 records available
only in corrected form as well because in large magnitude range. Think different correction procedures will not affect results.
Try both one-stage and two-stage regression method for horizontal PGA; nd large differences in b2 but very similar b3 . Find that (by examining cumulative frequency distribution
graphs for magnitude scaling of one-stage and two-stage methods) that two-stage better
represents large magnitude range than one-stage method. Examine plot of amplitude
factors from rst stage of two-stage method against Ms ; nd that amplitude factor of the
two Kocaeli (Ms = 7.8) records is far below least squares line through the amplitude
factors. Remove the two Kocaeli records and repeat analysis; nd b2 from two-stage
method is changed by a lot but b2 from one-stage method is not. Conclude two-stage
method is too greatly inuenced by the two records from Kocaeli and hence use onestage method.
Find b2 and b3 signicantly different than 0 at 5% level but bA and bS not signicant.
120
2.158
Bozorgnia et al. (2000)
ln Y
= c1 + c2 Mw + c3 (8.5 Mw )2
+ c4 ln({Rs2 + [(c5 SHS + c6 {SP S + SSR } + c7 SHR )
exp(c8 Mw + c9 {8.5 Mw }2 )]2 }1/2 ) + c10 FSS + c11 FRV + c12 FT H
+ c13 SHS + c14 SP S + c15 SSR + c16 SHR

HS Holocene soil: recent alluvium SHS = 1, SP S = 0, SSR = 0, SHR = 0.
PS Pleistocene soil: older alluvium SP S = 1, SHS = 0, SSR = 0, SHR = 0.
SR Soft rock SSR = 1, SHS = 0, SP S = 0, SHR = 0.
HR Hard rock SHR = 1, SHS = 0, SP S = 0, SSR = 0.
Consider all records to be free-eld.
All earthquakes occurred in shallow crustal tectonic environment.
Consider three source mechanisms: strike-slip (FSS = 1, FRV = 0, FT H = 0) 20+
earthquakes (including 1+ normal faulting shock), reverse (FRV = 1, FSS = 0, FT H = 0)
7+ earthquakes and thrust (FT H = 1, FSS = 0, FRV = 0) 6+ earthquakes.
Coefcients not given, only predictions.
2.159
Campbell & Bozorgnia (2000)
ln Y
= c1 + c2 Mw + c3 (8.5 Mw )2 + c4 ln({Rs2 + [(c5 + c6 {SP S + SSR } + c7 SHR )

exp(c8 Mw + c9 {8.5 Mw }2 )]2 }1/2 ) + c10 FSS + c11 FRV + c12 FT H
+ c13 SHS + c14 SP S + c15 SSR + c16 SHR
where Y is in g, for horizontal uncorrected PGA c1 = 2.896, c2 = 0.812, c3 = 0,

c4 = 1.318, c5 = 0.187, c6 = 0.029, c7 = 0.064, c8 = 0.616, c9 = 0, c10 = 0,
c11 = 0.179, c12 = 0.307, c13 = 0, c14 = 0.062, c15 = 0.195, c16 = 0.320
and = 0.509, for horizontal corrected PGA c1 = 4.033, c2 = 0.812, c3 = 0.036,
c4 = 1.061, c5 = 0.041, c6 = 0.005, c7 = 0.018, c8 = 0.766, c9 = 0.034, c10 = 0,
c11 = 0.343, c12 = 0.351, c13 = 0, c14 = 0.123, c15 = 0.138, c16 = 0.289 and
= 0.465, for vertical uncorrected PGA c1 = 2.807, c2 = 0.756, c3 = 0, c4 = 1.391,
c5 = 0.191, c6 = 0.044, c7 = 0.014, c8 = 0.544, c9 = 0, c10 = 0, c11 = 0.091,
c12 = 0.223, c13 = 0, c14 = 0.096, c15 = 0.212, c16 = 0.199 and = 0.548 and
for vertical corrected PGA c1 = 3.108, c2 = 0.756, c3 = 0, c4 = 1.287, c5 = 0.142,
c6 = 0.046, c7 = 0.040, c8 = 0.587, c9 = 0, c10 = 0, c11 = 0.253, c12 = 0.173,
c13 = 0, c14 = 0.135, c15 = 0.138, c16 = 0.256 and = 0.520.
121
HS Holocene soil: soil deposits of Holocene age (11,000 years or less), generally described on geological maps as recent alluvium, approximate average shear-wave
velocity in top 30 m is 290 m/s SHS = 1, SP S = 0, SSR = 0, SHR = 0.
PS Pleistocene soil: soil deposits of Pleistocene age (11,000 to 1.5 million years) ,
generally described on geological maps as older alluvium or terrace deposits, approximate average shear-wave velocity in top 30 m is 370 m/s SP S = 1, SHS =
0, SSR = 0, SHR = 0.
SR Soft rock: primarily includes sedimentary rock deposits of Tertiary age (1.5 to 100
million years), approximate average shear-wave velocity in top 30 m is 420 m/s
SSR = 1, SHS = 0, SP S = 0, SHR = 0.
HR Hard rock: primarily includes older sedimentary rock deposits, metamorphic rock
and crystalline rock, approximate average shear-wave velocity in top 30 m is 800 m/s
SHR = 1, SHS = 0, SP S = 0, SSR = 0.
Earthquakes from shallow crustal active tectonic regions.
Most earthquakes with 6 Mw 7.
Use three source mechanism categories:
SS Strike-slip: primarily vertical or near-vertical faults with predominantly lateral slip
(includes only normal faulting earthquake in set), FSS = 1, FRV = 0, FT H = 0.
RV Reverse: steeply dipping faults with either reverse or reverse-oblique slip, FRV =
1, FSS = 0, FT H = 0.
TH Thrust: shallow dipping faults with predominantly thrust slip including blind-thrust
shocks, FT H = 1, FSS = 0, FRV = 0.
Consider all records to be free-eld. Records from ground level in instrument shelter or
a building <3 storeys high (<7 if located on hard rock). Include records from dam abutments to increase number of rock records. Exclude data from basements of buildings of
any size or at toe or base of dams.
Exclude data from Rs > 60 km to avoid complicating problems related to arrival of multiple reections from lower crust. Distance range is believed to include most ground
shaking amplitudes of engineering interest, except for possibly long period spectral accelerations on extremely poor soil.
Equations for uncorrected (Phase 1 standard level of processing) and corrected (Phase
2 standard level of processing).
Find sediment depth (depth to basement rock) has signicant effect on amplitude of
ground motion and should be taken into account; it will be included once its mathematical
form is better understood.
2.160 Field (2000)

2
(M, rjb , Vs ) = b1 + b2 (M 6) + b3 (M 6)2 + b5 ln[(rjb
+ h2 )0.5 ] + bv ln(Vs /Va )
122
(M, rjb , Vs ) is natural logarithm of ground-motion parameter (e.g. ln(PGA) where

PGA is in g), b1,ss = 0.853 0.28, b1,rv = 0.872 0.27, b2 = 0.442 0.15, b3 =
0.067 0.16, b5 = 0.960 0.07, bv = 0.154 0.14, h = 8.90 km, Va = 760 m/s,
= 0.47 0.02 (intra-event) and = 0.23 (inter-event). Also gives overall = (0.93
0.10Mw )0.5 for Mw 7.0 and overall = 0.48 for Mw > 7.0.
Uses six site classes (from Wills et al. (2000)):
B 760 Vs 1500 m/s. Uses Vs = 1000 m/s in regression. 12 records.
BC Boundary between B and C. Uses Vs = 760 m/s in regression. 36 records.
C 360 Vs 760 m/s. Uses Vs = 560 m/s in regression. 16 records.
CD Boundary between C and D. Uses Vs = 360 m/s in regression. 166 records.
D 180 Vs 360 m/s. Uses Vs = 270 m/s in regression. 215 records.
DE Boundary between D and E. Uses Vs = 180 m/s in regression. 2 records.
Uses data from the SCEC Phase III strong-motion database.
Uses three faulting mechanism classes:
Use b1,ss Strike-slip. 14 earthquakes, 103 records.
Use b1,rv Reverse. 6 earthquakes, 300 records.
Use 0.5(b1,ss + b1,rv ) Oblique. 8 earthquakes, 46 records.
Notes that data is unbalanced in that each earthquake has a different number of records
for each site type hence it is important to correct observations for the inter-event terms
before examining residuals for site effects.
Plots average site class residuals w.r.t. BC category and the residuals predicted by
equation and nds good match.
Uses 197 records with basin-depth estimates (depth dened to the 2.5 km/s shearwave velocity isosurface) to examine dependence of inter-event corrected residuals w.r.t.
basin depth. Plots residuals against basin depth and ts linear function. Finds that all
slopes are signicantly different than zero by more than two sigmas. Finds a signicant trend in subset of residuals where basin-depths are known w.r.t. magnitude hence
needs to test whether basin-depth effect found is an artifact of something else. Hence
derives Ground-motion models (coefcients not reported) using only subset of data for
which basin-depth estimates are known and examines residuals w.r.t. basin-depth for
this subset. Finds similar trends as before hence concludes found basin effect is truly an
effect of the basin. Notes that basin-depth coefcients should be derived simultaneously
with other coefcients but because only a subset of sites have a value this could not be
done.
Tests for nonlinearity by plotting residuals for site class D w.r.t. predicted ground motion
for BC boundary. Fits linear equation. Finds slope for PGA is signicantly different than
zero.
Notes that due to large number of class D sites site nonlinearity could have affected
other coefcients in equation leading to less of a trend in residuals. Tests for this by
plotting residuals for site classes B and BC combined w.r.t. predicted ground motion for
BC boundary. Fits linear equation. Finds non-signicant slopes. Notes that nonlinearity
123
may lead to rock ground motions being underestimated by model but not enough data
to conclude.
Investigates inter-event variability estimate through Monte Carlo simulations using 250
synthetic databases because uncertainty estimate of was considered unreliable possibly due to limited number of events. Find that there could be a problem with the
regression methodology adopted w.r.t. the estimation of .
Plots squared residuals w.r.t. magnitude and ts linear equations. Finds signicant
trends. Notes that method could be not statistically correct because squared residuals
are not Gaussian distributed.
Plots squared residuals w.r.t. Vs and does not nd a signicant trend.
Provides magnitude-dependent estimates of overall up to Mw 7.0 and constant overall
for larger magnitudes.
Tests normality of residuals using Kolmogorov-Smirnov test and nds that the null hypothesis cannot be rejected. Also examines theoretical quantile-quantile plots and nds
nothing notable.
2.161 Jain et al. (2000)

ln(PGA) = b1 + b2 M + b3 R + b4 ln(R)
where PGA is in g, for central Himalayan earthquakes b1 = 4.135, b2 = 0.647, b3 =
0.00142, b4 = 0.753 and = 0.59 and for non-subduction earthquakes in N.E. India
b1 = 3.443, b2 = 0.706, b3 = 0, b4 = 0.828 and = 0.44 (coefcients of other
equations not given here because they are for a particular earthquake).
Data from strong-motion accelerographs (SMA) and converted from structural response
recorders (SRR), which consist of six seismoscopes with natural periods 0.40, 0.75 and
1.25 s and damping levels 5 and 10%. Conversion achieved by deriving spectral amplication factors (ratio of response ordinate and PGA) using SMA recordings close to
SRR, checking that these factors were independent of distance. The mean of the six
estimates of PGA (from the six spectral ordinates) from each SRR are then used as PGA
values. Check quality of such PGA values through statistical comparisons and discard
those few which appear inconsistent.
Data split into four categories for which derive separate equations:
a Central Himalayan earthquakes (thrust): (32 SMA records, 117 SRR records), 3
earthquakes with 5.5 M 7.0, focal depths 10 h 33 km and 2 R
322 km.
b Non-subduction earthquakes in NE India (thrust): (43 SMA records, 0 SRR records),
3 earthquakes with 5.2 M 5.9, focal depths 33 h 49 km and 6 R
243 km.
c Subduction earthquakes in NE India: (33 SMA records, 104 SRR records), 1 earthquake with M = 7.3, focal depth h = 90 km and 39 R 772 km.
124
d Bihar-Nepal earthquake in Indo-Gangetic plains (strike-slip): (0 SMA records, 38

SRR records), 1 earthquake with M = 6.8, focal depth h = 57 km and 42 R
337 km.
Limited details of fault ruptures so use epicentral distance.
Use epicentral locations which give best correlation between distance and PGA.
Find PGA not well predicted by earlier equations.
Simple model and regression method because of limited data.
Remove one PGA value from category b equation because signicantly affecting equation and because epicentral location only approximate.
Constrain b3 for category b equation to zero because otherwise positive.
Category c originally contained another earthquake (14 SMA records, M = 6.1, 200
d 320 km) but gave very small b2 so exclude it.
Equations for category c and category d have b2 equal to zero because only one earthquake.
Find considerable differences between predicted PGA in different regions.
Note lack of data hence use equations only as rst approximation.
2.162
Kobayashi et al. (2000)
log10 y = aM bx log(x + c10dM ) + eh + Sk

where h is focal depth, y is in cm/s2 , a = 0.578, b = 0.00355, e = 0.00661
, S = 0.069,
SR = 0.210, SH = 0.114, SM = 0.023, SS = 0.237 and T = 2 + 2 where
= 0.213 and = 0.162.
Use four site categories (most data from medium and hard soils):
Sk = SR Rock
Sk = SH Hard soil
Sk = SM Medium soil
Sk = SS Soft soil
S is the mean site coefcient, i.e. when do not consider site category.
Records interpolated in frequency domain from 0.02 to 0.005 s interval and displacement
time history calculated using a fast Fourier transform (FFT) method having perpended
to beginning and appended to end at least 5 s of zeros to record. Number of samples in
FFT is large enough that duration used in FFT is at least twice that of selected duration
for processing window so that numerical errors are small. Bandpass Ormsby lter used,
with limits 0.2 and 24.5 Hz, and displacement time history plotted. If displacement in
pre- and appended portions is large then increase lower frequency limit in lter until
displacements are small, using smoothed Fourier spectral amplitudes from 0.05 to 25 Hz
to make choice.
125
Most earthquakes are intra-slab.

Note lack of near-eld data for all magnitudes, most data from > 100 km, therefore use
coefcients, c and d, from an early study.
Excludes data from distances greater than the distance at which an earlier study predicts
PGA < 0.02 m/s2 .
Consider residuals of earthquakes in western Japan (a small subset of data) and nd
small difference in anelastic coefcient and focal depth coefcient but note may be due
to small number of records or because type of source not modelled.
Note model predicts intraslab motions well but signicantly over predicts interface motions.
Plots site correction factors (difference between individual site factor and mean factor
for that category) and nd rock sites have largest variation, which suggest due to hard
and soft rock included.
Examine residual plots. Find no signicant bias.
2.163 Monguilner et al. (2000a)

log am = C00 + C1 M + C2 + C3 log + C40 Sr
where am is in unknown unit, = DE2 + H 2 + S 2 , DE is epicentral distance, H is

focal depth, S is fault area and C00 = 1.23, C1 = 0.068, C2 = 0.001 and C3 = 0.043
( is not given). Note that there are typographical inconsistencies in the text, namely Sr
maybe should be replaced by Sal .
Use two site categories (based on Argentinean seismic code):
Sr = 1 Stiff soil (IIA ).

Sr = 0 Intermediate stiff soil (IIB ).
Since there is no geotechnical data available, classify sites, assuming a uniform surface
layer, using the predominant period of ground motions estimated using Fourier spectra
to get an equivalent shear-wave velocity (mainly these are between 100 and 400 m/s).
Records from instruments located in basements or ground oors of relatively small buildings.
Records from SMAC and SMA-1 instruments.
Uniform digitisation and correction procedure applied to all records to reduce noise in
high and low frequency range.
Calculate fault area using log S = Ms + 8.13 0.6667 log(/) where is stress
drop, is average stress and is rigidity.
Most magnitudes between 5.5 and 6.0.
126
Most records from DE < 100 km.

Most focal depths, H 40 km. One earthquake with H = 120 km.
Use weighted regression because of a correlation between magnitude and distance of
0.35. Weight each record by i = (M + DH )/2 where (note there are typographical
errors in formulae in paper):
DH =
MT
log DHT
ns (is )M (ni )ne (ni , is )MT

ncat
ns (is ) log DH(ni )ne (ni , is ) log DHT
ncat
M (ni )
n
cat
log DH(ni )
ncat
where M (ni ) is the width of the ni th magnitude interval and log DH(ni ) is the width
of the ni th distance interval, ncat is total number of intervals, ni the index of the interval,
ne (ni , is ) is the number of records in interval ni from site classication is and ns is
the number of records from site classication is . Use two site classications, three
magnitude intervals and four epicentral distance intervals so ncat = 2 3 4 = 24.
First do regression on log ai = C0 + C1 M + C2 + C3 log and then regress residuals,
i , against C4 Sr + C5 Sal where Sal = 1 if site is intermediate stiff soil and Sal = 0
otherwise. Then C00 = C0 + C5 and C40 = C4 + C5 . Similar method to that used by
Ambraseys et al. (1996).
2.164
Sharma (2000)
Based on Sharma (1998), see 2.147.

A is in g and coefcients are: c1 = 2.87, c2 = 0.634, c3 = 0.62, b = 1.16 and
= 0.142.
Fit log A = b log X + c to data from each earthquake separately and nd average b
equal to 1.18. Then t log A = aM b log X + c to data from all earthquakes and nd
b = 0.405. Fit log A = b log X + di li to all data, where li = 1 for ith earthquake and
0 otherwise and nd b = 1.16, use this for rest of analysis.
2.165
Smit et al. (2000)
= a + bM log R + dR
D 2 + h2
where R =
log Y
where Y is in cm/s2 , a = 0.72, b = 0.44, d = 0.00231, h = 4.5 and = 0.28.

Records from soil or alluvium sites.
127
All records corrected.

Note that scatter can be reduced by increasing number of records used (especially in
near eld), improving all seismological and local site parameters and increasing number
of variables (especially in near eld and those modelling local site behaviour) but that
this requires much more information than is available.
2.166 Takahashi et al. (2000)

log10 [y] = aM bx log10 (x + c10dM ) + e(h hc )h + Sk

where y is in cm/s2 , a = 0.446, b = 0.00350, c = 0.012, d = 0.446
, e = 0.00665,
S = 0.941, SR = 0.751, SH = 0.901, SM = 1.003, SS = 0.995, T = 2 + 2 where
= 0.135 (intra-event) and = 0.203 (inter-event), hc is chosen as 20 km because
gave positive depth term.
Sk = SR Rock
Sk = SH Hard soil
Sk = SM Medium soil
Sk = SS Soft soil
Note site conditions for many stations are uncertain. S is the mean site term for all data.
Note ISC focal depths, h, signicant reduce prediction errors compared with JMA depths.
h = 1 for h hc and h = 0 otherwise.
Most Japanese data from x > 50 km.
Use 166 Californian and Chilean (from 2 earthquakes) records to control model in near
source.
Due to lack of multiple records from many sites and because c and d require nearsource records use a maximum likelihood regression method of two steps. Firstly, nd
all coefcients using all data except those from sites with only one record associated
with them and unknown site class. Next, use individual site terms for all sites so as to
reduce inuence of uncertainty because of approximate site classications and nd a, b,
e and site terms using c and d from rst step.
Intra-event and inter-event residuals decrease with increasing magnitude.
Conclude variation in residuals against distance is due to small number of records at
short and large distances.
Individual site factors means prediction error propagates into site terms when number of
records per station is very small.
Note model may not be suitable for seismic hazard studies because model prediction
errors are partitioned into T and mean site terms for a given site class. Suitable model
can be derived when accurate site classications are available.
128
2.167
Wang & Tao (2000)
log Y = C + ( + M ) log(R + R0 )
where Y is in cm/s2 , C = 4.053, = 2.797, = 0.251, R0 = 8.84 and = 0.257.
Use same data as Joyner & Boore (1981), see Section 2.31.
Use a two-stage method based on Joyner & Boore (1981). Firstly t data to log Y =
C + ni=1 (ai Ei ) log(Ri +R0 ), where Ei = 1 for records from ith earthquake and Ei = 0
otherwise, to nd C and ai for each earthquake. Next t a = + M to nd and
using ai from rst stage.
2.168
Chang et al. (2001)
Ground-motion model for shallow crustal earthquakes is:
ln A = c1 + c2 M c3 ln Dp (c4 c5 Dp ) ln De
where A is in cm/s2 , c1 = 2.8096, c2 = 0.8993, c3 = 0.4381, c4 = 1.0954, c5 = 0.0079
and = 0.60.
Ground-motion model for subduction earthquakes is:
ln A = c01 + c02 M c03 ln Dp c04 ln Dh

where A is in cm/s2 , c01 = 4.7141, c02 = 0.8468, c03 = 0.17451, c04 = 1.2972 and
= 0.56.
Note that there is limited site information available for strong-motion stations in Taiwan
so do not consider local site effects.
Use strong-motion data from Central Weather Bureau from 1994 to 1998 because it is
more numerous and of better quality than older data.
Separate earthquakes into shallow crustal and subduction earthquakes because of different seismic attenuation and seismogenic situation for the two types of earthquake.
Shallow crustal earthquakes are mostly due to continental deformation, shallow collision
or back-arc opening or are the uppermost interface earthquakes. Focal depths depth between 1.1 and 43.7 km with most shallower than 20 km. Most records from earthquakes
with 4.5 Mw 6.0.
Subduction earthquakes are located in the Wadati-Benioff zone or the deep lateral collision zone and are principally intraslab. Focal depth between 39.9 and 146.4 km.
Do not use records from earthquakes associated with coseismic rupture because they
have complex near-eld source effects.
To avoid irregularly large amplitudes at great distances reject distant data predicted to
be less than trigger level plus 1 standard deviation using this threshold formula: aMw
b ln D +c ln V , where V is geometric mean of PGA equal to threshold plus 1 standard
deviation. For shallow crustal earthquakes: a = 0.64, b = 0.83, c = 2.56 and V = 6.93
and for subduction earthquakes: a = 0.76, b = 1.07, c = 3.13 and V = 6.79.
129
For shallow crustal earthquakes examine effect of focal depth on seismic attenuation
by nding geometric attenuation rate using epicentral distance, De , for earthquakes
with 5 km depth intervals. Find that deeper earthquakes have slower attenuation than
shallow earthquakes. Therefore assume ground motion, A, is product of fsource (source
effects) and fgeometricalspreading (geometrical spreading effects) where
(c4 c5 Dp )
fsource = C1 exp(c2 M )/Dpc3 and fgeometricalspreading = De

depth.
where Dp is focal
For subduction earthquakes examine effect of focal depth in the same way as done
for shallow crustal earthquakes but nd no effect of focal depth on attenuation rate.
Therefore use fgeometricalspreading = Dhc4 .
Plot residuals of both equations against distance and nd no trend.
Note that it is important to separate subduction and shallow crustal earthquakes because
of the different role of focal depth and attenuation characteristics.
Plot residual maps of ground motion for Taiwan and nd signicant features showing the
important effect of local structures on ground motion.
2.169 Herak et al. (2001)

log amax
= c1 + c2 ML + c3 log c24 + D2
where amax is in g, for horizontal PGA c1 = 1.300 0.192, c2 = 0.331 0.040,

c3 = 1.152 0.099, c4 = 11.8 4.8 km and = 0.311 and for vertical PGA c1 =
1.5180.293, c2 = 0.3020.035, c3 = 1.0610.096, c4 = 11.05.5 and = 0.313.
Records from 39 sites. Records from instruments on ground oor or in basements of
relatively small structures.
Site information only available for a small portion of the recording sites and therefore is
not considered. Believe that most sites are rock or stiff soil.
All records from Kinemetrics SMA-1s.
Select records with ML 4.5 and D 200 km because of poor reliability of SMA-1
records for small earthquakes and to avoid problems related to a possible change of
geometrical spreading when surface waves start to dominate over body waves at large
distances.
Bandpass lter with passbands selected for which signal-to-noise ratio is > 1. Widest
passband is 0.0725 Hz.
Do not use rjb because do not accurately know causative fault geometry for majority of
events.
Do not include an anelastic decay term because data is inadequate to independently
determine geometric and anelastic coefcients.
130
Note correlation between magnitude and distance in data distribution therefore use twostage regression. Because many earthquakes have only a few records data is divided
into classes based on magnitude (details not given).
Most data from ML < 5.5, particularly data from D < 20 km.
Find all coefcients signicantly different than 0 at levels exceeding 0.999.
Also regress using one-stage method and nd practically equal coefcients and larger
standard errors.
Find residuals are approximately lognormally distributed with slight asymmetry showing
longer tail on positive side. Relate this to site amplication at some stations balanced by
larger than expected number of slightly negative residuals.
Find no distance or magnitude dependence of residuals.
Compute ratio between larger and average horizontal component as 1.15.
Believe that higher than normal is due to lack of consideration of site effects and due
to the use of repi rather than rjb .
2.170
Lussou et al. (2001)
log PSA(f ) = a(f )M + b(f )R log R + c(i, f )

where PSA(f ) is in cm/s2 , a(f ) = 3.71 101 , b(f ) = 2.54 103 , c(A, f ) = 0.617,
c(B, f ) = 0.721, c(C, f ) = 0.845, c(D, f ) = 0.891 and = 3.13 101 .
Use four site categories, based on Vs,30 (average shear-wave velocity in top 30 m) as
proposed in Eurocode 8:
A Vs,30 > 800 m/s. Use c(A, f ). 14 records.
B 400 < Vs,30 800 m/s. Use c(B, f ). 856 records.
C 200 < Vs,30 400 m/s. Use c(C, f ). 1720 records.
D 100 < Vs,30 200 m/s. Use c(D, f ). 421 records.
Good determination of site conditions between shear-wave velocities have been measured down to 10 to 20 m at every site. Extrapolate shear-wave velocity data to 30 m to
nd Vs,30 . Vs,30 at stations is between about 50 m/s and about 1150 m/s.
Use data from Kyoshin network from 1996, 1997 and 1998.
All data from free-eld sites.
No instrument correction needed or applied.
Use data from earthquakes with MJMA > 3.5 and focal depth < 20 km because want
to compare results with Ambraseys et al. (1996) and Boore et al. (1997). Also this
criteria excludes data from deep subduction earthquakes and data that is not signicant
for seismic hazard studies.
131
Homogeneous determination of JMA magnitude and hypocentral distance.

Roughly uniform distribution of records with magnitude and distance.
Assume pseudo-spectral acceleration for 5% damping at 0.02 s equals PGA.
Note equation valid for 3.5 MJMA 6.3 and 10 rhypo 200 km.
Find inclusion of site classication has reduced standard deviation.
2.171 Sanchez & Jara (2001)

log(Amax ) = aMs + b log R + c

where the units of Amax are not given7 , a = 0.444, b = 2.254 and c = 4.059 ( is not
given).
Use one site category: rm ground.
2.172 Wu et al. (2001)

log10 (Y ) = C1 + C2 Mw log10 (rrup + h) + C3 rrup

where Y is in cm/s2 , C1 = 0.00215, C2 = 0.581, C3 = 0.00414, h = 0.00871
100.5Mw from the square root of the expected rupture area and = 0.79 (in terms of
natural logarithms not common logarithms).
Select data from events with ML > 5 and focal depths < 35 km to restrict interest to
large shallow earthquakes, which cause most damage.
Relocate events using available data.
Develop empirical relationship to convert ML to Mw .
Develop relation for use in near real-time (within 2 min) mapping of PGA following an
earthquake.
Select records from the Taiwan Rapid Earthquake Information Release System (TREIRS)
and records from the TSMIP if rrup < 30 km so as not to bias the results at larger distances by untriggered instruments.
Most data from 50 dr 200 km and 5 Mw 6.
Compute site correction factors for TSMIP stations (since these sites have not been well
classied), S , by averaging residuals between observed and predicted values. After
applying these site amplications in regression analysis obtain reduced of 0.66.
Display inter-event residuals w.r.t. Mw before and after site correction.
7
There could be a typographical error in the article since the use of common (base ten) logarithms leads to very
large ground motions the authors may mean natural logarithms.
132
2.173
Chen & Tsai (2002)
log10 PGA = 0 + 1 M + 2 M 2 + 3 R + 4 log10 (R + 5 106 M )

where PGA is in cm/s2 , 0 = 4.3662.020, 1 = 2.5400.714, 2 = 0.1720.0611,
3 = 0.001730.000822, 4 = 1.8450.224, 5 = 0.07460.411, 6 = 0.2210.405,
e2 = 0.0453 0.0113 (earthquake-specic variance), s2 = 0.0259 0.00699 (sitespecic variance) and r2 = 0.0297 0.00235 (record-specic variance). signies the
estimated standard errors.
Records from 45 stations on rock and rm soil. All sites have more than two records.
Use a new estimation procedure where the residual variance is decomposed into components due to various source of deviations. Separate variance into earthquake-toearthquake variance, site-to-site variance and the remainder.
Proposed method does not require additional regression or searching procedures.
Perform a simulation study and nd proposed procedure yields estimates with smaller
biases and take less computation time than do other similar regression techniques.
Visually examine the equation for various magnitude values before regressing.
2.174
Gregor et al. (2002a)
Ground-motion model is (their model D):
ln GM = 1 + 2 M + (3 + 4 M ) ln[D + exp(5 )] + 6 (1 S) + 7 (M 6)2 + 8 F

+ 9 / tanh(D + 10 )
where GM is in g, 1 = 4.31964, 2 = 0.00175, 3 = 2.40199, 4 = 0.19029, 5 =
2.14088, 6 = 0.09754, 7 = 0.21015, 8 = 0.38884, 9 = 2.29732, 10 = 448.88360,
= 0.5099 (intra-event) and = 0.4083 (inter-event) for horizontal PGA using the static
dataset without the Chi-Chi data and 1 = 1.50813, 2 = 0.15024, 3 = 2.52562, 4 =
0.17143, 5 = 2.12429, 6 = 0.10517, 7 = 0.16655, 8 = 0.22243, 9 = 0.11214,
10 = 19.85830, = 0.5141 (intra-event) and = 0.4546 (inter-event) for vertical PGA
using the static dataset without the Chi-Chi data. Coefcients are also given for the
three other models and for both the dynamic and the static datasets but are not reported
here due to lack of space.
S = 0 Soil: includes sites located on deep broad and deep narrow soil deposits.
S = 1 Rock: includes sites that are located on shallow stiff soil deposits;
Use three rupture mechanism categories:
F = 0 Strike-slip, 39 earthquakes, 387 records;

F = 0.5 Reverse/oblique, 13 earthquakes, 194 records;
F = 1 Thrust, 16 earthquakes, 412 records.
133
Process records using two procedures as described below.

1. Use the standard PEER procedure with individually chosen lter cut-offs.
2. Fit the original integrated velocity time-history with three different functional forms
(linear in velocity; bilinear, piecewise continuous function; and quadratic in velocity). Choose the best-t result and view it for reasonableness. Differentiate the
velocity time-history and then low-pass lter with a causal Butterworth lter with
cut-offs about 50 Hz.
PGA values from the two processing techniques are very similar.
Investigate using a nonlinear model for site response term but the resulting models did
not improve the t.
Also try three other functional forms:
ln(GM) = 1 + 2 M + 3 ln[D + 4 exp(5 M )] + 6 (1 S) + 7 F

ln(GM) = 1 + 2 M + (3 + 4 M ) ln[D + exp(5 )] + 6 (1 S) + 7 (M 6)2 + 8 F
ln(GM) = 1 + 2 M + 3 ln[D + exp(5 M )] + 6 (1 S) + 7 F + 8 / tanh(D + 9 )
which all give similar standard deviations and predictions but prefer model D.
Models oversaturate slightly for large magnitudes at close distances. Therefore recommend that the PGA equations are not used because this oversaturation is based on very
little data.
Because the Chi-Chi short period ground motions may be anomalous relative to California they develop equations including and excluding the Chi-Chi data, which only affects
predictions for large magnitudes (M > 7.5).
2.175 Glkan & Kalkan (2002)

ln Y
= b1 + b2 (M 6) + b3 (M 6)2 + b5 ln r + bV ln(VS /VA )
2
where r = (rcl
+ h2 )1/2
where Y is in g, b1 = 0.682, b2 = 0.253, b3 = 0.036, b5 = 0.562, bV = 0.297,

VA = 1381, h = 4.48 and = 0.562.
Soft soil Average shear-wave velocity, VS , is 200 m/s. 40 records.
Soil Average shear-wave velocity, VS , is 400 m/s. 24 records.
Rock Average shear-wave velocity, VS , is 700 m/s. 29 records.
Actual shear-wave velocities and detailed site descriptions are not available for most
stations in Turkey. Therefore estimate site classication by analogy with information in
similar geologic materials. Obtain type of geologic material in number of ways: consultation with geologists at Earthquake Research Division of Ministry of Public Works and
Settlement, various geological maps, past earthquake reports and geological references
prepared for Turkey.
134
Only used records from small earthquakes recorded at closer distances than large earthquakes to minimize the inuence of regional differences in attenuation and to avoid the
complex propagation effects coming from longer distances.
Only use records from earthquakes with Mw & 5.0 to emphasize ground motions of
engineering signicance and to limit analysis to more reliably recorded earthquakes.
During regression lock magnitudes within 0.25 magnitude unit bands centred at halves
or integer magnitudes to eliminate errors coming from magnitude determination.
Note that use of epicentral distance for small earthquakes does not introduce signicant
bias because dimensions of rupture area of small earthquakes are usually much smaller
than distance to recording stations.
Examine peak ground motions from the small number of normal- (14 records) and
reverse-faulting (6 records) earthquakes in set and nd that they were not signicantly
different from ground motions from strike-slip earthquakes (73 records). Therefore combine all data.
Records mainly from small buildings built as meteorological stations up to three stories
tall. Note that this modies the recorded accelerations and hence increases the uncertainty.
Exclude data from aftershocks (mainly of the Kocaeli and Duzce earthquakes) because
it was from free-eld stations and did not want to mix it with the data from the non-freeeld records.
Exclude a few records for which PGA of mainshock is . 0.04 g.
Note that there is limited data and the data is poorly distributed. Also note that there
is near-total lack of knowledge of local geology and that some of the records could be
affected by the building in which the instrument was housed.
More than half the records (49 records, 53% of total) are from two Mw > 7 earthquakes
(Kocaeli and Duzce) so the results are heavily based on the ground motions recorded in
these two earthquakes.
2.176
Khademi (2002)
Y = C1 exp(C2 M )((R + C3 exp(C4 M ))C5 ) + C6 S

where Y is in g, C1 = 0.040311, C2 = 0.417342, C3 = 0.001, C4 = 0.65, C5 =
0.351119 and C6 = 0.035852 for horizontal PGA and C1 = 0.0015, C2 = 0.8548,
C3 = 0.001, C4 = 0.4, C5 = 0.463 and C6 = 0.0006 for vertical PGA.
S = 0 Rock, site categories I and II of Iranian building code.

S = 1 Soil, site categories III and IV of Iranian building code.
135
Selection criteria are: i) causative earthquake, earthquake fault (if known) and respective
parameters are determined with reasonable accuracy, ii) PGA of at least one component
> 50 gal, iii) records from free-eld conditions or ground level of low-rise buildings (<
three stories), iv) some aftershocks have been eliminated to control effect of a few large
earthquakes and v) records have been processed with acceptable lter parameters.
Regresses directly on Y not on logarithm of Y . Therefore does not calculate standard
deviation in normal way. Considers the deviation of individual records from predictive
equations as being PGA dependent. Finds that a sigmoidal model ts the data well.
Therefore Y = (ab + cxd )/(b + xd ) where Y is the error term and x is the predicted
ground motion, a = 0.038723, b = 0.00207, c = 0.29094 and d = 4.97132 for horizontal
PGA and a = 0.00561, b = 0.0164, c = 0.1648 and d = 1.9524 for vertical PGA.
2.177 Margaris et al. (2002a) & Margaris et al. (2002b)

ln Y = c0 + c1 Mw + c2 ln(R + R0 ) + c3 S
where Y is in cm/s2 , c0 = 4.16, c1 = 0.69, c2 = 1.24, R0 = 6, c3 = 0.12 and
= 0.70.
S = 0 NEHRP and UBC category B. 145 records.

S = 1 NEHRP and UBC category C. 378 records.
S = 2 NEHRP and UBC category D. 221 records.
Selection criteria are: a) earthquake has Mw 4.5, b) PGA 0.05 g and c) PGA <
0.05 g but another record from same earthquake has PGA 0.05 g.
Records mainly from normal faulting earthquakes.
Exclude data recorded in buildings with four stories or higher.
Automatically digitize records and process records homogenously, paying special attention to the lters used.
Correlation between Mw and R in set of records used. For 4.5 Mw 5.0 records
exist at R 40 km and for larger magnitudes records exist at intermediate and long
distances. For Mw > 6.0 there is a lack of records for R < 20 km.
Use a two step regression method. In rst step use all records to nd c1 . In second step
use records from earthquakes with Mw 5.0 to nd c0 , c2 and c3 .
Adopt R0 = 6 km because difcult to nd R0 via regression due to its strong correlation
with c2 . This corresponds to average focal depth of earthquakes used.
Also try Ground-motion model: ln Y = c00 +c01 Mw +c02 ln(R2 +h20 )1/2 +c03 S . Coefcients
are: c00 = 3.52, c01 = 0.70, c02 = 1.14, h0 = 7 km (adopted), c03 = 0.12 and = 0.70.
Find no apparent trends in residuals w.r.t. distance.
Due to distribution of data, equations valid for 5 R 120 km and 4.5 Mw 7.0.
136
2.178
Saini et al. (2002)
Ground-motion model is unknown.
2.179
Schwarz et al. (2002)
log10 aH(V ) = c1 + c2 ML + c4 log10 (r) + cR SR + cA SA + cS SS
where r =
Re2 + h20
where aH(V ) is in g, c1 = 3.0815, c2 = 0.5161, c4 = 0.9501, cR = 0.1620, cA =
0.1078, cS = 0.0355, h0 = 2.0 and = 0.3193 for horizontal PGA and c1 = 2.8053,
c2 = 0.4858, c4 = 1.1842, cR = 0.1932, cA = 0.0210, cS = 0.0253, h0 = 2.5 and
= 0.3247 for vertical PGA.
R Rock, subsoil classes A1, (A2) Vs > 800 m/s (according to E DIN 4149) or subsoil
class B (rock) 760 < Vs 1500 m/s (according to UBC 97). SR = 1, SA = 0,
SS = 0. 59 records.
A Stiff soil, subsoil classes (A2), B2, C2 350 Vs 800 m/s (according to E DIN
4149) or subsoil class C (very dense soil and soft rock) 360 < Vs 760 m/s
(according to UBC 97). SA = 1, SR = 0, SS = 0. 88 records.
S Soft soil, subsoil classes A3, B3, C3 Vs < 350 m/s (according to E DIN 4149) or
subsoil class D (stiff clays and sandy soils) 180 < Vs 360 m/s (according to UBC
97). SS = 1, SR = 0, SA = 0. 536 records.
KOERI stations classied using UBC 97 and temporary stations of German TaskForce
classied using new German code E DIN 4149. Classify temporary stations of German
TaskForce using microtremor H/V spectral ratio measurements by comparing shapes of
H/V spectral ratios from microtremors to theoretical H/V spectral ratios as well as with
theoretical transfer functions determined for idealized subsoil proles.
Use Kocaeli aftershock records from temporary German TaskForce stations (records
from earthquakes with 1 . ML < 4.9 and distances Re < 70 km, 538 records) and
from mainshock and aftershocks records from Kandilli Observatory (KOERI) stations
(4.8 ML 7.2 and distances 10 Re 250 km, 145 records).
Visually inspect all time-histories and only use those thought to be of sufciently good
quality.
Baseline correct all records.
Use technique of Ambraseys et al. (1996) to nd the site coefcients cR , cA and cS , i.e.
use residuals from regression without considering site classication.
Note that equations may not be reliable for rock and stiff soil sites due to the lack of data
and that equations probably only apply for 2 ML 5 due to lack of data from large
magnitude earthquakes.
137
2.180 Stamatovska (2002)

ln PGA = b0 + bM M + bR ln
[
( R )2
e
]1/2
+ h2
+C
where PGA is in cm/s2 . For Bucharest azimuth b0 = 0.21056, bM = 1.29099, bR =

0.80404, C = 40 and = 0.52385, for Valeni azimuth b0 = 1.52412, bM = 1.42459,
bR = 0.70275, C = 40 and = 0.51389 and for Cherna Voda b0 = 4.16765, bM =
1.11724, bR = 1.44067, C = 40 and = 0.47607.
Incomplete data on local site conditions so not included in study.
Some strong-motion records are not from free-eld locations.
Uses to characterise the non-homogeneity of region. Includes effect of instrument
location w.r.t. the main direction of propagation of seismic energy,
as well as the non
homogeneous attenuation in two orthogonal directions. = (1 + tg 2 )/(a2 + tg 2 )
where is angle between instrument and main direction of seismic energy or direction
of fault projection on surface and a is parameter dening the non-homogeneous attenuation in two orthogonal directions, or relation between the semi-axes of the ellipse of
seismic eld.
Uses a two step method. In rst step derive equations for each earthquake using
ln PGA = b00 + b1 ln(Re /). In the second step the complete Ground-motion model
is found by normalizing separately for each earthquake with a value of dened for that
earthquake according to the location for which the equation was dened.
Notes that there is limited data so coefcients could be unreliable.
Strong-motion records processed by different institutions.
2.181 Tromans & Bommer (2002)

log y = C1 + C2 Ms + C4 log r + CA SA + CS SS
where r =
d2 + h20
where y is in cm/s2 , C1 = 2.080, C2 = 0.214, h0 = 7.27, C4 = 1.049, CA = 0.058,
CS = 0.085 and = 0.27.
S Soft soil, Vs,30 360 m/s. SS = 1, SA = 0. 25% of records.
A Stiff soil, 360 < Vs,30 < 750 m/s. SA = 1, SS = 0. 50% of records.
R Rock, Vs,30 750 m/s. SS = 0, SA = 0. 25% of records.
If no Vs,30 measurements at station then use agency classications.
138
Supplement dataset of Bommer et al. (1998) with 66 new records using same selection
criteria as Bommer et al. (1998) with a lower magnitude limit of Ms = 5.5. Remove 3
records from Bommer et al. (1998) with no site classications.
Roughly uniform distribution of records w.r.t. magnitude and distance. New data contributes signicantly to large magnitude and near-eld ranges.
Correct records using an elliptical lter selecting an appropriate low-frequency cut-off,
fL , individually for each record using the criterion of Bommer et al. (1998).
Plot PGA against fL for two pairs of horizontal components of ground motion from the
BOL and DZC stations from the Duzce earthquake (12/11/1999). Record from BOL
was recorded on a GSR-16 digital accelerograph and that from DZC was recorded on a
SMA-1 analogue accelerograph. Find PGA is stable for low-frequency cut-offs up to at
least 0.4 Hz for the selected records.
2.182
Zonno & Montaldo (2002)
log10 (Y ) = a + bM + c log10 (R2 + h2 )1/2 + e

where Y is in g, a = 1.632, b = 0.304, c = 1, h = 2.7, e = 0 and = 0.275.
Soil Vs,30 750 m/s, = 0.
Rock Vs,30 > 750 m/s, = 1.
Note that amount of data available for the Umbria-Marche area in central Italy is sufciently large to perform statistical analysis at regional scale.
Focal depths between 2 and 8.7 km. Exclude data from an earthquake that occurred at
47 km.
Select only records from earthquakes with ML 4.5 recorded at less than 100 km.
Exclude data from Nocera Umbra station because it shows a strong amplication effect
due to the presence of a sub-vertical fault and to highly fractured rocks.
Uniformly process records using BAP (Basic strong-motion Accelerogram Processing
software). Instrument correct records and band-pass lter records using a high-cut
lter between 23 and 28 Hz and a bi-directional Butterworth low-cut lter with corner
frequency of 0.4 Hz and rolloff parameter of 2.
Note that can use ML because it does not saturate until about 6.5 and largest earthquake in set is ML = 5.9.
More than half of records are from earthquakes with ML 5.5.
State that equations should not be used for ML > 6 because of lack of data.
Use similar regression method as Ambraseys et al. (1996) to nd site coefcient, e.
139
2.183 Alarcn (2003)

Ground-motion model is (his model 2):
log(a) = A + BM + Cr + D log(r)
where a is in gal, A = 5.5766, B = 0.06052, C = 0.0039232, D = 2.524849 and
= 0.2597.
Due to lack of information classify stations as soil or rock (stations with 10 m of soil).
Only derives equation for rock.
Uses data from National Accelerometer Network managed by INGEOMINAS from 1993
to 1999.
Exclude data from subduction zone, focal depths h > 60 km.
Focal depths, 11.4 h 59.8 km.
Exclude data from earthquakes with ML < 4.0.
Exclude data with PGA < 5 gal. 5 PGA 100.1 gal.
Derive equations using four different models:
a = C1 eC2 M (R + C3 )C4
log(a) = A + BM + Cr + D log(r)
log(y) = C0 + C1 (M 6) + C2 (M 6) + C3 log(r) + C4 r
ln(a) = a + bM + d ln(R) + qh
2.184 Alchalbi et al. (2003)

log A = b0 + b1 Mc + br log r
where A is in g, b0 = 1.939, b1 = 0.278, b2 = 0.858 and = 0.259 for horizontal
PGA and b0 = 2.367, b1 = 0.244, b2 = 0.752 and = 0.264 for vertical PGA.
Use two site categories: bedrock (S = 0) and sediments (S = 1) but found the coefcient b3 in the term +b3 S is close to zero so repeat analysis constraining b3 to 0.
Records from SSA-1 instruments.
Carefully inspect and select records.
Do not use record from the Aqaba (M = 7.2) earthquake because it is very far and was
only recorded at one station.
Do not use records from buildings or dams because they are affected by response of
structure.
Instrument correct records. Apply bandpass lter (0.1 to 25 Hz) to some low-quality
records.
140
Do regression using only records from earthquakes with 4.8 M 5.8 and also using
only records from earthquakes with 3.5 M 4.5.
Most data from M 5 and r 100 km.
Note that use a small set of records and so difcult to judge reliability of derived equation.
2.185
Atkinson & Boore (2003)
= c1 + c2 M + c3 h + c4 R g log R + c5 slSC + c6 slSD + c7 slSE
2
where R =
Dfault
+ 2
log Y
= 0.00724 100.507M
(f
1)(PGA
100)
rx
1
400
sl =
1 (f 1)
rx 100
1 PGA400
for
for
for
for
for
PGArx 100 cm/sorf 1 Hz

100 < PGArx < 500 cm/s&1 Hz < f < 2 Hz
PGArx 500 cm/s&1 Hz < f < 2 Hz
100 < PGArx < 500 cm/s&f 2 Hz
PGArx 500 cm/s&f 2 Hz)
where Y is in cm/s2 , f is frequency of interest, PGArx is predicted PGA on NEHRP

B sites, c1 = 2.991, c2 = 0.03525, c3 = 0.00759, c4 = 0.00206, 1 = 0.20 (intraevent) and 2 = 0.11 (inter-event) for interface events and c1 = 0.04713, c2 = 0.6909,
c3 = 0.01130, c4 = 0.00202, 1 = 0.23 and 2 = 0.14 for in-slab events and c5 =
0.19, c6 = 0.24, c7 = 0.29 for all events. g = 101.20.18M for interface events and
g = 100.3010.01M for in-slab events. Recommended revised c1 for interface events
in Cascadia is 2.79 and in Japan 3.14, recommended revised c1 for in-slab events in
Cascadia is 0.25 and in Japan 0.10.
B NEHRP site class B, Vs,30 > 760 m/s. SC = 0, SD = 0 and SE = 0.
C NEHRP site class C, 360 < Vs,30 760 m/s. SC = 1, SD = 0 and SE = 0.
D NEHRP site class D, 180 Vs,30 360 m/s. SD = 1, SC = 0 and SE = 0.
E NEHRP site class E, Vs,30 < 180 m/s. SE = 1, SC = 0 and SD = 0.
Stations in KNET were classied using shear-wave velocity proles using an statistical
method to extrapolate measured shear-wave velocities to depths up to 1020 m to 30 m.
Stations in Guerrero array assumed to be on rock, i.e. site class B. Broadband stations
in Washington and British Columbia sited on rock (Vs,30 1100 m/s), i.e. site class B.
Strong-motion stations in Washington classied using map of site classes based on correlations between geology and Vs,30 in Washington, and veried at 8 stations using actual borehole measurements. Converted Youngs et al. (1997) Geomatrix classications
by assuming Geomatrix A=NEHRP B, Geomatrix B=NEHRP C, Geomatrix C/D=NEHRP
D and Geomatrix E=NEHRP E using shear-wave velocity and descriptions of Geomatrix
classication.
Note that cannot develop equations using only Cascadia data because not enough data.
Combine data of Crouse (1991) and Youngs et al. (1997) with additional data from
Cascadia (strong-motion and broadband seismographic records), Japan (KNET data),
Mexico (Guerrero array data) and El Salvador data.
141
Classify event by type using focal depth and mechanism as:

In-slab All earthquakes with normal mechanism. Earthquakes with thrust mechanism at
depths > 50 km or if occur on steeply dipping planes.
Interface Earthquakes with thrust mechanism at depths < 50 km on shallow dipping planes.
Exclude events of unknown type.
Exclude events with focal depth h > 100 km.
Exclude events that occurred within crust above subduction zones.
Use many thousands of extra records to explore various aspects of ground motion scaling with M and Dfault .
Data relatively plentiful in most important M -Dfault ranges, dened according to deaggregations of typical hazard results. These are in-slab earthquakes of 6.5 M 7.5 for
40 Dfault 100 km and interface earthquakes of M 7.5 for 20 Dfault 200 km.
Data from KNET from moderate events at large distances are not reliable at higher
frequencies due to instrumentation limitations so exclude KNET data from M < 6 at
Dfault > 100 km and for M 6 at Dfault > 200 km. Excluded data may be reliable at
low frequencies.
Estimate Dfault for data from Crouse (1991) and for recent data using fault length versus
M relations of Wells & Coppersmith (1994) to estimate size of fault plane and assuming epicentre lies above geometric centre of dipping fault plane. Veried estimates for
several large events for which fault geometry is known.
Perform separate regressions for interface and in-slab events because analyses indicated extensive differences in amplitudes, scaling and attenuation between two types.
Experiment with a variety of functional forms. Selected functional form allows for magnitude dependence of geometrical spreading coefcient, g ; the observed scaling with
magnitude and amplitude-dependent soil nonlinearity.
For h > 100 km use h = 100 km to prevent prediction of unrealistically large amplitudes
for deeper earthquakes.
R is approximately equal to average distance to fault surface. is dened from basic
fault-to-site geometry. For a fault with length and width given by equations of Wells &
Coppersmith (1994), the average distance to the fault for a specied Dfault is calculated
(arithmetically averaged from a number of points distributed around the fault), then used
to determine . Magnitude dependence of R arises because large events have a large
spatial extent, so that even near-fault observation points are far from most of the fault.
Coefcients in were dened analytically, so as to represent average fault distance,
not be regression. Although coefcients in were varied over a wide range but did not
improve accuracy of model predictions.
Determine magnitude dependence of g by preliminary regressions of data for both interface and in-slab events. Split data into 1 magnitude unit increments to determine
slope of attenuation as a function of magnitude using only 1 and 2 s data and records
with 50 Dfault 300 km (50 km limit chosen to avoid near-source distance saturation effects). Within each bin regression was made to a simple functional form:
142
log Y 0 = a1 + a2 M g log R + a3 S where Y 0 = Y exp(0.001R), i.e. Y corrected

for curvature due to anelasticity, and S = 0 for NEHRP A or B and 1 otherwise. g is
far-eld slope determined for each magnitude bin.
Nonlinear soil effects not strongly apparent in database on upon examination of residuals from preliminary regressions, as most records have PGA < 200 cm/s2 , but may be
important for large M and small Dfault . To determine linear soil effects perform separate
preliminary regressions for each type of event to determine c5 , c6 and c7 assuming linear
response. Smooth these results (weighted by number of observations in each subset)
to x c5 , c6 and c7 (independent of earthquake type) for subsequent regressions. sl
was assigned by looking at residual plots and from consideration of NEHRP guidelines.
Conclude that there is weak evidence for records with PGArx > 100 cm/s2 , for NEHRP
E sites at periods < 1 s. Use these observations to x sl for nal regression.
Final regression needs to be iterated until convergence because of use of PGArx in
denition of dependent variable.
To optimize t for M -Dfault range of engineering interest limit nal regression to data
within: 5.5 M < 6.5 and Dfault 80 km, 6.5 M < 7.5 and Dfault 150 km
and M 7.5 and D 300 km for interface events and 6.0 M < 6.5 and Dfault
100 km and M 6.5 and Dfault 200 km for in-slab events. These criteria rened
by experimentation until achieved an optimal t for events that are important for seismic
hazard analysis. Need to restrict M -Dfault for regression because set dominated by
records from moderate events and from intermediate distances whereas hazard is from
large events and close distances.
Lightly smooth coefcients (using a weighted 3-point scheme) over frequency to get
smooth spectral shape and allows for reliable linear interpolation of coefcients for frequencies not explicitly used in regression.
In initial regressions, use a M2 term as well as a M term leading to a better t over
a linear magnitude scaling but lead to a positive sign of the M2 rather than negative
as expected. Therefore to ensure the best t in the magnitude range that is important
for hazard and constrained by data quadratic source terms ret to linear form. Linear
model constrained to provide same results in range 7.0 M 8.0 for interface events
and 6.5 M 7.5 for in-slab events. To ensure that non-decreasing ground motion
amplitudes for large magnitudes: for M > 8.5 use M = 8.5 for interface events and for
M > 8.0 use M = 8.0 for in-slab events.
Calculate based on records with M 7.2 and Dfault 100 km for interface events
and M 6.5 and Dfault 100 km for in-slab events. These magnitude ranges selected
to obtain the variability applicable for hazard calculations. Do not use KNET data when
computing because data appear to have greater high-frequency site response than
data from same soil class from other regions, due to prevalence of sites in Japan with
shallow soil over rock.
Determine 1 using data for several well-recorded
large events and determining average
2
2
value. Then calculate 2 assuming = 1 + 2 .
Examine residuals w.r.t. Dfault using all data from M 5.5 and Dfault 200 km and
M 6.5 and Dfault 300 km. Find large variability but average residuals near 0 for
Dfault 100 km.
143
Find signicantly lower variability for M 7.2 events ( = 0.20.35 for larger events
and = 0.250.4 for smaller events).
Examine graphs and statistics of subsets of data broken down by magnitude, soil type
and region. Find signicant positive residuals for M < 6.6 due to use of linear scaling
with magnitude. Accept positive residuals because small magnitudes do not contribute
strongly to hazard.
Find large positive residuals for class C sites for interface events (most records are
from Japan) whereas residuals for class C sites for in-slab events (which are from both
Japan and Cascadia) do not show trend. No other overwhelming trends. Differences in
residuals for Japan and Cascadia class C sites likely due to differences in typical soil
proles in the two regions within the same NEHRP class. Sites in Japan are typically
shallow soil over rock, which tend to amplify high frequencies, whereas in Cascadia most
soil sites represent relatively deep layers over rock or till. Provide revised c1 coefcients
for Japan and Cascadia to model these differences.
Note that debate over whether 1992 Cape Mendocino earthquake is a subduction zone
or crustal earthquake. Excluding it from regressions has a minor effect on results, reducing predictions for interface events for M < 7.5.
2.186 Boatwright et al. (2003)

log PGA = (M) log g(r) 0 (M)r

where
(M) = 1 + 2 (M 5.5) for M 5.5

= 1 + 3 (M 5.5) for M > 5.5
0
(M) = 1 for M 5.5

= 1 10(M5.5) M > 5.5
g(r) = r for r r0 = 27.5 km
= r0 (r/r0 )0.7 for r > r0 = 27.5 km
where PGA is in m/s2 , 1 = 1.45 0.24, 2 = 1.00 0.01, 3 = 0.31 0.09,
1 = 0.0073 0.0003, = 0.30 0.06, e = 0.170 (inter-earthquake) and r = 0.361
(intra-earthquake).
Classify station into four classes using the NEHRP categories using geological maps:
B Rock. Amplication from category C 0.79.
C Soft rock or stiff soil. Amplication from category C 1.00.
D Soft soil. Amplication from category C 1.35.
E Bay mud. Amplication from category D 1.64.
The amplications (from Boore et al. (1997)) are used to correct for site effects.
For some stations in the broadband Berkeley Digital Seismic Network, which are in
seismic vaults and mine adits and therefore have low site amplications, use one-half
the above site amplications.
144
Use data from August 1999 and December 2002 from the northern California ShakeMap
set of data. Extend set to larger earthquakes by adding data from nine previous large
northern California earthquakes.
Focal depths, 0.1 h 28..8 km.
Use hypocentral distance because this distance is available to ShakeMap immediately
after an earthquake. Note that this is a poor predictor of near-eld ground motion from
extended faults.
Plot decay of PGA with distance for two moderate earthquakes (M = 4.9, M = 3.9)
and nd decay is poorly t by a power-law function of distance and that tting such
an equation who require PGA r2 , which they believe is physically unrealistic for
body-wave propagation.
Find that PGAs atten or even increase at large distances, which is believed to be due
to noise. Hence use a magnitude-dependent limit of rmax = 100(M 2) 400 km,
determined by inspecting PGA and PGV data for all events, to exclude problem data.
Fit data from each event separately using log PGA = r log g(r) + log sBJF . Find
varies between four groups: events near Eureka triple junction, events within the Bay
Area, events near San Juan Bautista and those in the Sierras and the western Mojave
desert.
Use a numerical search to nd the segmentation magnitude M0 . Choose M0 = 5.5 as
the segmentation magnitude because it is the lowest segmentation magnitude within a
broad minimum in the 2 error for the regression.
Fit magnitude-dependent part of the equation to the PGA values scaled to 10 km and
site class C.
Note that the PGAs predicted are signicantly higher than those given by equations
derived by Joyner & Boore (1981) and Boore et al. (1997) because of use of hypocentral
rather than fault distance.
Recompute site amplications relative to category C as: for B 0.84 0.03, for D 1.35
0.05 and for E 2.17 0.15.
2.187
log y = C1 + C2 M + C4 log(
r2 + h2 ) + CA SA + CS SS + CN FN + CR FR
where y is in g, C1 = 1.482, C2 = 0.264, C4 = 0.883, h = 2.473, CA = 0.117,

CS = 0.101, CN = 0.088, CR = 0.021, 1 = 0.243 (intra-event) and 2 = 0.060
(inter-event).
Use four site conditions but retain three (because only three records from very soft (L)
soil which combine with soft (S) soil category):
145

L Very soft soil: Vs 180 m/s, SA = 0, SS = 1, 3 records.
Use same data as Ambraseys et al. (1996).
Use three faulting mechanism categories:
S Strike-slip: earthquakes with rake angles () 30 30 or 150 or

150 , FN = 0, FR = 0, 47 records.
N Normal: earthquakes with 150 < < 30 , FN = 1, FR = 0, 146 records.
R Reverse: earthquakes with 30 < < 150 , FR = 1, FN = 0, 229 records.
Earthquakes classied as either strike-slip or reverse or strike-slip or normal depending

on which plane is the main plane were included in the corresponding dip-slip category.
Some records (137 records, 51 normal, 10 strike-slip and 76 reverse) from earthquakes
with no published focal mechanism (80 earthquakes) were classied using the mechanism of the mainshock or regional stress characteristics.
Try using criteria of Campbell (1997) and Sadigh et al. (1997) to classify earthquakes
w.r.t. faulting mechanism. Also try classifying ambiguously classied earthquakes as
strike-slip. Find large differences in the faulting mechanism coefcients with more
stricter criteria for the rake angle of strike-slip earthquakes leading to higher CR coefcients.
Note that distribution of records is reasonably uniform w.r.t. to mechanism although

signicantly fewer records from strike-slip earthquakes.
Try to use two-stage maximum-likelihood method as employed by Ambraseys et al.

(1996) but nd numerical instabilities in regression.
Also rederive mechanism-independent equation of Ambraseys et al. (1996) using onestage maximum-likelihood method.
146
2.188
Campbell & Bozorgnia (2003d,a,b,c) & Bozorgnia & Campbell (2004b)
ln Y
= c1 + f1 (Mw ) + c4 ln
f2 (Mw , rseis , S) + f3 (F ) + f4 (S)
+f5 (HW, F, Mw , rseis )

where f1 (Mw ) = c2 Mw + c3 (8.5 Mw )2
2
f2 (Mw , rseis , S) = rseis
+ g(S)2 (exp[c8 Mw + c9 (8.5 Mw )2 ])2
g(S) = c5 + c6 (SV F S + SSR ) + c7 SF R

f3 (F ) = c10 FRV + c11 FT H
f4 (S) = c12 SV F S + c13 SSR + c14 SF R
f5 (HW, F, Mw , rseis ) = HWfHW (Mw )fHW (rseis )(FRV + FTH )
{
0 for rjb 5 km or > 70
HW =
(SV F S + SSR + SF R )(5 rjb )/5 for rjb < 5 km & 70
0 for Mw < 5.5
Mw 5.5 for 5.5 Mw 6.5

fHW (Mw ) =
1 for Mw > 6.5

{
c15 (rseis /8) for rseis < 8 km
fHW (rseis ) =
c15 for rseis 8 km
where Y is in g, rjb is the distance to the surface projection of rupture and is the
dip of the fault; for uncorrected horizontal PGA: c1 = 2.896, c2 = 0.812, c3 = 0.0,
c4 = 1.318, c5 = 0.187, c6 = 0.029, c7 = 0.064, c8 = 0.616, c9 = 0, c10 = 0.179,
c11 = 0.307, c12 = 0.062, c13 = 0.195, c14 = 0.320, c15 = 0.370 and =
c16 0.07Mw for Mw < 7.4 and = c16 0.518 for Mw 7.4 where c16 = 0.964 or
= c17 + 0.351 for PGA 0.07 g, = c17 0.132 ln(PGA) for 0.07 g < PGA < 0.25 g
and = c17 + 0.183 for PGA 0.25 g where c17 = 0.263; for corrected horizontal
PGA: c1 = 4.033, c2 = 0.812, c3 = 0.036, c4 = 1.061, c5 = 0.041, c6 = 0.005,
c7 = 0.018, c8 = 0.766, c9 = 0.034, c10 = 0.343, c11 = 0.351, c12 = 0.123,
c13 = 0.138, c14 = 0.289, c15 = 0.370 and = c16 0.07Mw for Mw < 7.4 and
= c16 0.518 for Mw 7.4 where c16 = 0.920 or = c17 + 0.351 for PGA
0.07 g, = c17 0.132 ln(PGA) for 0.07 g < PGA < 0.25 g and = c17 + 0.183
for PGA 0.25 g where c17 = 0.219; for uncorrected vertical PGA: c1 = 2.807,
c2 = 0.756, c3 = 0, c4 = 1.391, c5 = 0.191, c6 = 0.044, c7 = 0.014, c8 = 0.544,
c9 = 0, c10 = 0.091, c11 = 0.223, c12 = 0.096, c13 = 0.212, c14 = 0.199,
c15 = 0.630 and = c16 0.07Mw for Mw < 7.4 and = c16 0.518 for Mw 7.4
where c16 = 1.003 or = c17 + 0.351 for PGA 0.07 g, = c17 0.132 ln(PGA) for
0.07 g < PGA < 0.25 g and = c17 + 0.183 for PGA 0.25 g where c17 = 0.302; and
for corrected vertical PGA: c1 = 3.108, c2 = 0.756, c3 = 0, c4 = 1.287, c5 = 0.142,
c6 = 0.046, c7 = 0.040, c8 = 0.587, c9 = 0, c10 = 0.253, c11 = 0.173, c12 = 0.135,
c13 = 0.138, c14 = 0.256, c15 = 0.630 and = c16 0.07Mw for Mw < 7.4 and
= c16 0.518 for Mw 7.4 where c16 = 0.975 or = c17 + 0.351 for PGA 0.07 g,
= c17 0.132 ln(PGA) for 0.07 g < PGA < 0.25 g and = c17 + 0.183 for PGA
0.25 g where c17 = 0.274.
147
Firm soil Generally includes soil deposits of Holocene age (less than 11,000 years old) described on geological maps as recent alluvium, alluvial fans, or undifferentiated
Quaternary deposits. Approximately corresponds to Vs,30 = 298 92 m/s and
NEHRP soil class D. Uncorrected PGA: 534 horizontal records and 525 vertical records and corrected PGA: 241 horizontal records and 240 vertical records.
SV F S = 0, SSR = 0 and SF R = 0.
Very rm soil Generally includes soil deposits of Pleistocene age (11,000 to 1.5 million years old)
described on geological maps as older alluvium or terrace deposits. Approximately
corresponds to Vs,30 = 368 80 m/s and NEHRP soil class CD. Uncorrected PGA:
168 horizontal records and 166 vertical records and corrected PGA: 84 horizontal
records and 83 vertical records. SV F S = 1, SSR = 0 and SF R = 0.
Soft rock Generally includes sedimentary rock and soft volcanic deposits of Tertiary age (1.5
to 100 million years old) as well as softer units of the Franciscan Complex and
other low-grade metamorphic rocks generally described as melange, serpentine
and schist. Approximately corresponds to Vs,30 = 421 109 m/s and NEHRP soil
class CD. Uncorrected PGA: 126 horizontal records and 124 vertical records and
corrected PGA: 63 horizontal records and 62 vertical records. SSR = 1, SV F S = 0
and SF R = 0.
Firm rock Generally include older sedimentary rocks and hard volcanic deposits, high-grade
metamorphic rock, crystalline rock and the harder units of the Franciscan Complex generally described as sandstone, greywacke, shale, chert and greenstone.
Approximately corresponds to Vs,30 = 830 339 m/s and NEHRP soil class BC.
Uncorrected PGA: 132 horizontal records and 126 vertical records and corrected
PGA: 55 horizontal records and 54 vertical records. SF R = 1, SV F S = 0 and
SSR = 0.
Note that for generic soil (approximately corresponding to Vs,30 = 310 m/s and NEHRP
site class D) use SV F S = 0.25, SSR = 0, SF R = 0 and for generic rock (approximately
corresponding to Vs,30 = 620 m/s and NEHRP site class C) use SSR = 0.50, SF R =
0.50 and SV F S = 0.
Use four fault types but only model differences between strike-slip, reverse and thrust:
Normal Earthquakes with rake angles between 202.5 and 337.5 . 4 records from 1 earthquake.
Strike-slip Includes earthquakes on vertical or near-vertical faults with rake angles within 22.5
of the strike of the fault. Also include 4 records from 1975 Oroville normal faulting
earthquake. Uncorrected PGA: 404 horizontal records and 395 vertical records
and corrected PGA: 127 horizontal and vertical records. FRV = 0 and FT H = 0
Reverse Steeply dipping earthquakes with rake angles between 22.5 and 157.5 . Uncorrected PGA: 186 horizontal records and 183 vertical records and corrected PGA:
58 horizontal records and 57 vertical records. FRV = 1 and FT H = 0.
Thrust Shallow dipping earthquakes with rake angles between 22.5 and 157.5 . Includes
some blind thrust earthquakes. Uncorrected PGA: 370 horizontal records and
363 vertical records and corrected PGA: 258 horizontal records and 255 vertical
records. FT H = 1 and FRV = 0.
Note that for generic (unknown) fault type use FRV = 0.25 and FT H = 0.25.
Most records from 5.5 Mw 7.0.
148
Note that equations are an update to equations in Campbell (1997) because they used a
somewhat awkward and complicated set of Ground-motion models because there used
a mixture of functional forms. Consider that the new equations supersede their previous
studies.
Uncorrected PGA refers to the standard level of accelerogram processing known as
Phase 1. Uncorrected PGAs are either scaled directly from the recorded accelerogram or if the accelerogram was processed, from the baseline and instrument-corrected
Phase 1 acceleration time-history.
Corrected PGA measured from the Phase 1 acceleration time-history after it had been
band-pass ltered and decimated to a uniform time interval.
Restrict data to within 60 km of seismogenic rupture zone (rseis 60 km) of shallow
crustal earthquakes in active tectonic regions which have source and near-source attenuation similar to California. Most data from California with some from Alaska, Armenia,
Canada, Hawaii, India, Iran, Japan, Mexico, Nicaragua, Turkey and Uzbekistan. Note
some controversy whether this is true for all earthquakes (e.g. Gazli and Nahanni).
Exclude subduction-interface earthquakes.
Restrict earthquakes to those with focal depths < 25 km.
Exclude data from subduction-interface earthquakes, since such events occur in an entirely different tectonic environment that the other shallow crustal earthquakes, and it
has not been clearly shown that their near-source ground motions are similar to those
from shallow crustal earthquakes.
Restrict to rseis 60 km to avoid complications related to the arrival of multiple reections from the lower crust. Think that this distance range includes most ground-motion
amplitudes of engineering interest.
All records from free-eld, which dene as instrument shelters or non-embedded buildings < 3 storeys high and < 7 storeys high if located on rm rock. Include records from
dam abutments to enhance the rock records even though there could be some interaction between dam and recording site. Exclude records from toe or base of dam because
of soil-structure interaction.
Do preliminary analysis, nd coefcients in f3 need to be constrained in order to make
Y independent on Mw at rseis = 0, otherwise Y exhibits oversaturation and decreases
with magnitude at close distances. Therefore set c8 = c2 /c4 and c9 = c3 /c4 .
Functional form permits nonlinear soil behaviour.
Do not include sediment depth (depth to basement rock) as a parameter even though
analysis of residuals indicated that it is an important parameter especially at long periods. Do not think its exclusion is a serious practical limitation because sediment depth
is generally not used in engineering analyses and not included in any other widely used
attenuation relation.
Do not apply weights during regression analysis because of the relatively uniform distribution of records w.r.t. magnitude and distance.
To make regression analysis of corrected PGA more stable set c2 equal to value from
better-constrained regression of uncorrected PGAs.
149
Examine normalised residuals i = (ln Yi ln Y )/ln(Unc.PGA where ln Yi is the measured acceleration, Y is the predicted acceleration and ln(Unc.PGA is the standard deviation of the uncorrected PGA equation. Plot i against magnitude and distance and
nd models are unbiased.
Consider equations valid for Mw 5.0 and rseis 60 km. Probably can be extrapolated
to a distance of 100 km without serious compromise.
Note that should use equations for uncorrected PGA if only an estimate of PGA is required because of its statistical robustness. If want response spectra and PGA then
should use corrected PGA equation because the estimates are then consistent.
Note that should include ground motions from Kocaeli (17/8/1999, Mw = 7.4), Chi-Chi
(21/9/1999, Mw = 7.6), Hector Mine (16/10/1999, Mw = 7.1) and Duzce (12/11/1999,
Mw = 7.1) earthquakes but because short-period motions from these earthquakes was
signicantly lower than expected their inclusion could lead to unconservative estimated
ground motions for high magnitudes.
Prefer the relationship for in terms of PGA because statistically more robust. Note
that very few records to constrain value of for large earthquakes but many records to
constrain for PGA 0.25 g.
Find that Monte Carlo simulation indicates that all regression coefcients statistically
signicant at 10% level.
2.189 Halldrsson & Sveinsson (2003)

Ground-motion models are:
log A = aM b log R + c
where A is in g, a = 0.484, b = 1.4989, c = 2.1640 and = 0.3091, and:
log A = aM log R bR + c
a = 0.4805, b = 0.0049, c = 2.6860 and = 0.3415.
Vast majority of data from south Iceland (18 earthquakes in SW Iceland and 4 in N
Iceland).
Most data from less than 50 km and M < 5.5. 76% of data is from 5 to 50 km.
Examine residual plots against distance and nd no trends.
Recommend rst equation.
Most data from ve earthquakes (04/06/1998, 13/11/1998, 27/09/1999, 17/06/2000 and
21/06/2000).
2.190 Shi & Shen (2003)

log PGA = a1 + a2 Ms + a3 log[R + a4 exp(a5 Ms )]

where PGA is in cm/s2 , a1 = 1.3012, a2 = 0.6057, a3 = 1.7216, a4 = 1.126 and
a5 = 0.482 ( not reported).
150
2.191
Sigbjrnsson & Ambraseys (2003)
log10 (PGA) = b0 + b1 M log10 (R) + b2 R
R =
D2 + h2
where PGA is in g, b0 = 1.27800.1909, b1 = 0.28530.0316, b2 = 1.730103
2.132 104 and = 0.3368 ( indicates the standard deviation of the coefcients). h
was xed arbitrarily to 8 km.
Use data from ISESD (Ambraseys et al. , 2004). Select using de < 1000 km, 5 M 7
(where M is either Mw or Ms ).
Only use data from strike-slip earthquakes.
Note that coefcient of variation for b coefcients is in range 11 to 15%.
Note that b0 and b1 are very strongly negatively correlated (correlation coefcient of
0.9938), believed to be because PGA is governed by b0 + b1 M as D approaches
zero, but they are almost uncorrelated with b2 (correlation coefcients of 0.0679 and
0.0076 for b0 and b1 respectively), believed to be because of zero correlation between
M and D in the data used.
Also derive equation using log10 (PGA) = b0 + b1 M + b2 R + b3 log10 (R) (do not report
coefcients) and nd slightly smaller residuals but similar behaviour of the b parameters.
Plot distribution of residuals (binned into intervals of 0.25 units) and the normal probability density function.
2.192
Skarlatoudis et al. (2003)
log Y = c0 + c1 M + c2 log(R2 + h2 )1/2 + c3 F + c5 S

where Y is in cm/s2 , c0 = 0.86, c1 = 0.45, c2 = 1.27, c3 = 0.10, c5 = 0.06 and
= 0.286.
Use three site classes (from NEHRP):
S = 0 B: 19 stations plus 6 stations between A and B

S = 1 C: 68 stations
S = 2 D: 25 stations
No stations in NEHRP class A or E. Use geotechnical information where available and
geological maps for the other stations.
Focal depths, h, between 0.0 and 30.1 km.
Classify earthquakes into three faulting mechanism classes:
151
F = 0 Normal, 101 earthquakes

F = 1 Strike-slip, 89 earthquakes
F = 1 Thrust, 35 earthquakes
but only retain two categories: normal and strike-slip/thrust. Classify using plunges of
P and T axes and also knowledge of the geotectonic environment. Have fault-plane
solutions for 67 earthquakes.
Choose data that satises at least one of these criteria:
from earthquake with Mw 4.5;
record has PGA 0.05 g, independent of magnitude;
record has PGA < 0.05 g but at least one record from earthquake has PGA
0.05 g.
Relocate all earthquakes.
Redigitise all records using a standard procedure and bandpass lter using cut-offs chosen by a comparison of the Fourier amplitude spectrum (FAS) of the record to the FAS
of the digitised xed trace. Find that PGAs from uncorrected and ltered accelerograms
are almost identical.
Convert ML to Mw , for earthquakes with no Mw , using a locally derived linear equation
Most data from earthquakes with Mw < 6 and rhypo < 60 km.
Note correlation in data between Mw and rhypo .
Note lack of near-eld data (R < 20 km) for Mw > 6.0.
Plot estimated distance at which instruments would not be expected to trigger and nd
that all data lie within the acceptable distance range for mean trigger level and only
14 records fall outside the distance range for trigger level plus one . Try excluding
these records and nd no effect. Hence conclude that record truncation would not affect
results.
Use an optimization procedure based on the least-squares technique using singular
value decomposition because two-step methods always give less precise results than
one-step techniques. Adopted method allows the controlling of stability of optimization
and accurate determination and analysis of errors in solution. Also method expected
to overcome and quantify problems arising from correlation between magnitude and
distance.
Test assumption that site coefcient for site class D is twice that for C by deriving equations with two site terms: one for C and one for D. Find that the site coefcient for D is
roughly twice that of site coefcient for C.
Test effect of focal mechanism by including two coefcients to model difference between
normal, strike-slip and thrust motions. Find that the coefcients for difference between
strike-slip and normal and between thrust and normal are almost equal. Hence combine
strike-slip and thrust categories.
Try including quadratic M term but nd inadmissible (positive) value due to lack of data
from large magnitude events.
152
Also derive equations using this functional form: log Y = c0 + c1 M + c2 log(R + c4 ) +

c3 F + c5 S where c4 was constrained to 6 km from an earlier study due to problems in
deriving reliable values of c2 and c4 directly by regression.
Plot observed data scaled to Mw 6.5 against predictions and nd good t.
Find no systematic variations in residuals w.r.t. remaining variables.
Find reduction in w.r.t. earlier studies. Relate this to better locations and site classications.
2.193
Beauducel et al. (2004)
log(PGA) = aM + bR log(R) + c
where PGA is in g, a = 0.611377, b = 0.00584334, c = 3.216674 and = 0.5.
Do not include terms for site effects due to uncertainty of site classications (rock/soil).
Suggest multiplying predictions by 3 to estimate PGA at soil sites.
Derive model to better estimate macroseismic intensities rapidly after an earthquake.
Select data from 21/11/2004 to 28/12/2004, which mainly come from earthquakes in the
Les Saintes sequence but include some subduction events and crustal earthquakes in
other locations.
Data from 13 stations on Guadeloupe.
Vast majority of data from M < 4 and 20 < d < 100 km.
Remove constant offset from accelerations but do not lter.
Use resolved maximum because other denitions (e.g. larger) can underestimate PGA
by up to 30%.
Plot residuals against M and nd no trends. Observe some residuals of 1.5.
Apply model to other earthquakes from the region and nd good match to observations.
2.194
Beyaz (2004)
log PGA = a1 + a2 Mw2 + a3 log(r + a4 )

where PGA is in unknown unit (probably cm/s2 ), a1 = 2.581, a2 = 0.029, a3 = 1.305,
a4 = 7 and = 0.7128 .
Data from rock sites.
8
It is stated that common logarithms are used but this standard deviation is extremely high and hence it may
actually be in terms of natural logarithms.
153
2.195 Bragato (2004)

log10 (y) = a + (b + cm)m + (d + em) log10 ( r2 + h2 )

where y is in g, a = 0.46, b = 0.35, c = 0.07, d = 4.79, e = 0.60, h = 8.9 km and
= 0.33.
Investigates effect of nontriggering stations on derivation of empirical Ground-motion
model based on the assumption that the triggering level is known (or can be estimated
from data) but do not know which stations triggered (called left truncated data).
Develops mathematical theory and computational method (after trying various alternative methods) for truncated regression analysis (TRA) and randomly truncated regression analysis (RTRA) (where triggering level changes with time).
Tests developed methods on 1000 lognormally-distributed synthetic data points simulated using the equation of Ambraseys et al. (1996) for 4 Ms 7 and 1 df
100 km. A xed triggering threshold of 0.02 g is imposed. Regresses remaining 908
samples using TRA and RTRA. Finds a very similar equation using TRA but large differences for df > 20 km by using standard regression analysis (SRA) due to slower attenuation. Also apply TRA to randomly truncated synthetic data and nd a close match
to original curve, which is not found using SRA.
Applies method to 189 records from rock sites downloaded from ISESD with M >
4.5 (scale not specied) and
d < 80 km (scale not specied) using functional form:
log10 (y) = a + bm + c log10 ( r2 + h2 ). Uses these selection criteria to allow use of
simple functional form and to avoid complications due to crustal reections that reduce
attenuation. Discards the ve points with PGA < 0.01 g (assumed threshold of SMA-1s).
Applies TRA and SRA. Finds both M -scaling and distance attenuation are larger with
TRA than with SRA because TRA accounts for larger spread in original (not truncated)
data. Differences are relevant for M < 6 and d > 20 km.
Applies method to dataset including, in addition, non-rock records (456 in total). Finds
no differences between TRA and SRA results. Believes that this is due to lack of data
in range possibly affected by truncation (small M and large d). Finds similar results to
Applies method to NE Italian data from seven seismometric
and ten accelerometric dig
ital stations assuming: log10 (y) = a + bm + c log10 ( r2 + h2 ). Accelerometric stations
used usually trigger at 0.001 g. Seismometric stations used trigger based on ratio of
short-term and long-term averages (STA/LTA), which varies from station to station and
acts like a random threshold. Firstly neglects randomness and assumes trigger level
of each station equals lowest recorded PGA and applies TRA and SRA. Finds small
differences for d < 8 km and d > 30 km.
Applies method using functional form above, which believes is more physically justied.
SRA does not converge. Studies reason for this by regressing on data from M intervals
of 0.3 units wide. Finds behaviour of PGAs inverts for M < 3. Finds increasing
with decreasing M for M > 3. TRA does converge and shows stronger magnitude
saturation than SRA.
154
Notes that application of RTRA to model effect of STA/LTA for used data is not realistic
since probably not enough data to constrain all 23 parameters and to computational
expensive using adopted maximization technique for RTRA.
Estimates the random truncation parameters for one station (Zoufplan) and nds that
the xed threshold assumption made is acceptable since estimated random truncation
parameters predict that only 14% of observations are lost at the earlier assumed xed
threshold level (the lowest PGA recorded).
2.196
Gupta & Gupta (2004)
ln PGA = C1 + C2 M + C3 ln Rh + C4 Rh + C5 v
where PGA is in g, C1 = 7.515, C2 = 1.049, C3 = 0.105, C4 = 0.0211, C5 =
0.287 and = 0.511. v = 0 for horizontal PGA and 1 for vertical PGA.
Data from basalt sites (7 stations), thick hard lateritic soil underlain by basalt (1 station)
and dam galleries (4 stations).
Data from 13-station strong-motion network (AR-240 and RFT-250 instrument types)
close to Koyna Dam. Exclude data from dam top. Use data from foundation gallery
because believe they can be considered as ground acceleration data. Select set of 31
signicant records after scrutinizing all data.
Correct for instrument response and lter using cut-off frequencies based on a signalto-noise ratio > 1.
Use a 2-stage regression method. Firstly, nd C1 , C2 and C5 (magnitude and component dependencies) and then nd updated C1 , C3 and C4 (distance dependence) using
residuals from rst stage.
Find that equation matches the observed data quite well.
2.197
Kalkan & Glkan (2004a)
ln YV
= C1 + C2 (M 6) + C3 (M 6)2 + C4 (M 6)3 + C5 ln r + C6 1 + C7 2
2
r = (rcl
+ h2 )1/2
where Y is in g, C1 = 0.055, C2 = 0.387, C3 = 0.006, C4 = 0.041, C5 = 0.944,

C6 = 0.277, C7 = 0.030, h = 7.72 km, rock = 0.629, soil = 0.607 and softsoil = 0.575.
Use three site classes:
1 = 0, 2 = 0 Rock: average Vs = 700 m/s, 27 records

1 = 1, 2 = 0 Soil: average Vs = 400 m/s, 26 records
1 = 0, 2 = 1 Soft soil: average Vs = 200 m/s, 47 records
155
Classify using approximate methods due to lack of available information. Note that correspondence between average Vs values for each site class and more widely accepted
soil categories is tenuous.
Focal depths from 0 to 111.0 km. State that all earthquakes were shallow crustal events.
Only 4 records come from earthquakes with reported focal depths > 33 km.
Expand with data from after 1999 and update database of Glkan & Kalkan (2002).
Faulting mechanism distribution is: normal (12 earthquakes, 14 records), strike-slip (33
earthquakes, 81 records) and reverse (2 earthquakes, 5 records). Note that poor distribution w.r.t. mechanism does not allow its effect to be modelled.
Use only records from earthquakes with Mw 4.5 to emphasize motions having greatest engineering interest and to include only more reliably recorded events. Include data
from one Mw 4.2 earthquake because of high vertical acceleration (31 mg) recorded.
Data reasonably well distribution w.r.t. M and d for d < 100 km.
Data mainly recorded in small and medium-sized buildings 3 storeys. Note that these
buildings modify recorded motions and this is an unavoidable uncertainty of the study.
Data from main shocks. Exclude data from aftershocks, in particular that from the 1999
Kocaeli and Dzce aftershocks because these records are from free-eld stations, which
do not want to commingle with non-free-eld data.
Exclude a few records for which PGA caused by main shock is < 10 mg. Exclude data
from aftershocks from the same stations.
Note that data used is of varying quality and could be affected by errors.
Include cubic term for M dependence to compensate for the controversial effects of
sparsity of Turkish data. Find that it gives a better t.
Use two-step method of Ambraseys et al. (1996) to nd site coefcients C6 and C7
after exploratory analysis to nd regression method that gives the best estimates and
the lowest .
State equations can be used for 4.5 Mw 7.4 and df 200 km.
Find no signicant trends in residuals w.r.t. M or d for all data and for each site category
except for a few high residuals for soil and soft soil records at df > 100 km.
Compute individual s for each site class.
Find that observed ground motions for the Kocaeli earthquake are well predicted.
2.198 Kalkan & Glkan (2004b) and Kalkan & Glkan (2005)
ln Y
= b1 + b2 (M 6) + b3 (M 6)2 + b5 ln r + bV ln(VS /VA )
2
r = (rcl
+ h2 )1/2
where Y is in g, b1 = 0.393, b2 = 0.576, b3 = 0.107, b5 = 0.899, bV = 0.200,

VA = 1112 m/s, h = 6.91 km and = 0.612.
156

Rock Average Vs = 700 m/s, 23 records
Soil Average Vs = 400 m/s, 41 records
Soft soil Average Vs = 200 m/s, 48 records
Use Vs measurements where available (10 stations, 22 records) but mainly classify
using approximate methods. Note that correspondence between average Vs values for
each site class and more widely accepted soil categories is tenuous.
Focal depths from 0 to 111.0 km. State that all earthquakes were shallow crustal events.
Only 4 records come from earthquakes with reported focal depths > 33 km.
Expand with data from after 1999 and update database of Glkan & Kalkan (2002).
Faulting mechanism distribution is: normal (12 earthquakes, 14 records), strike-slip (34
earthquakes, 82 records), reverse (2 earthquakes, 5 records), unknown (9 earthquakes,
11 records). Note that poor distribution w.r.t. mechanism does not allow its effect to be
modelled.
Use only records from earthquakes with Mw 4.0 to include only more reliably recorded
events.
Data reasonably well distribution w.r.t. M and d for d < 100 km.
Data from main shocks. Exclude data from aftershocks, in particular that from the 1999
Kocaeli and Dzce aftershocks because of high nonlinear soil behaviour observed during the mainshocks near the recording stations.
Data mainly recorded in small and medium-sized buildings 3 storeys. Note that these
buildings modify recorded motions and this is an unavoidable uncertainty of the study.
State equations can be used for 4.0 Mw 7.5 and df 250 km.
Find no signicant trends in residuals w.r.t. M or d for all data and for each site category.
Find that observed ground motions for the Kocaeli earthquake are well predicted.
2.199
Lubkowski et al. (2004)
Ground-motion model is not reported. Use six functional forms.

Very soft soil Vs,30 < 180 m/s. 0 records.
Soft soil 180 Vs,30 < 360 m/s. 1 record.
Stiff soil 360 Vs,30 < 750 m/s. 34 records.
Rock Vs,30 750 m/s. 93 records.
Site conditions are unknown for 35 records. Classify mainly using description of local
site conditions owing to unavailability of Vs measurements.
157
Exclude data from Mw < 3.0 to exclude data from earthquakes that are likely to be associated with large uncertainties in their size and location and because ground motions
from smaller earthquakes are likely to be of no engineering signicance.
Exclude data from multi-storey buildings, on or in dams or on bridges.
Most data from Mw < 5.5 so believe use of repi is justied.
Records from: eastern N America (78 records), NW Europe (61 including 6 from UK)
and Australia (24).
Locations from special studies, ISC/NEIC or local network determinations.
Note distinct lack of data from < 10 km for Mw > 5.
Only retain good quality strong-motion data. No instrument correction applied because
of the lack of instrument characteristics for some records. Individually bandpass lter
each record with a Butterworth lter with cut-offs at 25 Hz and cut-off frequencies chosen
by examination of signal-to-noise ratio and integrated velocity and displacement traces.
Find use of different functional forms has signicant inuence on predicted PGA.
Regression on only rock data generally reduced PGA.
Predictions using the functional forms with quadratic M -dependence were unreliable
for Mw > 5.5 because they predict decrease PGA with increasing M since there was
insufcient data from large magnitude earthquakes to constrain the predictions.
Find different regression methods predict similar PGAs with differences of < 5% for a
Mw 5 event at 5 km when all records were used but differences up to 63% when using
only rock data. Prefer the one-stage maximum-likelihood method since allows for correlation between M and d in dataset and does not ignore earthquakes recorded by only a
single station (25% of data).
Find, from analysis of residuals, that equation generally underpredicts PGA of data from
eastern N America and Australia but overpredicts motions from Europe and UK.
Find no trends in residuals w.r.t. amplitude, distance, magnitude or fault mechanism.
Believe that large s found are due to: lack of data from close to large magnitude earthquakes, use of data from different regions with varying source and path characteristics
and use of much data from small earthquakes that are probably associated with higher
uncertainty w.r.t. magnitude and location since such earthquakes have not been as well
studied as large earthquakes and there is a lack of data with high signal-to-noise ratio
from which relocations can be made.
Do not recommend equations for practical use due to large uncertainties.
2.200 Marin et al. (2004)

log10 PGA = a1 + a2 ML + a3 log10 R

where PGA is in g, a1 = 3.93, a2 = 0.78, a3 = 1.5 and = 0.55.
158
All records from stiff bedrock. Shear-wave velocities estimated from geology gives:
12002000 m/s for carbonated formations and > 2500 m/s for eruptive formations (majority of data).
Derive equation since nd previous equations are not consistent with recent data recorded
in France and because of differences between ML of LDG and other ML scales.
Use data from the Alps, the Pyrenees and Armorican Massif recorded by LDG network
of vertical seismometers between 1995 and 1996. Convert vertical PGAs to horizontal
PGAs using empirical relation of Smit (1998).
11 records from 3 de 50 km, 34 from 50 < de 200 km and 18 from de > 200 km
(all from two largest earthquakes with ML 5.3 and ML 5.6).
Plot predictions and data from rock sites of all French earthquakes with ML 4 recorded
by RAP network (largest three earthquakes have ML 5.5, ML 5.7 and ML 5.9) and nd
good agreement. State that this agreement shows that equation can be extrapolated to
strongest earthquakes considered for France.
Note that it will be possible to establish a more robust equation using increasing number
of data from RAP, especially from near eld and large magnitudes.
2.201
Midorikawa & Ohtake (2004)
log A = b log(X + c) kX
for D 30 km
log A = b + 0.6 log(1.7D + c) 1.6 log(X + c) kX
for D > 30 km
where b = aMw + hD + di Si + e
where A is in gal, a = 0.59, c = 0.0060 100.5Mw (adopted from Si & Midorikawa
(2000)), d1 = 0.00 (for crustal earthquakes), d2 = 0.08 (for inter-plate earthquakes),
d3 = 0.30 (for intra-plate earthquakes), e = 0.02, h = 0.0023, k = 0.003 [adopted from
Si & Midorikawa (2000)], intraevent = 0.27 and interevent = 0.16.
Use two site categories [denitions of Joyner & Boore (1981)]:
Rock
Soil
Use Vs,30 where available. Multiply PGA values from rock sites by 1.4 to normalise them
w.r.t. PGA at soil sites.
All records from the free-eld or small buildings where soil-structure interaction is negligible.
Data from different types of instruments hence instrument correct and bandpass lter.
Classify earthquakes into these three types:
S1 = 1, S2 = S3 = 0 Crustal. 12 earthquakes, 1255 records. Focal depths, D, between 3 and 30 km.

159
S2 = 1, S1 = S3 = 0 Inter-plate. 10 earthquakes, 640 records. 6 D 49 km.

S3 = 1, S1 = S2 = 0 Intra-plate, 11 earthquakes, 1440 records. 30 D 120 km.
Most data from Mw < 7. No data between 6.9 and 7.6.
Use separate functional forms for D 30 km and D > 30 km because of signicantly
faster decay for deeper earthquakes.
Plot histograms of residuals and conclude that they are lognormally distributed.
Compute for 4 M ranges: 5.55.9, 6.06.5, 6.66.9 and 7.68.3. Find slight decrease
in w.r.t. M .
Compute for ranges of 20 km. Find signicantly smaller s for distances < 50 km and
almost constant s for longer distances.
Compute for ranges of PGA of roughly 50 km. Find much larger s for small PGA than
for large PGA.
Believe that main cause of M -dependent is that stress-drop is M -dependent and that
radiation pattern and directivity are not likely to be signicant causes.
Believe that distance-dependent is likely to be due to randomness of propagation path
(velocity and Q-structure).
Believe site effects do not contribute greatly to the variance.
Plot PGA versus distance and observe a saturation at several hundred cm/s2 , which
suggest may be due to nonlinear soil behaviour.
Plot w.r.t. PGA for three site categories: 100 Vs,30 300 m/s, 300 Vs,30
600 m/s and 600 Vs,30 2600 m/s. Find lower for soft soils than for stiff soils, which
believe may demonstrate that nonlinear soil response is a cause of PGA-dependent .
Note that because inter-event is signicantly smaller than intra-event , source effects
are unlikely to be the main cause for observed dependencies.
2.202 zbey et al. (2004)

log(Y ) = a + b(M 6) + c(M 6)2 + d log
R2 + h2 + eG1 + f G2
where Y is in cm/s2 , a = 3.287, b = 0.503, c = 0.079, d = 1.1177, e = 0.141,

f = 0.331, h = 14.82 km and = 0.260.
G1 = 0, G2 = 0 A: shear-wave velocity > 750 m/s, 4 records, and B: shear-wave velocity 360
750 m/s, 20 records.
G1 = 1, G2 = 0 C: shear-wave velocity 180360 m/s, 35 records.
G1 = 0, G2 = 1 D: shear-wave velocity < 180 m/s, 136 records.
Originally A and B were separate but combine due to lack of data for site class A.
160

Use Mw for M > 6 to avoid saturation effects.
Assume ML = Mw for M 6.
Select records from earthquakes with M 5.0.
Most (15 earthquakes, 146 records) data from earthquakes with M 5.8.
Only use data from the Earthquake Research Department of General Directorate of
Disaster Affairs from df 100 km.
Exclude record from Bolu because of possible instrument error.
Use mixed effects model to account for both inter-event and intra-event variability.
Find that the mixed effects model yields s lower than xed effects model.
Compare predictions with observed data from the Kocaeli and Dzce earthquakes and
nd reasonable t.
Plot coefcients and s against frequency and nd dependence on frequency.
Plot inter-event and intra-event residuals against distance and magnitude and nd not
systematic trends.
Find intra-event residuals are signicantly larger than inter-event residuals. Suggest that
this is because any individual events recordings used to develop model follow similar
trends with associated parameters.
Recommend that equations are only used for ground-motion estimation in NW Turkey.
2.203
Pankow & Pechmann (2004) and Pankow & Pechmann (2006)
log10 (Z) = b1 + b2 (M 6) + b3 (M 6)2 + b5 log10 D + b6

2
D = (rjb
+ h2 )1/2
where Z is in g, b1 = 0.237, b2 = 0.229, b3 = 0, b5 = 1.052, b6 = 0.174, h = 7.27 km

and log Z = 0.203 (see Spudich & Boore (2005) for correct value of 3 for use in
calculating for randomly-orientated component).
= 0 Rock: sites with soil depths of < 5 m.

= 1 Soil
Use data of Spudich et al. (1999).
Correct equations of Spudich et al. (1999) for 20% overprediction of motions for rock
sites, which was due either to underestimation of shear-wave velocities for rock sites
for extensional regimes (believed to be more likely) or an overestimation of shear-wave
velocities at soil sites. Correction based on adjusting b1 and b6 to eliminate bias in rock
estimates but leave soil estimates unchanged.
161
Verify that adjustment reduces bias in rock estimates.

Do not change log Z because changes to b1 and b6 have a negligible inuence on log Z
w.r.t. errors in determining log Z .
2.204 Sunuwar et al. (2004)

log Y (T ) = b1 (T ) + b2 (T )MJ b3 (T )D b4 (T ) log(R)

where Y (T ) is in cm/s2 , b1 (0) = 1.1064, b2 (0) = 0.2830, b3 (0) = 0.0076, b4 (0) =
0.6322 and = 0.303 for horizontal PGA and b1 (0) = 0.7134, b2 (0) = 0.3091, b3 (0) =
0.0069, b4 (0) = 0.7421 and = 0.301 for vertical PGA.
Records from 225 stations of K-Net network with 39.29 Vs,30 760.25 m/s (mean
Vs,30 = 330.80 m/s.
Select earthquakes that occurred within the region of the boundary of the OkhotskAmur plates (NE Japan bordering Sea of Japan) dened by its horizontal location and
vertically, to exclude earthquakes occurring in other plates or along other boundaries.
Focal depths, D, between 8 and 43 km with mean depth of 20.8 km.
Mean value of M is 4.72.
Mean repi is 84.67 km.
State that exclude records with PGA < 5 cm/s2 (although ranges of PGAs given include
records with PGA < 5 cm/s2 ).
Horizontal PGA range: 4.15411.56 cm/s2 . Vertical PGA range: 0.50163.11 cm/s2 .
Originally use this form: log Y (T ) = b1 (T ) + b2 (T )M b3 (T )D log(R) + b5 (T )R but
nd b5 (T ) > 0. Regress using the 379 records from sites with Vs,30 > 300 m/s and still
nd b5 (T ) > 0 but report results for investigating site effects.
Plot residuals w.r.t. rhypo and nd mean of residuals is zero but nd some high residuals.
Note that need to rene model to consider site effects.
2.205 Skarlatoudis et al. (2004)

log Y = c0 + c1 M + c2 log(R2 + h2 )1/2

where Y is in cm/s2 , c0 = 1.03, c1 = 0.32, c2 = 1.11, h = 7 km and = 0.34.
Classify stations into four NEHRP categories: A, B, C and D (through a site coefcient,
c4 ) but nd practically no effect so neglect.
Aim to investigate scaling of ground motions for small magnitude earthquakes.
162
Most earthquakes have normal mechanisms from aftershock sequences.

Records from permanent and temporary stations of ITSAK network. Many from EuroSeisTest array.
Records from ETNA, K2, SSA-1 and SSA-2 plus very few SMA-1 instruments.
Filter records based on a consideration of signal-to-noise ratio. For digital records use
these roll-off and cut-off frequencies based on magnitude (after studying frequency content of records and applying different bandpass lters): for 2 Mw < 3 fr = 0.95 Hz
and fc = 1.0 Hz, for 3 Mw < 4 fr = 0.65 Hz and fc = 0.7 Hz and for 4 Mw < 5
fr = 0.35 and fc = 0.4 Hz. Find that this method adequately removes the noise from
the accelerograms used.
Use source parameters computed from high-quality data from local networks. Note
that because focal parameters are from different institutes who use different location
techniques may mean data set is inhomogeneous.
Note that errors in phase picking in routine location procedures may lead to less accurate locations (especially focal depths) for small earthquakes as opposed to large
earthquakes due to indistinct rst arrivals.
To minimize effects of focal parameter uncertainties, x h as 7 km, which corresponds
to average focal depth in Greece and also within dataset used.
Exclude data from de > 40 km because only a few (3% of total) records exist for these
distances and also to exclude far-eld records that are not of interest.
Most records from de < 20 km and 2.5 Mw 4.5.
Also derive equations using this functional form: log Y = c0 + c1 M + c2 log(R + c3 )
where c3 was constrained to 6 km from an earlier study due to problems in deriving
reliable values of c2 and c3 directly by regression.
Use singular value decomposition for regression following Skarlatoudis et al. (2003).
Combined dataset with dataset of Skarlatoudis et al. (2003) and regress. Find significant number of data outside the 1 curves. Also plot average residual at each M
w.r.t. M and nd systematically underestimation of PGA for Mw 5. Conclude that this
shows the insufciency of a common relation to describe both datasets.
Find no trends in the residuals w.r.t. magnitude or distance.
Find that the predominant frequencies of PGAs are < 15 Hz so believe results not affected by low-pass ltering at 2527 Hz.
2.206
Ulusay et al. (2004)
PGA = a1 ea2 (a3 Mw Re +a4 SA +a5 SB )

where PGA is in gal, a1 = 2.18, a2 = 0.0218, a3 = 33.3, a4 = 7.8427, a5 = 18.9282
and = 86.4.
163
SA = 0, SB = 0 Rock, 55 records.
SA = 1, SB = 0 Soil, 94 records.
SA = 0, SB = 1 Soft soil, 72 records.
Classify by adopting those given by other authors, selecting the class reported by more
than one source.
Most data from instruments in small buildings.
Use records with PGA > 20 gal to avoid bias due to triggering.
PGAs of records between 20 and 806 gal.
Use records from earthquakes with Mw 4 because smaller earthquakes are generally
not of engineering signicance.
Derive linear conversion formulae (correlation coefcients > 0.9) to transform Ms (39),
mb (18), Md (10) and ML (6) to Mw (73 events in total).
Note that rupture surfaces have not been accurately dened for most events therefore
use repi .
Note that accurate focal depths are often difcult to obtain and different data sources
provide different estimates therefore do not use rhypo .
Use records from 5 km because of assumed average error in epicentral locations.
Use records from 100 km because this is the distance range where engineering signicant ground motions occur.
Most data from Mw 6 and de 50 km.
Do not consider faulting mechanism because focal mechanism solutions for most earthquakes not available.
Plot observed versus predicted PGA and nd that a few points fall above and below the
lines with slopes 1 : 0.5 and 1 : 2 but most are between these lines.
Note that to improve precision of equation site characterisation based on Vs measurements should be included. Also note that directivity, fault type and hanging wall effects
should be considered when sufcient data is available.
2.207 Ambraseys et al. (2005a)

log y = a1 + a2 Mw + (a3 + a4 Mw ) log d2 + a25 + a6 SS + a7 SA + a8 FN + a9 FT + a10 FO

where y is in m/s2 , a1 = 2.522, a2 = 0.142, a3 = 3.184, a4 = 0.314, a5 = 7.6,
a6 = 0.137, a7 = 0.050, a8 = 0.084, a9 = 0.062, a10 = 0.044, 1 = 0.6650.065Mw
(intra-event) and 2 = 0.222 0.022Mw (inter-event).
164
SS = 1, SA = 0 Soft soil (S), 180 < Vs,30 360 m/s. 143 records.
SS = 0, SA = 1 Stiff soil (A), 360 < Vs,30 750 m/s. 238 records.
SS = 0, SA = 0 Rock (R), Vs,30 > 750 m/s. 203 records.
Originally include a fourth category, very soft soil (Vs,30 180 m/s), but only included
11 records so combined with soft soil records. Note that measured Vs,30 only exist for
89 of 338 stations contributing 161 records so use descriptions of local site conditions to
classify stations. Exclude records from stations with unknown site conditions because
could not be handled by chosen regression method.
Use only data from Europe and Middle East because believe their databank is reasonably complete for moderate and large earthquakes that occurred in region. Also these
data have been carefully reviewed in previous studies. Finally based on a previous
study believe motions in California could be signicantly higher than those in Europe.
Note that including these data would increase the quantity of high-quality near-source
data available.
Combine data from all seismically active parts of Europe and the Middle East into a
common dataset because a previous study shows little evidence for regional differences
between ground motions in different regions of Europe.
Only use earthquakes with a M0 estimate for which to calculate Mw . Do not convert
magnitudes from other scales because this increases the uncertainty in the magnitude
estimates. Exclude records from earthquakes with Mw < 5 in order to have a good
distribution of records at all magnitudes. Note that this also excludes records from small
earthquakes that are unlikely to be of engineering signicance.
Use rjb because does not require a depth estimate, which can be associated with a
large error.
Exclude records from > 100 km because: excludes records likely to be of low engineering signicance, reduces possible bias due to non-triggering instruments, reduces
effect of differences in anelastic decay in different regions and it gives a reasonably uniform distribution w.r.t. magnitude and distance, which reduces likelihood of problems in
regression analysis.
Use only earthquakes with published focal mechanism in terms of trends and plunges
of T, B and P axes because estimating faulting type based on regional tectonics or to
be the same as the associated mainshock can lead to incorrect classication. Classify
earthquakes using method of Frohlich & Apperson (1992):
Thrust Plunge of T axis > 50 . 26 earthquakes, 91 records, FT = 1, FN = 0, FO = 0.
Normal Plunge of P axis > 60 . 38 earthquakes, 191 records, FT = 0, FN = 1, FO = 0.
Strike-slip Plunge of B axis > 60 . 37 earthquakes, 160 records, FT = 0, FN = 0, FO = 0.
Odd All other earthquakes. 34 earthquakes, 153 records, FT = 0, FN = 0, FO = 1.
Use this method because does not require knowledge of which plane is the main plane
and which the auxiliary.
165
Do not exclude records from ground oors or basements of large buildings because of
limited data.
Exclude records from instruments that triggered late and those that were poorly digitised.
Instrument correct records and then apply a low-pass lter with roll-off and cut-off frequencies of 23 and 25 Hz for records from analogue instruments and 50 and 100 Hz for
records from digital instruments. Select cut-off frequencies for high-pass bidirectional
Butterworth ltering based on estimated signal-to-noise ratio and also by examining displacement trace. For records from digital instruments use pre-event portion of records
as noise estimate. For those records from analogue instruments with an associated
digitised xed trace these were used to estimate the cut-offs. For records from analogue
instruments without a xed trace examine Fourier amplitude spectrum and choose the
cut-offs based on where the spectral amplitudes do not tend to zero at low frequencies.
Note that there is still some subjective in the process. Next choose a common cut-off
frequency for all three components. Use a few records from former Yugoslavia that were
only available in corrected form.
Only use records with three usable components in order that ground-motion estimates
are unbiased and that mutually consistent horizontal and vertical equations could be
derived.
Note lack of data from large (Mw > 6.5) earthquakes particularly from normal and
strike-slip earthquakes.
Data from: Italy (174 records), Turkey (128), Greece (112), Iceland (69), Albania (1), Algeria (3), Armenia (7), Bosnia & Herzegovina (4), Croatia (1), Cyprus (4), Georgia (14),
Iran (17), Israel (5), Macedonia (1), Portugal (4), Serbia & Montenegro (24), Slovenia
(15), Spain (6), Syria (5) and Uzbekistan (1).
Note that much strong-motion data could not be used due to lack of local site information.
Select one-stage maximum-likelihood regression method because accounts for correlation between ground motion from same earthquake whereas ordinary one-stage method
does not. Note that because there is little correlation between Mw and distance in the
data used (correlation coefcient of 0.23) ordinary one-stage and one-stage maximumlikelihood methods give similar coefcients. Do not use two-stage maximum-likelihood
method because underestimates for sets with many singly-recorded earthquakes (35
earthquakes were only recorded by one station). Do not use method that accounts for
correlation between records from same site because records are used from too many
different stations and consequently method is unlikely to lead to an accurate estimate of
the site-to-site variability (196 stations contribute a single record). Do not use methods
that account for uncertainty in magnitude determination because assume all magnitude
estimates are associated with the same uncertainty since all Mw are derived from published M0 values.
Apply pure error analysis of Douglas & Smit (2001). Divide dataspace into 0.2Mw units
by 2 km intervals and compute mean and unbiased standard deviation of untransformed
ground motion in each bin. Fit a linear equation to graphs of coefcient of variation
against ground motion and test if slope of line is signicantly different (at 5% signicance
level) than zero. If it is not then the logarithmic transformation is justied. Find that
slope of line is not signicantly different than zero so adopt logarithmic transformation of
ground motion.
166
Use pure error analysis to compute mean and unbiased standard deviation of logarithmically transformed ground motion in each 0.2Mw 2 km bin. Plot the standard deviations
against Mw and t linear equation. Test signicance (5% level) of slope. Find that it is
signicantly different than zero and hence magnitude-independent standard deviation
is not justied. Use the reciprocals of tted linear equations as weighting functions for
regression analysis.
Using the standard deviations computed by pure error analysis for each bin estimate
lowest possible for derived equations.
Investigate possible magnitude-dependence of decay rate of ground motions using ten
best-recorded earthquakes (total number of records between
13 and 26). Fit PGAs for
each earthquake with equation of form: log y = a1 + a2 log d2 + a23 . Plot decay rates
(a2 ) against Mw and t a linear equation. Find that the tted line has a signicant slope
and hence conclude that data supports a magnitude-dependent decay rate. Assume a
linear dependence between decay rate and Mw due to limited data.
Try including a quadratic magnitude term in order to model possible differences in scaling of ground motions for earthquakes that rupture entire seismogenic zone. Find that
term is not signicant at 5% level so drop.
Could not simultaneously nd negative geometric and anelastic decay coefcients so
assume decay attributable to anelastic decay is incorporated into geometric decay coefcient.
Test signicance of all coefcients at 5% level. Retain coefcients even if not signicant.
Note that there is not enough data to model possible distance dependence in effect of
faulting mechanism or nonlinear soil effects.
Compute median amplication factor (anti-logarithm of mean residual) for the 16 stations
that have recorded more than ve earthquakes. Find that some stations show large
amplications or large deamplications due to strong site effects.
Compute median amplication factor for the ten best recorded earthquakes. Find that
most earthquakes do not show signicant overall differences but that a few earthquakes
do display consistently lower or higher ground motions.
Plot residual plots w.r.t. weighted Mw and weighted distance and nd no obvious dependence of scatter on magnitude or distance.
Plot histograms of binned residuals.
Compare predicted and observed PGAs from the 2004 Parkeld earthquake and nd a
close match. Note that this may mean that the exclusion of data from California based
on possible differences in ground motions was not justied.
2.208
Ambraseys et al. (2005b)
log y = a1 + a2 Mw + (a3 + a4 Mw ) log

d2 + a25 + a6 SS + a7 SA + a8 FN + a9 FT + a10 FO
167

a6 = 0.078, a7 = 0.046, a8 = 0.126, a9 = 0.005, a10 = 0.082, 1 = 0.262 (intraevent) and 2 = 0.100 (inter-event).
Based on Ambraseys et al. (2005a). See Section 2.207.
2.209 Bragato (2005)

log10 (PGA) = c1 + c2 Ms + c3 r
where PGA is in m/s2 , c1 = 2.09, c2 = 0.47, c3 = 0.039 and = 0.3 (note that the
method given in the article must be followed in order to predict the correct accelerations
using this equation).
Uses data (186 records) of Ambraseys & Douglas (2000, 2003) for Ms 5.8. Add 57
records from ISESD (Ambraseys et al. , 2004) for 5.0 Ms 5.7.
Investigates whether magnitude-dependent attenuation, i.e. PGA saturation in response
to increasing magnitude, can be explained by PGA approaching an upper physical limit
through an accumulation of data points under an upper limit.
Proposes model with: a magnitude-independent attenuation model and a physical mechanism that prevents PGA from exceeding a given threshold. Considers a xed threshold
and a threshold with random characteristics.
Develops the mathematical models and regression techniques for the truncated and the
randomly clipped normal distribution.
Reduces number of parameters by not considering site conditions or rupture mechanism. Believes following results of Ambraseys & Douglas (2000, 2003) that neglecting
site effects is justied in the near-eld because they have little effect. Believes that the
distribution of data w.r.t. mechanism is too poor to consider mechanism.
Performs a standard one-stage, unweighted regression with adopted functional form
and also with form: log10 (PGA) = c1 + c2 M + c3 r + c4 M r + c5 M 2 + c6 r2 and nds
magnitude saturation and also decreasing standard deviation with magnitude.
Performs regression with the truncation model for a xed threshold with adopted functional form. Finds almost identical result to that from standard one-stage, unweighted
regression.
Performs regression with the random clipping model. Finds that it predicts magnitudedependent attenuation and decreasing standard deviation for increasing magnitude.
Investigates the effect of the removal of high-amplitude (PGA = 17.45 m/s2 ) record
from Tarzana of the 1994 Northridge earthquake. Finds that it has little effect.
168
2.210
Bragato & Slejko (2005)
log10 (Y ) = a + (b + cM )M + (d + eM 3 ) log10 (r)
r =
d2 + h2
where Y is in g, a = 3.27, b = 1.95, c = 0.202, d = 3.11, e = 0.00751, h = 8.9 km
and = 0.399 for horizontal PGA and repi , a = 3.37, b = 1.93, c = 0.203, d =
3.02, e = 0.00744, h = 7.3 km and = 0.358 for horizontal PGA and rjb , a = 2.96,
b = 1.79, c = 0.184, d = 3.26, e = 0.00708, h = 11.3 km and = 0.354 for
vertical PGA and repi and a = 3.18, b = 1.80, c = 0.188, d = 3.13, e = 0.00706,
h = 9.1 km and = 0.313 for vertical PGA and rjb .
Believe relation valid for rather rigid soil.
Use data from the Seismometric Network of Friuli-Venezia Giulia (SENF) (converted
to acceleration), the Friuli Accelerometric Network (RAF), data from the 1976 Friuli sequence and data from temporary seismometric (converted to acceleration) and accelerometric stations of Uprava RS za Geoziko (URSG) of the 1998 Bovec sequence.
Data from 1976 Friuli sequence is taken from ISESD. Records have been bandpass
ltered with cut-offs of 0.25 and 25 Hz. No instrument correction has been applied. Data
from other networks has been instrument corrected and high-pass ltered at 0.4 Hz.
Hypocentral locations and ML values adopted from local bulletins and studies.
Use running vectorial composition of horizontal time series because horizontal vector is
the actual motion that intersects seismic hazard. Find that on average running vectorial
composition is 8% larger than the larger horizontal peak and 27% larger than the geometric mean. Find that using other methods to combine horizontal components simply
changes a by about 0.1 downwards and does not change the other coefcients.
Use data from 19 earthquakes with ML 4.5 (161 vertical records, 130 horizontal
records).
Note that distribution w.r.t. magnitude of earthquakes used roughly follows log-linear
Gutenberg-Richter distribution up to about ML 4.5.
Few records available for d < 10 km and ML > 3.
Focal depths between 1.0 and 21.6 km. Average depth is 11.4 3.6 km.
Apply multi-linear multi-threshold truncated regression analysis (TRA) of Bragato (2004)
to handle the effect of nontriggering stations using the simplication that for SENF and
URSG data the random truncation level can be approximated by the lowest value available in the data set for that station. For data from the 1976 Friuli sequence use a
unique truncation level equal to the minimum ground motion for that entire network in
the dataset. Use same technique for RAF data.
Develop separate equations for repi and rjb (available for 48 records in total including all
from ML > 5.8). Note that physically rjb is a better choice but that repi is more similar
to geometric distance used for seismic hazard assessment.
169
Use ML because available for regional earthquakes eastern Alps since 1972.
Conduct preliminary tests and nd that weak-motion data shows higher attenuation than
strong-motion data. Investigate horizontal PGA using entire data set and data for 0.5wide magnitude classes. Find that attenuation is dependent on magnitude and it is not
useful to include a coefcient to model anelastic attenuation.
Since data is not uniformly distributed with magnitude, inversely weight data by number
of records within intervals of 0.1 magnitude units wide.
Because correlation between magnitude and distance is very low (0.03 and 0.02 for
vertical and horizontal components, respectively) apply one-stage method.
Note that large differences between results for repi and rjb are due to magnitudedependent weighting scheme used.
Plot predicted and observed ground motions binned into 0.3 magnitude intervals and
nd close match.
Plot residuals w.r.t. focal depth, rjb and ML . Find that it appears equation over-estimates
horizontal PGA for df > 80 km, ML < 3 and focal depths > 15 km but note that this is
due to the truncation of low amplitude data. Check apparent trend using TRA and nd
no signicant trend.
Note that difcult to investigate importance of focal depth on attenuation due to unreliability of depths particularly for small earthquakes. Find that focal depths seem to be
correlated to magnitude but believe that this is an artifact due to poor location of small
earthquakes. Try regression using rhypo and nd larger hence conclude that depth
estimates are not accurate enough to investigate effect of depth on ground motions.
Investigate methods for incorporation of site effect information using their ability to reduce as a criteria.
Note that largest possible reduction is obtained using individual average station residuals
for each site but that this is not practical because this method cannot be used to predict
ground motions at arbitrary site and that it requires sufcient number of observations
for each station. Using just those stations that recorded at least ve earthquakes obtain
estimate of lowest possible by adopting this method.
Try using a classication of stations into three site categories: rock (16 stations, 1020
records), stiff soil (9 stations, 117 records) and soft soil (4 stations, 27 records) and
nd no reduction in , which believe is due to the uneven distribution w.r.t. site class.
Find that the strong site effects at Tolmezzo has a signicant effect on the obtained site
coefcients.
Use Nakamura (H/V) ratios from ambient noise for a selection of stations by including
a term g(S) = cHV N (S), where N (S) is the Nakamura ratio at the period of interest
(0.1251 s for PGA), in the equation. Find large reductions in and high correlations
between Nakamura ratios and station residuals.
Use receiver functions from earthquake recordings in a similar way to Nakamura ratios.
Find that it is reduces more than site classication technique but less than using
the Nakamura ratios, which note could be because the geometry of the source affects
the computed receiver functions so that they are not representative of the average site
effects.
170
Believe equation is more appropriate than previous equations for ML < 5.8 and equivalent to the others up to ML 6.3. Discourage extrapolation for ML > 6.3 because it
overestimates PGA in the far-eld from about ML 6.5.
2.211
Frisenda et al. (2005)
log(Y ) = a + bM + cM 2 + d log(R) + eS
where Y is in g, a = 3.190.02, b = 0.870.01, c = 0.0420.002, d = 1.920.01,
e = 0.249 0.005 and = 0.316.
Use two site classes, because lack local geological information (e.g. average Vs ):
S = 0 Rock, eight stations, 3790 records.

S = 1 Soil, seven stations, 3109 records.
Classify station using geological reports, ML station corrections and H/V spectral ratios
computed over a 30 s wide time window of S waves for entire waveform data set.
Data from Regional Seismic Network of Northwestern Italy and Regional Seismic Network of Lunigiana-Garfagnana (ten Lennartz LE3D-5s and ve Guralp CMG-40 sensors
with Lennartz Mars88/MC recording systems). Sampling rate either 62.5 or 125 samples/s.
Records from broadband and enlarged band seismometers converted to acceleration
by: correcting for instrument response, bandpass ltering between 1 and 20 Hz and
then differentiating. Accuracy of conversion veried by comparing observed and derived PGA values at one station (STV2), which was equipped with both a Kinemetrics
K2 accelerometer and a Guralp CMG-40 broadband sensor.
Find strong attenuation for short distances (< 50 km) and small magnitudes (ML < 3.0).
ML calculated using a calibration formula derived for northwestern Italy using a similar
dataset.
Compute signal-to-noise (S/N) ratio for the S phase using windows of 3 s wide and nd
that data is good quality (85% of windows have S/N ratio greater than 10 dB. Only use
records with S/N ratio > 20 dB.
Most earthquakes are from SW Alps and NW Apennines.
Most records from earthquakes with 1 ML 3, small number from larger earthquakes
particularly those with ML > 4. ML < 1: 1285 records, 1 ML < 2: 2902 records,
2 ML < 3: 1737 records, 3 ML < 4: 693 records and ML 4: 282 records.
Data shows strong magnitude-distance correlation, e.g. records from earthquakes with
ML < 1 are from 0 R 100 km and those from earthquakes with ML > 4 are mainly
from R > 50 km. Distribution is uniform for 2 ML 4 and 0 R 200 km.
Originally include an anelastic decay term (d1 R) in addition but the value of d1 was
positive and not statistically signicantly different than zero so it was removed.
Regression in two-steps: rstly without site effect coefcient (e) and then with e added.
171
Compare data to estimated decay within one magnitude unit intervals and nd predictions are good up to ML = 4.0.
Find no systematic trends in the residuals.
2.212 Garca et al. (2005)

= c1 + c2 Mw + c3 R c4 log R + c5 H
2 + 2
R =
Rcld
log Y
= 0.00750 100.507Mw
where Y is in cm/s2 , for horizontal PGA: c1 = 0.2, c2 = 0.59, c3 = 0.0039, c4 = 1,
c5 = 0.008, r = 0.27, e = 0.10 and for vertical PGA: c1 = 0.4, c2 = 0.60, c3 =
0.0036, c4 = 1, c5 = 0.006, r = 0.25 and e = 0.11 where r is the intra-event
standard deviation and e is the inter-event standard deviation.
All data from 51 hard (NEHRP B) sites.
All stations in the Valley of Mexico omitted.
All data from free-eld stations: small shelters, isolated from any building, dam abutment, bridge, or structure with more than one storey.
Focal depths: 35 H 138 km, most records (13 earthquakes, 249 records) from
35 H 75 km.
Exclude data from Mw < 5.0 and R > 400 km.
Exclude data from deep earthquakes where wave paths cross the mantle edge.
All data from normal-faulting earthquakes.
Use about 27 records from velocity records from broadband seismograph network that
were differentiated to acceleration.
Adopt from Atkinson & Boore (2003).
Investigate a number of functional forms. Inclusion of substantially improves t, leading to a decrease in random variability at close distances, and an increase in c2 and c3
coefcients. Find worse correlation when add a quadratic magnitude term. A magnitudedependent c4 leads to higher s. Find unrealistically high ground motions at close distances using the form of c4 used by Atkinson & Boore (2003).
If exclude three deep earthquakes then little dependence on H .
Do not nd any noticeable bias in residuals w.r.t. distance, magnitude or depth (not
shown).
Note that decrease in variability w.r.t. magnitude is only apparent for frequencies < 1 Hz.
Discuss observed dependence of, particularly high-frequency, ground motions on focal
depth.
172
2.213
Liu & Tsai (2005)
ln Y = a ln(X + h) + bX + cMw + d
where Y is in cm/s2 for horizontal PGA (for whole Taiwan) a = 0.852, b = 0.0071,
c = 1.027, d = 1.062, h = 1.24 km and = 0.719 and for vertical PGA (for whole
Taiwan) a = 1.340, b = 0.0036, c = 1.101, d = 1.697, h = 1.62 km and = 0.687.
Also report coefcients for equations derived for three different sub-regions.
Do not differentiate site conditions.
Focal depths, h, between 2.72 and 29.98 km.
Data from high-quality digital strong-motion networks of Taiwan Strong Motion Instrumentation Program (TSMIP) and Central Mountain Strong Motion Array (CMSMA).
Select data from earthquakes with h 30 km and with records from 6 stations at
de 20 km.
Select events following the 1999 Chi-Chi earthquake (Mw 7.7) with ML > 6.
Do not use data from the Chi-Chi earthquake because: a) earlier analysis of Chi-Chi
data showed short-period ground motion was signicantly lower than expected and b)
the Chi-Chi rupture triggered two M 6 events on other faults thereby contaminating the
ground motions recorded at some stations.
Data uniformly distributed for Mw 6.5 and 20 rhypo 100 km. Signicant number
of records for rhypo > 100 km.
Use data from the Chi-Chi earthquake and the 2003 Cheng-Kung earthquake (Mw 6.8)
for testing applicability of developed equations.
For 32 earthquakes (mainly with Mw < 5.3) convert ML to Mw using empirical equation
developed for Taiwan.
Develop regional equations for three regions: CHY in SW Taiwan (16 earthquakes, 1382
records), IWA in NE Taiwan (14 earthquakes, 2105 records) and NTO in central Taiwan
(13 earthquakes, 3671 records) and for whole Taiwan to compare regional differences
of source clustering in ground-motion characteristics.
Use Mw since corresponds to well-dened physical properties of the source, also it
can be related directly to slip rate on faults and avoids saturation problems of other
M -scales.
Use relocated focal depths and epicentral locations.
Do not use rjb or rrup because insufcient information on rupture geometries, particularly those of small earthquakes, even though believe such distance metrics are justied.
However, for small earthquakes do not think using rhypo rather than rrup will introduce
signicant bias into the equations. Also use rhypo because it is quickly determined after
an earthquake hence early ground-motion maps can be produced.
173
From equations derived for different sub-regions and from site residual contour maps
that ground motions in CHY are about four times higher than elsewhere due to thick,
recent alluvial deposits.
Find predictions for Chi-Chi and Cheng-Kung PGAs are close to observations.
Plot contour maps of residuals for different sites and relate the results to local geology
(alluvial plains and valleys and high-density schist).
Divide site residuals into three classes: > 0.2 , 0.20.2 and < 0.2 for four
NEHRP-like site classes. Find the distribution of residuals is related to the site class
particularly for the softest class. Find residuals for C (very dense soil and soft rock)
and D (stiff soil) are similar so suggest combining them. Believe geomorphology may
also play an important role in site classication because a geomorphologic unit is often
closely related to a geologic unit.
2.214 McGarr & Fletcher (2005)

log(y) = a + bM + d log(R) + kR + s1 + s2
where y is in cm/s2 , a = 0.9892, b = 0.8824, d = 1.355, k = 0.1363, s1 = 0.337
(for stations on surface), s2 = 0 (for station at depth) and = 0.483.
Use data from seven stations, one of which (TU1) is located underground within the
mine. Determine site factors (constrained to be between 0 and 1) from PGV data. Originally group into three site categories: one for stations with close to horizontal straightline ray paths, one for stations with steeper ray paths and one for underground station.
Find site factors for rst two categories similar so combine, partly because there is no
precedent for topographic site factors in empirical ground-motion estimation equations.
Believe that low site factors found are because stations are on solid rock Vs > 1.5 km/s.
Most data from Trail Mountain coal mine from between 12/2000 and 03/2001 (maximum
MCL 2.17). Supplement with data (2 records) from a M 4.2 earthquake at Willow Creak
mine to provide data at much higher magnitude.
Most data from Mw < 1.7.
Lower magnitude limit dictated by need for adequate signal-to-noise ratio.
Focal depths between 50 and 720 m (relative to the ground surface).
Note that although data may be poorly suited to determine both d and k simultaneously
they are retained because both attenuation mechanisms must be operative. State that
d and k should be solely considered as empirical parameters due to trade-offs during
tting.
Do not include a quadratic M term because it is generally of little consequence.
Use rhypo because earthquakes are small compared to distances so can be considered
as point sources.
Selected events using these criteria:
174
event was recorded by 6 stations;

data had high signal-to-noise ratio;
to obtain the broadest M -range as possible; and
to have a broad distribution of epicentral locations.
Find that Mw (estimated for 6 events) does not signicantly differ from MCL .
Find that constrains must be applied to coefcients. Constrain k to range 20 because
otherwise nd small positive values. Believe that this is because data inadequate for
independently determining d and k .
2.215
Nowroozi (2005)
ln(A) = c1 + c2 (M 6) + c3 ln( EPD2 + h2 ) + c4 S
where A is in cm/s2 , c1 = 7.969, c2 = 1.220, c3 = 1.131, c4 = 0.212, h = 10 km (xed

after tests) and = 0.825 for horizontal PGA and c1 = 7.262, c2 = 1.214, c3 = 1.0949 ,
c4 = 0.103, h = 10 km (xed after tests) and = 0.773 for vertical PGA.
Uses four site categories (S equals number of site category):
1. Rock. 117 records.
2. Alluvial. 52 records.
3. Gravel and sandy. 70 records.
4. Soft. 39 records.
Does analysis combining 1 and 2 together in a rm rock category (S = 0) and 3 and 4
in a soft soil category (S = 1) and for all site categories combined. Reports coefcients
for these two tests.
Focal depths between 9 and 73 km. Most depths are shallow (depths xed at 33 km)
and majority are about 10 km. Does not use depth as independent parameter due to
uncertainties in depths.
Uses Mw because nearly all reported Ground-motion models use Mw .
Uses macroseismic distance for three events since no repi reported.
Believes that methods other than vectorial sum of both horizontal PGAs underestimates
true PGA that acts on the structure. Notes that vectorial sum ideally requires that PGAs
on the two components arrive at the same time but due to unknown or inaccurate timing
the occurrence time cannot be used to compute the resolved component.
Does not consider faulting mechanism due to lack of information for many events.
Most records from Mw 5.
Originally includes terms c5 (M 6)2 and c6 EPD but nds them statistically insignicant
so drops them.
9
There is a typographical error in Equation 12 of Nowroozi (2005) since this coefcient is reported as 1094.
175
Notes that all coefcients pass the t-test of signicance but that the site coefcients are
not highly signicant, which relates to poor site classication for some stations.
Compares observed and predicted PGAs with respect to distance. Notes that match to
observations is relatively good.
Compares observed PGAs during Bam 2003 earthquake to those predicted and nds
good match.
2.216 Ruiz & Saragoni (2005)

x=
AeBM
(R + C)D
where x is in cm/s2 , A = 4, B = 1.3, C = 30 and D = 1.43 for horizontal PGA,

hard rock sites and thrust earthquakes; A = 2, B = 1.28, C = 30 and D = 1.09 for
horizontal PGA, rock and hard soil sites and thrust earthquakes; A = 11, B = 1.11,
C = 30, D = 1.41 for vertical PGA, hard rock sites and thrust earthquakes; A = 18,
B = 1.31, C = 30, D = 1.65 for vertical PGA, rock and hard soil sites and thrust
earthquakes; A = 3840, B = 1.2, C = 80 and D = 2.16 for horizontal PGA, rock
and hard soil sites and intermediate-depth earthquakes; and A = 66687596, B = 1.2,
C = 80 and D = 4.09 for vertical PGA, rock and hard soil sites and intermediate-depth
earthquakes.
Hard rock Vs > 1500 m/s. 8 records.
Rock and hard soil 360 < Vs < 1500 m/s. 41 records.
Develop separate equations for interface and intraslab (intermediate-depth) events.
Baseline correct and bandpass lter (fourth-order Butterworth) with cut-offs 0.167 and
25 Hz.
8 records from between Ms 6.0 and 7.0, 13 from between 7.0 and 7.5 and 20 from between 7.5 and 8.0.
Values of coefcient D taken from previous studies.
2.217 Takahashi et al. (2005), Zhao et al. (2006) and Fukushima

et al. (2006)
loge (y) = aMw + bx loge (r) + e(h hc )h + FR + SI + SS + SSL loge (x) + Ck

where r = x + c exp(dMw )
where y is in cm/s2 , h = 1 when h hc and 0 otherwise, a = 1.101, b = 0.00564,
c = 0.0055, d = 1.080, e = 0.01412, SR = 0.251, SI = 0.000, SS = 2.607, SSL =
176
0.528, CH = 0.293, C1 = 1.111, C2 = 1.344, C3 = 1.355, C4 = 1.420, = 0.604

(intra-event) and = 0.398 (inter-event). Use hc = 15 km because best depth effect for
shallow events.
Use ve site classes (T is natural period of site):
Hard rock NEHRP site class A, Vs,30 > 1100 m/s. 93 records. Use CH .
SC I Rock, NEHRP site classes A+B, 600 < Vs,30 1100 m/s, T < 0.2 s. 1494 records.
Use C1 .
SC II Hard soil, NEHRP site class C, 300 < Vs,30 600 m/s, 0.2 T < 0.4 s. 1551
records. Use C2 .
SC III Medium soil, NEHRP site class D, 200 < Vs,30 300 m/s, 0.4 T < 0.6 s. 629
records. Use C3 .
SC IV Soft soil, NEHRP site classes E+F, Vs,30 200 m/s, T 0.6 s. 989 records. Use
C4 .
Site class unknown for 63 records.
Focal depths, h, between about 0 and 25 km for crustal events, between about 10 and
50 km for interface events, and about 15 and 162 km for intraslab events. For earthquakes with h > 125 km use h = 125 km.
Classify events into three source types:
1. Crustal.
2. Interface. Use SI .
3. Slab. Use SS and SSL .
and into four mechanisms using rake angle of 45 as limit between dip-slip and strikeslip earthquakes except for a few events where bounds slightly modied:
1. Reverse. Use FR if also crustal event.
2. Strike-slip
3. Normal
4. Unknown
Distribution of records by source type, faulting mechanism and region is given in following table.
Region
Japan
Iran and Western USA
All
Focal Mechanism
Reverse
Strike-slip
Normal
Unknown
Total
Reverse
Strike-slip
Total
Total
Crustal
250
1011
24
Interface
1492
13
3
1285
123
73
196
1481
1508
12
12
1520
Slab
408
574
735
8
1725
1725
Total
2150
1598
762
8
4518
135
73
208
4726
177
Exclude data from distances larger than a magnitude-dependent distance (300 km for
intraslab events) to eliminate bias introduced by untriggered instruments.
Only few records from < 30 km and all from < 10 km from 1995 Kobe and 2000 Tottori
earthquake. Therefore add records from overseas from < 40 km to constrain nearsource behaviour. Note that could affect inter-event error but since only 20 earthquakes
(out of 269 in total) added effect likely to be small.
Do not include records from Mexico and Chile because Mexico is characterised as a
weak coupling zone and Chile is characterised as a strong coupling zone (the two
extremes of subduction zone characteristics), which could be very different than those
in Japan.
Note reasonably good distribution w.r.t. magnitude and depth.
State that small number of records from normal faulting events does not warrant them
between considered as a separate group.
Note that number of records from each event varies greatly.
Process all Japanese records in a consistent manner. First correct for instrument response. Next low-pass lter with cut-offs at 24.5 Hz for 50 samples-per-second data
and 33 Hz for 100 samples-per-second data. Find that this step does not noticeably affect short period motions. Next determine location of other end of usable period range.
Note that this is difcult due to lack of estimates of recording noise. Use the following
procedure to select cut-off:
1. Visually inspect acceleration time-histories to detect faulty recordings, S-wave triggers or multiple events.
2. If record has relatively large values at beginning (P wave) and end of record, the
record was mirrored and tapered for 5 s at each end.
3. Append 5 s of zeros at both ends and calculate displacement time-history in frequency domain.
4. Compare displacement amplitude within padded zeros to peak displacement within
the record. If displacement in padded zeros was relatively large, apply a high-pass
lter.
5. Repeat using high-pass lters with increasing corner frequencies, fc , until the displacement within padded zeros was small (subjective judgement). Use 1/fc found
as maximum usable period.
Verify method by using K-Net data that contains 10 s pre-event portions.
Conduct extensive analysis on inter- and intra-event residuals. Find predictions are
reasonably unbiased w.r.t. magnitude and distance for crustal and interface events and
not seriously biased for slab events.
Do not smooth coefcients.
Do not impose constraints on coefcients. Check whether coefcient is statistically signicant.
178
Note that the assumption of the same anelastic attenuation coefcient for all types and
depths of earthquakes could lead to variation in the anelastic attenuation rate in a manner that is not consistent with physical understanding of anelastic attenuation.
Derive CH using intra-event residuals for hard rock sites.
Residual analyses show that assumption of the same magnitude scaling and nearsource characteristics for all source types is reasonable and that residuals not not have
a large linear trend w.r.t. magnitude. Find that introducing a magnitude-squared term
reveals different magnitude scaling for different source types and a sizable reduction
in inter-event error. Note that near-source behaviour mainly controlled by crustal data.
Derive correction function from inter-event residuals of each earthquake source type
separately to avoid trade-offs. Form of correction is: loge (SM Sst ) = Pst (Mw MC ) +
Qst (Mw MC )2 + Wst . Derive using following three-step process:
1. Fit inter-event residuals for earthquake type to a quadratic function of Mw MC
for all periods.
2. Fit coefcients Pst for (Mw MC ) and Qst for (Mw MC )2 (from step 1) where
subscript st denotes source types, to a function up to fourth oder of loge (T ) to get
smoothed coefcients.
3. Calculate mean values of differences between residuals and values of Pst (Mw
MC ) + Qst (Mw MC )2 for each earthquake, Wst , and t mean values Wst to a
function of loge (T ).
For PGA QC = WC = QI = WI = 0, C = 0.303, I = 0.308, PS = 0.1392,
QS = 0.1584, WS = 0.0529 and S = 0.321. Since magnitude-square term for crustal
and interface is not signicant at short periods when coefcient for magnitude-squared
term is positive, set all coefcients to zero. Find similar predicted motions if coefcients
for magnitude-squared terms derived simultaneously with other coefcients even though
the coefcients are different than those found using the adopted two-stage approach.
Compare predicted and observed motions normalized to Mw 7 and nd good match
for three source types and the different site conditions. Find model overpredicts some
near-source ground motions from SC III and SC IV that is believed to be due to nonlinear
effects.
2.218
Wald et al. (2005)
log10 (Y ) = B1 + B2 (M 6) B5 log10 R
2 + 62
where R =
Rjb
where Y is in cm/s2 , B1 = 4.037, B2 = 0.572, B5 = 1.757 and = 0.836.
2.219
Atkinson (2006)
= c0 + c1(M 5) + c2(M 5)2 + c3 log R + c4R + Si
R =
d2 + h2
log Y
179
where Y is in m/s2 , c0 = 2.007, c1 = 0.567, c2 = 0.0311, c3 = 1.472, c4 = 0.00000,

h = 5 km [from Boore et al. (1997)], (BJF) = 0.309, (emp amp) = 0.307 and
(NoSiteCorr) = 0.305. Individual station: with empirical-corrected amplitudes =
0.269 and with BJF-corrected amplitudes = 0.268.
Uses data from 21 TriNet stations with known Vs,30 values. 190 Vs,30 958 m/s.
Uses two approaches for site term Si . In rst method (denoted empirically-corrected
amplitudes, emp amp) uses empirical site amplication factors from previous study
of TriNet stations (for PGA uses site factor for PSA at 0.3 s because correction for PGA
is unavailable). In second method [denoted Boore-Joyner-Fumal (BJF)-corrected amplitudes, BJF] uses amplication factors based on Vs,30 from Boore et al. (1997) to
correct observations to reference (arbitrarily selected) Vs,30 = 760 m/s.
Uses only data with amplitudes > 0.01% g (100 times greater than resolution of data,
0.0001% g).
States that developed relations not intended for engineering applications due to lack of
data from large events and from short distances. Equations developed for investigation
of variability issues for which database limitations are not crucial.
Many records from Landers mainshock and aftershocks.
Uses standard linear regression since facilitates comparisons using regressions of different types of datasets, including single-station datasets.
Notes possible complications to functional form due to effects such as magnitude-dependent
shape are not important due to small source size of most events.
Truncates data at 300 km to get dataset that is well distributed in distance-amplitude
space.
Notes that small differences between s when no site correction is applied and when
site correction is applied could be due to complex site response in Los Angeles basin.
Fits trend-lines to residuals versus distance for each station and nds slope not signicantly different from zero at most stations except for Osito Audit (OSI) (lying in mountains
outside the geographical area dened by other stations), which has a signicant positive
trend.
Finds empirical-amplication factors give better estimate of average site response (average residuals per station closer to zero) than Vs,30 -based factors at short periods but
the reverse for long periods. Notes Vs,30 gives more stable site-response estimates, with
residuals for individual stations less than factor of 1.6 for most stations.
Finds standard deviations of station residuals not unusually large at sites with large
mean residual, indicating that average site response estimates could be improved.
Plots standard deviation of station residuals using Vs,30 -based factors and the average
of these weighted by number of observations per station. Compares with standard deviation from entire databank. Finds that generally standard deviations of station residuals
slightly lower (about 10%) than for entire databank.
Examines standard deviations of residuals averaged over 0.5-unit magnitude bins and
nds no apparent trend for M3.5 to M7.0 but notes lack of large magnitude data.
180
Restricts data by magnitude range (e.g. 4 M 6) and/or distance (e.g. 80 km)

and nd no reduction in standard deviation.
Finds no reduction in standard deviation using one component rather than both.
Performs separate analysis of residuals for Landers events (10 stations having 20
observations) recorded at > 100 km. Notes that due to similarity of source and path
effects for a station this should represent a minimum in single-station . Finds of
0.18 0.06.
2.220
Beyer & Bommer (2006)
Exact functional form of Ground-motion model is not given but note includes linear and
quadratic terms of magnitude and a geometric spreading term. Coefcients not given
but report ratios of using different denitions w.r.t. using geometric mean.
Distribution w.r.t. NEHRP site classes is:
A 8 records
B 37 records
C 358 records
D 534 records
E 11 records
Unspecied 1 record
Use data from Next Generation Attenuation (NGA) database.
Distribution w.r.t. mechanism is:
Strike-slip 333 records, 51 earthquakes
Normal 36 records, 12 earthquakes
Reverse 329 records, 21 earthquakes
Reverse-oblique 223 records, 9 earthquakes
Normal-oblique 25 records, 7 earthquakes
Undened 3 records, 3 earthquakes
Exclude records from Chi-Chi 1999 earthquake and its aftershocks to avoid bias due to
over-representation of these data (> 50% of 3551 records of NGA databank).
Exclude records with PGA (dened using geometric mean) < 0.05 g to focus on motions
of engineering signicance and to avoid problems with resolution of analogue records.
Exclude records with maximum usable period < 0.5 s.
Exclude records without hypocentral depth estimate since use depth in regression analysis.
Earthquakes contribute between 1 and 138 accelerograms.
Note data is from wide range of M , d, mechanism, site class and instrument type.
181
State aim was not to derive state-of-the-art Ground-motion models but to derive models
with the same data and regression method for different component denitions.
Assume ratios of s from different models fairly insensitive to assumptions made during
regression but that these assumptions affect values themselves.
Find ratios of s from using different denitions close to 1.
Note that results should be applied with caution to subduction and stable continental
regions since have not been checked against these data.
2.221 Bindi et al. (2006)

Ground-motion model is for repi :
log(y) = a + bM + c log
(R2 + h2 ) + e1 S1 + e2 S2 + e3 S3 + e4 S4
where y is in g, a = 2.487, b = 0.534, c = 1.280, h = 3.94, e1 = 0, e2 = 0.365, e3 =

0.065, e4 = 0.053, event = 0.117 and record = 0.241 (or alternatively station = 0.145
and record = 0.232). For rhypo :
log(y) = a + bM + c log Rh + e1 S1 + e2 S2 + e3 S3 + e4 S4
where y is in g, a = 2.500, b = 0.544, c = 1.284 and = 0.292 (do not report site
coefcients for rhypo ).
Use four site classes:
AC Lacustrine and alluvial deposits with thickness > 30 m (180 Vs,30 < 360 m/s).
Sites in largest lacustrine plains in Umbria region. S4 = 1 and others are zero.
BC Lacustrine and alluvial deposits with thickness 1030 m (180 Vs,30 < 360 m/s).
Sites in narrow alluvial plains or shallow basins. S3 = 1 and others are zero.
CE Shallow debris or colluvial deposits (310 m) overlaying rock (surface layer with
Vs < 360 m/s). Sites located on shallow colluvial covers or slope debris (maximum
depth 10 m) on gentle slopes. S2 = 1 and others are zero.
DA Rock (Vs,30 > 800 m/s). Sites on outcropping rock, or related morphologic features, such as rock crests and cliffs. S1 = 1 and others are zero.
Base classications on recently collected detailed site information from site investigations, census data, topographic maps, data from previous reports on depth of bedrock,
and data from public and private companies. Subscripts correspond to classication in
Eurocode 8.
Focal depths between 1.1 and 8.7 km except for one earthquake with depth 47.7 km.
Nearly all earthquakes have normal mechanism, with a few strike-slip earthquakes.
Select earthquakes with ML 4.0 and d < 100 km.
Use ML since available for all events.
Fault geometries only available for three events so use repi and rhypo rather than rjb .
Note that except for a few records differences between repi and rjb are small.
182
Correct for baseline and instrument response and lter analogue records to remove highand low-frequency noise by visually selecting a suitable frequency interval: average
range was 0.525 Hz. Filter digital records with bandpass of, on average, 0.340 Hz.
For ML < 5 no records from de > 50 km.
Use maximum-likelihood regression with event and record s and also one with station
and record s. Perform each regression twice: once including site coefcients and once
without to investigate reduction in s when site information is included.
Investigate difference in residuals for different stations when site coefcients are included or not. Find signicant reductions in residuals for some sites, particularly for
class CE .
Note that some stations seem to display site-specic amplications different than the
general trend of sites within one site class. For these sites the residuals increase when
site coefcients are introduced.
Find large negative residuals for records from the deep earthquake.
Find similar residuals for the four earthquakes not from the 19971998 Umbria-Marche
sequence.
183
2.222 Campbell & Bozorgnia (2006a) and Campbell & Bozorgnia

(2006b)
ln Y
= f1 (M ) + f2 (R) + f3 (F ) + f4 (HW) + f5 (S) + f6 (D)
M 5.5
c0 + c1 M
c0 + c1 M + c2 (M 5.5)
5.5 < M 6.5
f1 (M ) =
c0 + c1 M + c2 (M 5.5) + c3 (M 6.5) M > 6.5
2 + c2 )
f2 (R) = (c4 + c5 M ) ln( rrup
6
f3 (F ) = c7 FRV fF (H) + c8 FN
{
H H < 1 km
fF (H) =
1 H 1 km
f4 (HW) = c9 FRV fHW (M )fHW (H)
{
1
rjb = 0 km
fHW (R) =
1 (rjb /rrup ) rjb > 0 km
M 6.0
0
2(M 6.0) 6.0 < M < 6.5
fHW (M ) =
1
M 6.5
{
0
H 20 km
fHW (H) =
1 (H/20) H < 20 km
(
)
{ [
(
)n ]
}
c10 ln Vs30 + k2 ln PGAr + c Vs30
ln[PGA
+
c]
Vs30 < k1
r
k1
k1
(
)
f5 (S) =
(c10 + k2 n) ln Vs30
Vs30 k1
k1
D < 1 km
c11 (D 1)
0
1 D 3 km
f6 (D) =
c12 {k3 [0.0000454 exp(3.33D)] + k4 [0.472 exp(0.25D)]} D > 3 km

Do not report coefcients, only display predicted ground motions. H is the depth to top
of coseismic rupture in km, PGAr is the reference value of PGA on rock with Vs30 =
1100 m/s, D is depth to 2.5 km/s shear-wave velocity horizon (so-called sediment or
basin depth) in km.
Use Vs30 (average shear-wave velocity in top 30 m in m/s) to characterise site conditions.
Model developed as part of PEER Next Generation Attenuation (NGA) project.
State that model is not nal and articles should be considered as progress reports.
NGA database only includes records that represent free-eld conditions (i.e. records
from large buildings are excluded).
Include earthquake if: 1) it occurred within the shallow continental lithosphere, 2) it was
in a region considered to be tectonically active, 3) it had enough records to establish a
reasonable source term and 4) it had generally reliable source parameters.
Exclude records from earthquakes classied as poorly recorded dened by: M < 5.0
and N < 5, 5.0 M < 6.0 and N < 3 and 6.0 M < 7.0, rrup > 60 km and N < 2
184
where N is number of records. Include singly-recorded earthquakes with M 7.0 and

rrup 60 km because of importance in constraining near-source estimates.
Include records if: 1) it was from or near ground level, 2) it had negligible structural
interaction effects and 3) it had generally reliable site parameters.
Find two-step regression technique was much more stable than one-step method and
allows the independent evaluation and modelling of ground-motion scaling effects at
large magnitudes. Find random effects regression analysis gives very similar results to
two-step method.
Use classical data exploration techniques including analysis of residuals to develop functional forms. Develop forms using numerous iterations to capture observed trends. Select nal forms based on: 1) their simplicity, although not an overriding factor, 2) their
seismological bases, 3) their unbiased residuals and 4) their ability to be extrapolated
to parameter values important for engineering applications (especially probabilistic seismic hazard analysis). Find that data did not always allow fully empirical development
of functional form therefore apply theoretical constraints [coefcients n and c (periodindependent) and ki (period-dependent)].
Use three faulting mechanisms:
FRV = 1, FN = 0 Reverse and reverse-oblique faulting,30 < < 150 , where is the average rake
angle.
FN = 1, FRV = 1 Normal and normal-oblique faulting, 150 < < 30 .
FRV = 0, FRV = 0 Strike-slip, other s.
Find slight tendency for over-saturation of short-period ground motions at large magnitudes and short distances. Find other functional forms for magnitude dependence too
difcult to constrain empirically or could not be reliably extrapolated to large magnitudes.
Note transition depth for buried rupture (1 km) is somewhat arbitrary.
Find weak but signicant trend of increasing ground motion with dip for both reverse and
strike-slip faults. Do not believe that seismological justied therefore do not include such
a term.
Nonlinear site model constrained by theoretical studies since empirical data insufcient
to constrain complex nonlinear behaviour.
Use depth to 2.5 km/s horizon because it showed strongest correlation with shallow and
deep sediment-depth residuals.
Believe that aspect ratio (ratio of rupture length to rupture width) has promise as a source
parameter since it shows high correlation with residuals and could model change in
ground-motion scaling at large magnitudes.
Do not nd standard deviations are magnitude-dependent. Believe difference with earlier conclusions due to larger number of high-quality intra-event recordings for both small
and large earthquakes.
Find standard deviation is dependent on level of ground shaking at soft sites.
185
2.223 Costa et al. (2006)

log10 (PGA) = c0 + c1 M + c2 M 2 + (c3 + c4 M ) log(
d2 + h2 ) + cS S
where PGA is in g, c0 = 3.879, c1 = 1.178, c2 = 0.068, c3 = 2.063, c4 = 0.102,

cS = 0.411, h = 7.8 and = 0.3448 (for larger horizontal component), c0 = 3.401,
c1 = 1.140, c2 = 0.070, c3 = 2.356, c4 = 0.150, cS = 0.415, h = 8.2 and =
0.3415 (for horizontal component using vectorial addition), c0 = 3.464, c1 = 0.958,
c2 = 0.053, c3 = 2.224, c4 = 0.147, cS = 0.330, h = 6.1 and = 0.3137 (for
vertical).
Use two site classes (since do not have detailed information on geology at all considered
stations):
S = 0 Rock
S = 1 Soil
Use selection criteria: 3.0 M 6.5 and 1 de 100 km.
Bandpass lter with cut-offs between 0.1 and 0.25 Hz and between 25 and 30 Hz.
Compute mean ratio between recorded and predicted motions at some stations of the
RAF network. Find large ratios for some stations on soil and for some on rock.
2.224 Gmez-Sobern et al. (2006)

ln a = 0 + 1 M + 2 M 2 + 3 ln R + 5 R
where a is in cm/s2 , 0 = 1.237, 1 = 1.519, 2 = 0.0313, 3 = 0.844, 5 =
0.004 and = 0.780.
Exclude records from soft soil sites or with previously known site effects (amplication
or deamplication).
Also derive equation using functional form ln a = 0 + 1 M + 2 ln R + 4 R.
Select records from stations located along the seismically active Mexican Pacic coast.
Only use records from earthquakes with M > 4.5.
Exclude data from normal faulting earthquakes using focal mechanisms, focal depths,
location of epicentre and characteristics of records because subduction zone events are
the most dominant and frequent type of earthquakes.
Use Mw because consider best representation of energy release.
Visually inspect records to exclude poor quality records.
186
Exclude records from dams and buildings.

Exclude records from slow earthquakes, which produce smaller short-period ground
motions.
Correct accelerations by nding quadratic baseline to minimize the nal velocity then
lter using most appropriate bandpass lter (low cut-off frequencies between 0.05 and
0.4 Hz and high cut-off frequency of 30 Hz).
Use data from 105 stations: 7 in Chiapas, 6 in Oaxaca, 6 in Colima, 19 in Jalisco, 49 in
Guerrero, 14 in Michoacn and 6 near the Michoacn-Guerrero border.
2.225
Hernandez et al. (2006)
log(y) = aML log(X) + bX + cj

where y is in cm/s2 , a = 0.41296, b = 0.0003, c1 = 0.5120, c2 = 0.3983, c3 = 0.2576,
c4 = 0.1962, c5 = 0.1129 and = 0.2331.
Data from ARM1 and ARM2 vertical borehole arrays of the Hualien LSST array at: surface (use c1 ), 5.3 m (use c2 ), 15.8 m (use c3 ), 26.3 m (use c4 ) and 52.6 m (use c5 ).
Surface geology at site is massive unconsolidated poorly bedded Pleistocene conglomerate composed of pebbles varying in diameter from 5 to 20 cm, following 5 m is mainly
composed of ne and medium sand followed by a gravel layer of 35 m.
Apply these criteria to achieve uniform data: ML > 5, focal depth < 30 km and 0.42ML
log(X + 0.025100.42ML 0.0033X + 1.22 > log 10 from a previous study.
Most records from ML < 6.
Bandpass lter records with cut-offs at 0.08 and 40 Hz.
Propose Ms = 1.154ML 1.34.
Some comparisons between records and predicted spectra are show for four groups of
records and nd a good match although for the group ML 6.75 and X = 62 km nd a
slight overestimation, which believe is due to not modelling nonlinear magnitude dependence.
Coefcients for vertical equations not reported.
2.226
Kanno et al. (2006)
Ground-motion model is for D 30 km:
log pre = a1 Mw + b1 X log(X + d1 100.5Mw ) + c1

and for D > 30 km:
log pre = a2 Mw + b2 X log(X) + c2

where pre is in cm/s2 , a1 = 0.56, b1 = 0.0031, c1 = 0.26, d1 = 0.0055, a2 = 0.41,
b2 = 0.0039, c2 = 1.56, 1 = 0.37 and 2 = 0.40.
187
Use Vs,30 to characterise site effects using correction formula: G = log(obs/pre) =

p log Vs,30 + q . Derive p and q by regression analysis on residuals averaged at intervals
of every 100 m/s in Vs,30 . p = 0.55 and q = 1.35 for PGA. Note that the equation
without site correction predicts ground motions at sites with Vs,30 300 m/s.
Focal depths, D, for shallow events between 0 km and 30 km and for deep events between 30 km and about 180 km.
Note that it is difcult to determine a suitable model form due to large variability of strongmotion data, correlation among model variables and because of coupling of variables
in the model. Therefore choose a simple model to predict average characteristics with
minimum parameters.
Introduce correction terms for site effects and regional anomalies.
Originally collect 91731 records from 4967 Japanese earthquakes.
Include foreign near-source data (from California and Turkey, which are compressional
regimes similar to Japan) because insufcient from Japan.
High-pass lter records with cut-off of 0.1 Hz. Low-pass lter analogue records using
cut-offs selected by visual inspection.
Choose records where: 1) Mw 5.5, 2) data from ground surface, 3) two orthogonal
horizontal components available, 4) at least ve stations triggered and 5) the record
passed this Mw -dependent source distance criterion: f (Mw , X) log 10 (for data from
mechanical seismometer networks) or f (Mw , X) log 2 (for data from other networks)
where f (Mw , X) = 0.42Mw 0.0033X log(X + 0.025100.43Mw ) + 1.22 (from a consideration of triggering of instruments).
Examine data distributions w.r.t. amplitude and distance for each magnitude. Exclude
events with irregular distributions that could be associated with a particular geological/tectonic feature (such as volcanic earthquakes).
Do not include data from Chi-Chi 1999 earthquake because have remarkably low amplitudes, which could be due to a much-fractured continental margin causing different
seismic wave propagation than normal.
Data from 2236 different sites in Japan and 305 in other countries.
Note relatively few records from large and deep events.
Note that maybe best to use stress drop to account for different source types (shallow,
interface or intraslab) but cannot use since not available for all earthquakes in dataset.
Investigate effect of depth on ground motions and nd that ground-motions amplitudes
from earthquakes with D > 30 km are considerably different than from shallower events
hence derive separate equations for shallow and deep events.
Select 0.5 within function from earlier study.
Weight regression for shallow events to give more weight to near-source data. Use
weighting of 6.0 for X 25 km, 3.0 for 25 < X 50 km, 1.5 for 50 < X 75 km and
1.0 for X > 75 km. Note that weighting scheme has no physical meaning.
188
Note that amplitude saturation at short distances for shallow model is controlled by
crustal events hence region within several tens of kms of large (Mw > 8.0) interface
events falls outside range of data.
Note standard deviation decreases after site correction term is introduced.
Introduce correction to model anomalous ground motions in NE Japan from intermediate
and deep earthquakes occurring in the Pacic plate due to unique Q structure beneath
the island arc. Correction is: log(obs/pre) = (Rtr + )(D 30) where Rtr is shortest
distance from site to Kuril and Izu-Bonin trenches. and are derived by regression
on subset fullling criteria: hypocentre in Pacic plate, station E of 137 E and station
has Vs,30 measurement. For PGA = 6.73 105 and = 2.09 102 . Find
considerable reduction in standard deviation after correction. Note that Rtr may not be
the best parameter due to observed bias in residuals for deep events.
Examine normalised observed ground motions w.r.t. predicted values and nd good
match.
Examine residuals w.r.t. distance and predicted values. Find residuals decrease with
increasing predicted amplitude and with decreasing distance. Note that this is desirable
from engineering point of view, however, note that it may be due to insufcient data with
large amplitudes and from short distances.
Examine total, intra-event and inter-event residuals w.r.t. D for D > 30 km. When
no correction terms are used, intra-event residuals are not biased but inter-event residuals are. Find mean values of total error increase up to D = 70 km and then are
constant. Find depth correction term reduces intra-event residuals considerably but increases inter-event error slightly. Overall bias improves for D < 140 km. Find site
corrections have marginal effect on residuals.
Find no bias in residuals w.r.t. magnitude.
2.227
Laouami et al. (2006)
y = c exp(Ms )[Dk + a]R

where D is rhypo and R is repi , y is in m/s2 , c = 0.38778, = 0.32927, k = 0.29202,
a = 1.557574, = 1.537231, = 0.027024 and = 0.03 (note that this is additive).
All records except one at 13 km from distances of 20 to 70 km so note that lack information from near eld.
Compare predictions to records from the 2003 Boumerdes (Mw 6.8) earthquake and nd
that it underpredicts the recorded motions, which note maybe due to local site effects.
2.228
Luzi et al. (2006)
log10 Y = a + bM + c log10 R + s1,2

189
Y is in g, a = 4.417, b = 0.770, c = 1.097, s1 = 0, s2 = 0.123, event = 0.069

and record = 0.339 (for horizontal PGA assuming intra-event ), a = 4.367, b =
0.774, c = 1.146, s1 = 0, s2 = 0.119, station = 0.077 and record = 0.337 (for
horizontal PGA assuming intra-station ), a = 4.128, b = 0.722 ,c = 1.250, s1 = 0,
s2 = 0.096, event = 0.085 and record = 0.338 (for vertical PGA assuming intra-event
), a = 4.066, b = 0.729, c = 1.322, s1 = 0, s2 = 0.090, station = 0.105 and
record = 0.335 (for vertical PGA assuming intra-station ).
1. Rock, where Vs > 800 m/s. Use s1 .
2. Soil, where Vs < 800 m/s. This includes all kinds of supercial deposits from weak
rock to alluvial deposits. Use s2 .
Can only use two classes due to limited information.
Use 195 accelerometric records from 51 earthquakes (2.5 ML 5.4) from 29 sites.
Most records are from rock or stiff sites. Most data from rhypo < 50 km with few from
> 100 km. Also use data from velocimeters (Lennartz 1 or 5 s sensors and Guralp
CMG-40Ts). In total 2895 records with rhypo < 50 km from 78 events and 22 stations
available, most from 20 rhypo 30 km.
For records from analogue instruments, baseline correct, correct for instrument response and bandpass lter with average cut-offs at 0.5 and 20 Hz (after visual inspection
of Fourier amplitude spectra). For records from digital instruments, baseline correct and
bandpass lter with average cut-offs at 0.2 and 30 Hz. Sampling rate is 200 Hz. For
records from velocimeters, correct for instrument response and bandpass lter with average cut-offs at 0.5 and 25 Hz. Sampling rate is 100 Hz.
Select records from 37 stations with 10 rhypo 50 km.
Compare predictions and observations for ML 4.4 and nd acceptable agreement. Also
nd agreement between data from accelerometers and velocimeters.
2.229 Mahdavian (2006)

log(y) = a + bM + c log(R) + dR
where y is in cm/s2 . For horizontal PGA: a = 1.861, b = 0.201, c = 0.554, d =
0.0091 and = 0.242 (for Zagros, rock sites and Ms 4.5 or mb 5.0), a = 1.831,
b = 0.208, c = 0.499, d = 0.0137 and = 0.242 (for Zagros, rock sites and
3 < Ms < 4.6 or 4.0 mb < 5.0), a = 2.058, b = 0.243, c = 1.02, d = 0.000875
and = 0.219 (for central Iran and rock sites), a = 2.213, b = 0.225, c = 0.847,
d = 0.00918 and = 0.297 (for Zagros and soil sites), a = 1.912, b = 0.201, c =
0.790, d = 0.00253 and = 0.204 (for central Iran and soil sites). For vertical PGA:
a = 2.272, b = 0.115, c = 0.853, d = 0.00529 and = 0.241 (for Zagros, rock sites
and Ms 4.5 or mb 5.0), a = 2.060, b = 0.14710 , c = 0.758, d = 0.00847 and
= 0.270 (for Zagros, rock sites and Ms 3.0 or mb 4.0), a = 1.864, b = 0.232,
10
Assume that 147 reported in paper is a typographical error.
190
c = 1.049, d = 0.000372 and = 0.253 (for central Iran and rock sites), a = 2.251,
b = 0.14011 , c = 0.822, d = 0.00734 and = 0.29012 (for Zagros and soil sites) and
a = 1.76, b = 0.23213 , c = 1.013, d = 0.000551 and = 0.229 (for central Iran and
soil sites).
Uses two site classes:
1. Sedimentary. 55 records.
2. Rock. 95 records.
Bases classication on geological maps, station visits, published classications and
shape of response spectra from strong-motion records. Notes that the classication
could be incorrect for some stations. Uses only two classes to reduce possible errors.
Divides Iran into two regions: Zagros and other areas.
Select data with Ms or mb where mb > 3.5. Notes that only earthquakes with mb > 5.0
are of engineering concern for Iran but since not enough data (especially for Zagros)
includes smaller earthquakes.
Use Ms when mb 4.
Records bandpass ltered using Ormsby lters with cut-offs and roll-offs of 0.10.25 Hz
and 2325 Hz.
Notes that some data from far-eld.
Notes that some records do not feature the main portion of shaking.
To be consistent, calculates rhypo using S-P time difference. For some records P wave
arrival time is unknown so use published hypocentral locations. Assumes focal depth of
10 km for small and moderate earthquakes and 15 km for large earthquakes.
Does not recommend use of relation for Zagros and soil sites due to lack of data (15
records) and large .
Compares recorded and predicted motions for some ranges of magnitudes and concludes that they are similar.
2.230
McVerry et al. (2006)
Ground-motion model for crustal earthquakes is:
ln SA0A/B (T ) = C10 (T ) + C4AS (M 6) + C3AS (T )(8.5 M )2 + C50 (T )r
0
2
+ [C80 (T ) + C6AS (M 6)] ln r2 + C10AS
(T ) + C46
(T )rV OL
+ C32 CN + C33AS (T )CR + FHW (M, r)
11

13
12
191
Ground-motion model for subduction earthquakes is:

0
0
0
ln SA0A/B (T ) = C11
(T ) + {C12Y + [C15
(T ) C17
(T )]C19Y }(M 6)
0
0
+ C13Y (T )(10 M )3 + C17
(T ) ln[r + C18Y exp(C19Y M )] + C20
(T )Hc
0
0
+ C24
(T )SI + C46
(T )rV OL (1 DS)
0 (T ) = C
where C15
17Y (T ). For both models:
0
0
ln SA0C,D (T ) = ln SA0A/B (T ) + C29
(T )C + [C30AS (T ) ln(PGA0A/B + 0.03) + C43
(T )]D
where PGA0A/B = SA0A/B (T = 0). Final model given by:
SAA/B,C,D (T ) = SA0A/B,C,D (T )(PGAA/B,C,D /PGA0A/B,C,D )

where SAA/B,C,D is in g, rV OL is length in km of source-to-site path in volcanic zone
and FHW (M, r) is hanging wall factor of Abrahamson & Silva (1997). Coefcients for
PGA (larger component) are: C1 = 0.28815, C3 = 0, C4 = 0.14400, C5 = 0.00967,
C6 = 0.17000, C8 = 0.70494, C10 = 5.60000, C11 = 8.68354, C12 = 1.41400,
C13 = 0, C15 = 2.552000, C17 = 2.56727, C18 = 1.78180, C19 = 0.55400, C20 =
0.01550, C24 = 0.50962, C29 = 0.30206, C30 = 0.23000, C32 = 0.20000, C33 =
0.26000, C43 = 0.31769, C46 = 0.03279, M 6 = 0.4865, slope = 0.1261, where
= M 6 + slope (Mw 6) for 5 < Mw < 7, = M 6 slope for Mw < 5 and
= M 6 + slope for Mw > 7 (intra-event), and = 0.2687 (inter-event). Coefcients for
PGA0 (larger component) are: C1 = 0.18130, C3 = 0, C4 = 0.14400, C5 = 0.00846,
C6 = 0.17000, C8 = 0.75519, C10 = 5.60000, C11 = 8.10697, C12 = 1.41400, C13 =
0, C15 = 2.552000, C17 = 2.48795, C18 = 1.78180, C19 = 0.55400, C20 = 0.01622,
C24 = 0.41369, C29 = 0.44307, C30 = 0.23000, C32 = 0.20000, C33 = 0.26000,
C43 = 0.29648, C46 = 0.03301, M 6 = 0.5035, slope = 0.0635 and = 0.2598.
Use site classes (combine A and B together and do not use data from E):
A Strong rock. Strong to extremely-strong rock with: a) unconned compressive

strength > 50 MPa, and b) Vs,30 > 1500 m/s, and c) not underlain by materials
with compressive strength < 18 MPa or Vs < 600 m/s.
B Rock. Rock with: a) compressive strength between 1 and 50 MPa, and b) Vs,30 >
360 m/s, and c) not underlain by materials having compressive strength < 0.8 MPa
or Vs < 300 m/s.
C, C = 1, D = 0 Shallow soil sites. Sites that: a) are not class A, class B or class E sites, and
b) have low-amplitude natural period, T , 0.6 s, or c) have soil depths these
depths:
192
Cohesive soil
Very soft
Soft
Firm
Stiff
Very stiff or hard
Cohesionless soil
Very loose
Loose dry
Medium dense
Dense
Very dense
Gravels
Soil type
and description
Representative undrained
shear strengths ( kPa)
< 12.5
12.525
2550
50100
100200
Representative SPT N values
<6
610
1030
3050
> 50
> 30
Maximum
soil depth ( m)
0
20
25
40
60
0
40
45
55
60
100
D, D = 1, C = 0 Deep or soft soil sites. Sites that: a) are not class A, class B or class E sites, and
b) have a low-amplitude T > 0.6 s, or c) have soil depths > depths in table above,
or c) are underlain by < 10 m of soils with an undrained shear-strength < 12.5 kPa
or soils with SPT N-values < 6.
E Very soft soil sites. Sites with: a) > 10 m of very soft soils with undrained shearstrength < 12.5 kPa, b) > 10 m of soils with SPT N values < 6, c) > 10 m of
soils with Vs < 150 m/s, or d) > 10 m combined depth of soils with properties as
described in a), b) and c).
Categories based on classes in existing New Zealand Loadings Standard but modied
following statistical analysis. Note advantage of using site categories related to those
in loading standards. Site classications based on site periods but generally categories
from site descriptions.
Classify earthquakes in three categories:
Crustal Earthquakes occurring in the shallow crust of overlying Australian plate. 24 earthquakes. Classify into:
Strike-slip 33 33 , 147 180 or 180 147 where is the rake.
6 earthquakes. Centroid depths, Hc , 4 Hc 13 km. 5.20 Mw 6.31.
CN = 0, CR = 0.
Normal 146 34 . 7 earthquakes. 7 Hc 17 km. 5.27 Mw 7.09.
CN = 1, CR = 0.
Oblique-reverse 33 66 or 124 146 . 3 earthquakes. 5 Hc 19 km.
5.75 Mw 6.52. CR = 0.5, CN = 0.
Reverse 67 123 . 8 earthquakes. 4 Hc 13 km. 5.08 Mw 7.23.
CR = 1, CN = 0.
Interface Earthquake occurring on the interface between Pacic and Australian plates with
Hc < 50 km. 5 reserve and 1 strike-slip with reverse component. Use data with
15 Hc 24 km. Classify using location in 3D space. 6 earthquakes. 5.46
Mw 6.81. SI = 1, DS = 0.
Slab Earthquakes occurring in slab source zone within the subducted Pacic plate. Predominant mechanism changes with depth. 19 earthquakes. 26 Hc 149 km.
193
Split into shallow slab events with Hc 50 km (9 normal and 1 strike-slip, 5.17
Mw 6.23) and deep slab events with Hc > 50 km (6 reverse and 3 strike-slip,
5.30 Mw 6.69). SI = 0, DS = 1 (for deep slab events).
Note seismicity cross sections not sufcient to distinguish between interface and slab
events, also require source mechanism.
Find that mechanism is not a signicant extra parameter for motions from subduction
earthquakes.
State that model is not appropriate for source-to-site combinations where the propagation path is through the highly attenuating mantle wedge.
Note magnitude range of New Zealand is limited with little data for large magnitudes and
from short distances. Most data from d > 50 km and Mw < 6.5.
Only include records from earthquakes with available Mw estimates because correlations between ML and Mw are poor for New Zealand earthquakes. Include two earthquakes without Mw values (Ms was converted to Mw ) since they provide important data
for locations within and just outside the Central Volcanic Region.
Only include data with centroid depth, mechanism type, source-to-site distance and a
description of site conditions.
Only include records with PGA above these limits (dependent on resolution of instrument):
1. Acceleroscopes (scratch-plates): 0.02 g
2. Mechanical-optical accelerographs: 0.01 g
3. Digital 12-bit accelerographs: 0.004 g
4. Digital 16-bit accelerographs: 0.0005 g
Exclude data from two sites: Athene A (topographic effect) and Hanmer Springs (site
resonance at 1.51.7 Hz) that exhibit excessive amplications for their site class.
Exclude data from sites of class E (very soft soil sites with & 10 m of material with
Vs < 150 m/s) to be consistent with Abrahamson & Silva (1997) and Youngs et al.
(1997). Not excluded because of large amplications but because spectra appear to
have site-specic characteristics.
Exclude records from bases of buildings with > 4 storeys because may have been
inuenced by structural response.
Exclude data from very deep events with travel paths passing through the highly attenuating mantle were excluded.
Only use response spectral ordinates for periods where they exceed the estimated noise
levels of the combined recording and processing systems.
Lack of data from near-source. Only 11 crustal records from distances < 25 km with 7
of these from 3 stations. To constrain model at short distances include overseas PGA
data using same criteria as used for New Zealand data. Note that these data were not
intended to be comprehensive for 010 km range but felt to be representative. Note that
194
it is possible New Zealand earthquakes may produce PGAs at short distances different
that those observed elsewhere but feel that it is better to constrain the near-source
behaviour rather than predict very high PGAs using an unconstrained model.
In order to supplement limited data from moderate and high-strength rock and from the
volcanic region, data from digital seismographs were added.
Data corrected for instrument response.
Derive model from base models (other Ground-motion models for other regions). Select base model using residual analyses of New Zealand data w.r.t. various models.
Choose models of Abrahamson & Silva (1997) for crustal earthquakes and Youngs et al.
(1997). Link these models together by common site response terms and standard deviations to get more robust coefcients.
Apply constraints using base models to coefcients that are reliant on data from magnitude, distance and other model parameters sparsely represented in the New Zealand
data. Coefcients constrained are those affecting estimates in near-source region,
source-mechanism terms for crustal earthquakes and hanging-wall terms. Eliminate
some terms in base models because little effect on measures of t using Akaike Information Criterion (AIC).
Apply the following procedure to derive model. Derive models for PGA and SA using only
records with response spectra available (models with primed coefcients). Next derive
model for PGA including records without response spectra (unprimed coefcients). Finally multiply model for SA by ratio between the PGA model using all data and that using
only PGA data with corresponding response spectra. Apply this method since PGA estimates using complete dataset for some situations (notably on rock and deep soil and
for near-source region) are higher than PGA estimates using reduced dataset and are
more in line with those from models using western US data. This scaling introduces a
bias in nal model. Do not correct standard deviations of models for this bias.
Use rrup for 10 earthquakes and rc for rest. For most records were rc was used, state
that it is unlikely model is sensitive to use rc rather than rrup . For ve records discrepancy likely to be more than 10%.
Free coefcients are: C1 , C11 , C8 , C17 , C5 , C46 , C20 , C24 , C29 and C43 . Other coefcients xed during regression. Coefcients with subscript AS are from Abrahamson &
Silva (1997) and those with subscript Y are from Youngs et al. (1997). Try varying some
of these xed coefcients but nd little improvement in ts.
State that models apply for 5.25 Mw 7.5 and for distances 400 km, which is
roughly range covered by data.
Note possible problems in applying model for Hc > 150 km therefore suggest Hc is xed
to 150 km if applying model to deeper earthquakes.
Note possible problems in applying model for Mw < 5.25.
Apply constraints to coefcients to model magnitude- and distance-saturation.
Try including an anelastic term for subduction earthquakes but nd insignicant.
195
Investigate possibility of different magnitude-dependence and attenuation rates for interface and slab earthquakes but this required extra parameters that are not justied by
AIC.
Investigate possible different depth dependence for interface and slab earthquakes but
extra parameters not justied in terms of AIC.
Try adding additive deep slab term but not signicant according to AIC.
Cannot statistically justify nonlinear site terms. Believe this could be due to lack of nearsource records.
Find that if a term is not included for volcanic path lengths then residuals for paths crossing the volcanic zone are increasingly negative with distance but this trend is removed
when a volcanic path length term is included.
Compare predictions to observed ground motions in 21/08/2003 Fiordland interface
(Mw 7.2) earthquake and its aftershocks. Find ground motions, in general, underestimated.
2.231 Moss & Der Kiureghian (2006)

Ground-motion model is [adopted from Boore et al. (1997)]:
2 + 2 ) ln(V
ln(Y ) = 1 + 2 (Mw 6) + 3 (Mw 6)2 4 ln( Rjb
6
s,30 /7 )
5
Use Vs,30 to characterize site.
Use data of Boore et al. (1997).
Develop Bayesian regression method to account for parameter uncertainty in measured
accelerations (due to orientation of instrument) (coefcient of variation of 0.30, based
on analysis of recorded motions) and magnitudes (coefcient of variation of 0.10,
based on analysis of reported Mw by various agencies) to better understand sources of
uncertainty and to reduce model variance.
Do not report coefcients. Only compare predictions with observations and with predictions by model of Boore et al. (1997) for Mw 7.5 and Vs,30 = 750 m/s. Find slightly
different coefcients than Boore et al. (1997) but reduced model standard deviations.
2.232 Pousse et al. (2006)

log10 (PGA) = aPGA M + bPGA R log10 (R) + SPGA,k , k = 1, 2, . . . , 5

where PGA is in cm/s2 , aPGA = 0.4346, bPGA = 0.002459, SPGA,1 = 0.9259,
SPGA,2 = 0.9338, SPGA,3 = 0.9929, SPGA,4 = 0.9656, SPGA,5 = 0.9336 and =
0.2966.
Use ve site categories (from Eurocode 8):
196
A Vs,30 > 800 m/s. Use SPGA,1 . 43 stations, 396 records.

B 360 < Vs,30 < 800 m/s. Use SPGA,2 . 399 stations, 4190 records.
C 180 < Vs,30 < 360 m/s. Use SPGA,3 . 383 stations, 4108 records.
D Vs,30 < 180 m/s. Use SPGA,4 . 65 stations, 644 records.
E Site D or C underlain in rst 20 m with a stiffer layer of Vs > 800 m/s. Use SPGA,5 .
6 stations, 52 records.
Use statistical method of Boore (2004) with parameters derived from KiK-Net proles in
order to extend Vs proles down to 30 m depth.
Records from K-Net network whose digital stations have detailed geotechnical characterisation down to 20 m depth.
Retain only records from events whose focal depths < 25 km.
Convert MJMA to Mw using empirical conversion formula to be consist with other studies.
Apply magnitude-distance cut-off to exclude distant records.
Bandpass lter all records with cut-offs 0.25 and 25 Hz. Visually inspect records for
glitches and to retain only main event if multiple events recorded.
Find that one-stage maximum likelihood regression gives almost the same results.
Also derive equations for other strong-motion parameters.
2.233
Souriau (2006)
log10 (PGA) = a + bM + c log10 R

where y is in m/s2 , a = 2.50 0.18, b = 0.99 0.05 and c = 2.22 0.08 when
M = MLDG and a = 2.55 0.19, b = 1.04 0.05 and c = 2.17 0.08 when
M = MReNass ( is not given although notes that explained variance is of the order of
84%).
Most data from R < 200 km.
Uses PGAs from S-waves.
Finds that introducing an anelastic attenuation term does not signicantly improve explained variance because term is poorly constrained by data due to trade offs with geometric term and travel paths are short. When an anelastic term is introduced nds:
log10 (PGA) = 3.19(0.25)+1.09(0.05)MReNass 1.83(0.12) log10 R0.0013(0.0004)R.

197
2.234 Zare & Sabzali (2006)

log Sa(T ) = a1 (T )M + a2 (T )M 2 + b(T ) log(R) + ci (T )Si

where Sa is in g, a1 = 0.5781, a2 = 0.0317, b = 0.4352, c1 = 2.6224, c2 =
2.5154, c3 = 2.4654, c4 = 2.6213 and = 0.2768 (for horizontal PGA), a1 =
0.5593, a2 = 0.0258, b = 0.6119, c1 = 2.6261, c2 = 2.6667, c3 = 2.5633,
c4 = 2.7346 and = 0.2961 (for vertical PGA).
Use four site classes based on fundamental frequency, f , from receiver functions:
Class 1 f > 15 Hz. Corresponds to rock and stiff sediment sites with Vs,30 > 700 m/s. 22
records. S1 = 1 and other Si = 0.
Class 2 5 < f 15 Hz. Corresponds to stiff sediments and/or soft rocks with 500 <
Vs,30 700 m/s. 16 records. S2 = 1 and other Si = 0.
Class 3 2 < f 5 Hz. Corresponds to alluvial sites with 300 < V 500 m/s. 25 records.
S3 = 1 and other Si = 0.
Class 4 f 2 Hz. Corresponds to thick soft alluvium. 26 records. S4 = 1 and other Si = 0.
Separate records into four mechanisms: reverse (14 records), reverse/strike-slip (1
record), strike-slip (26 records) and unknown (48 records).
Select records that have PGA > 0.05 g on at least one component and are of good
quality in frequency band of 0.3 Hz or less.
Find results using one- or two-step regression techniques are similar. Only report results
from one-step regression.
Mw for earthquakes obtained directly from level of acceleration spectra plateau of records
used.
rhypo for records obtained from S-P time difference.
Most data from rhypo < 60 km.
Bandpass lter records with cut-offs of between 0.08 and 0.3 Hz and between 16 and
40 Hz.
Note that the lack of near-eld data is a limitation.
2.235 Akkar & Bommer (2007b)

log y = b1 + b2 M + b3 M 2 + (b4 + b5 M ) log
2 + b2 + b S + b S + b F + b F
Rjb
7 S
8 A
9 N
10 R
6
where y is in cm/s2 , b1 = 1.647, b2 = 0.767, b3 = 0.074, b4 = 3.162, b5 = 0.321,

b6 = 7.682, b7 = 0.105, b8 = 0.020, b9 = 0.045, b10 = 0.085, 1 = 0.557 0.049M
(intra-event) and 2 = 0.189 0.017M (inter-event) when b3 is unconstrained and b1 =
4.185, b2 = 0.112, b4 = 2.963, b5 = 0.290, b6 = 7.593, b7 = 0.099, b8 = 0.020,
b9 = 0.034, b10 = 0.104, 1 = 0.557 0.049M (intra-event) and 2 = 0.204 0.018M
(inter-event) when b3 is constrained to zero (to avoid super-saturation of PGA).
198

Soft soil SS = 1, SA = 0.
Stiff soil SA = 1, SS = 0.
Rock SS = 0, SA = 0.
Normal FN = 1, FR = 0.
Strike-slip FN = 0, FR = 0.
Reverse FR = 1, FN = 0.
Use same data as Akkar & Bommer (2007a), which is similar to that used by Ambraseys
et al. (2005a).
Individually process records using well-dened correction procedure to select the cut-off
frequencies (Akkar & Bommer, 2006).
Use pure error analysis to determine magnitude dependence of inter- and intra-event
variabilities before regression analysis.
2.236
Ghodrati Amiri et al. (2007a) & Ghodrati Amiri et al. (2007b)
ln y = C1 + C2 Ms + C3 ln[R + C4 exp(Ms )] + C5 R
where y is in cm/s2 , C1 = 4.15, C2 = 0.623, C3 = 0.96 and = 0.478 for horizontal
PGA, rock sites and Alborz and central Iran; C1 = 3.46, C2 = 0.635, C3 = 0.996 and
= 0.49 for vertical PGA, rock sites and Alborz and central Iran; C1 = 3.65, C2 = 0.678,
C2 = 0.95 and = 0.496 for horizontal PGA, soil sites and Alborz and central Iran;
C1 = 3.03, C2 = 0.732, C3 = 1.03 and = 0.53 for vertical PGA, soil sites and
Alborz and central Iran; C1 = 5.67, C2 = 0.318, C3 = 0.77, C5 = 0.016 and
= 0.52 for horizontal PGA, rock sites and Zagros; C1 = 5.26, C2 = 0.289, C3 = 0.8,
C5 = 0.018 and = 0.468 for vertical PGA, rock sites and Zagros; C1 = 5.51,
C2 = 0.55, C3 = 1.31 and = 0.488 for horizontal PGA, soil sites and Zagros; and
C1 = 5.52, C2 = 0.36, C3 = 1.25 and = 0.474 for vertical PGA, soil sites and
Zagros. Constrain C4 to zero for better convergence even though s are higher.
Use two site categories (derive individual equations for each):
Rock Roughly Vs 375 m/s.
Soil Roughly Vs < 375 m/s.
Divide Iran into two regions: Alborz and central Iran, and Zagros, based on tectonics
and derive separate equations for each.
Use S-P times to compute rhypo for records for which it is unknown.
Exclude data from earthquakes with Ms 4.5 to remove less accurate data and since
larger earthquakes more important for seismic hazard assessment purposes.
199
Most records from rhypo > 50 km.

Exclude poor quality records.
Instrument, baseline correct and bandpass lter records with cut-offs depending on instrument type and site class. For SSA-2 recommend: 0.150.2 Hz and 30 33 Hz
for rock records and 0.070.2 Hz and 3033 Hz for soil records. For SMA-1 recommend: 0.150.25 Hz and 2023 Hz for rock records and 0.150.2 Hz and 2023 Hz for
soil records. Apply trial and error based on magnitude, distance and velocity time-history
to select cut-off frequencies.
Test a number of different functional forms.
Often nd a positive (non-physical) value of C5 . Therefore, remove this term. Try removing records with rhypo > 100 km but nd little difference and poor convergence due to
limited data.
Do not include term for faulting mechanism because such information not available for
Iranian events.
2.237 Aydan (2007)

amax = F (Vs )G(R, )H(M )

Characterises sites by Vs (shear-wave velocity).
Considers effect of faulting mechanism.
Considers angle between strike and station, .
2.238 Bindi et al. (2007)

log10 Y = a + bM + (c + dM ) log10 Rhypo + s1,2

where Y is in m/s2 , a = 1.4580, b = 0.4982, c = 2.3639, d = 0.1901, s2 = 0.4683,
eve = 0.0683 (inter-event), sta = 0.0694 (inter-station) and rec = 0.2949 (record-torecord) for horizontal PGA; and a = 1.3327, b = 0.4610, c = 2.4148, d = 0.1749,
s2 = 0.3094, eve = 0.1212 (inter-event), sta = 0.1217 (inter-station) and rec = 0.2656
(record-to-record) for vertical PGA.
2
log10 Y = a + bM + (c + dM ) log10 (Repi
+ h2 )0.5 + s1,2
where Y is in m/s2 , a = 2.0924, b = 0.5880, c = 1.9887, d = 0.1306, h = 3.8653,

s2 = 0.4623, eve = 0.0670 (inter-event), sta = 0.0681 (inter-station) and rec = 0.2839
(record-to-record) for horizontal PGA; and a = 1.8883, b = 0.5358, c = 2.0869,
d = 0.1247, h = 4.8954, s2 = 0.3046, eve = 0.1196 (inter-event), sta = 0.0696 (interstation) and rec = 0.2762 (record-to-record). Coefcients not reported in article but in
electronic supplement.
200
s1 Rock. Maximum amplication less than 2.5 (for accelerometric stations) or than 4.5
(for geophone stations). Amplication thresholds dened after some trials.
s2 Soil. Maximum amplication greater than thresholds dened above.
Classify stations using generalized inversion technique.
Use aftershocks from the 1999 Kocaeli (Mw 7.4) earthquake.
Use data from 31 1 Hz 24-bit geophones and 23 12-bit and 16-bit accelerometers.
Records corrected for instrument response and bandpass ltered (fourth order Butterworth) with cut-offs 0.5 and 25 Hz for ML 4.5 and 0.1 and 25 Hz for ML > 4.5. Find
lters affect PGA by maximum 10%.
Only 13 earthquakes have ML < 1.0. Most data between have 1.5 < ML < 5 and from
10 de 140 km.
Geophone records from free-eld stations and accelerometric data from ground oors
of small buildings.
Use rhypo and repi since no evidence for surface ruptures from Turkey earthquakes with
ML < 6 and no systematic studies on the locations of the rupture planes for events
used.
Since most earthquakes are strike-slip do not include style-of-faulting factor.
Find differences in inter-event when using ML or Mw , which relate to frequency band
used to compute ML (about 110 Hz) compared to Mw (low frequencies), but nd similar intra-event s using the two different magnitudes, which expected since this not
source-related.
Investigate inuence of stress drop on inter-event for horizontal PGA relations using
repi and ML or Mw . Find inter-event errors range from negative (low stress drop) to
positive (high stress drop) depending on stress drop.
Regress twice: rstly not considering site classication and secondly considering. Find
site classication signicantly reduces inter-station errors for velometric stations but
inter-station errors for accelerometric stations less affected.
2.239
log10 [PSA(T )] = b1 + b2 Mw + b3 Mw2 + (b4 + b5 Mw ) log10
2 + b2 + b S + b S
Rjb
7 S
8 A
6
+ b9 FN + b10 FR
where PSA(T ) is in cm/s2 , b1 = 0.0031, b2 = 1.0848, b3 = 0.0835, b4 = 2.4423,
b5 = 0.2081, b6 = 8.0282, b7 = 0.0781, b8 = 0.0208, b9 = 0.0292, b10 = 0.0963, 1 =
0.5990.0410.0580.008Mw (intra-event) and 2 = 0.3230.0750.0310.014Mw
(inter-event).
201

Soft soil Vs,30 < 360 m/s. SS = 1, SA = 1. 75 records from 3 Mw < 5.
Stiff soil 360 < Vs,30 < 750 m/s. SA = 1, SS = 0. 173 records from 3 Mw < 5.
Rock Vs,30 750 m/s. SS = 0, SA = 0. 217 records from 3 Mw < 5.
Normal FN = 1, FR = 0. 291 records from 3 Mw < 5.
Strike-slip FN = 0, FR = 0. 140 records from 3 Mw < 5.
Reverse FR = 1, FN = 0. 24 records from 3 Mw < 5. 12% of all records. Note that
reverse events poorly represented.
Investigate whether Ground-motion models can be extrapolated outside the magnitude
range for which they were derived.
Extend dataset of Akkar & Bommer (2007b) by adding data from earthquakes with 3
Mw < 5. Search ISESD for records from earthquakes with Mw < 5, known site class
and known faulting mechanism. Find one record from a Mw 2 event but only 11 for
events with Mw < 3 therefore use Mw 3 as lower limit. Select 465 records from 158
events with 3 Mw < 5. Many additional records from Greece (mainly singly-recorded
events), Italy, Spain, Switzerland, Germany and France. Few additional records from
Iran and Turkey.
Data well distributed w.r.t. magnitude, distance and site class but for Mw < 4 data
sparse for distances > 40 km.
Additional data has been uniformly processed with cut-offs at 0.25 and 25 Hz.
Use same regression technique as Akkar & Bommer (2007b).
Observe that equations predict expected behaviour of response spectra so conclude
that equations are robust and reliable.
Compare predicted ground motions with predictions from model of Akkar & Bommer
(2007b) and nd large differences, which they relate to the extrapolation of models outside their range of applicability.
Investigate effect of different binning strategies for pure error analysis (Douglas & Smit,
2001). Derive weighting functions for published equations using bins of 2 km 0.2
magnitude units and require three records per bin before computing . Repeat using
1 km 0.1 unit bins. Find less bins allow computation of . Also repeat original analysis but require four or ve records per bin. Find more robust estimates of but note
that four or ve records are still small samples. Also repeating using logarithmic rather
than linear distance increments for bins since ground motions shown to mainly decay
geometrically. For all different approaches nd differences in computed magnitude dependence depending on binning scheme. None of the computed slopes are signicant
at 95% condence level.
Repeat analysis assuming no magnitude dependence of . Find predictions with this
model are very similar to those assuming a magnitude-dependent .
202
Find that compared to s of Akkar & Bommer (2007b) that inter-event s has greatly
increased but that intra-event s has not, which they relate to the uncertainty in the
determination of Mw and other parameters for small earthquakes.
Repeat analysis exclude data from (in turn) Greece, Italy, Spain and Switzerland to investigate importance of regional dependence on results. Find that results are insensitive
to the exclusion of individual regional datasets.
Compute residuals with respect to Mw for four regional datasets and nd that only for
Spain (the smallest set) is a signicant difference to general results found.
Examine total and intra-event residuals for evidence of soil nonlinearity. Find that evidence for nonlinearity is weak although the expected negative slopes are found. Conclude that insufcient data (and too crude site classication) to adjust the model for soil
nonlinearity.
Plot inter-event and intra-event residuals w.r.t. Mw and nd no trend and hence conclude
that new equations perform well for all magnitudes.
Do not propose model for application in seismic hazard assessments.
2.240
Boore & Atkinson (2007) & Boore & Atkinson (2008)
ln Y
= FM (M ) + FD (RJB , M ) + FS (VS30 , RJB , M )
FD (RJB , M ) = [c1 + c2 (M Mref )] ln(R/Rref ) + c3 (R Rref )
2 + h2
R =
RJB
e1 U + e2 SS + e3 NS + e4 RS + e5 (M Mh )+
e6 (M Mh )2 for M Mh
FM (M ) =
e1 U + e2 SS + e3 NS + e4 RS + e7 (M Mh ) for M > Mh
FS = FLIN + FN L
FLIN
FN L
= blin ln(VS30 /Vref )
bnl ln(pga_low/0.1) for pga4nl a1
bnl ln(pga_low/0.1) + c[ln(pga4nl/a1 )]2 +

=
d[ln(pga4nl/a1 )]3 for a1 < pga4nl a2
bnl ln(pga4nl/0.1) for a2 < pga4nl
c = (3y bnl x)/x2

d = (2y bnl x)/x3
x = ln(a2 /a1 )
y = bnl ln(a2 /pga_low)
b1 for VS30 V1
(b1 b2 ) ln(VS30 /V2 )/ ln(V1 /V2 ) + b2 for V1 < VS30 V2

bnl =
b ln(VS30 /Vref )/ ln(V2 /Vref ) for V2 < VS30 < Vref
2
0.0 for Vref VS30
where Y is in g, Mh = 6.75 (hinge magnitude), Vref = 760 m/s (specied reference
velocity corresponding to the NEHRP B/C boundary), a1 = 0.03 g (threshold for linear
203
amplifcation), a2 = 0.09 g (threshold for nonlinear amplication), pga_low = 0.06 g (for

transition between linear and nonlinear behaviour), pga4nl is predicted PGA in g for Vref
with FS = 0, V1 = 180 m/s, V2 = 300 m/s, blin = 0.360, b1 = 0.640, b2 = 0.14,
Mref = 4.5, Rref = 1 km, c1 = 0.66050, c2 = 0.11970, c3 = 0.01151, h = 1.35,
e1 = 0.53804, e2 = 0.50350, e3 = 0.75472, e4 = 0.50970, e5 = 0.28805,
e6 = 0.10164, e7 = 0.0; = 0.502 (intra-event); U = 0.265, M = 0.260 (inter-event);
T U = 0.566, T M = 0.560 (total).
Characterise sites using VS30 . Believe equations applicable for 180 VS30 1300 m/s
(state that equations should not be applied for very hard rock sites, VS30 1500 m/s).
Bulk of data from NEHRP C and D sites (soft rock and rm soil) and very few data from
A sites (hard rock). Use three equations for nonlinear amplication: to prevent nonlinear amplication increasing indenitely as pga4nl decreases and to smooth transition
from linear to nonlinear behaviour. Equations for nonlinear site amplication simplied
version of those of Choi & Stewart (2005) because believe NGA database insufcient
to simultaneously determine all coefcients for nonlinear site equations and magnitudedistance scaling due to trade-offs between parameters. Note that implicit trade-offs
involved and change in prescribed soil response equations would lead to change in
derived magnitude-distance scaling.
Focal depths between 2 and 31 km with most < 20 km.
Use data from the PEER Next Generation Attenuation (NGA) Flatle supplemented with
additional data from three small events (2001 Anza M 4.92, 2003 Big Bear City M 4.92
and 2002 Yorba Linda M 4.27) and the 2004 Parkeld earthquake, which were used only
for a study of distance attenuation function but not the nal regression (due to rules of
NGA project).
Use three faulting mechanism categories using P and T axes:
SS Strike-slip. Plunges of T and P axes < 40 . 35 earthquakes. Dips between 55 and
90 . 4.3 M 7.9. SS = 1, U = 0, NS = 0, RS = 0.
RS Reverse. Plunge of T axis > 40 . 12 earthquakes. Dips between 12 and 70 .
5.6 M 7.6. RS = 1, U = 0, SS = 0, NS = 0.
NS Normal. Plunge of P axis > 40 . 11 earthquakes. Dips between 30 and 70 .
5.3 M 6.9. NS = 1, U = 0, SS = 0, RS = 0.
Note that some advantages to using P and T axes to classify earthquakes but using categories based on rake angles with: within 30 of horizontal as strike-slip, from 30 to 150
as reverse and from 30 to 150 as normal, gives essentially the same classication.
Also allow prediction of motions for unspecied (U = 1, SS = 0, NS = 0, RS = 0)
mechanism (use s and s with subscript U otherwise use s and s with subscript M).
Exclude records from obvious aftershocks because believe that spectral scaling of aftershocks could be different than that of mainshocks. Note that this cuts the dataset
roughly in half.
Exclude singly-recorded earthquakes.
Note that possible bias due to lack of low-amplitude data (excluded due to non-triggering
of instrument, non-digitisation of record or below the noise threshold used in determining
low-cut lter frequencies). Distance to closest non-triggered station not available in NGA
204
Flatle so cannot exclude records from beyond this distance. No information available
that allows exclusion of records from digital accelerograms that could remove this bias.
Hence note that obtained distance dependence for small earthquakes and long periods
may be biased towards a decay that is less rapid than true decay.
Use estimated RJB s for earthquakes with unknown fault geometries.
Lack of data at close distances for small earthquakes.
Three events (1987 Whittier Narrows, 1994 Northridge and 1999 Chi-Chi) contribute
large proportion of records (7%, 10% and 24%).
Note that magnitude scaling better determined for strike-slip events, which circumvent
using common magnitude scaling for all mechanisms.
Seek simple functional forms with minimum required number of predictor variables.
Started with simplest reasonable form and added complexity as demanded by comparisons between predicted and observed motions. Selection of functional form heavily
guided by subjective inspection of nonparametric plots of data.
Data clearly show that modelling of anelastic attenuation required for distances > 80 km
and that effective geometric spreading is dependent on magnitude. Therefore, introduce
terms in the function to model these effects, which allows model to be used to 400 km.
Do not include factors for depth-to-top of rupture, hanging wall/footwall or basin depth
because residual analysis does not clearly show that the introduction of these factors
would improve the predictive capabilities of model on average.
Models are data-driven and make little use of simulations.
Believe that models provide a useful alternative to more complicated NGA models as
they are easier to implement in many applications.
Firstly correct ground motions to obtain equivalent observations for reference velocity
of 760 m/s using site amplication equations using only data with RJB 80 km and
VS30 > 360 m/s. Then regress site-corrected observations to obtain FD and FM with
FS = 0. No smoothing of coefcients determined in regression (although some of the
constrained coefcients were smoothed).
Assume distance part of model applies for crustal tectonic regimes represented by NGA
database. Believe that this is a reasonable initial approach. Test regional effects by
examining residuals by region.
Note that data sparse for RJB > 80 km, especially for moderate events, and, therefore,
difcult to obtain robust c1 (slope) and c3 (curvature) simultaneously. Therefore, use
data from outside NGA database (three small events and 2004 Parkeld) to dene c3
and use these xed values of c3 within regression to determine other coefcients. To
determine c3 and h from the four-event dataset set c1 equal to 0.5, 0.8 and 1.0
and c2 = 0 if the inclusion of event terms c0 for each event. Use c3 s when c1 = 0.8
since it is a typical value for this parameter in previous studies. Find that c3 and h are
comparable to those in previous studies.
Note that desirable to constrain h to avoid overlap in curves for large earthquakes at
very close distances. Do this by initially performing regression with h as free parameter
and then modifying h to avoid overlap.
205
After h and c3 have been constrained solve for c1 and c2 .
Constrain quadratic for magnitude scaling so that maximum not reached for M < 8.5
to prevent oversaturation. If maximum reached for M < 8.5 then perform two-segment
regression hinged at Mh with quadratic for M Mh and linear for M > Mh . If slope
of linear segment is negative then repeat regression by constraining slope above Mh
to 0.0. Find that data generally indicates oversaturation but believe this effect is too
extreme at present. Mh xed by observation that ground motions at short periods do
not get signicantly larger with increasing magnitude.
Plots of event terms (from rst stage of regression) against M show that normal-faulting
earthquakes have ground motions consistently below those of strike-slip and reverse
events. Firstly group data from all fault types together and solved for e1 , e5 , e6 , e7 and e8
by setting e2 , e3 and e4 to 0.0. Then repeat regression xing e5 , e6 , e7 and e8 to values
obtained in rst step to nd e2 , e3 and e4 .
Examine residual plots and nd no signicant trends w.r.t. M , RJB or VS30 although
some small departures from a null residual.
Examine event terms from rst stage of regression against M and conclude functional
form provides reasonable t to near-source data.
Examine event terms from rst stage of regression against M for surface-slip and nosurface-slip earthquakes. Find that most surface-slip events correspond to large magnitudes and so any reduction in motions for surface-slip earthquakes will be mapped into
reduced magnitude scaling. Examine event terms from strike-slip earthquakes (because
both surface- and buried-slip events in same magnitude range) and nd no indication
of difference in event terms for surface-slip and no-surface-slip earthquakes. Conclude
that no need to include dummy variables to account for this effect.
Examine residuals for basin depth effects. Find that VS30 and basin depth are highly correlated and so any basin-depth effect will tend to be captured by empirically-determined
site amplications. To separate VS30 and basin-depth effects would require additional
information or assumptions but since aiming for simplest equations no attempt made
to break down separate effects. Examine residuals w.r.t. basin depth and nd little
dependence.
Chi-Chi data forms signicant fraction (24% for PGA) of data set. Repeat complete
analysis without these data to examine their inuence. Find that predictions are not
dramatically different.
Note that use of anelastic coefcients derived using data from four earthquakes in central and southern California is not optimal and could lead to inconsistencies in hs.
206
2.241
Campbell & Bozorgnia (2007), Campbell & Bozorgnia (2008b)

& Campbell & Bozorgnia (2008a)
ln Y
fmag
fdis
= fmag + fdis + ff lt + fhng + fsite + fsed
c0 + c1 for M 5.5
c0 + c1 M + c2 (M 5.5) for 5.5 < M 6.5
=
c0 + c1 M + c2 (M 5.5) + c3 (M 6.5) for M > 6.5
2
2
= (c4 + c5 M ) ln( RRU
P + c6 )
ff lt = c7 FRV ff lt,Z + c8 FN M
{
ZT OR for ZT OR < 1
ff lt,Z =
1 for ZT OR 1
fhng = c9 fhng,R fhng,M fhng,Z fhng,
1 for RJB =
{max(RRU P , R2 + 1) RJB }/
JB
fhng,R =
2 + 1) for R
max(RRU P , RJB
JB > 0, ZT OR < 1
(RRU P RJB )/RRU P for RJB > 0, ZT OR 1
0 for M 6.0
2(M 6.0) for 6.0 < M < 6.5
fhng,M =
1 for M 6.5
{
0 for ZT OR 20
fhng,Z =
(20 ZT OR )/20 for 0 ZT OR < 20
{
1 for 70
fhng, =
(90 )/20 for > 70
(
)
{ [
(
)n ]
}
VS30
VS30
c
ln
+
k
ln
A
+
c
ln(A
+
c)
for VS30 < k1
2
1100
1100
k1
k1
10
)
(
fsite =
(c10 + k2 n) ln VkS30
for k1 VS30 < 1100
( 1 )
(c10 + k2 n) ln 1100
for VS30 1100
k1
c11 (Z2.5 1) for Z2.5 < 1

0 for 1 Z2.5 3
fsed =
c12 k3 e0.75 [1 e0.25(Z2.5 3) ] for Z2.5 > 3
2
2
2 2
ln
=
Y + ln AF + ln AB + 2ln YB ln AB
{
k2 A1100 {[A1100 + c(VS30 /k1 )n ]1 (A1100 + c)1 } for VS30 < k1
=
0 for VS30 k1
where Y is in g, c0 = 1.715, c1 = 0.500, c2 = 0.530, c3 = 0.262, c4 = 2.118,
c5 = 0.170, c6 = 5.60, c7 = 0.280, c8 = 0.120, c9 = 0.490, c10 = 1.058, c11 =
0.040, c12 = 0.610, k1 = 865, k2 = 1.186, k3 = 1.839, ln Y = 0.478 (intra-event),
ln Y = 0.219 (inter-event), C = 0.166, T = 0.526 (total), Arb = 0.551 and =
1.000 (correlation coefcient between intra-event residuals of ground-motion parameter
1/2 is standard deviation at base of site
2
2
of interest and PGA). ln YB = (ln
Y ln AF )
prole. Assume that ln AF 0.3 based on previous studies for deep soil sites. Arb =
2 for estimating aleatory uncertainty of arbitrary horizontal component.

T2 + C
207
Characterise sites using VS30 . Account for nonlinear effects using A1100 , median estimated PGA on reference rock outcrop (VS30 = 1100 m/s) in g. Linear part of fsite
is consistent with previous studies but with constraint for constant site term for VS30 >
1100 m/s (based on residual analysis) even though limited data for VS30 > 1100 m/s.
When only including linear part of shallow site response term nd residuals clearly exhibit bias when plotted against rock PGA, A1100 . Find that residuals not sufcient to
determine functional form for nonlinear amplication so use 1D equivalent-linear site
response simulations to constrain form and coefcients. Believe model applicable for
VS30 = 1501500 m/s.
Also use depth to 2.5 km/s shear-wave velocity horizon (basin or sediment depth) in km,
Z2.5 . Deep-basin term modelled based on 3D simulations for Los Angeles, San Gabriel
and San Fernando basins (southern California) calibrated empirically from residual analysis, since insufcient observational data for fully empirical study. Shallow-sediment
effects based on analysis of residuals. Note high correlation between VS30 and Z2.5 .
Provide relationships for predicting Z2.5 based on other site parameters. Believe model
applicable for Z2.5 = 010 km.
Use three faulting mechanism categories based on rake angle, :
RV Reverse and reverse-oblique. 30 < < 150 . 17 earthquakes. FRV = 1 and
FN M = 0.
NM Normal and normal-oblique. 150 < < 30 . 11 earthquakes. FN M = 1 and
FRV = 0.
SS Strike-slip. All other rake angles. 36 earthquakes. FRV = 0 and FN M = 0.
Use data from PEER Next Generation Attenuation (NGA) Flatle.
Select records of earthquakes located within shallow continental lithosphere (crust) in
a region considered to be tectonically active from stations located at or near ground
level and which exhibit no known embedment or topographic effects. Require that the
earthquakes have sufcient records to reliably represent the mean horizontal ground
motion (especially for small magnitude events) and that the earthquake and record is
considered reliable.
Exclude these data: 1) records with only one horizontal component or only a vertical
component; 2) stations without a measured or estimated VS30 ; 3) earthquakes without
a rake angle, focal mechanism or plunge of the P- and T-axes; 4) earthquakes with the
hypocentre or a signicant amount of fault rupture located in lower crust, in oceanic plate
or in a stable continental region; 5) LDGO records from the 1999 Dzce earthquake that
are considered to be unreliable due to their spectral shapes; 6) records from instruments
designated as low-quality from the 1999 Chi-Chi earthquake; 7) aftershocks but not
triggered earthquakes such as the 1992 Big Bear earthquake; 8) earthquakes with too
few records (N ) in relation to its magnitude, dened as: a) M < 5.0 and N < 5, b)
5.0 M < 6.0 and N < 3, c) 6.0 M < 7.0, RRU P > 60 km and N < 2 (retain singlyrecorded earthquakes with M 7.0 and RRU P 60 km because of their signicance);
9) records considered to represent non-free-eld site conditions, dened as instrument
located in a) basement of building, b) below the ground surface, c) on a dam except the
abutment; and 10) records with known topographic effects such as Pacoima Dam upper
left abutment and Tarzana Cedar Hill Nursery.
208
Functional forms developed or conrmed using classical data exploration techniques,

such as analysis of residuals. Candidate functional forms developed using numerous
iterations to capture the observed trends in the recorded ground motion data. Final functional forms selected according to: 1) sound seismological basis; 2) unbiased residuals;
3) ability to be extrapolated to magnitudes, distances and other explanatory variables
that are important for use in engineering and seismology; and 4) simplicity, although this
was not an overriding factor. Difcult to achieve because data did not always allow the
functional forms of some explanatory variables to be developed empirically. Theoretical
constraints were sometimes used to dene the functional forms.
Use two-stage maximum-likelihood method for model development but one-stage randomeffects method for nal regression.
Also perform statistical analysis for converting between selected denition of horizontal
component and other denitions.
Include depth to top of coseismic rupture plane, ZT OR , which nd important for reversefaulting events. Find that some strike-slip earthquakes with partial or weak surface
expression appeared to have higher-than-average ground motions but other strike-slip
events contradict this, which believe could be due to ambiguity in identifying coseismic
surface rupture in NGA database. Therefore, believe additional study required before
ZT OR can be used for strike-slip events. Believe model applicable for ZT OR = 015 km.
Include dip of rupture plane, . Believe model applicable for = 1590 .
Assume that is approximately equal to standard deviation of inter-event residuals,
ln Y , since inter-event terms are not signicantly affected by soil nonlinearity. Note that
if was subject to soil nonlinearity effects it would have only a relatively small effect
on T because intra-event dominates. takes into account soil nonlinearity effects.
Assume that ln Y and ln P GA represent aleatory uncertainty associated with linear site
response, reecting dominance of such records in database.
Based on statistical tests on binned intra-event residuals conclude that intra-event standard deviations not dependent on VS30 once nonlinear site effects are taken into account.
Use residual analysis to derive trilinear functional form for fmag . Piecewise linear relationship allows greater control of M > 6.5 scaling and decouples this scaling from that of
small magnitude scaling. Demonstrate using stochastic simulations that trilinear model
ts ground motions as well as quadratic model for M 6.5. Find that large-magnitude
scaling of trilinear model consistent with observed effects of aspect ratio (rupture length
divided by rupture width), which was abandoned as explanatory variable when inconsistencies in NGA database for this variable found.
Original unconstrained regression resulted in prediction of oversaturation at short periods, large magnitudes and short distances. Oversaturation not statistically signicant
nor is this behaviour scientically accepted and therefore constrain fmag to saturate
at M > 6.5 and RRU P = 0 when oversaturation predicted by unconstrained regression analysis. Constraint equivalent to setting c3 = c1 c2 c5 ln(c6 ). Inter- and
intra-event residual plots w.r.t. M show predictions relatively unbiased, except for larger
magnitudes where saturation constraint leads to overestimation of short-period ground
motions.
209
Examine inter-event residuals w.r.t. region and nd some bias, e.g. nd generally positive inter-event residuals at relatively long periods of M > 6.7 events in California but
only for ve events, which believe insufcient to dene magnitude scaling for this region.
Note that user may wish to take these dependences into account.
Note that adopted distance-dependence term has computational advantage since it
transfers magnitude-dependent attenuation term to outside square root, which significantly improves stability of nonlinear regression. Note that adopted functional form
consistent with broadband simulations for 6.5 and 7.5 between 2 and 100 km and with
simple theoretical constraints. Examine intra-event residuals w.r.t. distance and nd that
they are relatively unbiased.
Functional form for ff lt determined from residual analysis. Find coefcient for normal
faulting only marginally signicant at short periods but very signicant at long periods.
Believe long-period effects due to systematic differences in sediment depths rather than
source effects, since many normal-faulting events in regions with shallow depths to hard
rock (e.g. Italy, Greece and Basin and Range in the USA), but no estimates of sediment
depth to correct for this effect. Constrain normal-faulting factor found at short periods to
go to zero at long periods based on previous studies.
Functional form for fhng determined from residual analysis with additional constraints to
limit range of applicability so that hanging-wall factor has a smooth transition between
hanging and foot walls, even for small ZT OR . Include fhng,M , fhng,Z and fhng, to
phase out hanging-wall effects at small magnitudes, large rupture depths and large rupture dips, where residuals suggest that effects are either negligible or irresolvable from
data. Include hanging-wall effects for normal-faulting and non-vertical strike-slip earthquakes even those statistical evidence is weak but it is consistent with better constrained
hanging-wall factor for reverse faults and it is consistent with foam-rubber experiments
and simulations.
2.242 Danciu & Tselentis (2007a) & Danciu & Tselentis (2007b)
log10 Y = a + bM c log10
R2 + h2 + eS + f F
where Y is in cm/s2 , a = 0.883, b = 0.458, c = 1.278, h = 11.515, e = 0.038,

f = 0.116, = 0.109 (intra-event) and = 0.270 (inter-event).
B Rock, Vs,30 > 800 m/s. S = 0. 75 records.
C Stiff soil, 360 Vs 665 m/s. S = 1. 197 records.
D Soft soil, 200 Vs 360 m/s. S = 2. 63 records.
From initial analysis nd that ground-motions on D sites are double those on C sites.
Use three style-of-faulting categories:
Thrust F = 1
Strike-slip F = 1
210
Normal F = 0
From initial analysis nd that thrust and strike-slip ground motions are similar but greater
than normal motions.
Focal depths between 0 and 30 km with mean of 10.66 km.
Most records from earthquakes near the Ionian islands.
Use records from free-eld stations and from basements of buildings with < 2 storeys.
Note that some bias may be introduced by records from buildings but due to lack of data
from free-eld stations these records must be included.
Use corrected records from ISESD (bandpass ltered 0.25 and 25 Hz).
Use epicentral distance because most earthquakes are offshore and those that are onshore do not display evidence of surface faulting and therefore cannot use a fault-based
distance measure.
Data from large events recorded at intermediate and long distances and small events at
small distances. Correlation coefcient between magnitude and distance is 0.64.
Recommend that equation not used outside range of data used.
Analyse residuals normalized to have zero mean and unity variance (only display results
for PGA and SA at 1 s due to similar results for all periods). Find that residuals do not
show trends and are uncorrelated (at more than 99% condence level) w.r.t. independent
variables. Show normality of residuals through histograms for PGA and SA at 1 s.
Also derive equations for various other strong-motion parameters.
2.243
Douglas (2007)
log y = a1 + a2 M + a3 log
(d2 + 52 ) + a3+i Si
Coefcients not reported since purpose is not to develop models for seismic hazard
assessments but to derive condence limits on median PGA and thereafter to examine
possible regional dependence of ground motions.
Rederives models of Joyner & Boore (1981), Boore et al. (1993, 1997), Ambraseys
et al. (1996), Ambraseys et al. (2005a), Ulusay et al. (2004), Kalkan & Glkan (2004b)
and Sabetta & Pugliese (1987) to nd their complete covariance matrices in order to
compute condence limits of the predicted median PGA.
Uses same site classications as original studies. Si = 1 for site class i and 0 otherwise.
Adopts a simple linear functional form and standard one-stage regression method so
that the covariance matrices can be easily computed.
Assumes a xed coefcient of 5 km (a rough average value for this coefcient for most
models using adopted functional form) inside square root to make function linear.
211
Examines 95% condence limits on PGA since it is standard to use 5% signicance

levels when testing null hypotheses. Plots predicted median PGAs and their condence
limits for Mw 5, 6.5 and 8.0 up to 200 km to show effects of extrapolation outside range
of applicability of models. Finds that condence limits for models derived using limited
data (Ulusay et al. , 2004; Kalkan & Glkan, 2004b; Sabetta & Pugliese, 1987) are wider
than models derived using large well-distributed datasets (Joyner & Boore, 1981; Boore
et al. , 1993, 1997; Ambraseys et al. , 1996, 2005a). Notes that for 5.5 < Mw < 7
and 10 df 60 km the 95%-condence limits of the median are narrow and within
bands 1030% from the median but for other magnitudes and distances (away from the
centroid of data) they are much wider (bands of 100% from the median). Notes that
inclusion of data from large magnitude events decreases the width of the condence
limits of the model derived using the data of Boore et al. (1993, 1997) compared with
that derived using the data of Joyner & Boore (1981) and similarly that derived with
the data of Ambraseys et al. (2005a) compared with that derived using the data of
2.244 Hong & Goda (2007) & Goda & Hong (2008)
2
ln Y = b1 + b2 (M 7) + b3 (M 7)2 + [b4 + b5 (M 4.5)] ln[(rjb
+ h2 )0.5 ] + AFs
where Y is in g, b1 = 1.096, b2 = 0.444, b3 = 0.0, b4 = 1.047, b5 = 0.038, h = 5.7,

= 0.190 (inter-event) and = 0.464 (intra-event) for geometric mean.
AFs is the amplication factor due to linear and nonlinear soil behaviour used by Atkinson & Boore (2006), which is a function of Vs,30 and expected PGA at site with Vs,30 =
760 m/s, PGAref . Derive equation for PGAref of form ln PGAref = b1 + b2 (M 7) +
2 + h2 )0.5 ), where b = 0.851, b = 0.480, b = 0.884 and h = 6.3 km for
b4 ln((rjb
1
2
4
geometric mean ( not reported).
Use data from the PEER Next Generation Attenuation (NGA) database.
Investigate the spatial correlation of ground motions and their variabilities.
Generate datasets using normally distributed values of M (truncated at 2 standard deviations that are reported in the PEER NGA database) for earthquakes and lognormallydistributed values of Vs,30 (again using standard deviations from PEER NGA database)
for stations. Repeat regression analysis and nd coefcients very similar to those obtained ignoring the uncertainty in M and Vs,30 .
212
2.245
Graizer & Kalkan (2007) & Graizer & Kalkan (2008)
[(
]
)
R 2
2 R
ln(Y ) = ln(A) 0.5 ln 1
+ 4D0
R0
R0
(

)2
(
)
Vs,30
R
R
+ 4D12
+ bv ln
0.5 ln 1
R1
R1
VA
A = [c1 arctan(M + c2 ) + c3 ]F
R0 = c4 M + c5
D0 = c6 cos[c7 (M + c8 )] + c9
where Y is in g, c1 = 0.14, c2 = 6.25, c3 = 0.37, c4 = 2.237, c5 = 7.542, c6 =
0.125, c7 = 1.19, c8 = 6.15, c9 = 0.525, bv = 0.25, VA = 484.5, R1 = 100 km and
= 0.552.
Characterise sites by Vs,30 (average shear-wave velocity in upper 30 m). Note that approximately half the stations have measured shear-wave velocity proles.
Include basin effects through modication of D1 . For sediment depth (Z 1 km D1 =
0.35; otherwise D1 = 0.65.
Use three faulting mechanism classes:
Normal 13 records
Strike-slip 1120 records. F = 1.00.
Reverse 1450 records. F = 1.28 (taken from previous studies).
but only retain two (strike-slip and reverse) by combining normal and strike-slip categories.
Only use earthquakes with focal depths < 20 km. Focal depths between 4.6 and 19 km.
Exclude data from aftershocks.
Use data from: Alaska (24 records), Armenia (1 record), California (2034 records), Georgia (8), Iran (7 records) Italy (10 records), Nevada (8 records), Taiwan (427 records),
Turkey (63 records) and Uzbekistan (1 record).
Most data from 5.5 Mw 7.5.
Adopt functional form to model: a constant level of ground motion close to fault, a slope
of about R1 for > 10 km and R1.5 at greater distances (> 100 km) and observation
(and theoretical results) that highest amplitude ground motions do not always occur
nearest the fault but at distances of 310 km.
Choose functional form based on transfer function of a SDOF oscillator since this has
similar characteristics to those desired.
Note that magnitude scaling may need adjusting for small magnitudes.
213
Firstly regress for magnitude and distance dependency and then regress for site and
basin effects.
Examine residual w.r.t. magnitude and distance and observe no signicant trends.
Compare predictions to observations for 12 well-recorded events in the dataset and nd
that the observations are well predicted for near and far distances.
Demonstrate (for the 2004 Parkeld earthquake) that it is possible to add an additional
lter term in order to predict ground motions at large distances without modifying the
other terms.
2.246 Massa et al. (2007)

log10 (Y ) = a + bML + c log(R) + dSsoil

where Y is in g, a = 3.2191 0.16, b = 0.7194 0.025, c = 1.7521 0.075,
d = 0.1780 and = 0.282.
Originally use three site classes based on Eurocode 8:
A Rock, Vs,30 > 800 m/s. Marine clay or other rocks (Lower Pleistocene and Pliocene),
volcanic rock and deposits. 11 stations. 833 records.
B Stiff soil, 360 < Vs,30 < 800 m/s. Colluvial, alluvial, lacustrine, beach, uvial
terraces, glacial deposits and clay (Middle-Upper Pleistocene). Sand and loose
conglomerate (Pleistocene and Pliocene). Travertine (Pleistocene and Holocene).
6 stations. 163 records.
C Soft soil, Vs,30 < 360 m/s. Colluvial, alluvial, lacustrine, beach and uvial terrace
deposits (Holocene). 3 stations. 67 records.
Classify stations using geological maps. Find that results obtained using this classication are not realistic because of some stations on very thick (> 1000 m) sedimentary
deposits whose amplication factors are small. Therefore, use two site classes using
H/V ratios both using noise and earthquake records. Conrm H/V results by computing
magnitude residuals at each station.
Final site classes are:
Rock Site amplication factors < 2 at all considered frequencies from H/V analysis. 422
records. Ssoil = 0.
Soil Site amplication factors > 2. 641 records. Ssoil = 1.
Use data from velocimeters (31 stations) and accelerometers (2 stations) from 33 sites
with sampling rates of 62.5 samples/s.
Relocate events and calculate ML .
Exclude data from ML < 2.5 and rhypo > 300 km.
Few near-source records (rhypo < 150 km) from ML > 4 but for ML < 4 distances from
0 to 300 km well represented.
214
Exclude records with signal-to-noise ratios < 10 dB.

Correct for instrument response and bandpass lter between 0.5 and 25 Hz and then the
velocimetric records have been differentiated to obtain acceleration.
Visually inspect records to check for saturated signals and noisy records.
Compare records from co-located velocimetric and accelerometric instruments and nd
that they are very similar.
Compare PGAs using larger horizontal component, geometric mean of the two horizontal components and the resolved component. Find that results are similar and that the
records are not affected by bias due to orientation of sensors installed in eld.
Try including a quadratic magnitude term but nd that it does not reduce uncertainties
and therefore remove it.
Try including an anelastic attenuation term but nd that the coefcient is not statistically
signicant and that the coefcient is positive and close to zero and therefore remove this
term.
2 + h2 rather than c log (R) but nd that h is not well

Repi
10
constrained and hence PGAs for distances < 50 km underpredicted.
Try using a term c log10
Find that using a maximum-likelihood regression technique leads to very similar results
to the one-stage least-squares technique adopted, which relate to lack of correlation
between magnitudes and distances in dataset.
Find site coefcients via regression following the derivation of a, b and c using the 422
rock records.
Compare observed and predicted ground motions for events in narrow (usually 0.3 units)
magnitude bands. Find good match.
Examine residuals w.r.t. magnitude and distance and nd no signicant trends except for
slight underestimation for short distances and large magnitudes. Also check residuals
for different magnitude ranges. Check for bias due to non-triggering stations.
Compare predicted PGAs to observations for 69 records from central northern Italy from
magnitudes 5.06.3 and nd good match except for rhypo < 10 km where ground motions overpredicted, which relate to lack of near-source data.
2.247
Popescu et al. (2007)
log A = C1 Mw + C2 log R + C3
where A in in cm/s2 , C1 = 0.80 0.05, C2 = 0.30 0.08, C3 = 2.93 and = 0.314
using repi and C1 = 0.79 0.05, C2 = 0.89 0.38, C3 = 1.43 and = 0.341 using
rhypo .
Adjust observations by multiplicative factor S to account for site conditions (0.8 S 1
for hard rocks, 0.7 S 0.8 for thin sedimentary layers and 0.65 S 0.7 for thick
sedimentary cover.
215

Data from digital strong-motion network (K2 instruments) from 1997 to 2000 (4 Mw
6) plus data (SMA-1) from 30th August 1986 (Mw 7.1) and 30th and 31st May 1990
(Mw 6.9 and 6.4) earthquakes.
Regression in two steps: a) dependence on Mw found and then b) dependence on R is
found (details on this procedure are not given).
Also regress using just K2 data (log A = 0.94 0.09Mw 1.01 0.42 log R 1.84,
= 0.343) and using repi (log A = 0.890.09Mw 0.280.09 log 3.35, = 0.322).
Note that correlation coefcients are higher and s are lower when all data is used and
that match (based on relative residuals) to data from 1986 and 1990 earthquakes is
better when all data is used.
Present relative residuals for sites in epicentral area and in Bucharest. Note that for 63%
of earthquakes relative errors are < 50% for at least one station; for 43% of earthquake
relative errors are < 30% for at least one station; and for 9 earthquakes relative errors
are smaller than 10% for at least one station (BMG, the extreme site). Based on this
analysis it is concluded that predictions more reliable in far-eld than in epicentral area.
Also nd that largest absolute residuals are for MLR (stiff rock).
Note largest relative errors are for 4 Mw 4.5.
2.248 Sobhaninejad et al. (2007)

log y = a1 +a2 Mw +(a3 +a4 M w) log
2 + a2 +a S +a S +a F +a F +a F
rjb
6 S
7 A
8 N
9 T
10 O
5

a6 = 0.186, a7 = 0.125, a8 = 0.082, a9 = 0.012 and a10 = 0.038 (do not report
but unbiased mean square error) for horizontal PGA; and a1 = 0.495, a2 = 0.027,
a3 = 2.83, a4 = 0.235, a5 = 7.181, a6 = 1.150, a7 = 1.103, a8 = 0.074, a9 = 0.065
and a10 = 0.170 (do not report but unbiased mean square error).
Use four faulting mechanisms:
Normal FN = 1, FT = 0, FO = 0.
Strike-slip FN = 0, FT = 0, FO = 0.
Thrust FT = 1, FN = 0, FO = 0.
Odd FO = 1, FN = 0, FT = 0.
Use same data and functional form as Ambraseys et al. (2005a) and Ambraseys et al.
(2005b) but exclude six records that were not available.
216
Use genetic (global optimization) algorithm to nd coefcients so as to nd the global

(rather than a local) minimum. Use the unbiased mean square error as the error (cost
or tness) function in the algorithm. Use 20 chromosomes as initial population, besttness selection for offspring generation, uniform random selection for mutation of chromosomes and heuristic crossover algorithm for generation of new offspring.
Find smaller (by 26% for horizontal and 16.66% for vertical) unbiased mean square error
than using standard regression techniques.
2.249
Tavakoli & Pezeshk (2007)
log10 y = 1 + 2 M + 3 M 2 + 4 R + 5 log10 (R + 6 107 M )

where y is in cm/s2 , 1 = 3.4712, 2 = 2.2639, 3 = 0.1546, 4 = 0.0021, 5 =
1.8011, 6 = 0.0490, 7 = 0.2295, r = 0.2203 (intra-event) and e = 0.2028 (interevent).
All records from rock sites.
Strong correlation between magnitude and distance in dataset.
Use a derivative-free approach based on a hybrid genetic algorithm to derive the model.
Use a simplex search algorithm to reduce the search domain to improve convergence
speed. Then use a genetic algorithm to obtain the coefcients and uncertainties using
one-stage maximum-likelihood estimation. Believe that approach is able to overcome
shortcomings of previous methods in providing reliable and stable solutions although it
is slower.
In hybrid genetic algorithm an initial population of possible solutions is constructed in
a random way and represented as vectors called strings or chromosomes of length
determined by number of regression coefcients and variance components. Population
size is usually more than twice string length. Each value of population array is encoded
as binary string with known number of bits assigned according to level of accuracy or
range of each variable. Use three operations (reproduction/selection, crossover and
mutation) to conduct directed search. In reproduction phase each string assigned a
tness value derived from its raw performance measure given by objective function.
Probabilities of choosing a string is related to its tness value. Crossover or mating
combines pairs of strings to create improved strings in next population. In mutation one
or more bits of every string are altered randomly. The process is then repeated until a
termination criterion is met. Demonstrate approach using test function and nd small
maximum bias in results. Conclude that method is reliable.
Use Taiwanese dataset of Chen & Tsai (2002) to demonstrate method.
Compare results with those obtained using methods of Brillinger & Preisler (1985),
Joyner & Boore (1993) and Chen & Tsai (2002). Find differences in coefcients (although predictions are very similar except at edges of dataspace) and standard deviations (slightly lower for proposed method).
Compare predicted motions for ML 5.5 with observations for ML 56. Find good t.
217
Plot total residuals against magnitude and distance and nd no trends.

Note that residuals show that model is satisfactory up to 100 km but for larger distances
assumption of geometric spreading of body waves in not appropriate due to presence of
waves reected off Moho.
Note that near-source saturation should be included. Apply proposed method using a
complex functional form with different equations for three distance ranges and compare
results to those using simple functional form. Find differences at short and large distances.
2.250 Tejeda-Jcome & Chvez-Garca (2007)

ln A = c1 + c2 M c3 ln h c4 ln R
where A is in cm/s2 , c1 = 0.5342, c2 = 2.1380, c3 = 0.4440, c4 = 1.4821 and
= 0.28 for horizontal PGA and c1 = 0.5231, c2 = 1.9876, c3 = 0.5502, c4 = 1.4038
and = 0.27 for vertical PGA.
Most stations on rock or rm ground. 4 instruments (from close to coast) installed on
sandy or silty-sandy soils. Not enough data to correct for site effects or derive site
coefcients. Check residuals (not shown) for each station and nd no systematic bias.
Focal depths h between 3.4 and 76.0 km (most < 40 km). No correlation between h and
repi .
Use data from 12 (5 Etnas and 7 GSR-18s) temporary and 5 permanent strong-motion
stations.
Since data from digital instruments only apply baseline correction.
Exclude data from 3 events only recorded at 3 stations.
Relocate earthquakes because of poor locations given by agencies. Recompute ML
from accelerograms.
Inclusion of h leads to less scatter but note need for larger database to better understand
effect of h.
Examine residuals w.r.t. distance and nd no trend or bias.
218
2.251
Abrahamson & Silva (2008) & Abrahamson & Silva (2009)

ln Sa( g)
\
f1 (M, Rrup ) + a12 FRV + a13 FN M + a15 FAS + f5 (PGA
1100 , VS30 )
+FHW f4 (Rjb , Rrup , Rx , W, , ZT OR , M ) + f6 (ZT OR ) + f8 (Rrup , M )
f1 (M, Rrup )
\
f5 (PGA
1100 , VS30 )
where VS30
and V1
f4 (Rjb , Rrup , , ZT OR , M, W )
where T1 (Rjb )
T2 (Rx , W, )
T3 (Rx , ZT OR )
T4 (M )
T5 ()
f6 (ZT OR )
f8 (Rrup , M )
where T6 (M )
f10 (Z1.0 , VS30 )
1.0 (VS30 )]
ln[Z
a21
e2
a22
where
+f10 (Z1.0 , VS30 )

{
a1 + a4 (M c1 ) + a8 (8.5 M )2 + [a2 + a3 (M c1 )] ln(R) for M c1
a1 + a5 (M c1 ) + a8 (8.5 M )2 + [a2 + a3 (M c1 )] ln(R) for M > c1
2
Rrup
+ c24
( )
V
\
a10 ln V S30 b ln(PGA
1100 + c)
( LIN
( )n )
V
S30
\
+b ln PGA
for VS30 < VLIN
1100 + c V
( ) LIN
(a10 + bn) ln VS30

for VS30 VLIN
VLIN
{
VS30 for VS30 < V1
V1 for VS30 V1
1500 for T 0.50 s
exp[8.0 0.795 ln(T /0.21)] for 0.50 < T 1 s

exp[6.76 0.297 ln(T )] for 1 < T < 2 s
700 for T 2 s
a14 T1 (Rjb )T2 (Rx , W, )T3 (Rx , ZT OR )T4 (M )T5 ()
{
Rjb
for Rjb < 30 km
1 30
0 for Rjb 30 km
{
x
0.5 + 2W Rcos()
for Rx W cos()
1 for Rx > W cos() or = 90
{
1 for Rx ZT OR
Rx
for Rx < ZT OR
ZT OR
0 for M 6
M 6 for 6 < M < 7
1 for M 7
{
1 30
for 30
60
1 for < 30
{ a Z
16 T OR
for ZT OR < 10 km
10
a16 for ZT OR 10 km
{
0 for Rrup < 100 km
a18 (Rrup 100)T6 (M ) for Rrup 100 km
1 for M < 5.5

0.5(6.5 M ) + 0.5 for 5.5 M 6.5
0.5 for M > 6.5

(
) {
)
(
1.0
Z1.0 + c2
for Z1.0 200
a22 ln Z200
a21 ln
+
1.0 (VS30 ) + c2
Z
0 for Z1.0 < 200
6.745 for VS30
( < 180
) m/s
S30
6.745 1.35 ln V180
for 180 VS30 500 m/s
(
)
VS30
5.394 4.48 ln 500

for VS30 > 500 m/s
)
0 for VS30( 1000
VS30
(a10 +bn) ln min(V

)
(
)
(
,1000)
V
Z1.0 +c2
(
)1
<0
for (a10 + bn) ln min(VS30
+
e
ln
2
,1000)
Z
+c
Z1.0 +c2
1
ln 1.0 2
Z1.0 +c2
e2 otherwise
0 for T( < 0.35
) s (or V)S30 > 1000
S30
0.25 ln V1000
ln T
for 0.35 T 2 s
(
) ( 0.35)
0.25 ln VS30 ln 2
for T > 2 s
1000
0.35
{
0 for T < 2 s
0.0625(T 2) for T 2 s
219
The model for the standard deviation is:
2
B (M, T ) =
02 (M, T ) Amp
(T )
1/2
2
2
B (M, T ) + Amp
(T )
(
)
2
\
PGA
1100 ,VS30 )
2
+ ln Amp(T,
B
(M, PGA)
ln PGA1100
\
)
(
(T, M, PGA1100 , VS30 ) =
\
PGA
1100 ,VS30 )
+2 ln Amp(T,
ln PGA1100
B (M, T )B (M, PGA)/ (T, PGA)
0 for VS30 VLIN

\
ln Amp(T, PGA
1100 , VS30 )
\
\
b(T )PGA
b(T )PGA
=
1100
( 1100 ) n
+ \
for VS30 < VLIN
V
PGA
ln PGA1100
\
PGA1100 +c V S30
1100 +c
LIN
s1 for
( M1 )< 5
s1 + s2 s
(M 5) for 5 M 7
0 (M ) =
2
s2 for M > 7
s3 for
( M3 )< 5
s3 + s4 s
(M 5) for 5 M 7
0 (M ) =
2
s4 for M > 7
1100 is median peak acceleration for VS30 = 1100 m/s, B and

where Sa is in g, PGA
B (= 0 (M, T )) are intra-event and inter-event standard deviations, 0 and 0 are intraevent and inter-event standard deviations of the observed ground motions for low levels
of outcrop rock motions (directly from regression), amp is intra-event variability of the
site amplication factors (assumed equal to 0.3 for all periods based on 1D site response
results), c1 = 6.75, c4 = 4.5, a3 = 0.265, a4 = 0.231, a5 = 0.398, N = 1.18,
c = 1.88, c2 = 50, VLIN = 865.1, b = 1.186, a1 = 0.804, a2 = 0.9679, a8 =
0.0372 ,a10 = 0.9445, a12 = 0.0000, a13 = 0.0600, a14 = 1.0800, a15 = 0.3500,
a16 = 0.9000, a18 = 0.0067, s1 = 0.590 and s2 = 0.470 for VS30 estimated, s1 = 0.576
and s2 = 0.453 for VS30 measured, s3 = 0.470, s4 = 0.300 and (T, PGA) = 1.000.
Characterise sites using VS30 and depth to engineering rock (Vs = 1000 m/s), Z1.0 .
Prefer Vs,30 to generic soil/rock categories because it is consistent with site classication
in current building codes. Note that this does not imply that 30 m is key depth range for
site response but rather that Vs,30 is correlated with entire soil prole.
Classify events in three fault mechanism categories:
FRV = 1, FN M = 0 Reverse, reverse/oblique. Earthquakes dened by rake angles between 30 and

150 .
FRV = 0, FN M = 1 Normal. Earthquakes dened by rake angles between 60 and 120 .
FRV = 0, FN M = 0 Strike-slip. All other earthquakes.
Believe that model applicable for 5 Mw 8.5 (strike-slip) and 5 Mw 8.0 (dip-slip)
and 0 dr 200 km.
Use simulations for hard-rock from 1D nite-fault kinematic source models for 6.5
Mw 8.25, 3D basin response simulations for sites in southern California and equivalentlinear site response simulations to constrain extrapolations beyond the limits of the empirical data.
Select data from the Next Generation Attenuation (NGA) database (at-le version 7.2).
Include data from all earthquakes, including aftershocks, from shallow crustal earthquakes in active tectonic regions under assumption that median ground motions from
220
shallow crustal earthquakes at dr < 100 km are similar. This assumes that median
stress-drops are similar between shallow crustal events in: California, Alaska, Taiwan,
Japan, Turkey, Italy, Greece, New Zealand and NW China. Test assumption by comparing inter-event residuals from different regions to those from events in California. Since
aim is for model for California and since difference in crustal structure and attenuation
can affect ground motions at long distances exclude data from dr > 100 km from outside
western USA.
Also exclude these data: events not representative of shallow crustal tectonics, events
missing key source metadata, records not representative of free-eld motion, records
without a Vs,30 estimate, duplicate records from co-located stations, records with missing
horizontal components or poor quality accelerograms and records from western USA
from dr > 200 km.
Classify earthquakes by event class: AS (aftershock) (FAS = 1); MS (mainshock), FS
(foreshock) and swarm (FAS = 0). Note that classications not all unambiguous.
Use depth-to-top of rupture, ZT OR , fault dip in degrees, and down-dip rupture width,
W.
Use rjb and Rx (horizontal distance from top edge of rupture measured perpendicular
to fault strike) to model hanging wall effects. For hanging wall sites, dened by vertical
projection of the top of the rupture, FHW = 1. T1 , T2 and T3 constrained by 1D rock
simulations and the Chi-Chi data. T4 and T5 constrained by well-recorded hanging wall
events. Only a14 was estimated by regression. State that hanging-wall scaling is one of
the more poorly-constrained parts of model14 .
Records well distributed w.r.t. Mw and rrup .
For four Chi-Chi events show steep distance decay than other earthquakes so include
a separate coefcient for the ln(R) term for these events so they do not have a large
impact on the distance scaling. Retain these events since important for constraining
other aspects of the model, e.g. site response and intra-event variability.
Only used records from 5 M 6 to derive depth-to-top of rupture (ZT OR ) dependence to limit the effect on the relation of the positive correlation between ZT OR and
M.
Constrain (outside the main regression) the large distance (Rrup > 100 km) attenuation
for small and moderate earthquakes (4 M 5) using broadband records of 3 small
(M 4) Californian earthquakes because limited data for this magnitude-distance range in
NGA data set.
Note difcult in developing model for distinguishing between shallow and deep soil sites
due to signicant inconsistencies between VS30 and depth of soil (Z1.0 ), which believe
to be unreliable in NGA Flat-File. Therefore, develop soil-depth dependence based on
1D (for Z1.0 < 200 m) and 3D (for Z1.0 > 200 m) site response simulations. Motion for
shallow soil sites do not fall below motion for VS30 = 1000 m/s.
14
Model for T5 reported here is that given in 2009 errata. In original reference: T5 = 1 ( 70)/20 for 70
and 1 otherwise).
221
TD denotes period at which rock (VS30 = 1100 m/s) spectrum reaches constant displacement. Using point-source stochastic model and 1D rock simulations evaluate magnitude dependence of TD as log10 (TD ) = 1.25 + 0.3M . For T > TD compute rock
spectral acceleration at TD and then scale this acceleration at TD by (TD /T )2 for constant spectral displacements. The site response and soil depth scaling is applied to
T2
1100 , VS30 , T ) +
this rock spectral acceleration, i.e. Sa(TD , VS30 = 1100) TD2 + f5 (PGA
f10 (Z1.0 , VS30 , T ).
Reduce standard deviations to account for contribution of uncertainty in independent
parameters M , Rrup , ZT OR and VS30 .
Note that regression method used prevents well-recorded earthquakes from dominating
regression.
Examine inter-event residuals and nd that there is no systemic trend in residuals for
different regions. Find that residuals for M > 7.5 are biased to negative values because
of full-saturation constraint. Examine intra-event residuals and nd no signicant trend
in residuals.
Although derive hanging-wall factor only from reverse-faulting data suggest that it is
applied to normal-faulting events as well.
State that should use median PGA1100 for nonlinear site amplication even if conducting
a seismic hazard analysis for above median ground motions.
State that if using standard deviations for estimated VS30 and VS30 is accurate to within
30% do not need to use a range of VS30 but if using measured-VS30 standard deviations
then uncertainty in measurement of VS30 should be estimated by using a range of VS30
values.
State that if do not know Z1.0 then use median Z1.0 estimated from equations given and
do not adjust standard deviation.
2.252 gstsson et al. (2008)

log10 (acceleration) = a log10 (R) + b log10 (M ) + c

where acceleration is in m/s2 , a = 1.95600, b = 9.59878, c = 4.87778 and =
0.4591, and:
log10 (acceleration) = a log10 (R) + bM + c

where a = 1.96297, b = 0.89343, c = 2.65660 and = 0.4596.
Select data from SIL database with MLw > 3.5 in latitude range 63.5 to 64.3 N and
longitude range 18 to 23.5 W between July 1992 and April 2007.
Exclude data where several earthquakes are superimposed and retain only clean waveforms.
Most data from 5 Hz Lennarz seismometers. Some from 1 Hz and long-period instruments. Sampling frequency is 100 Hz.
222
Use data from SW Iceland plus data from Reykjanes Ridge and Myrdalsjokull volcano.
Investigate decay in several individual earthquakes and t equations of form log y =
a log R + b. Note that relations are well behaved so t entire dataset.
223
2.253 Aghabarati & Tehranizadeh (2008)

ln y
where for
and for
Mw c0
f1 (Mw )
= c1 + f1 (Mw ) + f2 (Mw )f3 (R) + f4 (F ) + FRf5 (ZF R ) +

FSf6 (ZF R ) + f7 (HW, RJB , Mw , DIP) +
f8 (Vs,30 , Vlin , PGAnonlin , PGArock ) + f9 (Vs,30 , Z1.5 )
= c3 (Mw c0 ) + c8 (T )(8.5 Mw )n
f2 (Mw ) = c2 (T ) + c4 (Mw c0 )
Mw > c0
f1 (Mw ) = c5 (Mw c0 ) + c8 (T )(8.5 Mw )n
= c2 (T ) + c6 (Mw c0 )
2
f3 (R) = ln Rrup
+ c7 (T )2
f2 (Mw )
f4 (F )
f5 (ZFR )
f6 (ZFS )
g1 (RJB )
g2 (Mw )
g3 (DIP)
= c9 (T )FR + c10 (T )FS + c11 (T )FN
0
Ztop 2 km
2 < Ztop 5 km
c12 (T )(Ztop 2)/3
c12 (T )
5 < Ztop 10 km
=
c12 (T )[1 (Ztop 10)/5] 5 < Ztop 10 km(sic)
0
Ztop > 10 km
c13 (T )Ztop /2
0 < Ztop 2 km
c13 (T )
2 < Ztop 4 km
=
c13 (T )[1 (Ztop 4)/2] 4 < Ztop 6 km
0
Ztop > 6 km
0 RJB < 15 km
1 RJB /45
2
(2 RJB /15) 15 RJB < 30 km
=
3
0
RJB 30 km
Mw < 6.0
0
2(Mw 6) 6.0 Mw < 6.5
=
1
Mw 6.5
{
1 (DIP 70)/20 DIP 70
=
1
DIP < 70
f7 (HW, RJB , Mw , DIP) = c14 (T )HWg1 (RJB )g2 (Mw )g3 (DIP)
f8 (Vs,30 , Vlin , PGAnonlin , PGArock ) = g4 (Vs,30 , Vlin ) + g5 (PGAnonlin , PGArock )
g4 (Vs,30 , Vlin ) = c15 (T ) ln(Vs,30 /Vlin )
c16 (T ) ln(PGAmin /0.1)

PGAnonlin < a1
c16 (T )[ln(PGAmin /0.1)

+a ln(PGAnonlin /a1 )
g5 (PGAnonlin , PGArock ) =
+b(ln(PGAnonlin /a1 ))2 ] a1 PGAnonlin a2
c16 (T ) ln(PGAnonlin /0.1) PGAnonlin a2

+ g7 (ZD , Z1.5 )
f9 (Vs,30 , Z1.5 ) = g6 (Vs,30 , Z1.5 , Z)
= c17 (T )(1/Z)
ln(Vs,30 /1500) ln(Z1.5 )
g6 (Vs,30 , Z1.5 , Z)
g7 (Z1.5 , ZD ) = ZD c18 (T )K1 (1 exp((Z1.5 200)/300)) +
ZD c19 (T )K2 (1 exp((Z1.5 200)/4000))
where y is in g, c1 = 1.81, c2 = 1.18, c7 = 8.647, c8 = 0.028, c9 = 0.176,

c10 = 0.266, c11 = 0.476, c12 = 0.52, c13 = 0.32, c14 = 0.4, c15 = 0.36, c17 = 0,
c18 = 0, c19 = 0, c20 = 0.496, c21 = 0.427, K1 = 2.260, K2 = 1.04, Vlin = 760,
= c20 (T ) + [c21 (T ) c20 (T )]Mw for 5.0 Mw < 7.0 and = c21 (T ) for Mw 7.0.
224
Use Vs,30 to characterize site conditions.

Characterize basin by depth to Vs = 1500 m/s, Z1.5 , since more likely to be obtained for
engineering projects.
Use three mechanism classes:
1. Normal. 34 records. FN = 1, FS = FR = 0.
2. Strike-slip. 184 records. FS = 1, FN = FR = 0.
3. Reverse. Originally classify as thrust, reverse and reverse oblique but combine.
423 records. FR = 1, FN = FS = 0.
Note lack of records from normal earthquakes.
Use data from earthquakes with focal depths 15 km.
Only use data from instrument shelters, non-embedded buildings with < 3 stories (< 7
if located on rm rock) and dam abutments (to enhance database even though could be
some interaction with dam).
Not sufcient data to investigate effect of tectonic environment. Exclude data from subduction zones because that is different tectonic regime than for shallow crustal earthquakes.
Data well distributed in magnitude-distance space so do not use special statistical procedures to decouple source and path effects. Do not use weights due to uniform distribution w.r.t. Mw and distance.
Exclude data from > 60 km to avoid records with multiple reections from lower crust.
Vast majority of data from western USA. Some from Alaska, Canada, Greece, Iran, Italy,
Japan, Mexico, New Zealand and Turkey.
Constrain c7 (T ) to be monotonically varying with period because otherwise can have
large changes in spectral shape at very short distances.
Note that for Mw < 5.8 magnitude dependence may be due to depth-to-top (ZFR and
ZFS ) effects since small earthquakes have on average larger depth-to-top than larger
earthquakes. Inter-event residuals from preliminary regression are functions of rake
and depth-to-top (stronger than rake dependency) particularly for reverse earthquakes.
These observations inuence functional form of f5 (Z).
Use residuals from 1D simulations to dene functional form for hanging wall effect
(HW = 1).
Coefcients for nonlinear soil effects determined from analytical results because of correlations between other parameters and nonlinearity and since analytical results better
constrained at high amplitudes than empirical data. Set a1 = 0.04 g, a2 = 0.1 g and
PGAmin = 0.06 g. PGAnonlin is expected PGA on rock (Vs,30 = 760 m/s). c15 (T ),
c16 (T ) and Vlin taken from Choi & Stewart (2005) and are not determined in regression.
Applied limited smoothing (using piecewise continuous linear ts on log period axis) to
avoid variability in predicted spectral ordinates for neighbouring periods particularly at
large magnitudes and short distances.
225
Examine normalized inter- and intra-event residuals w.r.t. Mw and distance (shown).
Find no bias nor trends. Also plot against mechanism, site and other parameters and
nd no bias nor trends (not shown).
2.254 Cauzzi & Faccioli (2008), Cauzzi (2008) & Cauzzi et al. (2008)
log10 = a1 + a2 Mw + a3 log10 R + aB SB + aC SC + aD SD
where y is in m/s2 , a1 = 1.296, a2 = 0.556, a3 = 1.582, aB = 0.22, aC = 0.304,
aD = 0.332 and = 0.344 for horizontal PGA.
Use four site categories based on Eurocode 8:
A Rock-like. Vs,30 800 m/s. SB = SC = SD = 0.
B Stiff ground. 360 Vs,30 < 800 m/s. SB = 1, SC = SD = 0.
C 180 Vs,30 < 360 m/s. SC = 1, SB = SD = 0.
D Very soft ground. Vs,30 < 180 m/s. SD = 1, SB = SC = 0.
Try to retain only records from stations of known site class but keep records from stations
of unknown class (4% of total), which assume are either B or C classes. Use various
techniques to extend 20 m proles of K-Net down to 30 m. Vast majority of data with
Vs,30 500 m/s.
Use mechanism classication scheme of Boore & Atkinson (2007) based on plunges of
P-, T- and B-axes:
Normal 16 earthquakes. 5 Mw 6.9.
Strike-slip 32 earthquakes. 5 Mw 7.2.
Reverse 12 earthquakes. 5.3 Mw 6.6.
Develop for use in displacement-based design.
Select records with minimal long-period noise so that the displacement ordinates are
reliable. Restrict selection to digital records because their displacement spectra are not
signicantly affected by correction procedure and for which reliable spectral ordinates
up to at least 10 s are obtainable. Include 9 analogue records from 1980 Irpinia (Mw 6.9)
earthquake after careful scrutiny of long-period characteristics.
Use approach of Paolucci et al. (2008) to estimate cut-off frequencies for bandpass
ltering. Compute noise index IV for each record based on PGV and average value
computed from coda of velocity time-history. Compare IV with curves representing as a
function of Mw the probability P that the long-period errors in the displacement spectrum
are less than a chosen threshold. Use probability P 0.9 and drifts in displacement
spectrum < 15% using IV from geometric mean. Rejections closely correlated with
instrument type (less data from high-bit instruments rejected than from low-bit instruments). Process records by removing pre-even offset from entire time-history. Following
this 57% of records satised criterion of Paolucci et al. (2008). Remaining records ltered using fourth-order acausal lter with cut-off 0.05 Hz after zero padding and cosine
tapering. After this step records pass criterion of Paolucci et al. (2008). Note that
ltering of 43% of records may affect reliability beyond 15 s.
226
Use data from K-Net and Kik-Net (Japan) (84%); California (5%); Italy, Iceland and
Turkey (5%); and Iran (6%). Try to uniformly cover magnitude-distance range of interest.
All data from M > 6.8 are from events outside Japan.
Exclude data from Mw < 5 because probabilistic seismic hazard deaggregation analyses show contribution to spectral displacement hazard from small events is very low.
Exclude data from Mw > 7.2 because 7.2 is representative of the largest estimated
magnitude in historical catalogue of Italy. Most records from Mw 6.6.
Exclude data from subduction zone events.
Focal depths between 2 and 22 km. Exclude earthquakes with focal depth > 22 km to
be in agreement with focal depths of most Italian earthquakes.
Use rhypo for greater exibility in seismic hazard analyses where source zones have
variable depth. Exclude data from rhypo > 150 km based on deaggregation results.
Test regional dependence of ground motions using analysis of variance. Divide dataset
into intervals of 10 km 0.3Mw units and consider only bins with 3 records. Apply
analysis for 18 bins on logarithmically transformed ground motions. Transform observed
motions to site class A by dividing by site amplication factor derived by regression. Find
no strong evidence for regional dependence.
Apply pure error analysis to test: i) standard logarithmic transformation, ii) magnitudedependence of scatter and iii) lower bound on standard deviation using only M and
rhypo . Divide dataset into bins of 2 km 0.2Mw units and consider only bins with 2
records (314 in total). Compute mean and standard deviation of untransformed ground
motion and calculate coefcient of variation (COV). Fit linear equation to plots of COV
against mean. Find no signicant trend for almost all periods so conclude logarithmic
transformation is justied for all periods. Compute standard deviation of logarithmicallytransformed ground motions and t linear equations w.r.t. Mw . Find that dependence
of scatter on magnitude is not signicant. Compute mean standard deviation of all bins
and nd limit on lowest possible standard deviation using only Mw and rhypo .
Aim for simplest functional form and add complexity in steps, checking the statistical
signicance of each modication and its inuence on standard error. Try including an
anelastic term, quadratic Mw dependence and magnitude-dependent decay term but
nd none of these is statistically signicant and/or leads to a reduction in standard deviation.
Try one-stage maximum likelihood regression but nd higher standard deviation so reject
it. Originally use two-stage approach of Joyner & Boore (1981).
Find that coefcients closely match a theoretical model at long periods.
Consider style-of-faulting by adding terms: aN EN +aR ER +aS ES where Ex are dummy
variables for normal, reverse and strike-slip mechanisms. Find that reduction in standard
deviation is only appreciable for limited period ranges but keep terms in nal model.
Replace terms: aB SB + aC SC + aD SD by bV log10 (Vs,30 /Va ) so that site amplication
factor is continuous. Vs,30 available for about 85% of records. To be consistent between
both approaches constrain Va to equal 800 m/s. Find bV closely matches theoretical
values 1 close to resonance period and 0.5 at long periods.
227
Examine residuals w.r.t. rhypo and Mw . Find no trends.
2.255 Chiou & Youngs (2008)

[ (
) ]
VS30
ln(y) = ln(yref ) + 1 min ln
,0
1130
+ 2 {e
{
+ 5 1
3 [min(VS30 ,1130)360]
3 (1130360)
(
} ln
yref e + 4
4
}
1
cosh[6 max(0, Z1.0 7 )]
8
+
cosh[0.15 max(0, Z1.0 15)]
ln(yref ) = c1 + [c1a FRV + c1b FN M + c7 (ZT OR 4)](1 AS)
c2 c3
+ [c10 + c7a (ZT OR 4)]AS + c2 (M 6) +
ln[1 + ecn (cM M ) ]
cn
+ c4 ln{RRU P + c5 cosh[c6 max(M cHM , 0)]}
2
2
+ (c4a c4 ) ln( RRU
P + cRB )
}
{
1
RRU P
+ c1 +
cosh[max(M c3 , 0)]
)
(
2 + Z2
2
RJB
T OR
RX cos
1
+ c9 FHW tanh
c9a
RRU P + 0.001
2 1
= 1 +
[min{max(M, 5), 7} 5]
2
}
{
2 1
[min(max(M, 5), 7) 5] + 4 AS
=
1 +
2
(3 FInf erred + 0.7FM easured ) + (1 + NL)2

(
)
yref e
where NL =
b
yref e + c
T2
2
= (1 + NL0 )2 2 + NL
0
where y is in g, c2 = 1.06, c3 = 3.45, c4 = 2.1, c4a = 0.5, cRB = 50, cHM = 3, c3 =

4, c1 = 1.2687, c1a = 0.1, c1b = 0.2550, cn = 2.996, cM = 4.1840, c5 = 6.1600,
c6 = 0.4893, c7 = 0.0512, c7a = 0.0860, c9 = 0.7900, c9a = 1.5005, c10 = 0.3218,
c1 = 0.00804, c2 = 0.00785, 1 = 0.4417, 2 = 0.1417, 3 = 0.007010,
4 = 0.102151, 5 = 0.2289, 6 = 0.014996, 7 = 580.0, 8 = 0.0700, 1 = 0.3437,
2 = 0.2637, 1 = 0.4458, 2 = 0.3459, 3 = 0.8 and 4 = 0.0663 ( is the inter-event
residual). T is the total variance for ln(y) and is approximate based on the Taylor series
expansion of the sum of the inter-event and intra-event variances. NL0 is the equation
for evaluated for = 0. Check approximate using Monte Carlo simulation and nd
good (within a few percent) match to exact answer.
Characterise sites using VS30 . FInf erred = 1 if VS30 inferred from geology and 0 otherwise. FM easured = 1 if VS30 is measured and 0 otherwise. Believe model applicable for
150 VS30 1500 m/s.
228
Use depth to shear-wave velocity of 1.0 km/s, Z1.0 , to model effect of near-surface sediments since 1 km/s similar to values commonly used in practice for rock, is close to
reference VS30 and depth to this velocity more likely to be available. For stations without
8
8
Z1.0 use this empirical relationship: ln(Z1.0 ) = 28.5 3.82
8 ln(VS30 + 378.7 ).
Use PEER Next Generation Attenuation (NGA) database supplemented by data from
TriNet system to provide additional guidance on functional forms and constraints on
coefcients.
Consider model to be update of Sadigh et al. (1997).
Focal depths less than 20 km and ZT OR 15 km. Therefore note that application to
regions with very thick crusts (e.g. 20 km) is extrapolation outside range of data used
to develop model.
Develop model to represent free-eld motions from shallow crustal earthquakes in active
tectonic regions, principally California.
Exclude data from earthquakes that occurred in oceanic crust offshore of California or
Taiwan because these data have been found to be more consistent with ground motions
from subduction zones. Include data from 1992 Cape Mendocino earthquakes because
source depth places event above likely interface location. Exclude data from four 1997
NW China earthquakes because of large depths ( 20 km) and the very limited information available on these data. Exclude data from the 1979 St Elias earthquake because
believe it occurred on subduction zone interface. Include data from the 1985 Nahanni
and 1992 Roermond because believe that they occurred on boundary of stable continental and active tectonic regions.
Assume that ground motions from different regions are similar and examine this hypothesis during development.
Include data from aftershocks, because they provide additional information on site model
coefcients, allowing for systematic differences in ground motions with mainshock motions. AS = 1 if event aftershock and 0 otherwise.
Exclude data from large buildings and at depth, which removes many old records. Include sites with known topographic effects since the effect of topography has not been
systematically studied for all sites so many other stations may be affected by such effects. Topographic effects are considered to be part of variability of ground motions.
Exclude records with only a single horizontal component.
Exclude records from more than 70 km (selected by visual inspection) to remove effects
of bias in sample.
To complete missing information in the NGA database estimate strike, dip ( ) and rake
() and/or depth to top of rupture, ZT OR , from other associated events (e.g. mainshock
or other aftershock) or from tectonic environment. For events unassociated to other
earthquake assigned based on known or inferred mechanisms: 90 for strike-slip, 40
for reverse and 55 for normal. For events without known fault geometries RRU P and
RJB estimated based on simulations of earthquake ruptures based on focal mechanisms, depths and epicentral locations.
229
Use Mw since simplest measure for correlating the amount of energy released in earthquake with ground motions. Develop functional form and constrain some coefcients for
magnitude dependence based on theoretical arguments on source spectra and some
previous analyses. Note that data are not sufcient to distinguish between various forms
of magnitude-scaling.
Exploratory analysis indicates that reverse faulting earthquakes produce larger highfrequency motions than strike-slip events. It also shows that style-of-faulting effect is
statistically signicant (p-values slightly less than 0.05) only when normal faulting was
restricted to in range 120 to 60 with normal-oblique in strike-slip class. Find styleof-faulting effect weaker for aftershocks than main shocks hence effect not included for
aftershocks.
Preliminary analysis indicates statistically-signicant dependence on depth to top of rupture, ZT OR and that effect stronger for aftershocks therefore model different depth dependence for aftershocks and main shocks. Find that aftershocks produce lower motions than main shocks hence include this in model.
Examine various functional forms for distance-scaling and nd all provide reasonable
ts to data since to discriminate between them would require more data at distances
< 10 km. Find that data shows magnitude-dependence in rate of attenuation at all distances but that at short distances due to effect of extended sources and large distances
due to interaction of path Q with differences in source Fourier spectra as a function of
magnitude. Choose functional form to allow for separation of effect of magnitude at
small and large distances.
Examine distance-scaling at large distances using 666 records from 3 small S. Californian earthquakes (2001 Anza, M 4.92; 2002 Yorba Linda, M 4.27; 2003 Big Bear City,
M 4.92) by tting ground motions to three functional forms. Find that two-slope models
t slightly better than a one-slope model with break point between 40 and 60 km. Other
data and simulations also show this behaviour. Prefer a smooth transition over broad
distance range between two decay rates since transition point may vary from earthquake
to earthquake. Constrain some coefcients based on previous studies.
Initially nd that anelastic attenuation coefcient, , is 50% larger for Taiwan than other
areas. Believe this (and other similar effects) due to missing data due to truncation
at lower amplitudes. Experiments with extended datasets for 21 events conrm this.
Conclude that regression analyses using NGA data will tend to underestimate anelastic attenuation rate at large distances and that problem cannot be solved by truncated
regression. Develop model for based on extended data sets for 13 Californian events.
To model hanging-wall effect, use RX , site coordinate (in km) measured perpendicular
to the fault strike from the surface projection of the updip edge of the fault rupture with
the downdip direction being positive and FHW (FHW = 1 for RX 0 and 0 for RX < 0.
Functional form developed based on simulations and empirical data.
Choose reference site VS30 to be 1130 m/s because expected that no signicant nonlinear site response at that velocity and very few records with VS30 > 1100 m/s in NGA
database. Functional form adopted for nonlinear site response able to present previous
models from empirical and simulation studies.
Develop functional form for Z1.0 -dependence based on preliminary analyses and residual plots.
230
Model variability using random variables i (inter-event) and ij (intra-event). Assume

inter-event residuals independent and normally distributed with variance 2 . Assume
intra-event error components independent and normally distributed with variances P2
2 (remaining). Assume total intra-event variance to be normally
(path), S2 (site) and X
distributed with variance 2 . Show that 2 is function of soil nonlinearity. Note that
complete model difcult to use in regression analysis due to lack of repeatedly sampled
paths and limited repeatedly sampled sites and unavailability of inference method capable of handling complicated data structure introduced by path error being included as
predictor of soil amplication. Therefore apply simplication to solve problem.
Find inter-event residuals do not exhibit trend w.r.t. magnitude. Residuals for Californian
and non-Californian earthquakes do not show any trends so both sets of earthquakes
consistent with model. Note that inter-event term for Chi-Chi approximately 2 below
population mean.
Find intra-event residuals do not exhibit trends w.r.t. M , RRU P , VS30 or yref . Note that
very limited data suggests slight upward trend in residuals for VS30 > 1130 m/s, which
relate to lower kappa attenuation for such sites.
Preliminary analyses based on visual inspection of residuals suggested that standard
errors did not depend on M but statistical analysis indicated that signicant (p-values
< 0.05) magnitude dependence is present [using test of Youngs et al. (1995)]. Find that
magnitude dependence remains even when accounting for differences in variance for
aftershocks and main shocks and for nonlinear site amplication.
Note that in regions where earthquakes at distances > 50 km are major contribution to
hazard adjustments to c1 and c2 may be warranted.
2.256
Cotton et al. (2008)
log[PSA(f )] = a(f ) + b(f )Mw + c(f )M 2 + d(f )R log10 [R + e(f ) 100.42Mw ] + Si (f )

where PSA(f ) is in m/s2 , a = 5.08210, b = 2.06210, c = 0.11966, d = 0.00319,
e = 0.00488, S = 0.01145 and = 0.32257 for borehole stations (S applies for
stations at 200 m) and a = 4.884, b = 2.18080, c = 0.12964, d = 0.00397, e =
0.01226, SB = 0.16101, SC = 0.27345, SD = 0.45195 and = 0.35325 for surface
stations.
Experiments on magnitude dependency of decay and reported below conducted using:
log10 [SAi,j (f )] = a(f )Mi + b(f )Rrup,j log10 (Rrup,j ) + S(f )

Do not report coefcients of these models.
Use four site classes (based on Eurocode 8) for surface stations:
Class A Vs,30 > 800 m/s.
Class B 360 < Vs,30 < 800 m/s. Use coefcient SB .
Class C 180 < Vs,30 < 360 m/s. Use coefcient SC .
Class D Vs,30 < 180 m/s. Use coefcient SD .
231
Use data from boreholes to reduce inuence of nonlinear site effects for investigating
magnitude-dependent decay. Also derive models using surface records.
Only use data from < 100 km.
Only retain events with depth < 25 km to exclude subduction earthquakes.
Note relatively good magnitude-distance coverage.
Visually inspect records to retain only main event if multiple events recorded and to
check for glitches. Bandpass Butterworth (four poles and two passes) lter records with
cut-offs 0.25 and 25 Hz. Longest usable period of model is less than 3 s due to ltering.
Derive equations using data from small (Mw 5) earthquakes (3376 records from 310
events) and large (Mw 5) earthquakes (518 records from 27 events) to examine ability of models to predict ground motions outside their magnitude range of applicability.
Find ground motions from small events attenuate faster than from large events. Predict
ground motions for Mw 4.0, 5.0 and 6.5 and 10, 30 and 99 km. Find overestimation of
ground motions for Mw 4.0 using model derived using data from Mw 5 and overestimation of ground motions for Mw 6.5 using model derived using data from Mw 5.
Predictions for Mw 5.0 are similar for both models. Also compare predictions from both
models and observations for Mw 4.1, 4.6, 5.2, 5.7, 6.5 and 7.3 and nd similar results.
Also derive models for 11 magnitude ranges: 4.04.2, 4.24.4, 4.44.6, 4.64.8, 4.85.0,
5.05.2, 5.25.4, 5.65.8, 5.86.8 and 6.87.3. Compare predictions with observations
for each magnitude range and nd good match. Find that decay rate depends on Mw
with faster decay for small events. Plot s from each model w.r.t. Mw and nd that it
has a negative correlation with Mw .
Examine residuals w.r.t. distance. Find slight increase at large distances, which relate
to magnitude dependency of attenuation.
Note that goal of analysis was not to compete with existing models but to compare
magnitude dependency of ground motions at depth and surface.
Examine residuals w.r.t. distance and magnitude of nal model. Find no trends.
Find that s for surface motions are larger (by about 9%) than those for motions at depth.
2.257 Humbert & Viallet (2008)

log(PGA) = aM + bR log(R) + c
where PGA is in cm/s2 , a = 0.31, b = 0.00091, c = 1.57 and = 0.23.
Use data of Berge-Thierry et al. (2003).
Plot rhypo , epicentral location and Ms from ISC against those used by Berge-Thierry
et al. (2003). Derive standard deviation, skewness and kurtosis based on these plots.
232
Account for estimated uncertainties of M and R in fuzzy regression and nd same

coefcients as standard regression but with estimated uncertainties and lower than in
standard regression.
Find that epistemic uncertainties increase at edge of magnitude-distance space.
2.258
Idriss (2008)
ln[PSA(T )] = 1 (T )+2 (T )M [1 (T )+2 (T )M ] ln(Rrup +10)+(T )Rrup +(T )F

where PSA is in g, 1 = 3.7066 and 2 = 0.1252 for M 6.75, 1 = 5.6315
and 2 = 0.4104 for 6.75 < M 8.5, 1 = 2.9832, 2 = 0.2339, = 0.00047,
= 0.12 and = 1.28 + 0.05 ln(T ) 0.08M . for M < 5 equals at M 5 and
for M > 7.5 equals at M 7.5. for T < 0.05 equals for T = 0.05 s. Correction
factor for VS30 > 900 m/s 1 (T ) = ln[(1 + 11T + 0.27T 2 )/(1 + 16T + 0.08T 2 )] for
0.05 T 10 s [1 (T ) for T < 0.05 s equals 1 (0.05)].
Use two site classes (may derive model for 180 VS30 < 450 m/s in future):
1. VS30 > 900 m/s. 45 records. Since not enough records from stations with VS30 >
900 m/s derive correction factor, 1 (T ), to 1 based on residuals for these 45
records. Find no trends in residuals w.r.t. M , R or VS30 .
2. 450 VS30 900 m/s. 942 records (333 from stations with measured VS30 ).
Notes that only 29% of stations have measured VS30 ; the rest have inferred VS30 s. Examine distributions of measured and inferred VS30 s and concluded no apparent bias by
using inferred values of VS30 .
Uses two mechanism categories:
Strike-slip Rake within 30 of horizontal. Includes records from normal events (rake within 30
of vertical downwards) because insufcient data to retain as separate category.
F = 0.
Reverse Rake within 30 of vertical upwards. Includes records from reverse oblique and
normal oblique events (remaining rake angles) because insufcient data to retain
as separate categories. F = 1.
Uses the PEER Next Generation Attenuation (NGA) database (Flat-File version 7.2).
Excludes (to retain only free-eld records): i) records from basements of any building; ii)
records from dam crests, toes or abutments; and iii) records from rst oor of buildings
with 3 storeys.
Excludes records from deep events, records from distances > 200 km and records
from co-located stations.
Only retains records with 450 VS30 900 m/s for regression. Notes that initial analysis indicated that ground motions not dependent on value of VS30 in this range so do not
include a dependency on VS30 .
233
Uses 187 records from California (42 events), 700 records from Taiwan (Chi-Chi, 152
records, and 5 aftershocks, 548 records) and 55 records from 24 events in other regions
(USA outside California, Canada, Georgia, Greece, Iran, Italy, Mexico and Turkey).
Only 17 records from R 5 km and 33 from R 10 km (for M 7 only 3 records from
California for these distance ranges) (all site classes). Therefore, difcult to constrain
predictions at short distances, particularly for large magnitudes.
States that, from a geotechnical engineering perspective, use of VS30 bins is more appropriate than use of VS30 as an independent parameter.
Does not investigate the inuence of other parameters within the NGA Flat-File on
ground motions.
Uses PSA at 0.01 s for PGA (checked difference and generally less than 2%).
Divides data into magnitude bins 0.5 units wide and conducts one-stage regression
analysis for each. Compares observed and predicted PGAs at distances of 3, 10, 30
and 100 km against magnitude. Find that results for each magnitude bin generally well
represent observations. Find oversaturation for large magnitudes due to presence of
many records (152 out of 159 records for M > 7.5) from Chi-Chi. Does not believe
that this is justied so derive 1 and 2 for M > 6.75 by regression using the expected
magnitude dependency based on previous studies and 1D simulations.
Examines residuals w.r.t. M , R and VS30 and concludes that for 5.2 M 7.2 model
provides excellent representation of data. Examine residuals for 5 Chi-Chi aftershocks
and nd that for R > 15 km there is no bias but for shorter distances some negative
bias.
Compares predictions to observations for Hector Mine (M 7.1), Loma Prieta (M 6.9),
Northridge (M 6.7) and San Fernando (M 6.6) events w.r.t. R. Finds good match.
Comments on the insufciency of VS30 as a parameter to characterise site response due
to soil layering and nonlinear effects.
2.259 Lin & Lee (2008)

ln(y) = C1 + C2 M + C3 ln(R + C4 eC5 M ) + C6 H + C7 Zt

where y is in g, C1 = 2.5, C2 = 1.205, C3 = 1.905, C4 = 0.516, C5 = 0.6325,
C6 = 0.0075, C7 = 0.275 and = 0.5268 for rock sites and C1 = 0.9, C2 = 1.00,
C3 = 1.90, C4 = 0.9918, C5 = 0.5263, C6 = 0.004, C7 = 0.31 and = 0.6277 for soil
sites.
Use two site categories (separate equations for each):
Rock B and C type sites
Soil D and E type sites
Use two earthquake types:
234
Interface Shallow angle thrust events occurring at interface between subducting and overriding plates. Classied events using 50 km maximum focal depth for interface
events. 12 events from Taiwan (819 records) and 5 from elsewhere (54 records).
Zt = 0.
Intraslab Typically high-angle normal-faulting events within the subducting oceanic plate. 32
events from Taiwan (3865 records) and 5 from elsewhere (85 records). Zt = 1.
Focal depths, H , between 3.94 and 30 km (for interface) and 43.39 and 161 km (for
intraslab).
Develop separate ML -Mw conversion formulae for deep (H > 50 km) and shallow
events.
Use data from TSMIP and the SMART-1 array.
Lack data from large Taiwanese earthquake (especially interface events). Therefore,
add data from foreign subduction events (Mexico, western USA and New Zealand).
Note that future study should examine suitability of adding these data.
Exclude poor-quality records by visual screening of available data. Baseline correct
records.
Weight data given the number of records from different sources (Taiwan or elsewhere).
Focus on data from foreign events since results using only Taiwanese data are not reliable for large magnitudes. Note that should use maximum-likelihood regression method.
Compare predicted and observed PGAs for the two best recorded events (interface
Mw 6.3 H = 6 km and intraslab Mw 5.9 H = 39 km) and nd good t.
Examine residuals and nd that a normal distribution ts them very well using histograms.
From limited analysis nd evidence for magnitude-dependent but do not give details.
Note that some events could be mislocated but that due to large distances of most data
this should not have big impact on results.
2.260
Massa et al. (2008)
log10 (Y ) = a + bM + c log(R2 + h2 )1/2 + s1 SA + s2 S(B+C)

where Y is in g; a = 2.66, b = 0.76, c = 1.97, d = 10.72, s1 = 0, s2 = 0.13,
eve = 0.09 (inter-event) and rec = 0.27 (intra-event) for horizontal PGA and ML ;
a = 2.66, b = 0.76, c = 1.97, d = 10.72, s1 = 0, s2 = 0.13, sta = 0.09 (inter-site)
and rec = 0.28 (intra-site) for horizontal PGA and ML ; a = 2.59, b = 0.69, c = 1.95,
d = 11.16, s1 = 0, s2 = 0.12, eve = 0.09 (inter-event) and rec = 0.26 (intra-event) for
vertical PGA and ML ; a = 2.59, b = 0.69, c = 1.95, d = 11.16, s1 = 0, s2 = 0.12,
eve = 0.08 (inter-site) and rec = 0.26 (intra-site) for vertical PGA and ML ; a = 3.62,
b = 0.93, c = 2.02, d = 11.71, s1 = 0, s2 = 0.12, eve = 0.10 (inter-event) and
rec = 0.28 (intra-event) for horizontal PGA and Mw ; a = 3.62, b = 0.93, c = 2.02,
235
d = 11.71, s1 = 0, s2 = 0.12, sta = 0.11 (inter-site) and rec = 0.29 (intra-site)

for horizontal PGA and Mw ; a = 3.49, b = 0.85, c = 1.99, d = 11.56, s1 = 0,
s2 = 0.11, eve = 0.09 (inter-event) and rec = 0.29 (intra-event) for vertical PGA and
Mw ; a = 3.49, b = 0.85, c = 1.99, d = 11.56, s1 = 0, s2 = 0.11, eve = 0.12
(inter-site) and rec = 0.30 (intra-site) for vertical PGA and Mw .
Also use functional form: log10 (Y ) = a + bM + (c + eM ) log(R2 + h2 )1/2 + s1 SA +
s2 S(B+C) but do not report coefcients since nd small values for e.
Use three site classications based on Eurocode 8 for the 77 stations:
A Rock, Vs,30 > 800 m/s: marine clay or other rocks (Lower Pleistocene and Pliocene)
and volcanic rock and deposits. 49 stations. SA = 1 and S(B+C) = 0.
B Stiff soil, 360 < Vs,30 < 800 m/s: colluvial, alluvial, lacustrine, beach, uvial terraces, glacial deposits and clay (Middle-Upper Pleistocene); sand and loose conglomerate (Pleistocene and Pliocene); and travertine (Pleistocene and Holocene).
19 stations. S(B+C) = 1 and SA = 0.
C Soft soil, Vs < 360 m/s: colluvial, alluvial, lacustrine, beach and uvial terraces
deposits (Holocene). 9 stations. S(B+C) = 1 and SA = 0.
Because of limited records from class C combine classes B and C in regression. Note
that the classication of some stations in class A could not be appropriate due to site
amplication due to structure-soil interaction and topographic effects. Also note that
class C is not appropriate for some stations on Po Plain due to deep sediments but that
there are few data from these sites so no bias.
Use data from various analogue and digital strong-motion (Episensor, K2, Etna, SSA-1
or SMA-1 instruments) and digital velocimetric (Mars-Lite, Mars88-MC, Reftek 130 or
other instruments) networks in northern Italy, western Slovenia and southern Switzerland.
Originally collect about 10 000 records but reduce by careful selection. Exclude data
with de > 100 km and with ML < 3.5. Consider earthquakes down to ML 3.5 because
such earthquakes could damage sensitive equipment in industrial zones.
216 components (both horizontal and vertical combined) from earthquakes with ML >
4.5.
Focal depths between 1.9 and 57.9 km. Most less than 15 km.
Bandpass lter using fourth-order acausal Butterworth lter with cut-offs of 0.4 and 25 Hz
for ML 4.5 and 0.2 and 25 Hz for ML > 4.5. Check using some records that PGA
is not affected by ltering nor are spectral accelerations in the period range of interest.
Check ltering of analogue records by visually examining Fourier amplitude spectra.
Check conversion of velocimetric records to acceleration is correct by examining records
from co-located instruments of different types. Exclude clipped records or records affected by noise.
Try including a quadratic magnitude term but nd that the coefcient is not statistically
signicant.
Try including an anelastic attenuation term but nd that coefcient is not statistically
signicant.
236
Do not use rjb since not sufcient information on rupture locations. Do not use rhypo so
as not to introduce errors due to unreliable focal depths.
Do not include style-of-faulting terms because most data from reverse-faulting earthquakes (often with strike-slip component).
Apply simple tests to check regional dependence and do not nd signicant evidence
for regional differences in ground motions. Since records from similar earthquakes of
similar mechanisms conclude that models appropriate for whole of northern Italy (6
15 E and 43 47 N).
Examine residuals (against earthquake and station indices, as box and whisker plots and
against distance and magnitude) for sites A and sites B & C and for ML 4.5 and ML >
4.5. Also compare predicted and observed ground motions for various magnitudes and
events. Find good results.
Suggest that for de < 10 km and ML > 5.5 10 km is considered the distance at which
distance saturation starts (since little data with de < 10 km to constrain curves and
predictions for shorter distances unrealistically high).
Also derive equations for other strong-motion intensity parameters.
2.261
Mezcua et al. (2008)
ln Y = C1 + C2 M + C3 ln R
where Y is in cm/s2 , C1 = 0.125, C2 = 1.286, C3 = 1.133 and = 0.69. Only
derive equation for rm soil sites due to insufcient data for other classes. For compact
rock sites propose using ratio between PGA on rm soil and rock derived by Campbell
(1997).
Use three site classications:
1 Compact rock. Crystalline rocks (granite and basalt), metamorphic rocks (e.g. marble, gneiss, schist and quartzite) and Cretaceous and older sedimentary deposits
following criteria of Campbell (1997). Similar to Spanish building code classes I
and II with 400 Vs 750 m/s. 23 stations.
2 Alluvium or rm soil. Quaternary consolidated deposits. Similar to Spanish building
code class III with 200 Vs 400 m/s. 29 stations.
3 Soft sedimentary deposits. 52 stations.
Classify using crude qualitative descriptions.
Most stations in basements of small buildings (e.g. city council ofces) and therefore
records are not truly free-eld.
Only consider data with 5 de 100 km and M 3.
237
Most data from 3 M 4 and de 50 km. Only one record with M > 5 and
de < 20 km.
Use hypocentral distance because no information on locations of rupture planes and
since using hypocentral distance automatically limits near-source ground motions.
Do not consider style-of-faulting since no reported mechanisms are available for most
events.
Compare predicted PGA for Mw 5 with observations for 4.9 Mw 5.1. Find reasonable t.
2.262 Morasca et al. (2008)

log10 Y = a + bM + c log10 R + s1,2

where Y is in g, a = 4.417, b = 0.770, c = 1.097, D = 0, D1 = 0.123, eve = 0.069
and rec = 0.339 for horizontal PGA and intra-event sigma; a = 4.128, b = 0.722,
c = 1.250, D = 0, D1 = 0.096, eve = 0.085 and rec = 0.338 for vertical PGA and
intra-event sigma; a = 4.367, b = 0.774, c = 1.146, D = 0, D1 = 0.119, sta =
0.077 and rec = 0.337 for horizontal PGA and intra-station sigma; and a = 4.066,
b = 0.729, c = 1.322, D = 0, D1 = 0.090, sta = 0.105 and rec = 0.335.
Use two site categories (s1,2 ) because insufcient information to use more:
D Rock. Average Vs > 800 m/s. 10 stations.

D1 Soil. Average Vs < 800 m/s. Includes all kinds of supercial deposits, from weak
rocks to alluvial deposits although they are mainly shallow alluvium and soft rock
(600-700 m/s) sites. 27 stations.
Use data from the 20022003 Molise sequence from various agencies.
Use data from accelerometers (SMA-1, 3 stations; RFT-250, 2 stations; Episensor, 10
stations) and velocimeters (CMG-40T, 4 stations; Lennartz 1 s, 5 stations; Lennartz 5 s,
13 stations).
Select data with M > 2.7.
Baseline and instrument correct records from analogue accelerometric instruments and
lter in average bandpass 0.520 Hz after visual inspection of the Fourier amplitude
spectra. Baseline correct records from digital accelerometric instruments and lter in
average bandpass 0.230 Hz after visual inspection of the Fourier amplitude spectra.
Instrument correct records from digital velocimetric instruments and lter in average
bandpass 0.525 Hz after visual inspection of the Fourier amplitude spectra.
Most data from rhypo < 40 km and almost all velocimetric data from 20-30 km.
Most focal depths between 10 and 30 km.
Relocate events using manual picks of P and S phases and a local velocity model.
Compute ML s using velocimetric data.
238
Note that small value of eve suggests that the calibrated local magnitudes and relocated
hypocentral locations are accurate.
Note that small value of sta suggests that the site classication is correct.
Note that records from accelerometric and velocimetric instruments are similar.
2.263
Slejko et al. (2008)
log10 P GA = a + (b + cMs )Ms + (d + eMs ) log10 r

where r2
= D2 + h2
where P GA is in g, a = 2.14, b = 0.98, c = 0.06, d = 1.88, e = 0.0009, h = 13.4

and = 0.35.
Only use data for de < 100 km because data from larger distances only available for
large earthquakes.
Only eight records have PGA< 0.005 g (standard trigger level).
Use truncated regression analysis (Bragato, 2004) to account for bias due to nontriggering stations.
2.264
Srinivasan et al. (2008)
log(A) = c1 + c2 M b log(X + ec3 M )

where A is in cm/s2 , c1 = 1.3489, c2 = 1.0095, b = 0.1956, c3 = 0.1272 and = 0.20.
Use data from one station.
Data from rockbursts in mines in the Kolar Gold Fields.
Exclude records with rhypo < 1 km due to large change in PGAs in near-source region.
Regress data using log(A) = b log(X) + c for data binned in 5 0.2 magnitude unit bins
from 2.0 upwards.
Also regress data using log(A) = aM b log(X) + c.
Also regress using log(A) = c1 + c2 M bc4 log(X + ec3 M ) (sic) but nd c4 has a very
large standard error so remove it.
Compare predictions and observations for M 2.1, 2.3, 2.5, 2.7 and 2.9.
239

ln y
c1 + f1 (Mw ) + f2 (Mw )f3 (R) + f4 (F ) + FRf5 (ZF R ) +

FSf6 (ZF R ) + f7 (HW, RJB , Mw , DIP) +
f8 (Vs,30 , Vlin , PGAnonlin , PGArock ) + f9 (Vs,30 , Z1.5 )
where for
Mw c0
f1 (Mw )
c3 (Mw c0 ) + c8 (T )(8.5 Mw )n
f2 (Mw )
c2 (T ) + c4 (Mw c0 )
f1 (Mw )
c5 (Mw c0 ) + c8 (T )(8.5 Mw )n
f2 (Mw )
f3 (R)
c2 (T ) + c6 (Mw c0 )
2
ln Rrup
+ c7 (T )2
f4 (F )
f5 (ZFR )
f6 (ZFS )
g1 (RJB )
g2 (Mw )
g3 (DIP)
f7 (HW, RJB , Mw , DIP)
c14 (T )HWg1 (RJB )g2 (Mw )g3 (DIP)
f8 (Vs,30 , Vlin , PGAnonlin , PGArock )
g4 (Vs,30 , Vlin ) + g5 (PGAnonlin , PGArock )
g4 (Vs,30 , Vlin )
g5 (PGAnonlin , PGArock )
c15 (T ) ln(Vs,30 /Vlin )
c16 (T ) ln(PGAmin /0.1)

PGAnonlin < a1
c16 (T )[ln(PGAmin /0.1)

+a ln(PGAnonlin /a1 )
+b(ln(PGAnonlin /a1 ))2 ] a1 PGAnonlin a2
c16 (T ) ln(PGAnonlin /0.1) PGAnonlin a2
f9 (Vs,30 , Z1.5 )
g6 (Vs,30 , Z1.5 , Z)
and for
Mw > c0
=
=
c9 (T )FR + c10 (T )FS + c11 (T )FN
Ztop 2 km
2 < Ztop 5 km
c12 (T )(Ztop 2)/3
c12 (T )
5 < Ztop 10 km
c
(T
)[1
(Z
10)/5]
10
< Ztop 15 km
12
top
0
Ztop > 15 km
c13 (T )Ztop /2
0 < Ztop 2 km
c13 (T )
2 < Ztop 4 km
c
(T
)[1
(Z
4)/2]
4
< Ztop 6 km
13
top
0
Ztop > 6 km
0 RJB < 15 km
1 RJB /45
2
(2
R
/15)
15 RJB < 30 km
JB
3
0
RJB 30 km
Mw < 6.0
0
2(Mw 6) 6.0 Mw < 6.5
1
Mw 6.5
{
1 (DIP 70)/20 DIP 70
1
DIP < 70
+ g7 (ZD , Z1.5 )
g6 (Vs,30 , Z1.5 , Z)
c17 (T )(1/Z) ln(Vs,30 /1500) ln(Z1.5 )
7.154 (4.4657.154)[ln(V /1500)ln(360/1500)]

s,30
7.154 +
ln(600/1500)ln(360/1500)
4.465 (2.7724.465)[ln(Vs,30 /1500)ln(600/1500)]

ln(600/1500)
g7 (Z1.5 , ZD )
Vs,30 360 m/s

360 < Vs,30 600 m/s
Vs,30 > 600 m/s
ZD c18 (T )K1 (1 exp((Z1.5 200)/300)) +

ZD c19 (T )K2 (1 exp((Z1.5 200)/4000))
where y is in g and c0 = 6.5; c1 = 2.033, c2 = 1.180, c7 = 8.647, c8 = 0.028,

c9 = 0.176, c10 = 0.266, c11 = 0.476, c12 = 0.520, c13 = 0.320, c14 = 0.400,
c15 = 0.360, c17 = 0, c18 = 0, c19 = 0, c20 = 0.412, c21 = 0.427, K1 = 2.260,
K2 = 1.040, Vlin = 760 for horizontal PGA; and c1 = 2.983, c2 = 1.616, c7 = 9.101,
240
c8 = 0.043, c9 = 0.253, c10 = 0.463, c11 = 0.706, c12 = 0.132, c13 = 0.171,
c14 = 0.513, c15 = 0.360, c17 = 0, c18 = 0, c19 = 0, c20 = 0.522, c21 = 0.537,
K1 = 2.260, K2 = 1.040, Vlin = 760 for vertical PGA; = c20 (T )+[c21 (T )c20 (T )]Mw
for 5.0 Mw < 7.0 and = c21 (T ) for Mw 7.0.
Almost identical to Aghabarati & Tehranizadeh (2008) (see Section 2.253) but some
coefcients are slightly different and they are also provided for the vertical components.
Set a1 = 0.04 g, a2 = 0.1 g and PGAmin = 0.06 g. PGAnonlin is expected PGA on
rock (Vs,30 = 760 m/s). c15 (T ), c16 (T ) and Vlin taken from Choi & Stewart (2005) and
are not determined in regression.
2.266
Akyol & Karagz (2009)
log y = a1 + a2 (M 6) + b log r + cS
where y is in g, a1 = 1.330095 0.068, a2 = 0.640047 0.066, b = 1.65663
0.055
, c = 0.14963 0.098, 1 = 0.196 (intra-event), 2 = 0.191 (inter-event) and
= 12 + 22 = 0.274.
Initially use four site classes:
1. Rock. 6 stations, 20 records.
2. Stiff soil. 11 stations, 57 records.
3. Soil. 11 stations, 32 records.
4. Deep soil. 9 stations, 59 records.
Sites classied using horizontal/vertical spectral ratios of the S-wave window of records
grouped by station (details not given). Only use data with S/N ratio > 3 and smooth
spectra using a nine-point moving average. Since data insufcient to obtain coefcients
for all classes, combine classes 1 and 2 and classes 3 and 4 to produce categories A
(S = 0) and B (S = 1) based on 77 and 91 records, respectively. Display average H/V
spectral ratios for each category.
Note that ideally would account for faulting mechanism but for many earthquakes this
parameter is unknown and also dataset is not large enough to assess its impact.
Use data from the Turkish National Strong Motion Network of the Earthquake Research
Department of the General Directorate of Disaster Affairs and the temporary Western
Anatolia Seismic Recording Experiment (WASRE).
Use rhypo because fault geometries unknown for most earthquakes.
Initially use 2123 records from all regions of Turkey. Discard records with unknown and
poor estimates of magnitude, distance and/or site conditions and those outside western
Anatolia. Select data with: rhypo < 200 km, Mw 4.0 and PGA > 0.0015 g.
241
Check low- and high-frequency noise for all records. Find that much data from SMA1s have signicant long-period noise (especially records of small earthquakes at large
distances). Do not lter data but eliminate suspect records. Apply correction for instrument response. Numerically differentiate data from velociometers of WASRE network.
Baseline correct all data.
Most data from 4.5 Mw 5.5 and 25 rhypo 125 km.
Note that due to lack of records from < 10 km cannot include ctitious depth in functional
form.
Initially include a quadratic magnitude term but this term does not improve match so
drop this term.
Test signicance of site coefcients and nd that they are generally signicant at more
than 90% condence level.
Plot residuals w.r.t. distance, magnitude and predicted log P GA. Find systematic trends,
especially for site B residuals versus Mw . Derive linear site coefcient correction terms
to remove these trends (not clear how they are applied), which relate to nonlinear site
response.
Compare predictions and observations for selected earthquakes.
Discuss reasons for differences in site effects in western Anatolia and in other regions.
Based on results, suggest that number of stations on rock should be increase and site
classications should be re-evaluated.
2.267 Bindi et al. (2009a)

log10 Y
= a + b1 (Mw Mref ) + b2 (Mw Mref )2
+[c1 + c2 (Mw Mref ] log10 (RJB + h2 ) + ei Si + fj Fj
where Y is in cm/s2 and Mref = 5.5 (to reduce trade-offs between attenuation and
source parameters), a = 3.0761, b1 = 0.1587, b2 = 0.0845, c1 = 1.0504, c2 =
0.0148, h = 7.3469, e1 = 0, e2 = 0.2541, e3 = 0.1367, f1 = 0, f2 = 0.0059,
f3 = 0.0168, event = 0.1482, station = 0.2083, record = 0.1498 and = 0.2963 for
larger horizontal component; a = 3.0191, b1 = 0.1643, b2 = 0.0674, c1 = 1.0284,
c2 = 0.0041, h = 6.8963, e1 = 0, e2 = 0.2275, e3 = 0.0774, f1 = 0, f2 = 0.0138,
f3 = 0.0005, event = 0.1465, station = 0.2184, record = 0.1345 and = 0.2930 for
geometric mean of horizontal components; and a = 3.0421, b1 = 0.3762, b2 = 0.0925,
c1 = 1.2350, c2 = 0.0891, h = 9.3012, e1 = 0, e2 = 0.1787, e3 = 0.1146, f1 = 0,
f2 = 0.0073, f3 = 0.0222, event = 0.1266, station = 0.2114, record = 0.1394 and
= 0.2831 for vertical component.
Use three site classes following Sabetta & Pugliese (1987, 1996):
Class 0 Rock: rock outcrops or deposits thinner than 5 m.. 98 records. S1 = 1 and S2 =
S3 = 0 .
242
Class 1 Shallow alluvium: deposits thinner than or equal to 20 m and thicker than 5 m. Vs
of alluvium between 400 and 800 m/s. 62 records. S2 = 1 and S1 = S3 = 0.
Class 2 Deep alluvium: deposits thicker than 20 m. 81 records. S3 = 1 and S1 = S2 = 0.
Site classication performed using veried geological, geophysical and geotechnical information, which altered the previous categorization of some stations. Data from 146
different stations. Note that only 6% of 600 Italian stations are associated with a Vs
prole.
Use data from Italian Accelerometric Archive (ITACA) from between 1972 and 2004,
which have been carefully revised during a project funded by the Italian Department of
Civil Protection. Records individually processed using individually-selected lters. Analogue records corrected for linear trend and instrument response and then band-pass ltered, selecting high-pass frequency from visual inspection of Fourier spectra (generally
between 0.3 and 0.5 Hz) and low-pass frequency chosen close to instrument frequency
(generally between 20 and 25 Hz). Digital records corrected for linear trend using entire
trace (because few records have usable pre-event portion) and then band-pass ltered
in the same way as analogue data (but with generally lower cut-offs, 0.10.3 Hz and 25
30 Hz). Use raised cosine lter for analogue records, which often triggered on S-phase,
and acausal fourth-order Butterworth for digital signals, which were padded with zeros
at both ends.
Normal F1 = 1 and F2 = F3 = 0.
Strike-slip F2 = 1 and F1 = F3 = 0.
Reverse F3 = 1 and F1 = F2 = 0.
Most earthquakes on normal faults in central and southern Apennines.
Number of records per earthquake ranges from two (Ancona, 14/06/1972) to 25 (UmbriaMarche, 14/10/1997). Most earthquakes recorded by four stations or more.
Near-source records are poorly represented: 11 records from 3 earthquakes have rjb <
5 km (none with Mw > 6.4 for which shortest rjb is 7 km).
Most data from 10 rjb 100 km and 5 Mw 6.
For Irpinia mainshock (23/11/1980), which is composed of three sub-events, used magnitude, location and time-histories of rst sub-event because it can be clearly recognized.
Assess the standard error of each coefcient using bootstrap technique based on randomly resampling, with replacement, the original dataset to obtain datasets of the same
size as original (500 times). Note the coefcients using this technique are very similar.
Note that some coefcients are not signicantly different than zero (e.g. c2 and fj )
because of the distribution of data w.r.t. Mw and mechanism.
Examine residual plots w.r.t. Mw and rjb and nd no signicant bias or trends.
243
Examine inter-event residuals and nd them within range 0.2 except for two earthquakes (2002 Molise second mainshock and 1990 eastern Sicily), which note could be
due to inaccuracies in magnitudes and locations for these events. Find inter-event residuals for normal earthquakes show smallest dispersion, while largest variability affects
strike-slip events.
Examine inter-station residuals. Note that most are within range 0.3 with few with
absolute values larger than 0.4. Discuss the possible reasons for these large residuals
in terms of local site proles.
Undertake other analyses to understand the source of observed variability in ground
motions.
Also derive model for larger horizontal component using hypocentral distance and no
style-of-faulting terms: log10 Y = 3.4192 + 0.4672(Mw 5.5) + 0.1231(Mw 5.5)2 +
[1.2221 0.1643(Mw 5.5)] log10 rhypo + 0.2474S2 + 0.1435S3 .
Note that unmodelled site effects are contributing a signicant proportion of the observed variability and that a more sophisticated classication scheme using depth of soil
deposit, average Vs of soil deposit and resonance period could signicantly reduce the
inter-station variability component.
2.268 Bindi et al. (2009b)

log10 y = a + bM + c log10
R2 + h2 + ei Si
where y is in cm/s2 ; a = 1.344, b = 0.328, c = 1.09, h = 5, e0 = 0, e1 = 0.262,

e2 = 0.096 and = 0.32 using repi ; and a = 1.954, b = 0.193, c = 1.01, h = 5.88,
e0 = 0, e1 = 0.264, e2 = 0.144 and = 0.300 using rjb .
Class 0 Rock: rock outcrops or deposits thinner than 5 m. 95 records. S1 = 1 and S2 =
S3 = 0 .
Class 1 Shallow alluvium: deposits thinner than or equal to 20 m and thicker than 5 m. Vs
of alluvium between 400 and 800 m/s. 61 records. S2 = 1 and S1 = S3 = 0.
Class 2 Deep alluvium: deposits thicker than 20 m. 79 records. S3 = 1 and S1 = S2 = 0.
Site classication performed using veried geological, geophysical and geotechnical information, which altered the previous categorization of some stations. Data from 137
different stations.
Focal depths from 2 to 29 km.
Civil Protection, plus some data from the Northern Italy Strong Motion network (RAIS).
Records individually processed. Analogue records corrected for linear trend and instrument response and then band-pass ltered, selecting high-pass frequency from visual
inspection of Fourier spectra (generally between 0.3 and 0.5 Hz) and low-pass frequency
244
chosen close to instrument frequency (generally between 20 and 25 Hz). Digital records
corrected for linear trend using entire trace (because few records have usable pre-event
portion) and then band-pass ltered in the same way as analogue data (but with generally lower cut-offs, 0.10.3 Hz and 2530 Hz). Use raised cosine lter for analogue
records, which often triggered on S-phase, and acausal fourth-order Butterworth for
digital signals, which were padded with zeros at both ends. Find PGAs are consistent
with those of Sabetta & Pugliese (1987, 1996) for common records.
Very similar data to that used by Bindi et al. (2009a) (see Section 2.267).
State that GMPEs are updates of those by Sabetta & Pugliese (1987, 1996).
Examine goodness of t of the GMPEs of Sabetta & Pugliese (1987, 1996) to the data
and nd that they do not adequately t because of a too small and non-zero bias.
Therefore, derive new GMPEs.
Use the data from the 17 earthquakes used by Sabetta & Pugliese (1987, 1996) plus
data from ten events that occurred from 1990 to 2002 with Mw > 5.3 and one earlier
shock (Ancona 1972) that was not used by Sabetta & Pugliese (1987, 1996).
Most new earthquakes on normal faults in central and southern Apennines with a few
on strike-slip faults.
Best sampled areas are: eastern Alps (Friuli), central-southern Apennines from Marche
to Pollino and north and east Sicily.
Majority of earthquakes recorded by more than four stations (minimum two, maximum
24).
For Irpinia mainshock (23/11/1980), which is composed of three sub-events, used magnitude, location and time-histories of rst sub-event because it can be clearly recognized.
Only seven records from < 5 km. Earthquakes with Mw > 6 recorded at distances
> 20 km. Best-sampled interval is 10100 km and Mw 56.
Compare observed and predicted PGAs for Mw 5.5 and 6.9 and nd good agreement.
Calculate inter-event and inter-station residuals and relate observed large under- or overestimation for particular events to deep focal depths or other source characteristics.
Compute eve = 0.174 and sta = 0.222 as inter-event and inter-station standard deviations.
Repeat regression using 17 earthquakes of Sabetta & Pugliese (1987, 1996) but including data from additional stations that were not used by Sabetta & Pugliese (1987, 1996)
and using the updated site classes. Find signicant differences for Mw 6.5 at 20 km.
2.269
Bragato (2009)
log(PGA) = c1 + c2 ML + c3 log(depi ) +
Ns
Sk kj
k=1
245
where PGA is in g, c1 = 0.45 0.44, c2 = 0.85 0.09, c3 = 2.39 0.20 and

= 0.27 for Italy with station correction and c1 = 0.49 0.38, c2 = 0.86 0.08,
c3 = 2.41 0.16 and = 0.38 for Italy without station correction. Sk is correction term
for k th station and kj is Kroneker delta and Ns is number of stations in a geographical
cluster. Also provides coefcients for different zones but these are not reported here.
Uses individual site terms for each station. Data from 137 different stations.
Investigates theoretical improvement of GMPEs for ShakeMap purposes in Italy, obtainable by accounting for regional dependencies and site effects. Notes that presented
GMPEs are explorative tools rather than proposals for ShakeMap implementation because of limited data and narrow magnitude range.
Uses data from INGV stations from between December 2005 and July 2008. Stations
give homogeneous coverage in central and southern Italy and eastern Sicily but more
sparse elsewhere and not existent in NE Italy.
To exclude possible outliers, performs preliminary regression on all data and removes
those records with absolute normalised standard deviations greater than three. Also
excludes data from stations that have recored only one earthquake.
Data distribution roughly uniform w.r.t. magnitude and distance.
Tries using rhypo but nds a slightly worse t ( = 0.39 rather than = 0.38), which relates to poor estimates of focal depths for some earthquakes (even though theoretically
rhypo should be better since it includes more information).
Considers various partitions of available stations into different geographical zones using
Delaunay triangulation. Derive a GMPE for each zone with station correction terms.
Applies a genetic algorithm to minimise the standard deviation, based on the Bayesian
information criterion, over the set of possible partitions. Note that this approach cannot
recognise regionalised site effects. Also this method uses some data from earthquakes
occurring outside the zone where the station is located. Notes that considering these
complexities is not possible with current data but that most earthquakes occur in the
same zone as the station. Finds that the optimal zonation has four zones.
Investigates source and focal depth characteristics of different zones to understand the
possible causes of regional variations. Concludes that observed differences are attributable to crustal structure and anelastic attenuation.
Computes GMPEs for the six regions used in ShakeMap implementation.
Computes GMPEs for all of Italy after correction for site amplication modelled by Vs,30 based amplication factors of Borcherdt (1994), used by ShakeMap. Find is unchanged. Also regress using site classes based on Vs,30 s estimated from geology.
Concludes that site effects contribute about 30% of overall standard deviation and that
regional differences contribute only 4%.
Find that station correction terms are weakly correlated to Vs,30 -based amplication factors of Borcherdt (1994) used in ShakeMap to model site effects.
246
2.270
Hong et al. (2009b)
Ground-motion models are, for interface:
log10 Y = c1 + c2 Mw + c3 R (1.82 0.16Mw ) log10 (R + c5 10c6 Mw ) + c7 H

where Y is in cm/s2 , c1 = 2.594, c2 = 0.112, c3 = 0.0037, c5 = 0.0075, c6 = 0.474,
c7 = 0.0033, e = 0.20 (inter-event), r = 0.27 (intra-event) and = 0.33 (total) for
maximum response; c1 = 2.545, c2 = 0.108, c3 = 0.0037, c5 = 0.0075, c6 = 0.474,
c7 = 0.0024, e = 0.20 (inter-event), r = 0.27 (intra-event); c = 0.10 (randomorientation variability) and = 0.35 (total) for geometric mean; and, for inslab:
= c1 + c2 Mw + c3 R c4 log10 R + c5 H
2 + (0.0075 100.507Mw )2
Rcld
where R =
log10 Y
where c1 = 0.014, c2 = 0.562, c3 = 0.0039, c5 = 0.0071, e = 0.10, r = 0.28

and = 0.30 for maximum response; and c1 = 0.109, c2 = 0.569, c3 = 0.0039,
c5 = 0.0070, e = 0.10, r = 0.28, c = 0.07 and = 0.30 for geometric mean.
All data from rm soil sites (NEHRP class B).
Similar analysis to that of Hong & Goda (2007) (see Section 2.244) concerning orientation of major response axis but for data from Mexican subduction zone.
Use data of Garca et al. (2005) (see Section 2.212) for inslab earthquakes.
Focal depths, H , for interplate earthquakes are between 8 and 29 km and depths for
inslab earthquakes are between 35 and 138 km.
Examine correlation of ratio of response along an arbitrary direction to the maximum
response in direction of major axis w.r.t. dependent and independent parameters and
nd that as an approximation there is no dependency.
Provide statistical models to describe the ratio of response along an arbitrary direction
to the maximum response in direction of major axis.
Term expressing magnitude-dependency of decay (i.e. 1.82 0.16Mw ) taken from previous study as is near-source saturation term (i.e. 0.0075 100.507Mw ).
2.271
Hong et al. (2009a)

2
ln Y = b1 + b2 (M 7) + b3 (M 7)2 + [b4 + b5 (M 4.5)] ln[(rjb
+ h2 )0.5 ] + AFs
where Y is in g, b1 = 1.143, b2 = 0.398, b3 = 0.0, b4 = 1.125, b5 = 0.064, h = 5.6,

= 0.150 (inter-event), = 0.438 (intra-event) and T = 0.463 (total) for geometric
mean and considering spatial correlation in regression analysis; b1 = 1.059(0.074),
b2 = 0.383(0.095), b3 = 0.006(0.014), b4 = 1.083(0.068), b5 = 0.056(0.028),
h = 5.7(0.40), = 0.187(0.014) (inter-event), = 0.463(0.008) (intra-event) and
T = 0.500(0.008) (total) for randomly-orientated component ignoring spatial correlation
(based on 50 runs); and b1 = 1.087(00.072), b2 = 0.337(0.096), b3 = 0.011(0.018),
247
b4 = 1.144(0.069), b5 = 0.077(0.027), h = 5.6(0.38), = 0.151(0.015) (inter-event),

= 0.467(0.008) (intra-event) and T = 0.491(0.008) (total) for randomly-orientated
component considering spatial correlation (based on 50 runs). Numbers in brackets are
standard deviations of coefcients.
Use Vs,30 directly within amplication factor AFs of Boore & Atkinson (2008) (see Section 2.240).
Use same data and functional form as Hong & Goda (2007) (see Section 2.244).
Modify the one- and two-stage maximum-likelihood regression methods of Joyner &
Boore (1993) to consider spatial correlation of residuals from stations close together.
The spatial correlation is incorporated into the covariance matrix of residuals, associated
with both inter- and intra-event variability, via an empirical parametric spatial correlation
model.
Report results using two-stage approach but verify them using the one-stage method
(not shown).
Find that predictions of median ground motion not signicantly affected by accounting for
spatial correlation but s do change. When spatial correlation is considered, inter-event
decreases, intra-event increases and total decreases.
2.272 Kuehn et al. (2009)

Ground-motion model is a nonphysical function (subsymbolic) (polynomial) of predictor
variables (Mw , rjb , Vs,30 , fault mechanism and depth to top of rupture) with 48 coefcients (not reported) (14 for Mw , 5 for rjb , 4 for Vs,30 , 6 for rupture depth, 15 for combination of Mw and rjb , intercept parameter, pseudo-depth and 2 for mechanism). Use
polynomials because simple, exible and easy to understand.
Characterize sites using Vs,30 .
Reverse Rake angle between 30 and 150 . 19 earthquakes and 1870 records.
Normal Rake angle between 150 and 30 . 11 earthquakes and 49 records.
Strike slip Other rake angle. 30 earthquakes and 741 records.
Use data from NGA project because best dataset currently available. Note that signicant amount of metadata are missing. Discuss the problems of missing metadata.
Assume that metadata are missing at random, which means that it is possible to perform unbiased statistical inference. To overcome missing metadata only select records
where all metadata exist, which note is only strictly valid when metadata are missing
completely at random.
Select only records that are representative of free-eld conditions based on Geomatrix
classication C1.
Exclude some data from Chi-Chi sequence due to poor quality or co-located instruments.
248
Exclude data from rjb > 200 km because of low engineering signicance and to reduce
correlation between magnitude and distance. Also note that this reduces possible bias
due to different attenuation in different regions.
In original selection one record with Mw 5.2 and the next at Mw 5.61. Record with Mw 5.2
had a dominant role for small magnitudes so it was removed.
Discuss the problem of over-tting (modelling more spurious details of sample than are
supported by data generating process) and propose the use of generalization error (estimated using cross validation), which directly estimates the average prediction error for
data not used to develop model, to counteract it. Judge quality of model primarily in
terms of predictive power. Conclude that approach is viable for large datasets.
State that objective is not to develop a fully-edged alternative NGA model but to present
an extension to traditional modelling strategies, based on intelligent data analysis from
the elds of machine learning and articial intelligence.
For k -fold cross validation, split data into k roughly equal-sized subsets. Fit model to
k 1 subsets and compute prediction error for unused subset. Repeat for all k subsets.
Combine k prediction error estimates to obtain estimate of generalization error. Use
k = 10, which is often used for this approach.
Use rjb because some trials with simple functional form show that it gives a smaller
generalization error than, e.g., rrup .
Start with simple functional form and add new terms and retain those that lead to a
reduction in generalization error.
Note that some coefcients not statistically signicant at 5% level but note that 5% is an
arbitrary level and they result in lower generalization error.
Compare generalization error of nal model to that from tting the functional form of
Akkar & Bommer (2007b) and an over-t polynomial model with 58 coefcients and nd
they have considerably higher generalization errors.
After having found the functional form, ret equation using random-effects regression.
Note that little data for rjb < 5 km.
Note that weakness of model is that it is not physically interpretable and it cannot be
extrapolated. Also note that could have problems if dataset is not representative of
underlying data generating process.
Note that problem with magnitude scaling of model since available data is not representative of underlying distribution.
2.273
Mandal et al. (2009)

2
ln(Y ) = a + bMw ln(rjb
+ c2 )1/2 + dS
where Y is in g, a = 7.9527, b = 1.4043, c = 19.82, d = 0.0682 and = 0.8243.

249
S = 0 Rock/stiff. Relatively compact Jurassic formations. Believe that Vs,30 > 760 m/s.
S = 1 Soil. Alluvium or fragile Tertiary and Quaternary formations. Believe that 250
Vs,30 < 760 m/s.
Classify using geological information.
Fault ruptures mainly less than 40 km depth.
Use data from engineering seismoscopes (SRR) from 2001 Mw 7.7 Bhuj earthquake and
from strong-motion (20) and broadband (8) instruments of its aftershocks (3.1 Mw
5.6), which correct for instrument response. Earthquakes recorded at 3 to 15 stations.
All data from aftershocks from repi < 80 km and all data from mainshock from rjb
44 km.
Relocate earthquakes using local 1D velocity model. Report average error of 1 km in
epicenter and 1.5 km in focal depth.
Estimate seismic moments (from which compute Mw ) and other source parameters,
assuming Brune spectra, using spectral analysis of SH waves from transverse components. Report uncertainty of 0.050.1 units.
Report that faults well mapped so believe rjb s are quite reliable.
Plot residuals w.r.t. rjb . Find greater scatter in residuals for 0 rjb 30 km, which
could be related to amplication/noise in data from stations in Kachchh sedimentary
basin. Note lower scatter for range 100 rjb 300 km is unreliable due to lack of data.
State equation less reliable for 100 rjb 300 km due to lack of data.
Plot observations and predictions for Mw 3.5, 4.1, 4.5, 5.6 and 7.7 and nd fair match.
Note that insufcient data to judge relation between Mw 5.6 and 7.7. Find reasonable
match to six records from 29 March 1999 Chamoli earthquake (Mw 6.5) but poor match
(predictions lower than observations) to single record from 10 December 1967 Koyna
earthquake (Mw 6.3).
2.274 Moss (2009)

Ground-motion model is that of Chiou & Youngs (2008) (see Section 2.255). Also uses
same data. This model selected since sufciently complete and readily available at time
of analysis.
Notes that most GMPEs treat input variables as exact, neglecting uncertainties associated with measurements of Vs , Mw and r. These uncertainties propagate through
regression and result in model overestimating inherent variability in ground motion.
Presents method to estimate uncertainty of input parameters and incorporate it into regression procedure using Bayesian framework.
Follows on from Moss & Der Kiureghian (2006) (see Section 2.231).
Presents the Bayesian framework used for regression. This procedure is iterative and
leads to results that are slightly non-unique.
250
Uses the functional form and data of Boore et al. (1997) for feasibility study. Repeat
analysis of Boore et al. (1997) and conrm published results. Then assumes uncertainties on Vs,30 and rjb of coefcient of variation (COV) of 15% and nd that intra-event
reduces by 15 and 17% respectively. Also introduces uncertainty of standard deviation
of 0.1 on Mw and nds inter-event reduces by 20%. Overall nds reduction of 37%.
Finds that coefcients obtained are similar to those found with standard regression.
Discusses in detail the epistemic uncertainties associated with measurements of Vs
and the procedures and data used to quantify intra- and inter-method variabilities of
measurement techniques. Conclusions are used to estimate standard deviations for
each measurement of Vs,30 based on the measurement method, soil type and Vs,30 and
possible bias in measurements are corrected using derived empirical formulae.
Briey discusses epistemic uncertainties associated with estimates of Mw . Plots standard deviations of Mw estimates w.r.t. Mw for NGA database. Finds negative correlation, which relates to a number of factors. Regression on data gives M _M =
0.1820 ln(M ) + 0.4355, which is combined
with reported time component of standard
2
M
_M + Mt to give the overall uncertainty in
Mw . Notes that more work is needed to quantify uncertainty in Mw . Does not include
the uncertainty in Mw in regression results.
deviation Mt = 0.081 thus: M =
Discusses epistemic uncertainties in source-to-site distances and estimates different

components of uncertainty. Notes that more work is needed to quantify uncertainties
and, therefore, does not account for this uncertainty in regression.
Replicates results reported by Chiou & Youngs (2008). Then assumes an average Vs,30
measurement uncertainty of COV 27% and reports the decrease in (4%).
Compare results to approximate solutions from rst-order second-moment and Monte
Carlo techniques, which are useful since they are quicker than the full Bayesian regression. Find reasonable match in results.
Notes that the smaller s could have a large impact on PSHAs for long return periods.
2.275
Ptursson & Vogfjrd (2009)
Ground-motion model is
2
log10 (PGA) = a log10 (r + k10gM +eM ) + bM + c + dM 2

where PGA is in g, a = 2.26, b = 1.28, c = 2.85, d = 0.0437, e = d/a =
0.0194, g = b/a = 0.569, k = 0.0309 and = 0.302.
Detailed information on site conditions is not available hence do not include site terms
in model.
Focal depths between 0.04 and 9.49 km with most 6 km.
Use data from SIL national seismic network (3-component velocimeters) converted to
acceleration. Most instruments are short-period Lennartz sensors (7 with corner frequency of 1 Hz and 35 with corner frequency of 0.2 Hz). 6 to 8 broadband sensors
(CMG-3T, CMG-40T, CMG-ESP and STS2 with corner frequencies at 0.008 and 0.033 Hz).
251
Full-scale amplitude of stations between 0.3 cm/s and 1.25 cm/s. Hence, at near-source
distances records are often saturated and unusable. Most data have sampling rate of
100 Hz but some records are sampled at 20 Hz. First, remove instrument response.
Next, high-pass lter (for short-period records use cut-off of 0.15 Hz and for broadband
used 0.1 Hz). Finally, differentiate velocity to obtain acceleration. Do not use data sampled at 20 Hz nor data from distances > 100 Hz from Lennartz 1 Hz sensors.
Note that magnitudes of earthquaks with M > 3 are generally underestimated by SIL
system, which is designed to monitor microseismicity. Therefore, use 5 of 6 largest
earthquakes with teleseismic (Global CMT) Mw estimates to calibrate the local moment
magnitudes MLw used for study.
Develop model for use in ShakeMap and real-time aftershock hazard mapping applications.
Most earthquakes from the Hengill region in 1997 and 1998. 7 are on Reykjanes Peninsula and 6 in the South Iceland Seismic Zone (mainly from sequence in 2000, which
provides three largest earthquakes used).
Note that model of gstsson et al. (2008) is signicantly awed. Use same data
but remove data from Reykjanes Ridge and Myrdalsjokull because of uncertainties in
magnitude estimates for these earthquakes.
Data selected based on magnitude and number and quality of usable waveforms.
Most data from MLw 5 and repi > 20 km and distribution shows effect of saturation
of records for larger (MLw > 5) earthquakes for repi < 20 km. Correlation coefcient
between MLw and log repi is 0.24. 39% of data is from 5 to 50 km.
Also derive most using simpler functional form: log10 (PGA) = 2.08 log10 (r)0.0431M 2 +
1.21M 2.96 with = 0.304.
In SW Iceland large earthquakes usually occur on NS faults. Hence, examine effect of
radiation pattern. Add radiation pattern variable to model so that all earthquakes were
assumed to take place on NS-striking vertical strike-slip faults. Find that, as predicted by
theory, the coefcient multiplying this term was close to unity and standard deviation was
signicantly reduced. However, nd that this term led to worse t for some earthquakes
and so it was dropped.
Examine effect of instrument type using residual plots. Find that data from Lennartz
1 Hz sensors and Nanometrics RD3 0.5 Hz digitizers from > 100 km were lower than
predicted, which led to them being excluded.
Find that observations from hve station are consistently lower than predicted, which
relate to strong attenuation in Western Volcanic Zone. Make similar observations for
ada, bru and mok, which relate to propagation through crust and upper mantle of Eastern
Volcanic Zone. Find data from snb station is consistently higher due to strong Moho
relections from Hengill region earthquakes at about 130 km.
2 + k 2 + bM + c but nd very small k . Also try form

repi
of Fukushima & Tanaka (1990) but nd higher standard deviations.
Try form log10 (PGA) = a log10
252
Discuss the theoretical basis of coefcient g and its constraints w.r.t. a and b. Initial
regression with g as free parameter led to coefcients very close to g = b/a (PGA
independent of M at source) and, therefore, impose this as constraint.
Try weighted regression to correct for uneven magnitude and distance distribution but
these are dropped since data follows magnitude distribution expected in SW Iceland and
also run risk of putting too much emphasis on erroneous recordings.
Find that residuals are approximately normally (in terms of log10 ) distributed, using normal Q-Q plots.
Compare predictions and observations for some magnitude ranges and for each earthquake grouped by geographical region.
Fit log10 (PGA) = a log repi + . . . using only data from < 150 km and MLw > 4.7 and
nd a = 1.70. Relate difference in distance scaling to lack of far-eld data.
Believe that model can be used between 0 and 380 km.
2.276
Rupakhety & Sigbjrnsson (2009)
log10 (Sa ) = b1 + b2 Mw + b3 log10
d2 + b24 + b5 S
where Sa is in g, b1 = 1.038, b2 = 0.387, b3 = 1.159, b4 = 2.600, b5 = 0.123 and

= 0.287.
Rock Eurocode 8 site class A, Vs30 > 800 m/s. 64 records. S = 0.
Stiff soil Eurocode 8 site class B (21 records) or C (8 records), 180 < Vs30 < 800 m/s.
S = 1.
Most records from Mw < 6.6.
Assume magnitude-independent decay rate, linear magnitude dependency and no anelastic term because insufcient data to do otherwise.
Data primarily from south Iceland supplemented with records from Greece, Turkey and
Slovenia.
Exclude distant records because of low engineering signicance and to minimise differences in anelastic decay between regions.
Records from strike-slip earthquakes except for data from one oblique-faulting Icelandic
earthquake. Select earthquakes from extensional regimes.
Do not exclude data from buildings because of limited records. Exclude data from Thjorsarbru Bridge because they show clear structural effects and site dependent conditions
not characteristic of study area as a whole.
Records processed using individually-chosen lters.
253
Show comparisons between predicted and observed normalized ground motions w.r.t.
distance and conclude that selected functional form ts the data sufciently well.
Note that correlation matrix shows strong multi-collinearity between coefcients, which
implies imprecise estimates of regression coefcients meaning that outside the range of
the data predictions could be unreliable.
2.277 Akkar & Bommer (2010)

log y = b1 + b2 M + b3 M 2 + (b4 + b5 M ) log
2 + b2 + b S + b S + b F + b F
Rjb
7 S
8 A
9 N
10 R
6
where y is in cm/s2 , b1 = 1.04159, b2 = 0.91333, b3 = 0.08140, b4 = 2.92728, b5 =

0.28120, b6 = 7.86638, b7 = 0.08753, b8 = 0.01527, b9 = 0.04189, b10 = 0.08015,
1 = 0.2610 (intra-event) and 2 = 0.0994 (inter-event).
Normal FN = 1, FR = 0.
Strike-slip FN = 0, FR = 0.
Reverse FR = 1, FN = 0.
Use same data as Akkar & Bommer (2007b) (see Section 2.235) but repeat regression
analysis for pseudo-spectral acceleration (rather than for spectral displacement), assuming homoscedastic variability, reporting the coefcients to ve decimal places and
not applying any smoothing. These changes made due to shortcomings revealed in
GMPEs of Akkar & Bommer (2007b) after their use in various projects that required, for
example, extrapolation outside their magnitude range of applicability and work reported
in Bommer et al. (2007) (see Section 2.239) and other studies.
Examine total, inter- and intra-event residuals w.r.t. Mw and rjb and found no apparent
trends (shown for a selection of periods). Note that some plots suggest magnitudedependent variability but insufcient data to constrain it.
254
2.278
Akkar & agnan

(2010)
Ground-motion model is [based on base model of Abrahamson & Silva (1997, 2008)]:
ln(Y ) = a1 + a2 (M c1 ) + a4 (8.5 M)2 + [a5 + a6 (M c1 )] ln

+a8 FN + a9 FR + FS
for M c1
ln(Y ) = a1 + a3 (M c1 ) + a4 (8.5 M)2 + [a5 + a6 (M c1 )] ln

+a8 FN + a9 FR + FS
where FS
FLIN
FN L
FN L
FN L
2 + a2
Rjb
7
2 + a2
Rjb
7
for M > c1
= FLIN + FN L
)
(
VS30
= blin ln
Vref
( pga )
low
= bnl ln
for pga4nl 0.03 g
0.1
[ (
)]
[ (
)]
( pga )
pga4nl 2
pga4nl 3
low
+ c ln
+ d ln
= bnl ln
0.1
0.03
0.03
for 0.03 < pga4nl 0.09 g
(
)
pga4nl
= bnl ln
for pga4nl > 0.09 g
0.1
where Y is in cm/s2 , a1 = 8.92418, a2 = 0.513, a3 = 0.695, a4 = 0.18555,

a5 = 1.25594, a6 = 0.18105, a7 = 7.33617, a8 =0.02125, a9 = 0.01851, = 0.6527
(intra-event), = 0.5163 (inter-event) and T ot = 2 + 2 = 0.8322 and blin = 0.36,
b1 = 0.64 and b2 = 0.14 [taken from Boore & Atkinson (2008)]. Fix c1 = 6.5. pga4nl
is predicted PGA in g for Vs,30 = 760 m/s. See Boore & Atkinson (2008) for bnl , c and d
[not repeated by Akkar & agnan

(2010)].
Characterise sites using Vs,30 and use the site response terms of Boore & Atkinson
(2008) because of their simplicity and fairly good performance for data (demonstrated
by intra-event residual plots and their distributions that do not show clear trends, except
perhaps for Vs,30 > 720 m/s). Majority of records from NEHRP C (360 Vs,30
760 m/s) and D (180 Vs,30 < 360 m/s) sites with very few from sites with VS30
760 m/s. All sites have measured Vs,30 values.
Normal FN = 1, FR = 0. 28% of records.
Strike-slip FN = 0, FR = 0. 70% of records.
Reverse/thrust FN = 0, FR = 1. 2% of records.
Focal depths between about 0 and 50 km with most between 5 and 20 km.
Use data from the recently compiled Turkish strong-motion database (Akkar et al. ,
2010), for which the independent parameters were carefully reassessed.
Note that there are many singly-recorded earthquakes.
Vast majority of data from Mw < 6 and rjb > 10 km.
Explore several functional forms (not shown). Try to keep balance between rigorous
model (for meaningful and reliable estimations) and a robust expression (for wider implementation in engineering applications).
255
Data from 102 mainshocks (346 records) and 35 aftershocks (88 records).
Bandpass lter records using method of Akkar & Bommer (2006).
Compare PGAs from unltered and lter records and nd negligible differences.
Note that aim of study is not to promote the use of poorly-constrained local models.
Use pure error analysis (Douglas & Smit, 2001) to investigate magnitude-dependence
of . Find strong dependence of results on binning strategy (including some bins that
suggest increase in with magnitude) and, therefore, disregard magnitude dependency.
Derive GMPEs using data with minimum thresholds of Mw 3.5, Mw 4.0, Mw 4.5 and
Mw 5.0 to study inuence of small-magnitude data on predictions. Find that equation
using Mw 5.0 threshold overestimates PGAs derived using lower thresholds; however,
ranking of predictions from GMPEs using thresholds of Mw 3.5, Mw 4.0 and Mw 4.5 is
not systematic.
Note that due to limited records from reverse-faulting earthquakes, the coefcient a9
needs rening using additional data.
Examine inter-event residuals for PGA, 0.2 s and 1 s w.r.t. Mw and intra-event residuals
w.r.t. rjb and Vs,30 . Fit straight lines to residuals and also compute bias over ranges
of independent variables. Test signicance of trends at 5% level. Find no signicant
bias w.r.t. Mw nor w.r.t. rjb . For Vs,30 for 1 s nd signicant overestimation for Vs,30 >
450 m/s, which relate to linear site term. Suggest linear site term needs adjustment
using Turkish data.
Compute inter-station residuals and identify 9 outlier stations, which are those with residuals mainly outside range generally observed.
Examine bias of residuals for mainshock and aftershock records. Find weak evidence
for overestimation of aftershock motions but this is not signicant at the 5% level.
Combine Turkish and Italian data from ITACA (1004 records) and derive GMPEs using
same functional form, except using site classes rather than Vs,30 directly, to test observed
differences between local and global GMPEs.
Compare focal depth distributions, using histograms with intervals of 5 km, of the datasets
for various GMPEs. Compute mean and standard deviations of Mw for each depth bin.
Find that records from Turkey and Italian are on average deeper than those for other
GMPEs, which seems to explain lower observed motions. Conclude that focal depth
can be important in explaining regional differences.
2.279 Arroyo et al. (2010)

E1 (4 (T )R) E1 (4 (T ) R2 + r02 )
ln SA(T ) = 1 (T ) + 2 (T )Mw + 3 (T ) ln
r02
r02 = 1.4447 105 e2.3026Mw
256
where SA is in cm/s2 , E1 (x) is the exponential integral function, 1 = 2.4862, 2 =

0.9392, 3 = 0.5061, 4 = 0.0150, b = 0.0181, = 0.7500 (total), e = 0.4654
(inter-event) and r = 0.5882 (intra-event).
All data from rock (NEHRP B) sites. Data from stations with known, signicant site
amplication and those located in volcanic belt are excluded. Use H/V ratios to verify
that stations are all on generic rock. Data from 56 different stations.
Functional form is based on the analytical solution of a circular nite-source model and
body waves, which also denes expression for r0 (the radius of the circular fault based
on Brunes model) using a stress drop of 100 bar in order to keep functional form as
simple as possible. Note that functional form allows for oversaturation, whose existence
is questionable.
Select data of interplate, thrust-faulting events (interface) from permanent networks between 1985 and 2004 on the Pacic coast between Colima and Oaxaca (majority of data
from Guerrero but some data from other regions, especially Oaxaca). Data from neartrench earthquakes whose high-frequency radiation is anomalously low are excluded. To
focus on ground motions of engineering interest, exclude data from small (Mw 5.5)
with few records that are only from distant stations (R > 100 km). Exclude data from
> 400 km (use a larger distance than usual because of previously observed slow decay). To reduce potential variability of data, select only one record from two stations
recording the same earthquake at less than 5 km (based on visual inspection of data).
Data from 1219 bit digital accelerographs (66% of data), which have at response down
to less than 0.1 Hz, and 24 bit broadband seismographs (34% of data), which have at
response for velocities between 0.01 and 30 Hz. Broadband data mainly from Mw < 6
and distances > 100 km. Sampling rates between 80 and 250 Hz. Instrumental responses and sampling rates mean data reliable up to 30 Hz.
Roughly 45% of records from 20100 km. Only 16 records from < 25 km and only 5
from 3 earthquakes with Mw > 7 and, therefore, note that any anomalous records will
strongly inuence results in this distance range. State that more near-source data from
large Mexican interplate earthquakes needed.
Use Bayesian regression that accounts, for linear functions, for these correlations: 1)
intra-event, 2) between coefcients and 3) between different periods. To linearize function perform regression as: for a given period and value of 4 , compute coefcients
1 , 2 and 3 through Bayesian analysis and iterate for different values of 4 to nd
the value that gives best t to data. This is repeated for each period. Note that this
means the regression is not fully Bayesian. To obtain prior information on coefcients
1 , 2 and 3 use random vibration theory and theoretical expression for Fourier amplitude spectrum. Dene other required prior parameters (covariances etc.) using previous
studies. Smooth 4 w.r.t. period. Discuss differences between prior and posterior values
and not that nal results not over-constrained to mean prior values.
Find that model systematically overestimates in whole period range but since less than
5% consider bias acceptable.
Plot residuals w.r.t. Mw , distance and depth and nd no signicant trend. Note that even
though focal depth is not included in model there is no signicant dependence on it.
257
Adjust observed near-source PGAs to a common distance of 16 km and include data

from Mw 2.54.9 from rhypo between 16 and 37 km. Compare to predictions. Note the
large scatter (more than an order of magnitude) so note that statistical signicance is
low. Note that model matches observations reasonably well.
2.280 Bindi et al. (2010)

log10 Y
= a + b1 (Mw Mref ) + b2 (Mw Mref )2
+[c1 + c2 (Mw Mref )] log10 (R2 + h2 ) + ei Si + fj Fj
where Y is in cm/s2 and Mref = 4.5; a = 3.7691, b1 = 0.0523, b2 = 0.1389, c1 =

1.9383, c2 = 0.4661, h = 10.1057, C0 = 0, C1 = 0.2260, C2 = 0.1043, eve = 0.2084,
sta = 0.2634 and = 0.3523 for horizontal PGA using rjb ; a = 3.2191, b1 = 0.1631,
b2 = 0.0765, c1 = 1.7613, c2 = 0.3144, h = 9.1688, C0 = 0, C1 = 0.1938,
C2 = 0.1242, eve = 0.2080, sta = 0.1859 and = 0.3384 for vertical PGA using
rjb ; a = 3.750, b1 = 0.1180, b2 = 0.1147, c1 = 1.9267, c2 = 0.4285, h = 10.0497,
C0 = 0, C1 = 0.2297, C2 = 0.1022, eve = 0.2103, sta = 0.2666 and = 0.3555 for
horizontal PGA using repi ; and a = 3.2015, b1 = 0.2482, b2 = 0.0428, c1 = 1.7514,
c2 = 0.2588, h = 9.1513, C0 = 0, C1 = 0.1983, C2 = 0.1230, eve = 0.1917, sta =
0.1877 and = 0.3241 for vertical PGA using repi 15 .
C0 Rock. Corresponding to NEHRP A and B categories. 104 stations. S1 = 1 and

S2 = S3 = 0 .
C1 Shallow sediment: deposits thinner than or equal to 20 m and thicker than 5 m. Vs
of sediment lower than 800 m/s. 47 stations. S2 = 1 and S1 = S3 = 0.
C2 Deep sediment: deposits thicker than 20 m. 55 stations. S3 = 1 and S1 = S2 = 0.
Site classication performed using veried geological, geophysical and geotechnical information but of varying detail. Note that classication between C1 and C2 is a simple
but efcient method to identify sites with amplications at frequencies larger or smaller
than 25 Hz. Data from 206 different stations.
Use four faulting mechanism classes:
Normal 50 earthquakes.
Strike-slip 12 earthquakes.
Reverse 17 earthquakes.
Unknown 28 earthquakes.
Find that mechanism coefcients are not signicantly different than zero and, therefore,
remove them.
Focal depths between 0 and 29.21 km.
15
There is an inconsistency between the names given to the site coefcients in the tables of this article (C0 , C1
and C2 ) and those used to describe the functional form (e0 , e1 and e2 ).
258
Civil Protection, plus some data from the Northern Italy Strong Motion network (RAIS).
Records individually processed using acausal fourth-order Butterworth lters with cutoffs selected by visual inspection of Fourier spectra.
Select records with rjb < 100 km from earthquakes with Mw 4 recorded at two or
more stations.
Mw 6 are well sampled for rjb > 5 km, particularly for 4 Mw 4.6. No data from
rjb < 10 km from earthquakes with Mw > 6.
Compare PGAs and rjb for common records with Sabetta & Pugliese (1987). For PGA
nd similar values, indicating that the different processing applied results in consistent
results. For rjb nd signicant differences for distances shorter than 20 km, which attribute to improvements in knowledge of source geometries.
Examine inter-event and inter-station residuals. Find most inter-event errors are between 0.2 and 0.2 with a few events (e.g. 2002 Molise) with largely or over- or underestimated.
When comparing observations and predictions for Irpina (Mw 6.9) 1980 earthquake state
that comparisons unreliable for rjb < 10 km due to lack of data.
Compare predictions and observations for the 23/12/2008 (Mw 5.4) northern Apennines
earthquake mainshock to Parma and its Mw 4.9 aftershock (both with focal depth >
20 km and reverse mechanism), which were not used to develop GMPEs. 33 records
(32 repi 217 km) of mainshock and 26 (9 repi 217 km) records of aftershock.
Find the most observations fall within 1 but some for 30 repi 60 km are overestimated by up to one order of magnitude.
Note importance of improving site categorization to reduce sta .
2.281
Cua & Heaton (2010)
log Y
R1
= aM + b[R1 + C(M )] + d log[R1 + C(M )] + e
=
R2 + 9
C(M ) = c1 exp[c2 (M 5)][tan1 (M 5) + /2]

where Y is in cm/s2 , a = 0.73, b = 7.2 104 , c1 = 1.16, c2 = 0.96, d = 1.48,
e = 0.42 and = 0.31 for rock and a = 0.71, b = 2.38 103 , c1 = 1.72, c2 = 0.96,
d = 1.44, e = 2.45 102 and = 0.33 for soil.
Use two site classes using southern California site classication map based on Vs,30 of
Wills et al. (2000):
Rock Class BC and above, Vs,30 > 464 m/s. 35 SCSN stations with 958 records. 50
records from NGA.
Soil Class C and below, Vs,30 464 m/s. No data from very soft soils. 129 SCSN
stations with 2630 records. 1557 records from NGA.
259
and develop independent equations for each since sufcient data.

Use data from the Southern California Seismic Network (SCSN) (150 stations) and
COSMOS (6 events) supplemented by the Next Generation Attenuation (NGA) dataset.
Mainly used broadband data from SCSN except when clipped, when accelerometric
data is used instead.
Correct records for gain and baseline and convert to acceleration using differentiation, if
needed.
For SCSN data use S-wave envelope amplitudes and not PGAs directly. Note that
should be comparable to true PGAs.
Constrain c2 to be approximately unity within regression.
Develop conversion factors for converting between different denitions of horizontal
component and their s.
Compare predicted and observed PGAs for ranges: 6.5 < M < 7.5 (predictions for
M 7.0), 4.0 < M < 6.0 (predictions for M 5.0) and M < 3.0 (predictions for M 2.5) and
nd good match.
Examine residuals and nd no signicant trends w.r.t. distance or magnitude.
Compute station-specic site corrections for SCSN stations that recorded more than 3
times. Applying these corrections for rock PGA produces a 20% reduction in (to 0.24).
2.282 Douglas & Halldrsson (2010)

Ground-motion model is the same as Ambraseys et al. (2005a) (see Section 2.207) with
the addition of a term b11 AS, where AS = 1 for an aftershock record and 0 otherwise.
Find predicted motions from aftershocks slightly smaller than from mainshocks.
Examine total residual plots, biases and standard deviations of rederived GMPEs of
Ambraseys et al. (2005a) with magnitude-independent with earthquake classied as
aftershock or other. Do not nd signicant differences in residuals between aftershocks
and the rest of the data.
Discuss the use of aftershock data when developing GMPEs.
2.283 Faccioli et al. (2010)

log10 DRS(T ) = a1 + a2 Mw + a3 log10 (Rrup + a4 10a5 Mw )

+aB SB + aC SC + aD SD + aN EN + aR ER + aS ES
where DRS(T ) is in cm/s2 , a1 = 1.18, a2 = 0.559, a3 = 1.624, a4 = 0.018,
a5 = 0.445, aB = 0.25, aC = 0.31, aD = 0.33, aN = 0.01, aR = 0.09, aS = 0.05,
k1 = 2.03, k2 = 0.138, k3 = 0.962 and = 0.3616 .
16
Typographical error in article (EI should be ES ).
260
Use four Eurocode 8 classes:

A Rock. SB = SC = SD = 0.
B Stiff soil. SB = 1, SC = SD = 0.
C Medium-dense soil deposits. SC = 1, SB = SD = 0.
D Soft soil deposits. SD = 1, SB = SC = 0.
Normal EN = 1, ER = ES = 0.
Reverse ER = 1, EN = ES = 0.
Strike-slip ES = 1, EN = ER = 0.
Update of Cauzzi & Faccioli (2008) (see Section 2.254) using more data and rrup rather
than rhypo because this is more appropriate close to large earthquakes.
Find that differences between rrup and rhypo are not statistically signicant for Mw 5.7
so use rhypo below this threshold.
Most data from Japan.
Use a subset of data to decide on the best functional form, including forms with Mw2
and/or distance-saturation terms and site classes or Vs,30 directly.
Carefully examine (not show) t between predicted and observed spectra in near-source
region and nd distance-saturation term provides best t.
Note that Mw2 term has negligible impact on but improves predictions for large Mw .
Drops Mw2 from nal functional form.
Find site terms signicantly reduce .
Effect of style of faulting terms on is minimal but does improve predictions.
Note that functional form means that one-step rather than two-step approach must be
used that means that effects of magnitude and distance cannot be decoupled and s
are larger.
Compare predictions and observations for two records and nd overprediction in one
case and underprediction in other, which relate to the approximation of the model and
not an error in determination of coefcients.
Test model against data (4.5 Mw 6.9, rrup < 150 km) from the Italian Accelerometric Archive (ITACA) using residual plots and method of Scherbaum et al. (2004).
Find that good ranking is obtained using approach of Scherbaum et al. (2004). Find
trends in residual plots, which correct using functions, with coefcients k1 , k2 and k3 , t
to the residuals. ki can be added to ai to obtain corrected coefcients (a4 and a5 are
unchanged).
Note that improvements to Cauzzi & Faccioli (2008) are still ongoing.
261
2.284 Graizer et al. (2010)

(

)2
R
R
R
R
ln(Y ) = ln(A) 0.5 ln 1
+ 4D22
0.5 ln 1
+ 4D32
+
R2
R2
R3
R3
(
)2

)
(
Vs,30
R
R
0.5 ln 1
+ 4D52
bv ln
VA
R5
R5
[(
)2
A = [c1 arctan(M + c2 ) + c3 ]F
R2 = c4 M + c5
D2 = c6 cos[c7 (M + c8 )] + c9
R5 = c11 M 2 + c12 M + c13
where Y is in g, c1 = 0.14, c2 = 6.25, c3 = 0.37, c4 = 3.67, c5 = 12.42, c6 =
0.125, c7 = 1.19, c8 = 6.15, c9 = 0.525, c10 = 0.16, c11 = 18.04, c12 = 167.9,
c13 = 476.3, D5 = 0.7, bv = 0.24, VA = 484.5, R3 = 100 km and = 0.83.
Coefcients c4 , c5 , c10 c13 and D5 are newly derived as is the others are adopted
from GMPE of Graizer & Kalkan (2007). D3 = 0.65 for Z < 1 km and 0.35 for Z 1 km.
Use sediment depth Z to model basin effects.
Strike-slip and normal F = 1.00
Reverse F = 1.28
Update of GMPE of Graizer & Kalkan (2007) (see Section 2.245) to model faster attenuation for R > 100 km using more data (from the USGS-Atlas global database).
Compare data binned into 9 magnitude ranges with interval 0.4 and nd good match.
Note that large due to variability in Atlas database.
Using data binned w.r.t. Mw and into 25 distance bins (with spacing of 20 km) derive
these models for : = 0.043M + 1.10 and = 0.0004R + 0.89.
Examine residual plots w.r.t. distance, Mw and Vs,30 and nd no trends.
2.285 Hong & Goda (2010)

Ground-motion models are the same as Hong et al. (2009b) (see Section 2.270) for
interplate and inslab Mexican earthquakes and Hong & Goda (2007) and Goda & Hong
(2008) (see Section 2.244) for intraplate Californian earthquakes. Coefcients are: b1 =
1.271, b2 = 0.337, b3 = 0.0, b4 = 1.119, b5 = 0.063, h = 5.9, = 0.190, = 0.463,
2 + 6.32 )0.5 ] for major
T = 0.501 and PGAref = exp[1.0 + 0.446(Mw 7) 0.888 ln(rjb
principal axis and b1 = 0.717, b2 = 0.454, b3 = 0.009, b4 = 1.000, b5 = 0.041,
h = 5.0, = 0.182 (inter-event), = 0.441 (intra-event), T = 0.477 (total) and
2 + 5.62 )0.5 ] for minor principal
PGAref = exp[0.532 + 0.518(Mw 7) 0.846 ln(rjb
axis for intraplate California; c1 = 3.005, c2 = 0.555, c3 = 0.00392, c4 = 0.0079,
262
= 0.106 (inter-event), = 0.285 (intra-event) and T = 0.304 (total) for major

principal axis and c1 = 3.253, c2 = 0.575, c3 = 0.00380, c4 = 0.0079, = 0.121,
= 0.270 (intra-event) and T = 0.296 (total) for minor principal axis for inslab Mexican
earthquakes; and d1 = 0.396, d2 = 0.113, d3 = 0.00361, d4 = 0.0075, d5 = 0.474,
d6 = 0.0040, = 0.193 (inter-event), = 0.264 (intra-event) and T = 0.327 (total)
for major principal axis and d1 = 0.653, d2 = 0.125, d3 = 0.00356, d4 = 0.0075,
d5 = 0.474, d6 = 0.00177, = 0.200 (inter-event), = 0.273 and T = 0.339 (total)
for minor principal axis for interface Mexican earthquakes.
Similar analysis to that of Hong & Goda (2007) (see Section 2.244) and Hong et al.
(2009b) (see Section 2.270) concerning orientation of major response axis.
Conduct analyses for intraplate Californian (Hong & Goda, 2007) and interface and inslab Mexican data (Hong et al. , 2009b).
Discuss impact of different denitions of horizontal component on predicted ground motions and s for the three types of earthquake.
2.286
Jayaram & Baker (2010)
Ground-motion model is that of Campbell & Bozorgnia (2008b) (see Section 2.241).
Use same data as Campbell & Bozorgnia (2008b) (see Section 2.241).
Modify the random-effects regression method of Abrahamson & Youngs (1992) to account for spatial correlation dened by a pre-dened empirical model dependent on
separation distance or derived during the regression analysis. Prefer the use of a predened empirical model for various reasons.
To provide baseline model for comparison, ret model of Campbell & Bozorgnia (2008b)
using random-effects regression ignoring spatial correlation. Find minor differences with
reported coefcients of Campbell & Bozorgnia (2008b), which relate to manual coefcient smoothing.
Find intra-event increases and inter-event decreases but total remains roughly
the same when spatial correlation is accounted for. Provide theoretical justication for
difference in s if spatial correlation between records is considered or not.
Do not report coefcients, only provide graphs of s.
State that, because regression coefcients are not signicant different if spatial correlation is accounted for, the regression procedure can be simplied.
Discuss the implications of ndings on risk assessments of spatially-distributed systems.
263
2.287 Montalva (2010)

Ground-motion model is for combined model using both surface and borehole records17
(same as Boore & Atkinson (2008)):
A
med = Fm + Fd + Fsite Surf f lag + F100 S100f lag + F200 S200f lag
Fm = e1 + e5 (Mw Mh ) + e6 (Mw Mh )2 for Mw Mh
Fm = e1 + e7 (Mw Mh ) for Mw Mh
Fd = [c1 + c2 (Mw Mh )] log(R/Rref ) + c3 (R Rref )
2
2
R =
RRU
P +h
Fsite = blin ln(Vs30 /Vref ) + bh800 ln(h800/href )
F100 = a100 + b100 ln(Vs30 /Vref ) + c100 ln(Vshole/Vsholeref )
F200 = a200 + b200 ln(Vs30 /Vref ) + c200 ln(Vshole/Vsholeref )
where A
med is in g, Mref = 4.5, Rref = 1 km, Vref = 760 m/s, href = 60 m and
Vsholeref = 3000 m/s (reference values); c1 = 1.2534, c2 = 0.4271, c3 = 0.0140,
e1 = 0.0663, e5 = 0.5997, e6 = 0.5012, e7 = 0, blin = 0.4665, bh800 =
0.1801, a100 = 1.4372, a200 = 1.6518, b100 = 0.0269, b200 = 0.1884, c100 =
0.2666, c200 = 0.3793, = 0.6293 (intra-event) for Mw < 5, = 0.6202 (intraevent) for Mw > 6.5, = 0.4929 (inter-event) for Mw < 5, = 0.9164 (inter-event) for
Mw > 6.5 (linear interpolation of and between Mw 5 and 6.5) and = 0.4981 for
Mw > 6.5 (computed using inter-event residuals corrected for the observed bias using
a linear term) 18 .
Characterise sites by Vs,30 , depth to reach Vs of 800 m/s (h800) and Vs at bedrock
(Vshole).
Uses an NGA functional form to reect state of the art in ground-motion prediction and
the form of Boore & Atkinson (2008) specically because it can be constrained by the
data.
Extension of analysis by Rodriguez-Marek & Montalva (2010) (see Section 4.173).
Analysis conducted to investigate single-site variability of ground motions.
Data from KiK-net on surface and at depth as processed by Pousse et al. (2005) and
Cotton et al. (2008) (see Sections 4.126 and 2.256 for details). Note that although
Cotton et al. (2008) state that spectral accelerations up to 3 s can be used, in fact some
spectral accelerations at long periods are less than the number of decimals used for
storing the data. Hence limit analysis to periods < 1.3 s.
Majority of data is for Mw 6.1, which will have an impact on the regression.
Presents histogram of Vs,30 at surface stations: peak around 500 m/s with very few
records for Vs,30 > 1000 m/s.
Presents histogram of borehole depths: almost all at 100 and 200 m. Use ag to indicate
borehole instrument depth 150 m or > 150 m.
17
Same functional form is used for separate models using only surface and only borehole records but without
the ags indicating surface or borehole stations.
18
Mh not clearly stated in report but could be 5.6 (p. 150).
264
Presents histogram of Vs,30 at borehole stations: roughly uniformly distributed between

1000 m/s and 2500 m/s with some higher and lower.
Notes that geographical distribution of earthquakes shows clusters that could enable a
further separation of source and path effects from site effects in future studies.
Only uses data from earthquake that were recorded by 5 stations to adequately constrain inter-event residuals.
Uses multiple step regression method. First, uses only data from earthquakes recorded
by > 100 stations to constrain c3 and h by maximum-likelihood regression (after xing c1
to a value between 0.2 and 1.1 and xing c2 to 0). Next nds Mh and e7 . Originally
nd that e7 is negative (oversaturation) but note that lack of data from large earthquakes
so constrain it to be positive. Mh is chosen by inspection. Rest of coefcients found by
random-effects regression.
Combined model assumes source and path terms are independent of near-surface layering, which note is desired from a phenomenological view.
Plots inter-event residuals against Mw and nd overestimation for Mw > 6.5 due to
constraint that ground motions do not oversaturate. Plots inter-event residuals against
depth and nd that motions from deeper events underestimated, which relate to less
attenuation than shallower events and possibly different stress drops in shallow and
deep earthquakes.
Plots intra-event residuals against Mw and rrup and site parameters and nd no trends.
However, nds trend in residuals from earthquakes recorded at rrup < 20 km, which relate to lack of near-fault-effects and nonlinear soil terms. Also nds decreasing variation
in the intra-event residuals for rrup > 200 km.
Examines correlation between normalised inter- and intra-event residuals and concludes
that they are uncorrelated.
Finds combined, surface and borehole inter-event residuals well correlated.
Recommends use of combined model rather than the surface or borehole only models.
Computes single-station residuals and s, by dening site terms for each station based
on intra-event residuals, from 131 stations that recorded > 10 earthquakes. Finds slight
magnitude dependence of residuals. Finds no correlation between intra-event residuals
corrected by site terms and inter-event residuals. Examine in detail single-station residuals, associated s and their various components w.r.t. to their use in PSHA without
the ergodic assumption. Report these single-station s: surface = 0.4967, borehole
= 0.5060, surface = 0.6725 and borehole = 0.6684.
Examine effect of selecting data from a station-to-event azimuthal bracket of 8 and nds
that sigma is reduced.
2.288
Ornthammarath et al. (2010b), Ornthammarath (2010) &

Ornthammarath et al. (2010a)
log y = b1 + b2 Mw + b3 log(
2 + b2 ) + b S
rjb
5 S
4
265
where y is in m/s2 , b1 = 2.622, b2 = 0.643, b3 = 1.249, b4 = 3.190, b5 = 0.344,

event = 0.0723, station = 0.1198 and record = 0.1640.
S Stiff soil, 360 Vs30 750 m/s, 3 stations, 13 records, SS = 1
R Rock, Vs30 > 750 m/s, 28 stations, 68 records, SS = 0
Vs30 at most stations unknown so use local site conditions to classify stations. Note that
there are no deep alluvium soil deposits in Iceland so basin effects are limited.
All earthquakes have strike-slip mechanisms since the South Iceland Seismic Zone is
transform zone.
Develop model to investigate source of variability in ground motions in Iceland not for
seismic hazard assessments.
Data well distributed with respect to Mw , rjb and earthquakes (between 9 and 18 records
per event).
Only use high-quality data, following visual inspection.
Dropped anelastic attenuation term since it was positive.
Dropped quadratic magnitude term since insufcient data to constrain it, as were magnitudedependent distance decay terms.
Note that low could be due to limited records from only six earthquake of similar sizes
and mechanisms and from a small geographical area and few stations.
Examines single-station residuals (site terms) and single-station s for a few stations.
2.289 Ulutas & Ozer (2010)

Ground-motion model is19 :
log A = C1 + C2 Mw log10 (rrup + 0.0183 100.4537Mw ) + log C3 rrup

where A is in g, C1 = 2.7809, C2 = 0.5344, C3 = 0.0015 and = 0.392 (stated to
be in terms of natural logarithms although GMPE presented in terms of log).
Purpose of develop GMPE is for rapid assessment of PGA following earthquake and,
therefore, no distinction made between rock and soil sites.
All records from 1999 Kocaeli (Izmit) and Dzce earthquakes and their aftershocks from
132 permanent and temporary stations.
Earthquakes are mainly strike-slip but some have normal mechanisms. Believe that
model should only be used for these types of mechanisms.
19
Although rrup is used in Equation 4 of the paper it is probable that the distance metric is actually rjb since
they default to repi when the fault geometric is not known.
266
Baseline and instrument correct records. Examine Fourier amplitude spectra to select
the high- and low-pass lters. Use the Basic strong-motion Accelerogram Processing
(BAP) software: high-cut ltering with a cosine shape and then low-cut bi-directional
second-order Butterworth ltering (after padding with zeros).
Select data with Mw 4.
Distance saturation term (0.0183 100.4537Mw ) within the log10 given by square root of
rupture area estimated by regression analysis on areas for the two mainshocks and the
equations of Wells & Coppersmith (1994) for other earthquakes.
Compare observed and predicted PGAs for different Mw .
State that GMPE can be used for 4 Mw 7.5 and distances 200 km.
Note that site effects should be included within the model but currently lack of information.
267
268
Chapter 3
General characteristics of GMPEs for

PGA
Table 1 gives the general characteristics of published attenuation relations for peak ground
acceleration. The columns are:
H Number of horizontal records (if both horizontal components are used then multiply by
two to get total number)
V Number of vertical components
E Number of earthquakes
Mmin Magnitude of smallest earthquake

Mmax Magnitude of largest earthquake
M scale Magnitude scale (scales in brackets refer to those scales which the main M values were
sometimes converted from, or used without conversion, when no data existed), where:
mb Body-wave magnitude
MC Chinese surface wave magnitude
MCL Coda length magnitude
MD Duration magnitude
MJMA Japanese Meteorological Agency magnitude
ML Local magnitude
MLw Local moment magnitude reported by the Icelandic Meterological Ofce
MbLg Magnitude calculated using Lg amplitudes on short-period vertical seismographs
Ms Surface-wave magnitude
Mw Moment magnitude
rmin Shortest source-to-site distance
rmax Longest source-to-site distance
r scale Distance metric, where (when available the de facto standard abbreviations of Abrahamson & Shedlock (1997) are used):
269
rc Distance to rupture centroid

repi Epicentral distance
rE Distance to energy centre
rjb Distance to projection of rupture plane on surface (Joyner & Boore, 1981)
rhypo Hypocentral (or focal) distance
rq Equivalent hypocentral distance (EHD) (Ohno et al. , 1993)
rrup Distance to rupture plane
rseis Distance to seismogenic rupture plane (assumes near-surface rupture in sediments is non-seismogenic) (Campbell, 1997)
S Number of different site conditions modelled, where:
C Continuous classication
I Individual classication for each site
C Use of the two horizontal components of each accelerogram [see Beyer & Bommer
(2006)], where:
1 Principal 1
2 Principal 2
A Arithmetic mean
B Both components
C Randomly chosen component
D50 GMrotD50 (Boore et al. , 2006).
G Geometric mean
I50 GMrotI50 (Boore et al. , 2006).
L Larger component
M Mean (not stated what type)
N Fault normal
O Randomly oriented component
P Fault parallel
Q Quadratic mean,
& Goda, 2007)
(a21 + a22 )/2, where a1 and a2 are the two components (Hong
R Resolved component
S
(a1 + a2 )/2, where a1 and a2 are the two components (Reyes, 1998)
U Unknown
V Vectorially resolved component, i.e. square root of sum of squares of the two
components
V3 Vectorially resolved component including vertical, i.e. square root of sum of squares
of the three components
R Regression method used, where:
1 Ordinary one-stage
270
1B Bayesian one-stage (Ordaz et al. , 1994)

1M Maximum likelihood one-stage or random-effects (Abrahamson & Youngs, 1992;
Joyner & Boore, 1993)
1W Weighted one-stage
1WM Weighted maximum-likelihood one-stage
2 Two-stage (Joyner & Boore, 1981)
2M Maximum likelihood two-stage (Joyner & Boore, 1993)
2W Two-stage with second staged weighted as described in Joyner & Boore (1988)
O Other (see section referring to study)
U Unknown (often probably ordinary one-stage regression)
M Source mechanisms (and tectonic type) of earthquakes (letters in brackets refer to those
mechanism that are separately modelled), where:
A All (this is assumed if no information is given in the reference)
AS Aftershock
B Interslab
C Shallow crustal
F Interface
HW Hanging wall
I Intraplate
M Mining-induced
N Normal
O Oblique or odd (Frohlich & Apperson, 1992)
R Reverse
S Strike-slip
T Thrust
U Unspecied
+ refers to extra records from outside region used to supplement data. (. . . ) refer either to magnitudes of supplementing records or to those used for part of analysis. * means
information is approximate because either read from graph or found in another way.
271
272
-
U
8
U
678
58
Europe
181
140
California
W. USA
181
25
W. USA
Yonki,
New
Guinea
Unknown
Mostly W. USA but
100+ foreign
Papua
New
Guinea
W. USA
U
U
California & Japan

W. USA
V
-
H
46*
Area
W. USA
State that it is Richter magnitude which assume to be ML

Ambraseys & Bommer (1995) state that uses 38 earthquakes.
3
Ambraseys & Bommer (1995) state that uses larger component.
4
Note only valid for R 20 km
5
Esteva & Villaverde

(1973)
&
Esteva
(1974)
Orphal & Lahoud
(1974)
Ambraseys (1975b),
Ambraseys (1975a) &
Ambraseys (1978a)
Trifunac
&
Brady
(1975),Trifunac (1976)
& Trifunac & Brady
(1976)
Denham et al. (1973)
Reference
Esteva & Rosenblueth
(1964)
Kanai (1966)
Milne & Davenport
(1969)
Esteva (1970)
Denham
&
Small
(1971)
Davenport (1972)
Donovan (1973)
7.7
5.0
7.0
8.0
U
>8
U
U
U
U
Mmax
U
continued on next page
3.8
3.5
U2
57
4.1
5.2
<5
U
U
U
U
Mmin
U
31
25
U
U
U
8
U
U
E
U
Mostly ML
64 *
15
ML
ML
15*
80*
ML
U
U
3*
15*
U
U
U
rmin
15*
U
U
ML 1
U
U
M scale
U
Illustration 1: Characteristics of published peak ground acceleration relations
rmax
rhypo
rhypo
1
1
rhypo
rhypo
rhypo
1
1
rhypo
rhypo
C
1
rrup
repi
S
1
r scale
rhypo
4005 * repi
35
350
150*
300
U
450*
500*
U
U
U
450*
U3
U
U
U
U
U
U
C
U
U
U
U
U
U
U
R
U
A
A
A
A
A
A
M
A
301
- (70*)
52
51
2.1
U
3.7
3.7
4.5*
4 (5.3)
4.9
5.0
5.3
3.5
3.0*
Mmin
U
7.6
U
6.3
6.3
7.7
7.7 (7.8)
7.8
7.7
7.6
7.7
6.6
Mmax
U
>5.0
<7.9
U (59)
U
5*
14
17+*
25 (23)
23
10
22
U
U
E
U
11*
U
U
10*
5
(repi )
U
ML
ML
ML
<20
ML 14
Ms
U10
0.1
U7
15
14
1
0*
ML
U
mb
U9
rmin
U
M scale
ML
1
1
rhypo
rhypo
1
1
2
2
2
S
2
(1)
1
1
1
rhypo
rhypo
rhypo
rhypo
rrup
rE , rrup
and rhypo
rhypo
rhypo
rhypo
rhypo
r scale
rhypo
>200 repi
449
U
30*
190
(repi )
U
210*
U
342
321
125
380
30*
rmax
U
C
B
B
B
B
B
U
L
C
B
1,
O
1
U
1
1
U
U
U
U
O
R
U
A (T, TS, S,
SN, N)13
A
A
A
A
A
A
A
A
A
M
A
Total earthquake components (does not need to be multiplied by two) for magnitude and distance dependence. Uses 2713 underground nuclear explosion records for site
dependence.
7
Idriss (1978) nds magnitudes to be mixture of ML and Ms .
8
Total earthquake components (does not need to be multiplied by two)
9
Idriss (1978) believes majority are Ms .
10
Idriss (1978) nds magnitudes to be mixture of ML , mb and Ms .
11
Reported in Idriss (1978).
12
Does not need to be multiplied by two.
13
Assume dip-slip means normal mechanism.
14
State that it is Richter magnitude which assume to be ML
Japan
Iwasaki et al. (1980)
Many
100s
816
Worldwide
W. USA
70
1912
66
W. USA
Friuli, Italy
Friuli, Italy
478
70
63 (32)
V
-
59
34
200*
162
H
7956
Mostly W. USA &

Japan, some foreign
W. USA
Worldwide
Area
California & W.
Nevada
W. USA
W. USA
Europe & Middle
East
W. USA
McGuire (1978)
A. Patwardhan et al.
(1978)11
Cornell et al. (1979)
Faccioli (1979)
Faccioli & Agalbato
(1979)
Aptikaev & Kopnichev
(1980)
Blume (1980)
Donovan & Bornstein

(1978)
Faccioli (1978)
McGuire (1977)
Milne (1977)
Ambraseys (1978b)
Reference
Blume (1977)
Illustration 1: continued
273
274
Schenk (1982)
Brillinger & Preisler
(1984)
Joyner & Fumal (1984)
and Joyner & Fumal
(1985)
Kawashima et al.
(1984) & Kawashima
et al. (1986)
McCann
Jr.
&
Echezwia (1984)
Schenk (1984)
Xu et al. (1984)
182
182
182
113
3500
182
182
W. N. America
Europe + USA +
others
Unknown
W. N. America
W. N. America
197
83
3500
19
Japan
N. America + foreign
Unknown
N. China
V
75
-
224
Chiaruttini
&
Siro
(1981)
Joyner & Boore (1981)
Bolt & Abrahamson
(1982)
Joyner
&
Boore
(1982b) & Joyner &
Boore (1988)
PML (1982)
H
61
75
116
Europe & Mid.

East
W. N. America
W. N. America
Area
New Zealand
Japan
W. USA+8 foreign
Reference
Matuschka (1980)
Campbell (1981)
U
10
18
90
23
U
23
32
23
23
23
117
E
U
U
27
6.5
7.8
5.0+
2.5
4.5
7.9
7.7
6.5
7.7
7.7
7.7
7.7
7.8
Mmax
U
7.4
7.7
5.0
5.0
2.5
5.0
4.3
5.0
5.0
5.0
2.7
Mmin
4.5*
4
5.0
0.5
0.1
2
0.5
0.5
Mw (ML )
Ms
Ms
Mw (ML )
Mw (ML )
Mw
(ML
for
M < 6.0,
Ms
for
M 6.0)
Mw
2
10.1
5*
0.5
0.5
Mw (ML )
Mw (ML )
MJMA
rmin
U
6
0.08
M scale
U
U
ML
for
M < 6.0
and
Ms
otherwise
ML (mb )
rmax
600
157
550*
370
600
370
330
370
370
370
480
120
500
47.7
or
1
1
1
rhypo
repi
1
2
rrup
repi
rjb
rhypo
rrup
rhypo
rjb
rjb
2
1
rhypo
rjb
rjb
S
U
1
1
r scale
dhypo
rhypo
rrup
U
L
U
L
L
L
C
U
U
M
O
1
O
1M
2
O
R
1
1
O
A
A
A
A
A
A
M
A
A
A
73
93
U
U
38917
95
N.E. China
Tangshan region,
China
USA + Europe +
others
E. Australia
S. California
Plate boundaries16
Italy
W. USA + others
Mexico
Vicinity of San Salvador
Japan
Japan
87
119
V
-
U
16
82
203
H
182
Area
W. N. America
45
U
16
U
17
U
U
U
46
20
19
90*
E
23
U
8.1
U
6.8
5.4
U
7.4*18
6.9
7.8
5.3
7.5*
Mmax
7.7
U
5.6
U
4.6
1.7
U
5.0*
3.1
3.7
2.9
5.0*
Mmin
5.0
0.1
2.5
U
7*20
1.5,
1.5
Ms
ML
Ms
M 19
Ms
for
M 5.5,
ML otherwise
Mw
Ms
Ms
U
2
2*
MC
ML
U
282
U
5*
MJMA
16
repi
r scale
rjb
rrup
U
466
U
179,
180
rrup
rrup
rhypo
repi
Both rjb &
134
rhypo
U
rrup
200*21 rhypo 22
40
442.5 repi
50*
repi
500*
rmax
370
rmin
0.5
M scale
Mw (ML )
Reported in Joyner & Boore (1988).

Also derive equations for Japan subduction zones.
17
195 for subduction zone equations.
18
>7.5 for subduction zone equations.
19
Call magnitude scale Richter magnitude, which note is equivalent to Mw for M 8.3, ML for M < 5.9 and Ms for 5.9 M 8.0.
20
About 15km for subduction zone equations.
21
About 400km for subduction zone equations.
22
repi for subduction zone equations.
23
15
K. Sadigh (1987)23
Singh et al. (1987)
Algermissen et al.
(1988)
Annaka & Nozawa
(1988)
McCue (1986)
C.B. Crouse (1987)15
Krinitzsky
et
al.
(1987) & Krinitzsky
et al. (1988)
Sabetta & Pugliese
(1987)
PML (1985)
Reference
Brillinger & Preisler
(1985)
Kawashima et al.
(1985)
Peng et al. (1985b)
Peng et al. (1985a)
2
1
1
1
1
2
1
1
S
2
B
U
M
U
B
B
U
L
C
L
U
1
U
U
U
O
1
2
R
1M
A (S, R)
A
A
A
A
A
A (S, T)
A
A
M
A
275
276
25
24

Consider equations valid for Mw 8
Ambraseys (1990)
Ordaz et al. (1989)

Alfaro et al. (1990)
W. N. America + 3
from Managua
Guerrero, Mexico
Guatemala,
Nicaragua
&
El Salvador
W. N. America
197+389
Worldwide
subduction zones
Campbell (1989)
Japan
75%+ California,
rest foreign
<227
Kanto (Japan)
Abrahamson & Litehiser (1989)
120
62
120
S.E. Australia
Europe
McCue et al. (1988)

Petrovski & Marcellini
(1988)
Tong & Katayama
(1988)
Yamabe
&
Kanai
(1988)
182
U
20
190
585
25+
182
S.W. W. Australia
W. N. America
Gaull (1988)
Joyner & Boore (1988)
585
486+200
W.
Japan+200
USA
Fukushima
et
al.
(1988) & Fukushima &
Tanaka (1990)
V
-
H
U
Campbell
Area
Worldwide
Reference
K.W.
(1988)24
23
U
12
91
76
60
22
<27
U
46
12+
23
U
4.1
2.9
5.0
5.3
4.5*
0.5*
3
2.6
5.0
4.6(5.0)
Mmin
5
U
7.5
5.0
8.1
8.1
(8.2)25
7.9
7.9*
6*
7
6.9
7.7
8.2(7.7)
Mmax
U
5.03
7.7
28+15
E
U
U
1
0.5
U
Ms
Mw (ML )
0.6
0.08
15*
(20*)
Mw
mb )
Ms
for
Ms 6.0,
ML (mb )
otherwise
ML
U
(Ms ,
5*
8
ML
U
10*
2.5
0.5
ML
Mw (ML )
16
(0.1)
rmin
U
M scale
ML
for
M < 6.0
and
Ms
otherwise
Ms
(MJMA )
rhypo
370
U
27
18.3
400
rjb
repi
U
1
repi
U
7.5
rrup
1
1
repi
repi
rhypo
rhypo ,
rrup for 2
Japanese
& all US
rhypo
rjb
1
2
S
2
r scale
rseis
450* rrup , rhypo

(450*) for Mw .
750*
833
200
175
370
303
(48)
rmax
<50
U
L
U
L,
O
U
L
C
M
U
U
1W
O
1
O
2W
R
U
A
A
A (R & RO,
I)
A (B,F)
A
A
A
A
M
A (S, R)
Worldwide
intraplate regions
E. N. America
Whittier Narrows
area
Iceland
Worldwide
Europe & Mid.
East
Dahle et al. (1990b) &

Dahle et al. (1990a)
Jacob et al. (1990)
Sen (1990)
960+4
California with 4
foreign
367
3.8 (6.8)
3.1
4.8
U
4.9*
4
2.2
1.8
2.9
Mmin
U
27
7.4 (7.4)
5.0
0.1
(3)
Mw
>8
U
3*
1
820
20
12*
rhypo
for
rhypo for
small ones
305
rrup
(172)30 some,
1
1
1
rhypo
U26
S
1
r scale
rseis
rjb
rrup
rjb
for
Ms & 6.0,
repi otherwise
>866 rE , rhypo
for M <
7.5
U
repi
U
150*
H:313,
V:214
21*
1300
rmax
U
rmin
U
mbLg
Mw (Ms ,
MJMA )
U
Mw
Ms
5.827
7.4
7.34
8.2
ML
M scale
ML
for
M < 6,
Ms
for
M 6
Ms (ML ,
mb , MCL )
mb
3.5
6.4
7.8
Mmax
U
119+2
<51
H:219,
V:191
11
56
E
U
Free (1996) believes it is rhypo .

This is Ms .
28
Total number of components, does not need to be multiplied by two.
29
Also present equations for SSE (using 140 records) and NE Iberia (using 107 records).
30
Equations stated to be for distances up to 100 km
26
57
Iberia29
Garca-Fernndez
& Canas (1991) &
Garcia-Fernandez &
Canas (1995)
Geomatrix
Consultants (1991), Sadigh
et al. (1993) & Sadigh
et al. (1997)
69728
<217
529
Worldwide
subduction zones
459
72*
V
-
87
H
U
Crouse (1991)
Sigbjrnsson (1990)
Tsai et al. (1990)
Ambraseys & Bommer
(1991) & Ambraseys &
Bommer (1992)
Area
Unknown
Reference
Campbell (1990)
U
M
L
C
U
U
U
1, 2
1M
R
U
A(R,S)
A
T (S,O)
A
A (T)
M
A
277
278
Worldwide
Rogers et al. (1991)
32
31
357
262
Iceland
504
48932
1241
234
80
V
-
USA + Europe +
others
Japan
Mainly Italy and

former Yugoslavia
Unknown
Distance to centre of array


Abrahamson
&
Youngs (1992)
Ambraseys
et
al.
(1992)
Kamiyama
et
al.
(1992) & Kamiyama
(1995)
Sigbjrnsson & Baldvinsson (1992)
236
SMART-1
Taiwan
array,
112
80
Taiwan
New Zealand
Loh et al. (1991)

Matuschka & Davis
(1991)
Niazi & Bozorgnia
(1991)
572
Unknown
I.M. Idriss (1991) reported in Idriss (1993)
H
383+25
Area
W. USA with 25
foreign
Reference
Huo & Hu (1991)
39
82
45
78
180*
12
63
30
30*
E
14+2
6.0
7.9
6.87
8*
8.1*
7.8
7.1
U
7.4
Mmax
7.4 (7.3)
2.0
4.1
3.1
3*
5.3*
3.6
4.0
U
4.6
5.0
Mmin
3.4
MJMA
0.5
Ms
10*
4*
rjb , repi for

some
rhypo
rrup
if
have,
rhypo
otherwise
80
rjb
413.3 rhypo
39
500*
400*
119.731 rhypo
1
3
rrup , rhypo
for M < 6
3.131
100
S
2
r scale
rjb
178.3 rhypo
U
U
rmax
227
(265)
5.0
U
rmin
0.1
ML
(MD ) for
ML < 6.6,
else Ms
ML
for
M 6,
Ms for 6 <
M < 8
and
Mw
for M 8
ML
M scale
ML or mb
for M <
6.0 and Ms
otherwise
ML
for
M < 6,
Ms
for
M 6
ML
U
B,L
L
B
C
B
1M
2W
U
U
R
O
A (U, U)
A
A
M
A
256
89
Israel
New Zealand
Nicaragua,
El
Salvador & Costa
Rica
Worldwide
New Zealand
271
27
31*
20
18
36+4
40
27
E
12
U35
34
7.6
7.3
5.1
3.0
5.1
3.9
7.7
5.134
7.4
7.0 (7.5)
6.6
7.6
Mmax
7.4
6.0
4.5 (7.2)
3.0
Mmin
6.1
Total number of components does not need to be multiplied by two

Boore et al. (1997) revise this magnitude to 5.87. New minimum magnitude is 5.2.
35
Considers equation valid for M 4.7.
36
Considers equation valid for d 300 km.
33
Dowrick & Sritharan

(1993)
Gitterman
et
al.
(1993)
& McVerry et al.
(1995)
Singh et al. (1993)
201
W. USA with 4 foreign

W. N. America
105+1633
Greece+16 foreign
89
V
-
H
136
137
Theodulidis & Papazachos (1992)

Abrahamson & Silva
(1993)
Boore et al. (1993),
Boore et al. (1997) &
Boore (2005)
Campbell (1993)
Tento et al. (1992)
Area
Nicaragua,
El
Salvador & Costa
Rica
Italy
Reference
Silva & Abrahamson
(1992)
Taylor Castillo et al.
(1992)
3.2
ML
U
U
13
ML
Mw
Ms
ML
for
M < 6.0
and
Ms
otherwise
U
Mw
1
(48)
0.6*
Ms
Ms , Mw ,
MJMA
Mw
rmin
3*
M scale
Mw
rrup
210
312
U36
rhypo
rc or rhypo
rjb
repi
rseis
S
2
rjb
for
ML 5.7,
repi otherwise
repi
rhypo
r scale
rseis
118.2 rjb
128
(236)
100*
170
210
rmax
100*
L,
G
C
G
2M
1M
R
1M
A, R
A (T,S)
A (S, R)
M
A (S,R)
279
280
250+
300
W. USA
Romania
UK + 30* foreign
84*
131+114
83
Iran
Yunnan, China +
114 W. N. America
Himalayan region
106
Romania
83
284
20
4
4+16
11
U
42
20 (9)
76
E
U
42+1
5.7
7.2
7.6*
7.2(ML )
or
7.5(Mw )
7.7
6.7(ML )
or
7.0(Mw )
5.1
2.5*
7.4
3.5 (6.4)
7.4
6.3
3 (3.7)
5.8
7.7 (7.4)
5.137
(5.3)
U
5.0
U
7.7
7.7
Mmax
5*
7.7
5.0
Mmin
U
4.1
3*
MB
350*
120*
2*
Ms (ML )
repi
rhypo for
most, rrup
for 1942
repi
180
U40
rjb
Ms 41
100
rhypo
>477.4 rhypo
(200*)
320* rhypo
Mw
38
1
1
1
I
rjb , repi
118.2 rjb
(109)
U
rhypo
400* rhypo
375
S
1
C
r scale
repi
repi
U
70*
(>1.3)
90*
U
60*
ML
MJMA
rmax
U
150*
Mw
ML
rmin
U
2*
Mw
M scale
U
ML
for
M < 6,
else Ms
Ms
Boore et al. (1997) revise this magnitude to 5.87. New minimum magnitude is 5.2.
Coefcients given in Boore et al. (1994b)
39
Free (1996) believes it is largest horizontal component.
40
It is not clear whether use Richter magnitude (ML ) or Mw .
41
Some may be mb because in their Table 1 some earthquakes to not have Ms given but do have mb . If so new minimum is 5.0.
42
They state it is closest distance from the exposure of ruptured part of the fault, instead of focal distances so may not be rupture distance.
37
Aman et al. (1995)
Xiang & Gao (1994)
Radu et al. (1994),

Lungu et al. (1995a) &
Lungu et al. (1996)
Ramazi & Schenk
(1994)
125
-
U
285
Unknown
3 vertical arrays in
Japan
15 + 30*
271 (70)
W. N. America
947
Worldwide
V
-
Ambraseys & Srbulov

(1994)
Boore et al. (1994a) &
Boore et al. (1997)
El Hassan (1994)
Fukushima
et
al.
(1994) & Fukushima
et al. (1995)
Lawson & Krawinkler
(1994)
Lungu et al. (1994)
Musson et al. (1994)
H
U
150+1
Area
Worldwide
Reference
Steinberg et al. (1993)
Sun & Peng (1993)
U
U39
L,
G
U
B
C
U
R
1
O
1M
1M,
2M
1
1,2
2W
R
1
1
A
A
A
A
A (R,S)38
M
A
A
Europe
East
Europe
East
Turkey
et
al.
Simpson
Ambraseys
(1996) &
(1996)
Ambraseys & Simpson (1996) & Simpson

(1996)
Aydan et al. (1996)
&
Mid.
27*
422
106
1926
23*
417
1926
V
620
19*
157
157
33
387
297
72
E
334
7.9
7.9
3.5*
7.6*
4.0
4.0
5.9
7.2(ML )
or
7.5(Mw )
7.8*
6.7(ML )
or
7.0(Mw )
4.1*
2.8
7.7
8*
Mmax
7.3
1.7
3*
Mmin
4.0
8*
MJMA
10*
Ms
(unspecied)
Ms
Mw (mb ,
ML , Ms )
Ms
(unspecied)
200+
44
rhypo
rhypo
r scale
rjb
for
Ms > 6.0,
repi otherwise
rhypo
350*
260
260
820
rjb
for
Ms > 6.0,
repi otherwise
rjb
for
M > 6.0,
repi otherwise
rhypo
rhypo
otherwise
rjb or repi
1000* rrup
for
2
earthquakes,
490*
6*
Mw (Ms ,
mb ,MD )
Usually
ML
for
M 6.5
and Ms for
M > 6.5
U43
rmax
260*
rmin
0*
M scale
Ms
It is not clear whether use Richter magnitude (ML ) or Mw .

Also derive equations for Australia and N. E. China
45
Ambraseys et al. (1996) state it is two-stage of Joyner & Boore (1981) but in fact it is two-stage method of Joyner & Boore (1988).
46
Ambraseys et al. (1996) state it is two-stage of Joyner & Boore (1981) but in fact it is two-stage method of Joyner & Boore (1988).
43
E. N. America44
Sarma & Free (1995)
Mid.
77
Japan
Molas &
(1995)
&
2166
Romania
Lungu et al. (1995b)
Yamazaki
W. N. America
Lee et al. (1995)
280
Cen. America
Dahle et al. (1995)
H
830
Area
Europe and Mid.
East
Reference
Ambraseys (1995)
9,
3
S
1
C
L
M
A
2W46 A
2W45 A
1B
R
2W
281
282
-
86
128
19047
645
51
Worldwide
Himalayas
Worldwide extensional regimes
Romania,
Bulgaria & former
Yugoslavia
Worldwide
Hawaii
Sarma
& Srbulov
(1996)
Singh et al. (1996)
Spudich et al. (1996)
& Spudich et al.
(1997)
48
47
95
225
Total number of components. Does not need to be multiplied by two.

Called Richter magnitude.
Campbell
(1997),
Campbell
(2000),
Campbell (2001) &
Campbell & Bozorgnia
(1994)
Munson & Thurber
(1997)
350
Italy
Romeo et al. (1996)
U
248
478
558
Turkey
California
238
Inan et al. (1996)

Ohno et al. (1996)
V
-
Crouse & McGuire

(1996)
Free (1996) & Free
et al. (1998)
H
36
Area
El Salvador &
Nicaragua
Cen. & S. California
Stable continental
regions
Reference
22
H:47,
V:26
5
30
114
17
U
17
4.0
4.7
6.1
5.7
5.10
3.9
4.6*
U
5.0
1.5
6.0
Mmin
3.7
7.2
H:8.0,
V:8.1
7.2
7.2
6.90
7.7
6.8*
U
7.5
6.8
7.7
Mmax
7.0
H: 222,
V: 189
16
E
20
Mw
rmax
Mw
Ms
for
Ms 6.1,
ML otherwise
ML 48
10*
88
60
310*
rjb
rseis
repi
33.15 340.97 rhypo

0
102.1 rjb
mb
Mw
Ms
179,
180
213
rjb
for
some, repi
for most
repi
rq for M >
5.3, rhypo
otherwise
Both rjb &
repi
rjb & repi
rrup
r scale
rhypo
1.5,
1.5
1
U
99.6
820
211
260
Mw
Mw (ML )
U
7.2
0.1
Ms
rmin
62
M scale
Ms
1
2
1
2
S
1
B,
L
U
G,
O
U
B
C
L
2M
1
2M
U
2M
1W
R
U
A(S,R,N)
A
NS
A
A
R,S (R,S)
M
A

-
<1546
69052
66
1546
Worldwide
Indian Himalayas
Switzerland
+
some from S.
Germany
50
Includes some not used for regression

Total number of components do not need to be multiplied by two.
51
52
Total number of components do not need to be multiplied by two.
49
Sarma &
Srbulov
(1998)
Sharma (1998)
Smit (1998)
Rinaldis et al. (1998)

Sadigh & Egan (1998)
80*
-
80*
27650
20+
Friuli
N.W. Balkans
University
City,
Mexico City
Italy & Greece
California with 4
foreign
137*
960+4
U
U
Korea
Algeria
46149 +66
Baag et al. (1998)

Bouhadad
et
al.
(1998)
Costa et al. (1998)
Manic (1998)
Reyes (1998)
Zhao et al. (1997)
V
-
476
182
200
W. N. America
Costa Rica
Worldwide
subduction zones
NZ with 66 foreign
H
U
Area
New Zealand
Reference
Pancha
&
Taber
(1997)
Rhoades (1997)
3.9
4.5
3.8
1.3*
4
U
U
5.6
5.08
5.0
5.0
3.3
Mmin
U
M scale
U
7.7
7
7.4
4.3*
7
U
U
6.1
7.7
7.6
7
0.1
Ms or Mw
Mw
ML
8
1
3*
U
U
MD
Ms
Mw
Ms (U)
U
20
11
(0.1)
8.5
0.5
6.1
rmin
U
Ms
Mw (ML )
Mw (Ms ,
mb , MD )
8.2
Mw
(Ms ,mb )
7.23(7.41) Mw
Mmax
U
5
5.5
6.6
H:
2.0
5.1
<120,
V: 120
113
24*
119+2
20*
56
20+
U
2
49+17
164
23
57
E
U
r scale
rhypo
rhypo
rrup
248
290
197
2
2
2
1
1
rhypo for
small ones
rjb , repi
rhypo
rhypo
1
2
I
for
138
repi
30551 rrup
some,
66*
U
U
1
1
550.9 rrup , rhypo

for some
573
rrup
for
(10)
some, rc
for most
U
repi
70
rhypo
S
1
1
3
370
rjb
182.1 rhypo
rmax
U
L
U
U
G
U
L,
M
U
B
S
L
L,
B
G
C
U
1W
2
O
U
1
1
U
U
1
1M
O
O
R
2
A
A
A (N,ST)
A(R,SN)
A
A
A
A
A
A(R)
NT
A
A
M
A
283
284
142

Tangshan,
N.
China
Iran
96058
447
S California
2823
Worldwide
Worldwide
186
Worldwide
468
44
8856
Iceland
94159
2823
183
468
304
610+66
W. N. America
NZ with 66 foreign
V
U
H
U
Area
Mediterranean region53
54
Also derive equations for Spain.

Also derive equations using ML .
55
Also derive equations using rhypo .
56
Total number of components. Does not need to be multiplied by two.
57
Equation given in terms of log M0 .
58
Equation for corrected PGA uses 443 records.
59
Equation for corrected PGA uses 439 records.
60
Equation for corrected PGA uses data from 36 earthquakes.
53
Ambraseys & Douglas (2000), Douglas

(2001b) & Ambraseys
& Douglas (2003)
Bozorgnia
et
al.
(2000)
(2000)
Field (2000)
Zar et al. (1999)
Wang et al. (1999)

Spudich et al. (1999)
Reference
Cabaas et al. (1999),
Cabaas et al. (2000)
& Benito et al. (2000)
Chapman (1999)
Cousins et al. (1999)
28
4960
48
44
47*
39
17
5.1
4.7
4.7
5.83
2.7
3.7
5.1
3.4
5.0
5.17
Mmin
2.5
M scale
Ms 54
7.5
7.7
7.7
7.8
7.4
4.9
7.2
5.9
Mw
Mw (Ms ,
mb , ML )
Ms
Mw
2.1
Ms (ML )
1*
Mw
Mw
0.1
0.1
rmin
0
Mw 57
7.7
Mw
7.09(7.41) Mw
Mmax
7.0
23
25+17
E
U
rmax
r scale
repi 55
148.9 rjb
rseis
rseis
60
60*
15
rhypo (rrup
for 2)
rjb
repi
rjb
224
41.3
99.4
189.4 rjb
400
rrup
for
some, rc
for most
112
repi
250
C
(6)
3
3
S
4
G,
O
L
G
U
C
L
1M
2M
1M
2M
U
R
1
A (R, S, O)
A (S,R,T)
A (R,S,T)
R, RS & S
NS
A
A(R)
M
A
424
993
993
48
68
60
46
102
U
23
4563 ,
1964
26
U+7*
5
21
1061
E
3
62
Assuming they use same data as Monguilner et al. (2000b).

Smit et al. (2000) give rhypo but this is typographical error (Smit, 2000).
63
Shallow crustal records.
64
Subduction records.
61
Taiwan
Shallow
crustal
worldwide (mainly
California)
Chen & Tsai (2002)

Gregor et al. (2002a)
1941
Dinarides
Japan
Pacic coast
Mexico
Taiwan
Herak et al. (2001)

Sanchez
&
Jara
(2001)
Wu et al. (2001)
of
182
472063 ,
252864
W. N. America
Taiwan
145
3011
-
84
1332
Caucasus
Japan+166 foreign
66
-
145
3011
U
856
5461
Indian Himalayas
Japan
W. Argentina
V
-
H
32 (117)
Sharma (2000)
Si
&
Midorikawa
(1999, 2000)
Smit et al. (2000)
Takahashi
et
al.
(2000)
Wang & Tao (2000)
Chang et al. (2001)
al.
et
Japan
al.
Kobayashi
(2000)
Monguilner
(2000a)
et
Area
Central Himalayas
Reference
Jain et al. (2000)
U
7.4
7.6
6.8
6.3
U
7.7
7.063 ,
6.364
7.1
8.3* (8*)
6.6
8.3
7.4
7.8
Mmax
7.0
U
4.4
4.8
4.5
3.7
U
5.0
4.163 ,
4.664
4.0
5* (5.8*)
5.5
5.8
4.361
5.0
Mmin
5.5
3*
4*
U
0.05*
U
0.1
Mw (ML )
ML
Mw
4
1*
(0.1*)
0.5
063 ,
40.264
Mw (ML )
Mw
(ML
for
ML < 6.5)
ML
MJMA
Ms
Ms
Mw
8
0*
1161
0.9*
Mw
Ms if ML
& Ms > 6,
ML otherwise
U
Mw
rmin
2 (4)
M scale
U
rmax
rrup (repi
for some)
repi
rhypo
repi
U
rhypo
267.3 rrup
400*
200*
600*
U
230
repi 62
300* rrup , rhypo
(100*) for some
370
rjb
264.463 ,repi 63 ,
272.464 rhypo 64
rrup
Both rq &
1
&
I
1
2
1
4
1
2
1
1
4
1
2
248
280*
rhypo
S
1
r scale
repi
35061 rhypo
152
(322)
400*
U
U
L
B
U
L
G
L
G
C
U
O
1M
2
2
U
O
2
2
O
1W
O
1W
1M
R
1
A
A (S, R/O,
T)
A
A
BF
A
A
A
A
A
A
M
T
285
286
66
65
161
47
49
1200+
4028
Umbria-Marche
Colombia
Syria
Subduction zones
N. California
422
U
683
19066
249
Indian Himalayas
N.W. Turkey
Romania
Europe
Mid.
744
Greece
&
160
Iran
Europe
East
H
9365
Area
Turkey
49
-
U
683
-
160
V
-
157
104
U
10
43*
15
U
U
4
51
142
28*
E
19
7.9
7.1
6.7
5.8
8.3
5.9
U
7.2
7.2
7.9
7.0
7.4
Mmax
7.4
4.0
3.3
4.0
3.5
5.5
4.5
U
0.9*
6.1
5.5
4.5
3.4*
Mmin
4.5
This is total number of horizontal components used. They come from 47 triaxial records.
This is total number of components. Does not need to be multiplied by two.
Margaris
et
al.
(2002a) & Margaris
et al. (2002b)
Saini et al. (2002)
Schwarz et al. (2002)
Stamatovska (2002)
Tromans & Bommer
(2002)
Zonno & Montaldo
(2002)
Alarcn (2003)
Alchalbi et al. (2003)
Atkinson & Boore
(2003)
Boatwright
et
al.
(2003)
Reference
Glkan
&
Kalkan
(2002)
Khademi (2002)
0.1*
Mw
(m b
for Ms < 5
and
Ms
otherwise)
Mw
49.7
21
11*
Ms
MCL
Mw
1*
2*
ML
Mainly
Mw , ML
for some
Ms
(unspecied)
U
0*
10*
1
U
ML
U
Ms
rmin
1.20
M scale
Mw
rmax
repi
rjb
for
Ms > 6.0,
repi otherwise
260
370*
rhypo
2
1
2
4
repi
repi
repi
rjb
U
3
1
3
rjb , repi for

M < 5.9
S
3
r scale
rjb , repi
322.4 rhypo
400
rhypo
550* rrup
100*
U
250*
310*
359
150
180*
150
U
U
C
U
U
B
L
C
L,
R
L
1M
U
1
1M
U
1
1
2
R
1
A (S, R, N)
A
A
F, B
N, O
A
A
A
A
M
A
&
Glkan
112
31
100
31
V
43968
57
47
192
398
225
U
U
12
E
3669
7.4
7.4
6.5
4.5
6.3
7.0
U
7*
6.6
Mmax
7.7
4.0
4.2
2.5
1.1
4.5
U
5*
4.1
Mmin
4.7
250
250
U
1*
1.5*
1.7
U
U
3*
1.2
1.2
Ms
Mw or Ms
Mw (ML )
MD (mb ,
Mw )
Mw
ML
ML
Mw (unspecied
scales)
Mw (unspecied
scales)
25*
450
150*
U
500*
300*
5*
MLw
rmax
60*
rmin
2*
M scale
Mw
rjb ,
repi
for
small
events
rjb ,
repi
for
small
events
rhypo
repi
repi
1
1
S
4
rhypo
rhypo
rjb if available, repi
otherwise
repi
r scale
rseis
L70
68
U
1
R
1
U
L
C
G
There are 960 components for uncorrected PGA.

There are 941 components for uncorrected PGA.
69
For horizontal corrected records. There are 49 for horizontal uncorrected PGA. There are 34 for vertical corrected records and 46 for vertical uncorrected PGA.
70
The caption of their Table 2 states that reported coefcients are for mean.
67
Kalkan
&
Glkan
(2004b) and Kalkan &
Glkan (2005)
Kalkan
(2004a)
Turkey
Unknown
(Turkey?)
NE
Italy
(45
46.5 N & 12
14 E)
Koyna region, India
Turkey
814
1430
Guadeloupe
al.
U
465
Shanghai region
Europe & Middle
East
1000
131
Iceland
Greece
H
44367
Area
Worldwide
al.
Gupta & Gupta (2004)
Bragato (2004)
Skarlatoudis et
(2003)
Beauducel
et
(2004)
Beyaz (2004)
Reference
(2003d), Campbell &
Bozorgnia (2003a) &
Bozorgnia & Campbell
(2004b)
Halldrsson & Sveinsson (2003)
Shi & Shen (2003)
Sigbjrnsson & Ambraseys (2003)
A (N, ST)
A
S
M
A (S & N,
R, T)
287
288
Europe & Middle

East
Ambraseys
(2005a)
Worldwide
E Alps (45.6
46.8 N & 12
14 E)
NW Italy
Bragato (2005)
Bragato
&
Slejko
(2005)
3168
689971
595
243
1402
667
V
-
5.0
2.5
5.0
5.0
4.1
1.7
4.0
5.0
5.1
2.6
5.5
Mmin
3.0
7.8
6.3
7.6
7.6
7.5
5.1
5.6
7.4
7.2
5.6
8.3
Mmax
6.8
>1152
0.0*
5.172
60*
240
135
135
122
423
42
17
39
14
33
E
U
72
71
Authors state in text that more than 14 000 values were used but their Table 1 gives 2 6899.
State equations valid to 4.5.
73
State equations valid up to 200 km.
Frisenda et al. (2005)
Europe & Middle

East
et
al.
Ambraseys
(2005b)
221
Turkey
595
819
Skarlatoudis et al.
(2004)
Ulusay et al. (2004)
al.
667
Okhotsk-Amur
plate boundary
Greece
et
195
142
H
163
NW Turkey
Area
Stable continental
regions
63
3335
al.
France
Japan
et
Marin et al. (2004)

Midorikawa & Ohtake
(2004)
zbey et al. (2004)
Pankow & Pechmann
(2004) and Pankow &
Pechmann (2006)
Sunuwar et al. (2004)
Reference
Lubkowski
(2004)
5*
0
>3
3
5.1
Mw (ML )
Mw
MJMA
Mw
Mw (Ms ,
mb , Md ,
ML )
Mw
0
0
Ms
ML
ML
Mw
5
0*
ML
Mw
rmin
0
M scale
Mw (ML )
rmax
rjb
rjb
rhypo
rrup
r scale
repi (rjb for
1 event)
repi
rjb
(repi
for
small
events)
rjb
(repi
for
small
events)
rjb
rjb & repi
1
1
3
2
1
2
S
1
repi
30073 rhypo
15
130
99
99
99.7
40
>264 rhypo
300*
99.4
700
300*
854
L
R
G
G,
O
L
L
C
U
NS
NS
A
A (C, B, F)
M
A
O
O
A
A
1WM A (N, T, S,
O)
1WM A (N, T, S,
O)
2M
1M
1M
R
1,
1M,
2,
2M
1
1
&
Bommer
75
74
Umbria-Marche
California
Los Angeles region
Shallow
crustal
(USA,
Taiwan,
Turkey and others)
239
949
U
4179
41
Call it quadratic mean, which is assumed to be geometric mean.

Also develop equations for hard rock sites and intraslab events.
Bindi et al. (2006)
Beyer
(2006)
4518+208
Japan+208
seas
over-
41
Chile
279
279
Ruiz
&
Saragoni
(2005)75
Takahashi
et
al.
(2005), Zhao et al.
(2006) and Fukushima
et al. (2006)
Wald et al. (2005)
Atkinson (2006)
7907
-
V
277
7907
72
Taiwan
Central Utah coalmining areas
Iran
Liu & Tsai (2005)

McGarr & Fletcher
(2005)
Nowroozi (2005)
H
277
Area
Central Mexico
Reference
Garca et al. (2005)
45
103
U
485+
4.0
4.3*
U
3.1*
5.0
6.4
3.0*
4.05
0.98
Mmin
5.2
5.9
7.9*
5.3*
7.1*
8.3
7.8
7.4
7.10
4.2
Mmax
7.4
249+20
45
51
12
E
16
5*
0.5*
Mw (ML )
Mw
(MCL )
Mw (Ms ,
mb , ML )
Ms
245
300*
10*
rmax
400*
repi
0*
U
5*
6*
1*
Mw
Mw
Mw
Mw
ML
100*
200*
U
300*
300*
repi
rhypo
&
rjb
repi (rjb for
some)
rhypo
rrup
1
I,
C
U
1
2
S
1
r scale
rrup
for
Mw > 6.5,
rhypo
otherwise
rhypo
rhypo
35.72 315.01 rhypo
rmin
4*
M scale
Mw
U
1
1M
2M
2M
R
1M
1,
1M
2,
A,
B,
C,
D50,
G,
I50,
L,
N,
P,
R
L
1M
L
B
M
L
C
G74
NS
A (U)
A
A
C (R, S/N)
& F, B
A
M
M
B
289
290
-
939077
Molise (Italy)
Central Iran76
New Zealand+66
overseas
W. N. America
Japan
France
Iran
Luzi et al. (2006)

Mahdavian (2006)
Moss & Der Kiureghian (2006)

Pousse et al. (2006)
Souriau (2006)
Zare & Sabzali (2006)
77
76
271
Algeria
89
150
-
55*
20
20
2.7
3.0
4.1
5.2
2.6*
3.1
5.08
(5.2)
5.6
&
5*
4
6
(0.1)
0
ML
Ms (mb )
Mw
Mw
7.4
5.4
7.3
7.7
5.7
7.4
7.23
(7.4)
6.0
10*
ML (ReNass
&
LDG)
Mw
5*
(Mw )
Ms
1*
(1.5*)
&
30*
13
Mw
(MJMA )
&
13.7
8.2*
(7.4)
8.0*
7.3
5*
Mw (Ms if
M > 6,
mb if M <
6)
ML
8.1*
rmin
0
1*
M scale
Mw
6.5*
Mmax
7.9
U
U
49+17
5.0*
(6.1)
5.5*
4.5*
3.0*
Mmin
4.2
Also develops equations for Zagros using 98 records from an unknown number of earthquakes.
89
175
886
150
535+66
73+10
& 111
3392+377
(shallow)
& 8150
(deep)
28
Laouami et al. (2006)
51
456
109
123
E
60+
456
Haulien LSTT (Taiwan)

Japan+some
foreign
900*
Hernandez
et
al.
(2006)
Kanno et al. (2006)
1983
900*
NE Italy & Slovenia
V
-
Mexico
H
1500+
Area
Worldwide
Gmez-Sobern et al.
(2006)
Reference
(2006a) and Campbell
& Bozorgnia (2006b)
Costa et al. (2006)
rmax
167
800*
250*
rhypo
2
2
3
rhypo (rrup
for some)
rhypo
118.2 rjb
55*
98
400
(10)
rhypo
rhypo
rhypo
rc (rrup )
rhypo (rrup
for some)
450* rrup (rhypo

(350*) for some)
&
450*
70
repi
&
repi
S
C
r scale
rrup
134.8 rhypo
800*
100*
200
L
A
L,
G
L,
V
U
C
G
1M
&
2M
2M
1M
1
1M
2M
R
2M
A
A
C (R, OR,
S & N) & F,
B
A
M
A (R, S, N)
1574
1561
335
592
Worldwide shallow
crustal
Worldwide shallow
crustal
Greece
California
See text
4047
-
200*
200*
U
4047
997
V
-
H
532
Turkey
NW Turkey
Europe and Middle
East
Area
Europe & Middle
East
Alborz and central
Iran78
39
151
64
58
U
528
289
50*
E
131
7.9085
4.2784
7.3*
5*
6.9
7.9082
4.2781
4.5
U
5.9
7.6
7.3*
Mmax
7.6
U
0.5
3
4.5*
Mmin
5.0
79
Also develop models for the Zagros region of Iran using about 100 records.
Also derive model using Mw .
80
Also derive model using repi .
81
Recommend that model is not extrapolated below 5 due to lack of data.
82
Believe that model can be used to 8.0.
83
Recommend that model is not used for distances 200 km.
84
Believe that model can be extrapolated down to 4.0.
85
Believe that model can be extrapolated up to 8.5 for strike-slip faulting and 8.0 for reverse faulting.
78
Boore & Atkinson

(2007) & Boore &
Atkinson (2008)
Campbell & Bozorgnia (2007), Campbell
& Bozorgnia (2008b) &
(2008a)
Danciu & Tselentis
(2007a) & Danciu &
Tselentis (2007b)
Douglas (2007)
Hong & Goda (2007) &
Goda & Hong (2008)
Reference
Akkar & Bommer
(2007b)
Ghodrati Amiri et al.
(2007a) & Ghodrati
Amiri et al. (2007b)
Aydan (2007)
Bindi et al. (2007)
C
2
3
rhypo
rhypo 80
rjb
(repi
for
small
events)
rjb
See text
0.2*
100*
Mw
Mw
136
0*
Mw
rjb
rjb
repi
I50
I50
U
L
G
C
G
See text
C
G,
Q,
R
S
3
rhypo
r scale
rjb
199.27 rrup
28083
0.07
U
200*
99
400*
rmax
99
Mw
Mw
ML 79
Mw
U
5*
0
5*
Ms (mb )
rmin
0
M scale
Mw
A
1
1M
1M
1M
2M
A
A
A (ST, N)
A (N, R, S,
HW)
A (N, R, S,
U)
U
A
1M
A
1WM A (N, S, R)
R
M
1WM A (N, S, R)
291
292
1063
Central northern
Italy
Romania
2754
1085
646
Worldwide shallow
crustal
South Iceland
Worldwide shallow
crustal
Worldwide shallow
crustal
1132
1085
162
589
V
-
87
Graizer & Kalkan (2007) state that valid down to 4.5.

Graizer & Kalkan (2007) state that valid up to 7.6.
88
Graizer & Kalkan (2007) state that valid up to 200 km.
89
90
Believe that model can be reliably extrapolated to 8.5.
91
Not clear from article if the authors mean rrup or rjb .
86
162
Colima, Mexico
1164
424
Taiwan
Tavakoli & Pezeshk

(2007)
Tejeda-Jcome
&
Chvez-Garca (2007)
Abrahamson & Silva
(2008) & Abrahamson
& Silva (2009)
gstsson
et
al.
(2008)
Aghabarati
&
Tehranizadeh (2008)
Cauzzi & Faccioli
(2008), Cauzzi (2008)
& Cauzzi et al. (2008)
589
Europe & Middle

East
H
2583
Area
Worldwide shallow
crustal
Sobhaninejad et al.
(2007)
Popescu et al. (2007)
Reference
Graizer
&
Kalkan
(2007) & Graizer &
Kalkan (2008)
Massa et al. (2007)
60
54
64
135
26
48
131
58
243
E
47
7.2
7.9
5.0
5.2
6.5
7.990
4.2789
3.5
5.2
7.3*
7.6
7.1
5.2
Mmax
7.987
3.3
4.3*
5.0
4*
2.5
Mmin
4.986
5*
5*
ML
ML
Mw
6*
3*
MLw
Mw
0.06*
Mw
150*
60
350*
200*
175
260*
227
99
70
0
Mw
Mw
300*
0*
ML
4
&
C
rrup 91
rhypo
S
C
repi
rrup
rhypo
rjb
(repi
for
small
events)
rhypo
repi , rhypo
rhypo
rmax r scale
349.688 rrup
rmin
0.1
M scale
Mw
V3
I50
C
U
2M
1M
1M
2M
R
O
A (N, R, S)
A (N, R, S)
A (N, R, S,
HW)
A (N, T, S,
O)
M
A (R,SN)
Europe & Middle

East
Worldwide shallow
crustal
Humbert &
(2008)
Idriss (2008)
Italy
Bindi et al. (2009b)
27
27
49
55
795
100
21
149
82
4.8
4.03
5.2
0.5
2.7
4.0*
3.1
3.5 & 4.0
4.1 (6.0)
4.5
4.0
Mmin
4.26592
6.9
6.40
7.9
3.0
5.7
8.1*
5.3
6.3 & 6.5
7.3 (8.1)
7.7
7.4
7.3
Mmax
7.9093
4.6
6.9
44+10
72
138
337
E
125
93

94
Believe that model valid to 0 km.
95
96
For stations on surface.
97
For borehole stations.
92
241
Italy
235
168
241
678
3090
-
678
3090
200
Morasca et al. (2008)

Slejko et al. (2008)
250
306
795
Mezcua et al. (2008)
306
Srinivasan
et
al.
(2008)
Aghabarati
&
Tehranizadeh (2009)
Akyol
&
Karagz
(2009)
Bindi et al. (2009a)
Spain
Massa et al. (2008)
942
4244+139
3894
960
V
-
H
1950
Molise
Caucasus
(36
46 N, 3852 E)
Kolar Gold Fields,
India
Worldwide shallow
crustal
Western Anatolia
NE Taiwan+10 foreign
Northern Italy
Lin & Lee (2008)
Viallet
Area
Worldwide shallow
crustal
Japan
Reference
Chiou
&
Youngs
(2008)
Cotton et al. (2008)
U
0.3
15
1*
1
0
15
0
0
Ms
Mw
Mw (ML )
Mw (ML )
& ML
Mw
(mb (Lg ))
ML
Ms (ML ,
Mw , mb )
ML
Mw
Mw (Md ,
ML )
Mw
Mw (ML )
12*
2*
5*
Mw
(MJMA )
rhypo
rrup (rhypo
for small)
r scale
rrup
183
190
200
60
4.76
60*
100*
100*
100*
630
2
1
1
rhypo
repi
rhypo
rjb (repi for

small)
rjb , repi
rhypo
rhypo
rrup
496
&
297
1
L,
G
L
L
U
I50
C
I50
S
C
repi
rhypo
199.3 rrup
100
rmin rmax
0.2*94 70*95
M scale
Mw
1M
1M
2M
1M
1M
O
1M
1W
2M
R
1M
A (N, S, R)
A (N, R, S)
A
A
A
(R/RO/NO,
S/N)
A (B, F)
M
A (N, R, S,
HW, AS)
A
293
294
California
Worldwide
Gujarat, India
Worldwide shallow
crustal
SW Iceland
Hong et al. (2009a)
Kuehn et al. (2009)
Mandal et al. (2009)

Moss (2009)
Douglas & Halldrsson (2010)
Bindi et al. (2010)

Cua & Heaton (2010)
of
Italy
Southern California+other shallow
crustal
Europe & Middle
East
Pacic coast
Mexico
South
Iceland+others
Europe & Middle
East
Turkey
Mexico (interface
& inslab)
Hong et al. (2009b)
Ptursson & Vogfjrd

(2009)
Rupakhety & Sigbjrnsson (2009)
Akkar & Bommer
(2010)
Akkar
&
agnan
(2010)
Arroyo et al. (2010)
Area
Italy
Reference
Bragato (2009)
595
561
561
3588+1607 -
418
532
433
823
-, -
V
-
64+29
823
248
1950
2660
592
418, 277
H
922
135
107
70
40
137
131
12
46
33
125
60
39
5.0
4.0
2 (5)
5.0
5.0
5.0
5.02
3.3
3.1
4.265
5.61
5*
5.0, 5.2
Mmin
2.7
7.6
6.9
7.3 (7.9)
8.0
7.6
7.6
7.67
6.5
7.7
7.90
7.9*
7.3*
8.0, 7.4
Mmax
4.5
40, 16
E
116
1*
0.2*
3
1
0
Mw
Mw
MLw
Mw
Mw
200*
99
97
380
300*
70*
200*
rjb
rjb (repi for

some)
rjb
repi
rjb
rrup
rjb
rjb
Mw
Mw
Mw
20
99
400
rjb
(repi
for
small
events)
rrup (rhypo
for Mw <
6)
1*
100* rjb , repi
0.8
200
rjb (repi for
(0.1)* (200)* M < 5)
0.1*
Mw
100*
rrup (rhypo
for small)
Mw
0.2*
Mw
r scale
repi
0*
Mw
rmax
100
Mw
rmin
6
M scale
ML
3
2
2
C
S
I,
3,
1,
C
1
L
G
V3
L
I50
G,
R,
Q
G,
R
C
U
A
A
A (N, S, R)
A (N, S, R)
S&O
A
A (N, R, S,
HW, AS)
A
A (N, R, S)
F, S
M
A
1WM A (N, T, S,
O, AS)
1M
1
1M
1M
1M,
2M,
O
1M
(O)
2
1M
1M
R
1
99
98
751
78
337
64
4.0
5.1
4.27
4.2
5.0
60
245
39
Mmin
4.5
7.4
6.5
7.3
7.90
7.9
7.28
Mmax
7.6
rmin
0.2*
0.1*
0.2*
0.07
1
1*
0.1
M scale
Mw
Mw
Mw
Mw
Mw
(MJMA )
Mw
Mw (Md )
rrup
rjb
r scale
rrup (rhypo
for small)
196.8 rjb 99 (repi

for
small
events)
80*
C,
I
2
S
4
&
C
C
C
rrup (rhypo
for small)
rjb (repi for
Mw < 6)
100
199.27 rrup
500*
100*
rmax
200*
Also derive models for inslab (273 records from 16 earthquakes) and interface (413 records from 40 earthquakes) Mexican earthquakes.
Not entirely clear in the article if rrup was actually used.
region,
Marmara
Turkey
81
South Iceland
Ornthammarath et al.
(2010b),
Ornthammarath
(2010)
&
Ornthammarath et al.
(2010a)
Ulutas & Ozer (2010)
3894
Jayaram & Baker

(2010)
Montalva (2010)
13992
592
V
-
1561
Worldwide
California98
Graizer et al. (2010)

Hong & Goda (2010)
H
1499
Worldwide shallow
crustal
Japan
Area
Worldwide shallow
crustal
Reference
Faccioli et al. (2010)
U
1,
2
I50
C
G
O
(1M)
1M
O
1M
R
1M
SN
A (N, R, S,
HW)
A
A (SN, R)
A
M
A (N, R, S)
295
296
Chapter 4
Summary of published GMPEs for

spectral ordinates
4.1
Johnson (1973)
PSRV = C10mb Rm
Response parameter is pseudo-velocity for 5% damping.
Most (76%) records from R < 70 km.
Uses only shallow focus earthquakes of normal or less depth, to minimize variables,
except for one record from deeper earthquake (mb = 6.5, R = 61.1 km) which produces
no distortion in statistical calculations.
4.2
McGuire (1974) & McGuire (1977)
See Section 2.10.

Response parameter is pseudo-velocity for 0, 2, 5 and 10% damping.
Residuals pass Kolmogorov-Smirnov goodness-of-t test at 5% signicance level for
normal distribution, so it is concluded that pseudo-velocities are lognormally distributed.
Feels that using 16 natural periods presents a very good picture of spectral trends
throughout entire period range.
Only gives graphs of coefcients not actual calculated values.
4.3
Kobayashi & Nagahashi (1977)
log10 SV 0 = a()M b() log10 x c()

Response parameter is velocity for unspecied1 damping.
1
It is probably 5%.
297
Do regression iteratively. Assume a(), b() and c(). Find amplication factors, Gi (),
for each response spectra,
Ri (): Gi = Ri ()/SV 0 . Calculate amplication factor, G,
for each site: G = n ni=1 Gi (). Estimate bedrock spectrum, Bi (), for each record:
Bi () = Ri ()/G(). Find a(), b() and c() by least squares. Repeat these steps
until convergence. Hence nd attenuation relation for bedrock and amplication function
for each site.
4.4 Trifunac (1977) & Trifunac & Anderson (1977)

log10 [SA(T ), p] = M + log10 A0 (R) a(T )p b(T )M c(T ) d(T )s e(T )v

f (T )M 2 g(T )R
where log A0 (R) is an empirically determined attenuation function from Richter (1958)
used for calculation of ML , p is condence level and v is component direction (v = 0 for
horizontal and 1 for vertical). log A0 (R) not given here due to lack of space.
s = 0 Alluvium. 63% of data.

s = 1 Intermediate. 23% of data.
s = 2 Basement rock. 8% of data.
Response parameter is acceleration for 0, 2, 5, 10 and 20% damping.
Note that do not believe the chosen independent parameters are the best physical characterization of strong shaking but they are based on instrumental and qualitative information available to the engineering community in different parts of the USA and the
world.
Data from free-eld stations and basements of tall buildings, which assume are not
seriously affected by the surroundings of the recording station. Note that detailed investigations will show that data from basements of tall buildings or adjacent to some other
large structure are affected by the structures but do not consider these effects.
Equation constrained to interval Mmin M Mmax where Mmin = b(T )/2f (T ) and
Mmax = [1 b(T )]/2f (T ). For M > Mmax replace f (T )M 2 by f (T )(M Mmax )2 and
for M < Mmin replace M by Mmin everywhere to right of log10 A0 (R).
Use almost same data as Trifunac (1976). See Section 2.13.
Use same regression method as Trifunac (1976). See Section 2.13.
Note that need to examine extent to which computed spectra are affected by digitization
and processing noise. Note that routine band-pass ltering with cut-offs of 0.07 and
25 Hz or between 0.125 and 25 Hz may not be adequate because digitisation noise does
not have constant spectral amplitudes in respective frequency bands and because noise
amplitudes depend on total length of record.
298
Find approximate noise spectra based on 13 digitisations of a diagonal line processed

using the same technique used to process the accelerograms used for the regression.
Linearly interpolate noise spectra for durations of 15, 30, 60 and 100 s to obtain noise
spectra for duration of record and then subtract noise spectrum from record spectrum.
Note that since SA(y1 + y2 ) 6= SA(y1 ) + SA(y2 ) this subtraction is an approximate
method to eliminate noise which, empirically, decreases the distortion by noise of the
SA spectra when the signal-to-noise ratio is small.
Note that p is not a probability but for values of p between 0.1 and 0.9 it approximates
probability that SA(T ),p will not be exceeded given other parameters of the regression.
g(T )R term represents a correction to average attenuation which is represented by
log10 A0 (R).
Do not use data ltered at 0.125 Hz in regression for T > 8 s.
Due to low signal-to-noise ratio for records from many intermediate and small earthquakes only did regression up to 12 s rather than 15 s.
Smooth coefcients using an Ormsby low-pass lter along the log10 T axis.
Only give coefcients for 11 selected periods. Give graphs of coefcients for other
periods.
Note that due to the small size of g(T ) a good approximation would be log A0 (R) +
R/1000.
Note that due to digitisation noise, and because subtraction of noise spectra did not
eliminate all noise, b(T ), c(T ) and f (T ) still reect considerable noise content for T >
1 2 s for M 4.5 and T > 6 8 s for M 7.5. Hence predicted spectra not
accurate for periods greater than these.
Note that could apply an optimum band-pass lter for each of the accelerograms used
so that only selected frequency bands remain with a predetermined signal-to-noise ratio.
Do not do this because many data points would have been eliminated from analysis
which already has only a marginal number of representative accelerograms. Also note
that such correction procedures would require separate extensive and costly analysis.
Note that low signal-to-noise ratio is less of a problem at short periods.
Compare predicted spectra with observed spectra and nd relatively poor agreement.
Note that cannot expect using only magnitude to characterise source will yield satisfactory estimates in all cases, especially for complex earthquake mechanisms. Additional
parameters, such as a better distance metric than epicentral distance and inclusion
of radiation pattern and direction and velocity of propagating dislocation, could reduce
scatter. Note, however, that such parameters could be difcult to predict a priori and
hence may be desirable to use equations no more detailed than those proposed so that
empirical models do not imply smaller uncertainties than those associated with the input
parameters.
Plot fraction of data points, pa which are smaller than spectral amplitude predicted
for p values between 0.1 and 0.9. Find relationship between pa and p. Note that
response spectral amplitudes should be nearly Rayleigh distributed, hence pa (T ) =
{1 exp[ exp((T )p + (T )]}N (T ) . Find , and N by regression and smoothed
299
by eye. N (T ) should correspond to the number of peaks of the response of a singledegree-of-freedom system with period T but best-t values are smaller than the value of
N (T ) derived from independent considerations.
4.5 Faccioli (1978)

See Section 2.18.
Plots all spectra. 2 records have abnormally high values in long period range, so remove
and repeat. Results practically unaffected so leave them in.
Notes that due to small size of sample, site and source correlation can introduce some
error in coefcients because all data treated as statistically independent. Assume correlations are small so neglect error.
4.6 McGuire (1978)

See Section 2.19.
4.7 Trifunac (1978) & Trifunac & Anderson (1978a)

Ground-motion model is (from denition of local magnitude scale):
log[PSV(T ),p ] = M + log A0 (R) a(T )p b(T )M c(T ) d(T )s e(T )v

f (T )M 2 g(T )R
Response parameter is pseudo-velocity for 0, 2, 5, 10 and 20% damping.

s = 1 Intermediate. 23% of data. Notes that ideally would not need but had to be introduced because in some cases difcult to make a choice in complex geological
environment or because of insufcient data.
Use same data as Trifunac & Anderson (1977). See Section 4.4.
Use same regression method as Trifunac & Anderson (1977). See Section 4.4.
300
Equation constrained to interval Mmin M Mmax where Mmin = b(T )/2f (T )

and Mmax = [1 b(T )]/2f (T ). For M > Mmax replace M by Mmax everywhere and
for M < Mmin replace M by Mmin in b(T )M and f (T )M 2 . This gives linear growth
for M < Mmin , parabolic growth for Mmin M Mmax and constant amplitude for
M > Mmax .
98 records from San Fernando earthquake (9/2/1971) but regression method eliminated
70% of these before computing the coefcients.
Epicentral distance used for simplicity, consistency with earlier studies and for lack of
signicantly better choice. Distance measure chosen has small effect whenever epicentral distance greater than several source dimensions.
Notes that recording and processing noise in signal means that quality of coefcients
diminishes for T > 2 s. Equations not recommended for periods longer than those for
which selected spectral amplitudes plotted.
Notes that equations should be considered only as preliminary and an empirical approximation to a complicated physical problem.
Notes that data are limited to narrow magnitude interval, most data comes from alluvium
sites and about half comes from one earthquake.
Only gives coefcients for 11 periods. Graphs of coefcients for other periods.
4.8
Trifunac & Anderson (1978b)
Ground-motion model is (from denition of local magnitude scale):
log[PSV(T ),p ] = M + log A0 (R) a(T )p b(T )M c(T ) d(T )s e(T )v

f (T )M 2 g(T )R
Response parameter is velocity for 0, 2, 5, 10 and 20% damping.

s = 1 Intermediate. 23% of data. Notes that ideally would not need but had to be introduced because in some cases difcult to make a choice in complex geological
environment or because of insufcient data.
Use same data as Trifunac & Anderson (1977). See Section 4.4.
Use same regression method as Trifunac & Anderson (1977). See Section 4.4.
301
Equation constrained to interval Mmin M Mmax where Mmin = b(T )/2f (T )

and Mmax = [1 b(T )]/2f (T ). For M > Mmax replace M by Mmax everywhere and
for M < Mmin replace M by Mmin in b(T )M and f (T )M 2 . This gives linear growth
for M < Mmin , parabolic growth for Mmin M Mmax and constant amplitude for
M > Mmax .
Only gives coefcients for 11 periods. Graphs of coefcients for other periods.
4.9 Cornell et al. (1979)

See Section 2.21.
Response parameter is pseudo-velocity for 0, 2 and 10% damping.
Consider different paths, e.g. going through intensities, Fourier spectra and PGA, to
predict PSV. Note that direct paths have minimum variance but that going through intermediate steps does not signicantly increase prediction uncertainty provided that intermediate parameters are representative of frequency band of structural system.
Do not give coefcients.
4.10 Faccioli & Agalbato (1979)

See Section 2.23.
4.11 Trifunac & Lee (1979)

log10 PSV(T ) = M + log10 A0 (R) b(T )M c(T ) d(T )h e(T )v f (T )M 2 g(T )R

used for calculation of ML and v is component direction (v = 0 for horizontal 1 for
vertical).
Use depth of sedimentary deposits, h, to characterise local geology.
Depths of sedimentary and alluvial deposits at stations used are between 0 and about
6 km and most are less than about 4 km.
Use data and regression technique of Trifunac & Anderson (1977), see Section 4.4.
Note no obvious physical reason why dependence of PSV on h should be linear. Try
including terms with h2 , h3 and higher powers of h but they lead to values which are
undistinguishable from zero at 95% condence level.
Approximate signicance tests show that coefcients are signicantly different from zero
in large subregions of the complete period range.
302
Only give coefcients for 11 periods. Graphs of coefcients for other periods.
Note results are only preliminary.
Note amount of data too small to include more sophisticated independent parameters.
4.12
Ohsaki et al. (1980b)
log Sv = a0 M b0 log x c0
Response parameter is velocity for 5% damping.

Use two soil conditions:
Group A Hard rock: geology consists of granite, andesite and shale of Miocene or earlier
geological age, having S wave velocity & 1500 m/s or P wave velocity & 3000 m/s,
60 records
Group B Rather soft rock: geology consists of mudstone of Pliocene or late Miocene age,
having S wave velocity of about 5001000 m/s, 35 records.
Use records where geological and geotechnical conditions investigated in detail and considered to represent free-eld rock motions. Exclude records suspected to be amplied
by surface soil or affected by high topographical relief.
Most records from 30 km.
Do regression on both site categories separately and give graphs of coefcients not
tables.
4.13
See Section 2.28.

Also give smoothed results using correction factors based on derived PGV equation.
4.14
Trifunac (1980)
log10 PSV(T ) =
M log10 A0 (R) b(T )Mmin c(T ) d(T )h e(T )v

2 g(T )R
f (T )Mmin
for M Mmin
M log10 A0 (R) b(T )M c(T ) d(T )h e(T )v
f (T )M 2 g(T )R
for Mmin < M < Mmax
Mmax log10 A0 (R) b(T )Mmax c(T ) d(T )h e(T )v
2
g(T )R
f (T )Mmax
for M Mmax
303
where log10 A0 (R) is an empirically determined attenuation function from Richter (1958)
used for calculation of ML , v is component direction (v = 0 for horizontal and 1 for
vertical), Mmin = b(T )/(2f (T )) and Mmax = (1 b(T ))/(2f (T )).
Characterises site condition by depth of sedimentary and alluvial deposits beneath station, h. Uses records with 0 h 6 km, with most < 4 km.
Performs analysis to minimize possible bias due to uneven distribution of data among
magnitude, site conditions and from abundance of data for some earthquakes.
Tries terms with higher powers of h but coefcients are undistinguishable from zero at
95% condence level.
Assumes probability that log10 PSV(T ) log10 PSV(T

) , where log10 PSV(T ) is
measured PSV and PSV(T

) is predicted PSV and is a probability, can be expressed as
p(, T ) = [1exp( exp((T )(T )+(T )))]N (T ) . This assumption passes KolmogorovSmirnov and 2 tests at 95% level.
Finds a(T ) through g(T ) signicantly different than zero for large subregions of whole
period range. d(T ) is only signicantly different than zero for T & 0.3 s.
Gives coefcients of smoothed results for 11 periods.
Notes only preliminary. Improvements should be based on physical nature of phenomenon using a functional form predicted by theory and experiment but due to lack
of data cannot be done.
4.15 Devillers & Mohammadioun (1981)

V (f ) = C10M Rn
Most records from between 20 and 40 km. No records from R < 10 km so equation
does not apply there.
Eliminate suspect and/or redundant (San Fernando) records.
Split data into intensity groups: VI (126 records), VII (56 records), V+VI (186 records),
VI+VII (182 records) and VII+ VIII (70 records) and calculates coefcients for each
group.
Note not adjusted for local site conditions. Try to distinguish effect but correlations do
not reveal signicant variations. Notes very few records on hard rock.
Do not give coefcients only graphs of results.
304
4.16
Joyner & Boore (1982a)
log y = + Mp log r + br + cS
r = (d2 + h2 )1/2
Rock S = 1
Soil S = 0
Test magnitude dependence of h by selecting data from < 10 km and plot residuals
against M . Do not nd any systematic relationship so conclude that data does not
support a magnitude-dependent shape.
Smooth coefcients using unspecied method.
No data from rock sites with d < 8 km and M > 6 so suggest caution in applying equations for rock sites at shorter distances and larger magnitudes. Also suggest caution in
applying equations for d < 25 km and M > 6.6 for either soil or rock because no data
in this range. Also do not recommend equations for M > 7.7.
4.17
Joyner & Boore (1982b)
See Section 2.33.

Use same data and method as Joyner & Boore (1982a).
Restrict regressions to T 4 s to avoid problems due to record-processing errors.
Find that coefcient for quadratic term is not statistically signicant at 90% level for most
periods but the values obtained at different periods are sufciently consistent to warrant
inclusion of this term. Note that maximum difference with and without quadratic term is
about 20%.
Include soil term at short periods even though not signicant at 90% level.
Smooth coefcients by plotting them against log T and drawing smooth curves.
4.18
Kobayashi & Midorikawa (1982)
log Sv0 (T ) = a(T )(log M0 c) b(T ) log X + d

where a(T ) = a1 + a2 log T
and: b(T )) =
b1 (log T )2 + b2 log T + b3 for: 0.1 T 0.3 s

b4 b5 log T for: 0.3 T 5 s
305

Magnitudes converted to seismic moment, M0 , by using empirical formula.
Observed surface spectra divided by amplication over bedrock (assumed to have shearwave velocity of 3 km/s), calculated for each of the 9 sites.
Note equation not for near eld because earthquake is not a point source.
4.19 Joyner & Fumal (1984), Joyner & Fumal (1985) & Joyner &
Boore (1988)
See Section 2.37.
Use data from Joyner & Boore (1982b).
shear-wave velocity not signicant, at 90%, for periods 0.1, 0.15 and 0.2 s but signicant
for longer periods.
Regression using shear-wave velocity and depth to rock shows signicant correlation
(decreasing ground motion with increasing depth) for long periods but small coefcients.
Short periods do not show signicant correlation.
State inappropriate to use depth to rock for present data due to limited correlation and
because San Fernando data is analysed on its own does not show signicant correlation.
4.20 Kawashima et al. (1984)

See Section 2.38.
Response parameter is acceleration for 5% damping.
4.21 Kawashima et al. (1985)

See section 2.43.
Variation of a and b with respect to T is due to insufcient number of records.

Ground-motion models are (if dene site in terms of local geological site classication):
log PSV(T ) = M + Att(, M, T ) + b1 (T )M + b2 (T )s + b3 (T )v + b5 (T ) + b6 (T )M 2

or (if dene site in terms of depth of sediment):
log PSV(T ) = M + Att(, M, T ) + b1 (T )M + b2 (T )h + b3 (T )v + b5 (T ) + b6 (T )M 2

306
where
A0 (T ) log10 for R Rmax

A0 (T ) log10 max (R Rmax )/200 for R > Rmax
)1/2
(
R2 + H 2 + S 2
= S ln 2
R + H 2 + S02
max = (Rmax , H, S)
1
Rmax =
( + 2 4H 2 )
2
Att(, M, T ) =
S0 = S0 (T ) represents the coherence radius of the source and can be approximated

by S0 Cs T /2, Cs is shear-wave velocity in source region (taken to be 1 km/s), T is
period, S is source dimension approximated by S = 0.2 for M < 3 and S = 25.34 +
8.151M for 3 M 7.25 and v is component direction (v = 0 for horizontal 1 for
vertical).
Use two types of site parameter:
Local geological site classication:
s = 0 Sites on sediments.
s = 1 Intermediate sites.
s = 2 Sites on basement rock.
Depth of sediments from surface to geological basement rock beneath site, h.
Equations only apply in range Mmin M Mmax where Mmin = b1 (T )/(2b6 (T ))
and Mmax = (1+b1 (T ))/(2b6 (T )). For M < Mmin use M only in rst term of equation
and Mmin elsewhere and for M > Mmax using Mmax everywhere.
Screen data to minimize possible bias in the model, which could result from uneven
distribution of data among the different magnitude ranges and site conditions, or from
excessive contribution to the database from several abundantly recorded earthquakes.
Originally include a term linear in , i.e. b4 (T )/100, but nd that b4 (T ) is insignicant
for most periods so deleted it.
Use method of Trifunac & Anderson (1977) for residuals, see Section 4.4.
4.23
Kamiyama & Yanagisawa (1986)
log10 V (T ) = a(T )MJ b(T ) log10 (+30)d(T )Dc(T )+A1 (T )S1 +. . .+AN 1 (T )SN 1
where Si = 1 for ith site and 0 otherwise.
Response parameters are acceleration, velocity and displacement for 0, 2, 5 and 10%
damping
Model site amplication of each of the 26 sites individually by using Si . Choose one site
as bed rock site, which has S-wave velocity of about 1000 m/s.
307
Use records with PGA> 20gal (0.2 m/s2 ).

Focal depths, D, between 0 and 130 km, with most between 10 and 50 km.
Find no signicant differences between site amplication spectra for different response
parameters or different damping levels.
Compare amplication spectra from regression for different sites with those predicted
using S-wave theory and nd good agreement.
Coefcients only given for velocity for 5% damping.
4.24 C.B. Crouse (1987) reported in Joyner & Boore (1988)

See Section 2.48.
4.25 Lee (1987) & Lee (1993)

[ )] = M< + Att(, M, T ) + b1 (T )M<> + b2 (T )h + b3 (T )v

log10 [PSV(T
(1)
(1)
(2)
(2)
+ b4 (T )hv + b5 (T ) + b6 (T )M 2 + b (T )S + b (T )S
<>
where M<
= min(M, Mmax )
M<> = max(Mmin , M< )

Mmin = b1 /(2b6 (T ))
Mmax = (1 + b1 (T ))/(2b6 (T ))
where v = 0 for horizontal component, 1 for vertical, h is depth of sedimentary deposits
beneath recording station and Att(, M, T ) is same as Trifunac & Lee (1989) (see
Section 4.33).
SL = 0 Rock: 1 sediment site (h > 0), 11 intermediate sites (h 0) and 13 bedrock sites
(1)
(2)
(h = 0) SL = 0 & SL = 0.
SL = 1 Stiff soil ( 45 60 m deep): 37 sediment sites (h > 0), 24 intermediate sites
(1)
(2)
(h 0) and 3 bedrock sites (h = 0) SL = 1 & SL = 0.
(1)
SL = 2 Deep soil: 44 sediment sites (h > 0) and 2 intermediate sites (h 0) SL = 0

(2)
& SL = 1 .
For M > 6.5 uses different (unspecied) magnitude scales because for seismic risk
analysis often catalogues do not specify scale and often estimates are not homogeneous.
Free-eld records with both soil and alluvial depth information.
308
Screens data to minimize possible bias due to uneven distribution of soil classication
or excessive contribution from several abundantly recorded earthquakes.
Gives smoothed coefcients for 12 periods.
Uses method of Trifunac (1980) for uncertainties.
(1)
(2)
Also uses method where site coefcients, b7 & b7 , are found from residues from
equation without site coefcients; nd similar results.
4.26
K. Sadigh (1987) reported in Joyner & Boore (1988)
See Section 2.51.

Response parameter is pseudo-acceleration for 5% damping.
4.27
Annaka & Nozawa (1988)
See Section 2.54.

Give only graphs of coefcients.
4.28
Crouse et al. (1988)
ln[PSV(T )] = a + bM + c ln[R] + dh
Most data from shallow stiff soil and sedimentary deposits between about 5 and 25 m
deep on Tertiary or older bedrock.
All earthquakes from Benioff-Wadati zones.
Exclude data with magnitudes or distances well outside range of most selected records.
Focal depths, h between 14 and 130 km.
No strong correlations between h, R and M .
Try terms eM 2 and f R but nd not signicant (using t-test).
Try term R + C1 exp(C2 M ) instead of R; nd similar standard errors.
Find d is insignicant for 0.6 to 2 s; nd d does not signicantly reduce standard errors.
Find residuals are normally distributed (by plotting on normal probability paper and by
Kolmogorov-Smirnov test).
309
Split data by fault mechanism (thrust: 49 records, normal: 11 records, strike-slip: 4

records) and nd attenuation equation for each subset; results are not signicantly different (at 95% using F test). Also check by examining normal deviates (normalised
residuals) for each subset and period; nd no signicant differences.
Use 131 records from six other subduction zones (Nankai, Kuril, Alaska, Peru/N. Chile,
Mexico and New Britain/Bougainville) to examine whether ground motions from all subduction zones are similar.
Examine normal deviates for residuals between other zones ground motion and N. Honshu equation. Find no signicant differences (although obtain signicant results for some
periods and focal mechanisms) between N. Honshu, Kuril and Nankai motions. Find differences for Alaskan and Mexican data but could be due to site effects (because some
data from soft soil sites). Find differences for Peru/N. Chile and New Britain/Bougainville
which are probably source effects.
Plot seismotectonic data (age, convergence rate, dip, contact width, maximum subduction depth, maximum historical earthquake (Mw ), maximum rupture length, stress drop
and seismic slip) against decreasing ground motion at stiff sites for T > 0.8 s. Find
weak correlations for stress drop and Mw (if ignore Mexican data) but due to variability
in stress drop estimates note lack of condence in results.
4.29 Petrovski & Marcellini (1988)

See Section 2.59.
Response parameter is relative pseudo-velocity for 0.5%, 2%, 5% and 10% damping.
4.30 Yokota et al. (1988)

log Sv (T ) = a(T )M + b(T ) log X + c(T )

Focal depths between about 20 and 100 km.
Records from two stations in lowlands of Tokyo 3.7 km apart.
Also analyse another region, using 26 records from 17 earthquakes with distances between 95 and 216 km. Note difference in results between regions.
Analyses vertical spectra from three small regions separately, one with 24 records with
4.0 M 6.1 and 60 X 100 km, one with 22 records with 4.2 M 6.0 and
68 X 99 km and one with 5 records with 4.4 M 6.0 and 59 X 82 km.
Give no coefcients, only results.
310
4.31
See Section 2.62.

ln(Sv /amax ) = C6 + C7 (C8 Mw )C9

Response parameter, Sv , is velocity2 for 5% damping
Develop relationships for ratio Sv /amax because there is a much more data for PGA
than spectral ordinates and use of ratio results in relationships that are consistent over
full range of magnitudes and distances.
Calculate median spectral shapes from all records with 7.8 Mw 8.1 (choose this because abundant data) and R < 150 km and one for R > 150 km. Find signicant difference in spectral shape for two distance ranges. Since interest is in near-eld ground motion use smoothed R < 150 km spectral shape. Plot ratios [Sv /amax (Mw )]/[Sv /amax (Mw =
8)] against magnitude. Fit equation given above, xing C8 = 10 (for complete saturation
at Mw = 10) and C9 = 3 (average value obtained for periods > 1 s). Fit C7 by a linear
function of ln T and then x C6 to yield calculated spectral amplications for Mw = 8.
Calculate standard deviation using residuals of all response spectra and conclude standard deviation is governed by equation derived for PGA.
4.32
Kamiyama (1989)
log10 V () = log10 M0 a() log10 r + b() log10 L + e()r + c() +
N
1
Aj ()Sj
j=1
where Sj = 1 for site j and Sj = 0 otherwise.

Uses same data as Kamiyama & Yanagisawa (1986).
Uses same regression method as Kamiyama & Yanagisawa (1986).
Uses fault length, L, for 52 records. For others
where such data does not exist uses
M0 = 10(1.5 log10 S+22.3) , S = 10M 4.07 and L = S/2 where S is fault area in km2 .
Chooses hard slate site with shear-wave velocity of 12 km/s as basic site.
Does not give coefcients, only graphs of coefcients.
2
In paper conversion is made between Sv and spectral acceleration, Sa , suggesting that it is pseudo-velocity.
311

log10 [PSV(T )] = M + Att(, M, T ) + b1 (T )M + b2 (T )h + b3 (T )v + b5 (T )

+ b6 (T )M 2
where Att(, M, T ) = A0 (T ) log10
0.732025 for: T > 1.8 s

0.767093 + 0.271556 log10 T 0.525641(log10 T )2
A0 (T ) =
for: T < 1.8 s

(
)
1/2
R2 + H 2 + S 2
= S ln 2
R + H 2 + S02
S = 0.2 + 8.51(M 5)
where v = 0 for horizontal component and 1 for vertical, is representative distance,

S0 is correlation radius of source function (or coherence size of source) (which can be
approximated by Cs T /2, where Cs is shear wave velocity), h is depth of sedimentary
deposits beneath recording station and H is focal depth.
Screen data to minimize possible bias due to uneven distribution of data among different
magnitude ranges and site conditions or from excessive contribution to database from
several abundantly recorded earthquakes.
Include term, b4 (T )/100, but insignicant for most periods so remove.
Equation only applies for Mmin M Mmax , where Mmin = b1 (T )/(2b6 (T )) and
Mmax = (1 + b1 (T ))/(2b6 (T )). For M Mmin use Mmin everywhere except rst
term. For M Mmax use Mmax everywhere.
Use method of Trifunac (1980) for uncertainties.
Note estimates should only be used where signal to noise ratio (based on estimated
digitisation noise) not much less than unity or slope in log-log scale is not signicantly
greater than 1.
Also t data to log10 PSV(T ) = M + Att(, M, T ) + b1 (T )M + b2 (T )s + b3 (T )v +
b5 (T ) + b6 (T )M 2 (where s = 0 for sediment sites, 1 for intermediate sites and 2 for
basement rock sites) because depth of sediment not always known.
4.34 Atkinson (1990)

log y = c1 + c2 (M 6) + c3 (M 6)2 log R c4 R

All data from rock sites.
Includes only if a reliable seismic moment estimate exists.
312
Converts ECTN vertical seismograms to equivalent horizontal component by multiplying

by 1.4.
Includes Nahanni (western Canada) earthquakes because exhibit dominant characteristics of eastern North American shocks (low seismicity area, high horizontal compressive
stress, thrust mechanisms dominant, no surface ruptures despite shallow focus and
rocks have high seismic velocity).
Excludes US digital strong-motion Saguenay records due to low resolution. Two effects
on response spectra: i) high frequencies contaminated by a mathematical noise oor,
ii) signicant errors in amplitudes of low to intermediate frequencies (severity dependent
on resolution degree). Inclusion of such data could lead to signicant misinterpretation
of these earthquakes.
Most records (66, 65%) from R 111 km and M 5.22.
Examines residuals from equations. Finds no persistent trends except for Saguenay
data (M = 6) between 63 R 158 km.
Notes data very limited in large magnitude range and that one or two earthquakes are
controlling predictions.
Notes different regression technique could change predictions for large magnitudes but
i) data too limited to warrant more sophisticated analysis and ii) may be other factors, in
addition to number of recordings, which should be considered in weighting each earthquake.
4.35
Campbell (1990)
See Section 2.68.

4.36
Dahle et al. (1990b) & Dahle et al. (1990a)
See Section 2.69.

Coefcients only given for 7 periods; graphs for others.
4.37
Tamura et al. (1990)
SA (Ti , GC) = a(Ti , GC)10b(Ti ,GC)M ( + 30)C(Ti ,GC)

Response parameter is acceleration for 2 and 5% damping.
Use three site categories (GC) for which perform separate regression:
Group 1 Ground characteristic index . 0.67, 29 records.
313
Group 2 Ground characteristic index between about 0.67 and 1.50, 46 records.
Group 3 Ground characteristic index & 1.50, 22 records.
where the ground characteristic index is calculated from statistical analysis of amplitude
of records. Thought to reect the characteristic of deep soil deposits at site (1.0 means
amplication is average for Japan, < 1.0 or > 1.0 means amplication is lower or greater,
respectively, than average for Japan).
Records from JMA low-magnication mechanical seismographs (natural period 6 s, damping ratio 0.55) which were instrument corrected (because sensitivity for periods > 10 s is
substantially suppressed) , ltered (cut-offs 1.32 s and 2030 s chosen from a study of
recording accuracy of instruments) and differentiated in frequency domain to nd ground
velocity and acceleration. Hence limit analysis to 2 to 20 s.
Do not use resultant of two horizontal components because two components not synchronous.
Find difference in predicted ground motion using derived equations and those from earlier equations for short periods. Find that b for earlier equations increases almost linearly
with logarithm of natural period, T , so nd equation, by least squares, connecting b and
log T . Assume this equation holds for 2 to 20 s and so x b and recalculate a and c; nd
predictions now agree.
Only give graphs for original coefcients for 5% damping. Give tables of coefcients for
preferred second analysis.
4.38 Tsai et al. (1990)

See Section 2.73.
Also give equations for average acceleration for 2 period bands 0.120.33 s and 0.07
0.2 s.
4.39 Crouse (1991)

See Section 2.75.
Notes that spectral database is biased to higher ground motions (because only higher
ground motions are digitised). Suggest either using a different form of equation or impose constraints. Do not do either because (1) consider sample adequate for regression
and (2) although overestimate smaller, more distant motion, it would properly estimate
larger motions which are of greater concern for design applications.
Sets p3 , p5 and p6 to those for PGA equation after trial regressions; does not appreciably
affect standard deviation.
314
Finds relatively larger standard deviation for 3.0 and 4.0 s which suggests form of equation may be inappropriate for longer periods.
Plots normalised residuals (not shown) which show uniform distribution.
4.40
Dahle et al. (1991)
ln A = c1 + c2 M + c4 R + ln G(R, R0 )
and: G(R, R0 ) =
R01
for R R0
R0
R
)5/6
for R > R0
this equation assumes spherical spreading (S waves) to R0 and cylindrical spreading

with dispersion (Lg waves) for larger distances.
All data from solid rock sites.
Follow-on study to Dahle et al. (1990b) and Dahle et al. (1990a) but remove Chinese
and Friuli data and data from border zone of Eurasian plate, so data is a more genuine
intraplate set.
Use 395 records from Norwegian digital seismograms. Require that the Lg displacement
amplitude spectra should have a signal-to-noise ratio of a least 4 in the frequency range
110 Hz, when compared to the noise window preceding the P-wave arrival.
For the selected seismograms the following procedure was followed. Select an Lg window, starting at a manually picked arrival time and with a length that corresponds to
a group velocity window between 2.6 and 3.6 km/s. Apply a cosine tapering bringing
the signal level down to zero over a length corresponding to 5% of the data window.
Compute a Fast Fourier Transform (FFT). Correct for instrument response to obtain true
ground motion displacement spectra. Bandpass lter the spectra to avoid unreasonable amplication of spectral estimates outside the main response of the instruments.
Passband was between 0.8 Hz and 15 or 20 Hz, dependent on
sampling rate. The amplitude spectra obtained using the direct method, using A = t ZZ where t is time
step and Z is Fourier transformed time-history and Z is its complex conjugate. Convert instrument corrected displacement Lg Fourier transforms to acceleration by double
differentiation and an inverse FFT.
Use 31 accelerograms from eastern N. America, N. Europe and Australia.
Use R0 = 100 km although note that R0 may be about 200 km in Norway.
Correlation in magnitude-distance space is 0.20.
Use a variant of the two-stage method to avoid an over-representation of the magnitude
scaling terms at small magnitudes. Compute average magnitude scaling coefcients
within cells of 0.2 magnitude units before the second stage.
315
Resample data to make sure all the original data is used in a variant of the one-stage
method. Compute new (resampled) data points as the average of one or more original
points within a grid of cells 160 km by 0.4 magnitude units. Correlation in resampled
magnitude-distance space is 0.10.
Find estimated ground motions from one-stage method systematically higher than those
from two-stage method particularly at short distances and large magnitudes. Effect more
signicant for low frequencies. Find that this is because one-stage method gives more
weight to supplementary accelerograph data from near eld of large earthquakes.
Standard deviations similar for one- and two-stage equations.
Scatter in magnitude scaling coefcients from rst stage of two-stage method is greater
for strong-motion data.
Try xing the anelastic decay coefcient (c4 ) using a previous studys results. Find
almost identical results.
Remove 1 record from Nahanni earthquake (Ms = 6.9) and recompute; only a small
effect.
Remove 17 records from Saguenay earthquake (Ms = 5.8) and recompute; nd significant effect for large magnitudes but effect within range of variation between different
regression methods.
4.41 Geomatrix Consultants (1991), Sadigh et al. (1993) & Sadigh

et al. (1997)
See Section 2.77
Ground-motion model for deep soil is:
ln y = C1 + C2 M C3 ln(rrup + C4 eC5 M ) + C6 + C7 (8.5 M )2.5

where C6 is different for reverse and strike-slip earthquakes.
Ground-motion model for rock is:
ln y = C1 + C2 M + C3 (8.5 M )2.5 + C4 ln(rrup + exp(C5 + C6 M )) + C7 ln(rrup + 2)

where C1 is different for reverse and strike-slip earthquakes.
Vertical equations do not include C7 .
Perform analysis on spectral amplication ln(SA/PGA).
Give smooth coefcients.
Find standard errors to be dependent on magnitude and t to a linear relation.
316
4.42
I.M. Idriss (1991) reported in Idriss (1993)
See section 2.79.

4.43
Loh et al. (1991)
See Section 2.80.

Response parameters are acceleration, velocity and displacement for 5% damping.
Only give coefcients for acceleration for periods 0.1 s.
4.44
Matuschka & Davis (1991)
See Section 2.81.

4.45
Mohammadioun (1991)
log PSV(f ) = k(f ) + a(f )M + n(f )R

Response parameter is pseudo-velocity for 5%.
Records not baseline corrected so no equations for periods > 2 s.
Does not split up data into subsets by intensity because risk of creating data populations
which are not statistically signicant.
Notes that could be inconsistency with using both rhypo and rrup .
Notes that results are preliminary.
Also analyses wide range of Californian data for 96 periods between 0.013 and 5 s split
into two intensity dependent subsets: those records with site intensities VI-VII (326
records) and those with site intensities VII+ (156 records). Uses rrup except for Imperial Valley earthquake where uses rE . Does not use include soil or other variables
because poorly dened and lead to selection of records that are not statistically valid.
4.46
See Section 2.84.

Response parameter is pseudo-velocity for 0, 0.5, 1, 2, 5, 7, 10 and 20% damping.
317
4.47 Benito et al. (1992)

ln
PSA
PSV
= c1 + c2 M + c3 ln(R + R0 ) + c4 (R + R0 )
Response parameters are pseudo-acceleration, PSA, and pseudo-velocity, PSV, for 5%

damping3 .
Use three soil conditions (revised when cross hole information was available):
S = 0 Hard and rock sites, 50 records.

S = 1 Intermediate soil, 10 records.
S = 2 Soft soil, 12 records.
Use ML because most suitable for distance range of majority of records.
Try including c5 S term but nd low signicance values for c5 . Repeat regression for
each soil category separately. Give results when coefcient of determination R2 > 0.80,
standard errors < 25% and coefcients have high signicance levels.
For PSA for S = 0 give coefcients for all periods, for S = 1 give coefcients for 0.17 to
0.2 s and for S = 2 give coefcients for 1 to 10 s.
Also consider Friuli records (4.2 ML 6.5, epicentral distances between 2 and
192 km, 14 records for S = 0, 23 records for S = 1 and 16 records for S = 2).
Note need to include term in model reecting explicitly local amplication dependent on
natural period of soil as well as predominant period of incident radiation to bed rock.
4.48 Niazi & Bozorgnia (1992)

See Section 2.82.
For some periods (0.20 s for vertical and 0.10 and 0.111 s for horizontal) constrain c2 to
zero so that predicted amplitude would not decrease with increasing magnitude at zero
distance. Note that does not affect uncertainty.
Note that long period lter cutoff may be too long for records from small shocks but if
a shorter period was used then information on long period spectral ordinates would be
lost. Note that insufcient data for well constrained results at M = 5 or M > 7.
Find evidence for long period noise in d and in Degree of Magnitude Saturation (DMS =
(c2 d/b) 100).
Examine median and normalized standard deviation (coefcient of variation) and nd
evidence for decreasing uncertainty with increasing magnitude.
3
Although coefcients should only differ by a constant because PSA = (2/T )PSV they do not; hence response parameters are probably not those stated.
318
4.49
Silva & Abrahamson (1992)
See Section 2.89.

Ground-motion model for PSA to PGA ratio is:
ln(Sa/pga) = c1 + c3 r + c4 {1 tanh[(r1.1 10)/3]}(1 F )

Regress on ratio of PSA to PGA ratio because more stable than regression on absolute
values.
Choice of functional form guided by numerical simulations and previous empirical studies. Numerical simulations suggest that strike-slip events maybe more likely to show
near-eld directivity effects at long periods than dip-slip events.
Data does not allow magnitude dependency to be reliably determined hence not modelled.
Judge whether long period motion is realistic based on consistency of amplitudes and
timing of long period energy and that of higher frequency motions. Expect that seismic
ground motions have consistent phase structure at long periods whereas noise will have
random phase. Examine the analytical derivative of the phase with respect to frequency
and chose the upper period of reliable PSAs based on the period at which the phase
derivative becomes more random.
Only use PSAs for frequencies greater than 1.25 times the high-pass lter corner frequency and for periods less than the shortest period at which phase derivative is not
well behaved. Note that these criteria tend to bias regression to larger spectral values
because these will be above noise level more often than smaller motions. Do not try to
correct for this bias.
For 10 s insufcient data to yield stable coefcients. Based on numerical simulations,
nd response spectra are approximately at for > 8 s and M < 7.5 and, therefore,
extend model to 20 s by assuming constant spectral displacement. Note that may not be
appropriate for M > 7.5.
Note that Loma Prieta is major contributor to dataset, which may explain strong distance
dependency of spectral shape.
4.50
Tento et al. (1992)
See Section 2.91.

Note that correction procedure signicantly affects results for T > 2 s. Correction procedure introduces dishomogeneity and errors due to subjectivity of choice of low frequency
lter limits.
319
4.51 Abrahamson & Silva (1993)

See Section 2.93.
Ground-motion model for PSA to PGA ratio is:
For M > 6.5:
ln(Sa/pga)soil = c1 + c2 (8.5 M )c8 + c6 r + c5 {1 tanh[(r c9 )/c10 ]}(1 F1 )

For M > 6.5:
ln(Sa/pga)rock = c3 + c4 (8.5 M )c8 + c7 r + c5 {1 tanh[(r c9 )/c10 ]}(1 F1 )

For 6 M 6.5:
ln(Sa/pga)soil = c1 + c2 (8.5 M )c8 + c6 r

+ 2(M 6)c5 {1 tanh[(r c9 )/c10 ]}(1 F1 )
For 6 M 6.5:
ln(Sa/pga)rock = c3 + c4 (8.5 M )c8 + c7 r

+ 2(M 6)c5 {1 tanh[(r c9 )/c10 ]}(1 F1 )
Regress on ratio of PSA to PGA ratio because more stable than regression on absolute
values.
Choice of functional form guided by numerical simulations and previous empirical studies. Numerical simulations suggest that strike-slip events maybe more likely to show
near-eld directivity effects at long periods than dip-slip events.
Interested in long-period motions. Apply new accelerogram processing procedure to
evaluate reliable long-period range based on Fourier phase spectra. Apply high-pass
lter in frequency domain and a polynomial baseline correction in time domain. Judge
whether long period motion is realistic based on consistency of amplitudes and timing
of long period energy and that of higher frequency motions. Expect that seismic ground
motions have consistent phase structure at long periods whereas noise will have random phase. Examine the analytical derivative of the phase with respect to frequency
and chose the upper period of reliable PSAs based on the period at which the phase
derivative becomes more random.
Only use PSAs for frequencies greater than 1.25 times the high-pass lter corner frequency and for periods less than the shortest period at which phase derivative is not
well behaved. Note that these criteria tend to bias regression to larger spectral values
because these will be above noise level more often than smaller motions. Do not try to
correct for this bias.
For 10 s insufcient data to yield stable coefcients. Based on numerical simulations,
nd response spectra are approximately at for > 8 s and M < 7.5 and, therefore,
extend model to 20 s by assuming constant spectral displacement. Note that may not be
appropriate for M > 7.5.
Compare predictions to spectrum of Landers 1992 (Mw 7.5) recorded at Lucerne station.
Find that model overpredicts observation.
320
4.52
Boore et al. (1993) & Boore et al. (1997)
See Section 2.94

Cutoff distance is lesser of distance to rst digitized record triggered by S wave, distance
to closest non-digitized recording, and closest distance to an operational nontriggered
instrument.
Note that can only use response spectral values between 0.1 and 2 s because of low
sampling rate of older data (sometimes only 50 samples/sec) and low signal to noise
ratios and lter cutoffs.
Site categories same as in Section 2.94 but due to smaller dataset number of records in
each category is less. Class A: 12 records, B: 51 records, C: 49 records.
Smoothed coefcients using a least-squares t of a cubic polynomial.
4.53
Caillot & Bard (1993)
ln y = 1 + 2 M + 3 ln HYPO + 4 S1
Consider three site conditions but only retain two:
1. Rock: ENEA/ENEL S0 classication S1 = 0, 49 records.
2. Thin alluvium: depth of soil between 5 and 20 m,ENEA/ENEL S1 classication
S1 = 1, 34 records.
Selected records have de < 60 km and focal depth less than 30 km. Data selected so
that mean and standard deviation of magnitude and hypocentral distance in each site
category are equal, in this case 5.1 and 20 km respectively.
All records processed using common procedure. High pass ltered with fl = 0.5 Hz,
instrument corrected and low pass ltered with fh = 30 Hz.
Considered three things when choosing method of analysis:
1. Attenuation equation must have some physical basis.
2. Parameters must be available for original data set.
3. Attenuation equation must be easy to use in a predictive manner.
Hypocentral distance used because rupture not known for most earthquakes. Note that
only important for magnitudes greater than about 6.5 and distances less than about
15 km.
Originally included another set of data (32 records) from thick soil with depth greater
than about 20 m (ENEA/ENEL classication S2) but note that results for this category
are much more uncertain, possibly due to diversity of geotechnical characteristics of
soils. Therefore excluded.
321
Regression was done using two-stage algorithm (Joyner & Boore, 1981) and a weighted
one-stage method. Weight by splitting the magnitude and distance ranges into four
intervals and weighting data in each interval inversely proportionally to number of points
in the bin. Thus gives roughly equal weight to each part of magnitude-distance space.
Note that results from two-stage regression for this set of data may be misleading because for some periods it does not bring any explanation to the variance of initial data.
The two-stage and normal one-stage and weighted one-stage yield signicant changes
in predictions.
Repeat analysis using only S0 subset and using only S1 subset but no signicant
changes in magnitude or distance scaling between the two subsets so consider complete set and include a constant scaling between rock and shallow soil. If set is reduced
to 53 records with similar spread of magnitude, distance and sites then difference between shallow soil and rock is not signicant.
Note that condence interval should be given by formula in Weisburg (1985) not normal
way of simply using standard deviation.
4.54 Campbell (1993)

See Section 2.95.
Notes that equation can predict smaller pseudo-acceleration than PGA for short periods,
which is impossible in practice. Hence pseudo-acceleration for periods 0.2 s should
be constrained to be PGA.
4.55 Electric Power Research Institute (1993a)

ln[y(f )] = C1 + C2 (M 6) + C3 (M 6)2 + C4 ln(R) + C5 R + C6 ZSS + C7 ZIS + C8 ZDS

Use three site classes
SS Shallow soil (depth to rock < 20 m). ZSS = 1, ZIS = 0 and ZDS = 0.
IS Intermediate soil (depth to rock between 20 and 100 m). Very limited data. ZIS = 1,
ZSS = 0 and ZDS = 0.
DS Deep soil (depth to rock more than 100 m). ZDS = 1, ZSS = 0 and ZIS = 0.
Cannot also examine effect of rock type (hard crystalline; hard sedimentary; softer,
weathered; soft over hard) because of lack of data from non-crystalline sites in SS and
IS classes.
Collect all data from strong-motion instruments in eastern North America (ENA) and all
seismographic network data from mb 5.0 at 500 km. Also include some data from
Eastern Canadian Telemetered Network (ECTN).
322
Most data from M < 5 and > 10 km.

Roughly half the data from aftershocks or secondary earthquakes in sequences.
Limit analysis to M 4 because focus is on ground motions of engineering interest.
Use geometric mean to avoid having to account for correlation between two components.
Note the large error bars on C3 , C5 shows that data does not provide tight constraints
on magnitude scaling and attenuation parameters.
Do not provide actual coefcients only graphs of coefcients and their error bars.
Find smaller inter-event standard deviations when using mLg than when using Mw .
Examine effect on standard deviation of not including site terms. Compute the statistical
signicance of the reduction using the likelihood ratio test. Conclude that the hypothesis
that the site terms are zero cannot be rejected at any period.
Split data by region: the Gulf Coast (no records), the rest of ENA or a subregion of ENA
that may have marginally different attenuation characteristics. Add dummy variable to
account for site location in one of the two zones with data and another dummy variable
for earthquake and site in different zones. Neither variable is statistically signicant due
to the limited and scattered data.
Try tting a bilinear geometric spreading term but nd that the reduction in standard
deviation is minimal.
4.56
Sun & Peng (1993)
See section 2.101.

Coefcients not given.
4.57
Boore et al. (1994a), Boore et al. (1997) & Boore (2005)
See Section 2.103

Find no evidence for magnitude dependent uncertainty for spectral values.
Find no evidence for amplitude dependent uncertainty for spectral values.
Note that effect of basin-generated surface waves can have an important effect but
probably not at periods between 0.1 and 2 s.
4.58
Climent et al. (1994)
Inspect observed and predicted values and conclude no clear difference between uppercrustal and subduction zone ground motions. Equations are for region regardless of
earthquake source type.
323
4.59 Fukushima et al. (1994) & Fukushima et al. (1995)

See Section 2.105.
Only give graphs of coefcients.
Note possible noise contamination, for periods < 0.1 s, in coefcients.
4.60 Lawson & Krawinkler (1994)

See Section 2.106.
4.61 Lee & Manic (1994) & Lee (1995)

2
[ = M< + Att + b1 M<> + b S (1) + b S (2) + b3 v + b4 + b5 M<>
log10 PSV
2
2
(1)
(2)
(1) (1)
+ b6 S L
M< = min(M, Mmax )

(1 + b1 )
where Mmax =
2b5
M<> = max(M< , Mmin )
b1
where Mmin =
2b5
{
A0 log10 for R R0
Att =
0)
A0 log10 0 (RR
for R > R0
200
{
0.761 for T 1.8 s
with: A0 =
0.831 + 0.313 log10 T 0.161(log10 T )2 for T < 1.8 s
[ ( 2
)] 12
R + H 2 + S2
= S ln
R2 + H 2 + S02
0 = (R0 )
{
[[
]}
]2
1 200A0 (1 S02 /S 2 )
200A0 (1 S02 /S 2 )
where R0 =
+
4H 2
2
ln 10
ln 10
where is representative distance, S is size of fault, S0 is coherence radius of source
and v is component orientation (v = 0 for horizontal, v = 1 for vertical).
Consider three geological site conditions:
s = 0 Sediment: S (1) = 0, S (2) = 0, 151 records.

s = 1 Intermediate sites: S (1) = 1, S (2) = 0, 106 records.
s = 2 Basement rock: S (1) = 0, S (2) = 1, 54 records.
324
Consider three local site categories but only retain two:

(1)
sL = 0 Rock: SL = 0, 100 records.

(1)
sL = 1 Stiff soil: SL = 1, 205 records.

Cannot include those records from deep soil sites (sL = 2) because only six records.
Most earthquakes are shallow, depth H < 25 km.
Most records have epicentral distances, R < 50 km.
Most have magnitudes between 3 and 6.
Only use records with high signal-to-noise ratio. Quality of records is not adequate for
response spectrum calculation outside range 0.04 to 2 s.
Analysis performed using residue 2-step method. In rst step use only records from
M 4.25 to force a concave form to magnitude scaling (if all records used then nd
a convex parabola), sL parameter is not included. In second step nd sL dependence
from residuals of rst stage including all magnitudes.
Give expressions to describe distribution of residuals so that can nd condence limits,
unlike normal standard deviation based method, see Trifunac (1980).
Note difference between western USA and Yugoslavian ground motions.
4.62
Mohammadioun (1994a)
log SR(f ) = k(f ) + (f )M + n(f ) log R

Uses records from rock sites (Vs 750 m/s).
Half of records from R < 30 km and signicant number from R < 10 km.
Most (82%) records from earthquakes with 6.2 M 7.0.
Coefcients not given, only results.
4.63
Mohammadioun (1994b)
log V (f ) = k(f ) + (f )M + n(f ) log R

Choose W. USA to make data as homogeneous as possible in terms of seismotectonic
context and parameter quality.
325
Notes recording site-intensities may only be average intensity values, thereby neglecting
possible microzoning effects.
Uses ML because generally available and uniformly determined. Notes may not be best
choice.
Records from free-eld and typical of different intensity classes.
Does regression for records associated with three different intensities: V (184 records,
5.5 . R . 200 km), VI (256 records, 3 . R . 250 km, VII (274 records, 1 . R .
150 km) and four different intensity groups: V-VI, VI-VII, VII and more (extra 25 records,
1 . R . 100 km) and V and less (extra 30 records, 25 . R . 350 km.
Graph of (f ) given for horizontal component for the four intensity groups and graph of
n(f ) for vertical component for intensity VI.
4.64 Musson et al. (1994)

See section 2.108.
More data because use analogue records as well.
4.65 Theodulidis & Papazachos (1994)

Use same data, equation and procedure as Theodulidis & Papazachos (1992), see
Section 2.92.
Note lack of near-eld data (R < 20 km, M > 6.2) to constrain R0 .
Only give graphs of original coefcients but give table of smoothed (using a ( 14 + 12 + 14
running average along log T ) coefcients for 13 periods and all 5 damping levels.
Note large residuals for T > 0.5 s due mainly to different digitising and processing procedures which signicantly affect long period spectral values.
Check histograms of residuals for 0.1, 0.3, 0.5, 1, 3 and 5 s and nd similar to normal
distribution.
Note no data from R < 30 km for M > 6.5 so state caution is required for use of
equations in that range. Also suggest do not use equations for M > 7.5 or for R >
130 km.
Note may not apply for very soft soils.
Note lack of data.
326
4.66
Dahle et al. (1995)
See Section 2.114.

Derive spectral attenuation relations for almost double number of periods given. Coefcients smoothed using a third degree polynomial.
4.67
Lee & Trifunac (1995)
Based on Lee et al. (1995). See Section 2.115.

Response parameter is pseudo-velocity for 5% damping (also use 0, 2, 10 and 20%
damping but do not report results).
Before regression, smooth the actual response spectral amplitudes along the log10 T
axis to remove the oscillatory (erratic) nature of spectra.
State that for small earthquakes (M 3) equations only valid up to about 1 s because
recorded spectra are smaller than recording noise for longer periods.
Only give coefcients for 0.04, 0.06, 0.10, 0.17, 0.28, 0.42, 0.70, 1.10, 1.90, 3.20, 4.80
and 8.00 s but give graphs for rest.
Assume that distribution of residuals from last step can be described by probability function:
p(, T ) = [1 exp( exp((T ) + (T )))]n(T )

[ ) (T ), n(T ) = min[10, [25/T ]],
where p(, T ) is probability that log PSV(T )log PSV(T
[25/T ] is integral part of 25/T . Arrange residuals in increasing order and assign an actual probability of no exceedance, p (, T ) depending on its relative order. Estimate
(T ) and (T ) by least-squares t of ln( ln(1 p1/n(T ) )) = (T )(T ) + (T ). Test
quality of t between p(, T ) and p (, T ) by 2 and Kolmogorov-Smirnov tests. For
some periods the 2 test rejects the t at the 95% level but the Kolmogorov-Smirnov test
accepts it.
4.68
Ambraseys et al. (1996) & Simpson (1996)
See Section 2.119.

Do no smoothing because if plotted on a normal scale then smoothing should be done
on T , but if on log-log plot then smoothing should be done on log T .
4.69
Ambraseys & Simpson (1996) & Simpson (1996)
See Section 2.120.

327
4.70 Bommer et al. (1996)

See section 2.122.
Response parameter is pseudo-velocity for unspecied damping.
4.71 Crouse & McGuire (1996)

See section 2.123.
Find k1 not signicantly different than 1 for T 0.15 s and k2 not signicantly different
than 1 for T 0.50 s.
4.72 Free (1996) & Free et al. (1998)

See Section 2.124.
Finds including focal depth, h, explicitly has dramatic effect on predicted spectra at short
distances but insignicant effect at large distances.
Repeats analysis using only E. N. American data. Finds signicantly larger amplitudes
than predictions from combined set for short and intermediate distances for periods
> 0.3 s but similar spectra for large distances.
4.73 Molas & Yamazaki (1996)

Based on Molas & Yamazaki (1995), see Section 2.88 of Douglas (2001a).
Response parameters are absolute acceleration and relative velocity for 5% damping.
4.74 Ohno et al. (1996)

See Section 2.126.
Plot amplitude factors from rst stage against Mw ; nd well represented by linear function.
Do not give table of coefcients only graphs of coefcients.
328
4.75
Sabetta & Pugliese (1996)
Ground-motion model used is:
log10 Y = a + bM log10
d2 + h2 + e1 S1 + e2 S2
Response parameter, Y , is pseudo-velocity for 5% damping

Use data from Sabetta & Pugliese (1987).
Remove anelastic decay term because it was not signicant at = 0.1 and sometimes
it was positive. Originally geometrical decay coefcient c was allowed to vary but nd it
is close to 1 so constrain.
S1 = 1, S2 = 0 Shallow: depth H 20 m alluvium 400 Vs 800 m/s.

S1 = 0, S2 = 1 Deep: depth H > 20 m alluvium 400 Vs 800 m/s.
S1 = 0, S2 = 0 Stiff: Vs > 800 m/s.
Accelerograms digitised at 400 samples/sec. Bandpass frequencies chosen by an analysis of signal and xed trace Fourier spectra. fmin between 0.2 and 0.7 Hz most about
0.3 Hz and fmax between 20 and 35 Hz most about 25 Hz. Instrument correction applied.
Use one-stage method although two-stage method yields similar results.
Also present smoothed coefcients.
4.76
Spudich et al. (1996) & Spudich et al. (1997)
See Section 2.130

Only use spectral values within the passband of the lter used to correct records hence
number of records used for each period varies, lowest number is 99 for periods between
1.7 and 2.0 s.
Smooth coefcients using cubics or quadratics.
329
4.77 Abrahamson & Silva (1997)

Ground-motion model is4 :
ln Sa = f1 + F f3 + HW fHW (M )fHW (Rrup ) + Sf5
a1 + a2 (M c1 ) + a12 (8.5 M )n + [a3 + a13 (M c1 )] ln R
for M c1
f1 =
n + [a + a (M c )] ln R
a
+
a
(M
c
)
+
a
(8.5
M
)
1
4
1
12
3
13
1
for M > c1
where R =
rrup + c24
a5 for M 5.8
a5
a5 + ca165.8
(M 5.8) for 5.8 < M < c1
f3 =
a6 for M c1
0 for M 5.5
5.5
for 5.5 < M < 6.5
fHW (M ) =
1 for M 6.5
0 for rrup < 4
rrup 4
a9 4
for 4 < rrup < 8
a
for
8 < rrup < 18
fHW (rrup ) =
)9
(
r
18
rup
for 18 < rrup < 24

a9 1
0 for rrup > 25

[ + c5 )
f5 = a10 + a11 ln(PGA
[ is expected peak acceleration on rock as predicted by the attenuation equawhere PGA
tion with S = 0.
Response parameter is acceleration for unspecied5 damping.
S = 0 Rock: rock (Vs > 600 m/s), very thin soil (< 5 m) over rock or soil 5 to 20 m thick
over rock.
S = 1 Deep soil: deep soil in narrow canyon (soil > 20 m thick and canyon < 2 km wide)
or deep soil in broad canyon (soil > 20 m thick and canyon > 2 km wide).
All records reprocessed using common procedure. Interpolated to 400 samples/sec,
low-pass ltering with corner frequency selected for each record based on visual examination of Fourier amplitude spectrum, instrument corrected, decimated to 100 to 200
samples/sec depending on low-pass corner frequency, baseline correction using 0 to 10
degree polynomial, high-pass ltered based on integrated displacements.
Only use response spectral data within frequency band 1.25fh to 0.8fl to avoid effects
of lter roll-off. Hence number of records used for regression at each period varies,
minimum number is less than 100 records for 0.01 s.
Well distributed dataset in terms of magnitude and distance.
4
f3 given in Abrahamson & Silva (1997) was modied to ensure homogeneity and a linear variation in f3 with
magnitude.
5
It is probably 5%.
330
Supplement data with records from Gazli, Friuli, Tabas, Taiwan, Nahanni and Spitak.
Consider source mechanism: reverse F = 1, reverse/oblique F = 0.5, others
(strike-slip and normal) F = 0).
Consider hanging wall effect: if over hanging wall HW = 1, otherwise HW = 0.
Note that interpretation of c4 is not clear for their distance measure but yields better t.
Model nonlinear soil response by f5 .
Model uncertainty as magnitude dependent.
Fix some coefcients to be independent of period so that response spectral values vary
smoothly with distance, magnitude and period.
Smooth coefcients using piecewise continuous linear ts on log period axis. For highly
correlated coefcients, smooth one coefcient and re-estimate other coefcients.
4.78
Atkinson (1997)
log PSA = c0 + c1 (Mw 6) + c2 (Mw 6)2 + c3 h ca1 log R ca2 R + cs S

with: ca2
= ca3 + ca4 h

Uses two site categories (no soil proles were available for Cascadia region):
S = 0 Rock: average Vs assumed to be about 2000 m/s

S = 1 Soil: average Vs assumed to be about 255 m/s (although includes some soft soil
sites with average Vs about 125 m/s).
Tectonic type of earthquakes used: crustal, subcrustal and subduction
Most Cascadia data is from seismograms. Converts vertical measurements from these
to one horizontal component.
Supplements in large magnitude range (6.7 < Mw 8.2) with data from 9 subduction
earthquakes in Alaska, Mexico, Japan and Chile
Most magnitudes below 5.3 and no data between 6.8 and 7.5.
Focal depths between 1 and 60 km
Only uses events recorded at 3 or more stations. Improves ability of regression to distinguish between magnitude and distance dependencies in data.
Most low magnitude events were recorded on rock and most high magnitude events
were on soil. Thus to stabilize regression takes the coefcients cs from Boore et al.
(1994a) and not derived from this data.
331
Magnitude partitioning, in rst step, into 0.5 unit intervals gave evidence for magnitude
dependent attenuation. Uses ca1 = 1 for 4.1 Mw 6.7 and ca1 = 0.5 (largest
which yielded positive ca2 ) for Mw 7.5. Thought to show breakdown of point source
assumption.
Demonstrates depth dependence in anelastic decay by performing regression in four
15 km deep subsets for range 4.1 Mw 6.7. ca3 and ca4 then nds by regression for
each period. No depth dependence for Mw 7.5 because of lack of different depths.
Includes depth dependence in second step because gave better t for short periods.
Checks dependence on crustal, interface and intra-slab events; nds no dependence.
4.79 Campbell (1997), Campbell (2000) & Campbell (2001)

See Section 2.132
Ground-motion model (horizontal component) is:
ln SAH
fSA
= ln AH + c1 + c2 tanh[c3 (M 4.7)] + (c4 + c5 M )RSEIS + 0.5c6 SSR

+ c6 SHR + c7 tanh(c8 D)(1 SHR ) + fSA
{
0 for D 1 km
=
c6 (1 SHR )(1 D)(1 0.5SSR ) for D < 1 km
Ground-motion model (vertical component) is:
ln SAV
= ln SAH + c1 + b2 M + c2 tanh[d1 (M 4.7)] + c3 tanh[d2 (M 4.7)]

+ b3 ln[RSEIS + b4 exp(b5 M )] + b6 ln[RSEIS + b7 exp(b8 M )] + b9 F
+ [c4 tanh(d3 D) + c5 tanh(d4 D)](1 SSR )

Notes importance of depth to basement rock, D, for modelling long period site response.
For shallow sediments denes D as depth to top of Cretaceous or older deposits, for
deep sediments determine D from crustal velocity proles where dene basement as
crystalline basement rock or sedimentary deposits having a P-wave velocity 5 km/s
or shear-wave velocity 3 km/s (called seismic basement by geophysicists).
Uses different data than for PGA equations hence: reverse (3), thrust (H:9, V:6), reverseoblique (2) and thrust-oblique (0), total (H:14, V:11) (H:140 records, V:85 records), strikeslip (H:124 records, V:88 records). Only two normal faulting earthquakes in horizontal
set of records (contributing 2 records) so a difference in not modelled although F = 0.5
is given as rst approximation (later revised to F = 0) to use as for PGA case.
Only excludes records from toe and base of dams, included those from buildings and
bridge columns which were excluded from PGA study, because of lack of data.
Uses weighted regression analysis. Assigns recordings from a given earthquake that
fell within the same distance interval (ten logarithmical spaced) same weight as those
recordings from other earthquakes that fell within the same distance interval. Gives
recordings from a given earthquake that occurred at the same site location the same
cumulative weight as a single recording at that distance, thus reducing the bias.
332
Performs analysis on spectral ratio ln(PSA/PGA) because of unacceptably large periodto-period variability in regression coefcients when direct regression is applied and
strongly correlated coefcients. Notes that are too many regression coefcients so it
was necessary to perform analysis in many steps, at each step different coefcients are
determined and detrended and residuals examined to nd appropriate functional forms
for trends present. Yields more stable results.
No consideration of nontriggering instruments made, unlike PGA study.
4.80
See Section 2.136.

4.81
See Section 2.137.

[
]
ln(SA/PGA) = B1 + B2 (10 M)3 + B3 ln rrup + e1 +2 M
where 1 and 2 are set equal to C4 and C5 of appropriate PGA equation.
Response parameter, SA, is acceleration for 5% damping.
Do analysis on response spectral amplication because digitised and processed accelerograms used for spectral attenuation is only a subset of PGA database and they
are often those with strongest shaking. Hence analysis directly on spectral accelerations
may be biased.
Smooth coefcients.
4.82
log(SD) = C1 + C2 M + C4 log r + CA SA + CS SS
r =
d2 + h20
Response parameter is displacement for 5, 10, 15, 20, 25 and 30% damping.
Use three site conditions:
333
Use subset of data of Ambraseys et al. (1996) (see 2.119) data with a few changes and
exclusion of records from earthquakes with Ms < 5.5 because ground motion at long
periods was of interest and to increase likelihood of acceptable single-to-noise ratio at
longer periods.
Each record individually ltered. Firstly lter record with sharp low cut-off at 0.1 Hz and
plot velocity and displacement time-histories. Check, visually, whether contaminated by
noise and if so increase cutoff frequency by small amount and repeat procedure until
resulting velocity and displacement time-histories are deemed acceptable and no signicant improvement is observed by further increase of cutoff frequency. Instrument
correction not applied because high frequency distortion caused by transducer characteristics not important for displacement spectra. Only use each record for regression for
periods up to 0.1 s less than lter cutoff used for that record to avoid distortion by lter,
hence as period increases number of data points decreases.
Regression procedure same as Ambraseys et al. (1996), see 2.119.
4.83 Perea & Sordo (1998)

ln Pa = 1 + 2 M + 3 ln(R + 25)
All records from ve medium soft soil sites.
Use mb for M < 6 and Ms otherwise, because mb is more representative of released
energy for small earthquakes and Ms better represents energy release for large earthquakes because mb saturates starting from M > 6.
Try including anelastic decay term, 4 R but it does not signicantly affect standard deviation.
Also repeat analysis for three other zones. Zone 1: 3 earthquakes, 3 records (5.0
M 6.4, 80 R 156 km) for which conclude has too limited data for reliable
equation. Zone 36 : 11 earthquakes, 13 records (4.5 M 7.7, 251 R 426 km)
for which nd ts spectra of medium sized shocks better than large shocks because of
lack of data for large earthquakes. Zone 4: 4 earthquakes, 7 records (5.1 M 6.2,
356 R 573 km) for which nd 2 is negative and 3 is positive for some periods
(which is nonphysical) which state is due to limited number of earthquakes and their
similar epicentral distances.
Find t spectra of medium sized earthquakes than large earthquakes because of lack of
data from large earthquakes.
Only give graphs of coefcients.
6
The following values are from their Table 1 which does not match with their Figure 3.
334
4.84
Reyes (1998)
See Section 2.143.

4.85
Shabestari & Yamazaki (1998)
log y(T ) = b0 (T ) + b1 (T )M + b2 (T ) log r + b4 (T )h + ci (T )

where ci (T ) is the station coefcient, reecting relative site effect for each period, assuming zero mean for all stations.
Response parameters are acceleration and velocity for 5% damping.
Include at least ve earthquakes with MJMA 7.2.
Exclude earthquakes with focal depths, h, equal to 0 km or greater than 200 km.
Exclude records with vectorial composition of PGA less than 0.01 m/s2 .
Use three-stage iterative partial regression method.
For T 6 s constrain horizontal anelastic coefcient to zero because get positive coefcient.
See Yamazaki et al. (2000) for examination of station coefcients.
4.86
Chapman (1999)
See Section 2.150.

Response parameter is pseudo-velocity for 2, 5 and 10% damping.
4.87
Spudich et al. (1999) & Spudich & Boore (2005)
See Section 2.154.

Use only use response spectral data within frequency band 1.25fh to 0.75fl to avoid effects of lter roll-off. Eight records were not processed like the rest so use only response
spectral values within 0.1 to 1 s. Hence number of records used for regression at each
period varies, minimum number used is 105 records for 2 s.
Give smoothed coefcients using cubic function.
335
4.88 Ambraseys & Douglas (2000), Douglas (2001b) & Ambraseys

& Douglas (2003)
See Section 2.157.
Find b2 and b3 signicantly different than 0 at 5% level for all periods but bA and bS not
signicant for many periods (especially for vertical component).
Find deamplication for vertical component on soft and stiff soil compared with rock.
Check by removing all 34 Northridge records (many of which were on soft soil) and
repeat analysis; nd little change.
Also derive equations for horizontal response under inuence of vertical acceleration
using a bending SDOF model; nd little change in response.
4.89 Bozorgnia et al. (2000)

See Section 2.158.
Different set of data than for PGA hence: strike-slip: 20 earthquakes (including one
normal faulting shock), reverse: 7 earthquakes and thrust: 6 earthquakes.
Find considerable period-to-period variability in coefcients causing predicted spectra
to be very jagged near limits of magnitude and distance ranges so carried out partial
smoothing of coefcients.
4.90 Campbell & Bozorgnia (2000)

See Section 2.159.
4.91 Chou & Uang (2000)

log Y = a + b(M 6) + c(M 6)2 + d log(D2 + h2 )1/2 + eGc + f Gd

Use three site categories (based on average shear-wave velocity, Vs , over top 30 m):
Classes A+B Hard rock or rock: Vs > 760 m/s, Gc = 0, Gd = 0, 35 records.
Class C Very dense soil and soft rock: 360 < Vs 760 m/s, Gc = 1, Gd = 0, 97 records.
Class D Stiff soil: 180 Vs 360 m/s, Gc = 0, Gd = 1, 141 records.
Records from free-eld or ground level of structures no more than two storeys in height.
336
Smooth coefcients using cubic polynomial.

Do not give coefcients for all periods.
Find cannot use equation to predict near-eld ground motions.
4.92
Field (2000)
See Section 2.160.

Distribution w.r.t. site class for 3.0 s is: B, 10 records; BC, 27 records; C, 13 records;
CD, 119 records; D, 187 records; DE, 1 record.
Constrains b3 for 1.0 and 3.0 s to zero because originally nds positive value.
151 records have basin-depth estimates.
Does not nd signicant slopes for residuals w.r.t. predicted ground motion at BC sites.
Plots squared residuals w.r.t. Vs and nds small signicant trends for 1.0 and 3.0 s.
4.93
Kawano et al. (2000)
log Si (T ) = a(T )M {b(T )Xeq + log Xeq } + ci (T )

where ci (T ) is an individual site amplication factor for each of 12 stations.
Use data either recorded at ground surface where 0.5 Vs 2.7 km/s (1.7 Vp
5.5 km/s) or obtained by analytically removing effects of uppermost surface layers of
ground from underground observation data (or by stripping-off analysis) using underground structure.
Use only ground motion after arrival of rst S wave because most important for aseismic
design.
Do not give table of coefcients, only graphs of coefcients.
Dene amplication factors, di (T ) = ci (T ) c0 (T ) for horizontal motion and di (T ) =
cv,i (T ) c0 (T ) for vertical motion, where c0 (T ) is the regression coefcient for data
observed at ground layer equivalent to seismic bedrock.
Find Sh (T ) = Sb (T )h (T )h (T ) where Sb (T ) is S0 (T ). h (T ) = (Vs /Vs,b )h (T ) for
T Ts,1 and h (T ) = h (Ts,1 ) for T > Ts,1 where Ts,1 is the primary predominant
period of surface layer. h (T ) = 1 for T Ts,1 , h (T ) = (T /Ts,1 ) log(h (Ts,1 )) for
10Ts,1 > T > Ts,1 and h (T ) = 10 log(h (Ts,1 )) for T 10Ts,1 . Vs,b = 2.2 km/s.
Similar relationships are dened for vertical motion, Sv (T ).
Note that relation does not include effect of source mechanism or rupture propagation,
so probably less valid in near-fault region.
337
4.94 Kobayashi et al. (2000)

See Section 2.162.
Use signicantly less records for T > 1.5 s.
4.95 McVerry et al. (2000)

Ground-motion model for crustal earthquakes is (using form from Abrahamson & Silva
(1997), see Section 4.77):
ln SA0 (T ) = C1 (T ) + C4AS (M 6) + C3AS (T )(8.5 M )2 + C5 (T )r

2
+ (C8 (T ) + C6AS (M 6)) ln(r2 + C10AS
(T ))1/2 + C46 (T )rVOL
2
+ {C2 (T )r + C44 (T ) + (C9 (T ) + C7 (T )(M 6))(ln(r2 + C10AS
(T ))1/2
ln C10AS )}
+ {C29 (T )}
+ {C30AS (T ) ln(PGA0W A + 0.03) + C43 (T )}
+ C32 CN + C33AS (T )CR
Also add on hanging wall term, see Section 4.77. Subscript AS denotes those coefcients from Abrahamson & Silva (1997). Three parts of equation within {. . .} are for
site conditions MA/SA, Class B and Class C respectively. PGA0W A is the predicted
PGA (SA0 (0)) for weak rock category. CN = 1 for normal mechanism and 0 otherwise. CR = 0.5 for reverse/oblique, 1.0 for reverse and 0 otherwise. Ground-motion
model for subduction zone earthquakes is (using form from Youngs et al. (1997), see
Section 4.81):
ln SA0 (T ) = C11 (T ) + [C12Y + (C17Y (T ) C17 (T ))C19Y ]

+ C13Y (T )(10 M )3 + C17 (T ) ln(r + C18Y exp(C19Y M )) + C20 (T )HC
+ C24 (T )SI + C46 (T )rVOL (1 DS)
+ {C44 (T ) + C16 (T )(ln(r + C18Y exp(C19Y M ))
ln(C18Y exp(C19Y M )))}
+ {C29 (T )}
+ {C30Y (T ) ln(PGA0W A + 0.03) + C43 (T )}
Subscript Y denotes those coefcients from Youngs et al. (1997). Three parts of equation within {. . .} are for site conditions MA/SA, Class B and Class C respectively. SI = 1
for subduction interface and 0 otherwise. DS = 1 for deep slab and 0 otherwise. rVOL
is length of path that lies in the volcanic zone.
Use four site conditions (mostly based on geological descriptions rather than measured
shear-wave velocity):
WA Weak rock sites, or sites with soil layer of thickness 3 m overlying weak rock.
338
MA/SA Moderate-strength or strong rock sites, or sites with soil layer of thickness 3 m
overlying moderate-strength or strong rock.
Class B Intermediate soil sites or sites with soil layer of thickness > 3 m overlying rock.
Class C Flexible or deep soil sites with natural periods > 0.6 s.
Justify soil categories using statistical studies of residuals at early stage. Exclude response spectra from very soft soil sites (Vs < 150 m/s for depths of & 10 m).
Use data for PGA equation from Zhao et al. (1997), see Section 2.138.
Exclude records from bases of buildings with >4 storeys.
Use less records for long periods because noise.
Lack of data prevent development of robust model purely from NZ data. Plot residuals
of predicted response using published attenuation relations (base models) for other areas to nd relations which gave good representations of NZ data. Then modify some
coefcients to improve match; imposing constraints so that the selected models control
behaviour at short distances where NZ data lacking. Require crustal and subduction
zone expressions for rock sites to match magnitude dependence of base models at
r = 0 km. Constrain coefcients that occur nonlinearly and nonlinear site response
coefcient for Class C to base model values.
Find anelastic attenuation term and additive terms for shallow slab earthquakes for subduction earthquakes not statistically signicant. Also differences in attenuation rates for
shallow slab, deep slab and interface earthquakes not statistically signicant.
Exclude deep slab earthquakes because of high attenuation in mantle; note equation
should not be used for such earthquakes.
Different attenuation rate for site category MA/SA because of magnitude dependence
apparent in residuals for simpler model.
Eliminate nonlinear site response term for Class B because nd unacceptable (positive)
values of coefcient and constraining to negative values produces poorer t.
Predicted PGA (SA0 (0)) from response spectrum set of records considerably smaller
than those, SA(0), from the complete PGA set of records. Thus scale SA0 (T ) by ratio
SA(0)/SA0 (0).
Standard error has a magnitude dependent intra-event component and a magnitude
independent inter-event component.
Note lack of data for large magnitude subduction zone earthquakes and large magnitude
near source data for crustal earthquakes.
Do not give coefcients, only predictions.
4.96
Monguilner et al. (2000b)
log SA (T ) = A(, T ) + M + b1 (T ) + b2 (T )M + b3 (T )s + b4 (T )v + b5 (T )M 2 + ep (i)

339
where A(DE, H, S, T ) = A0 (T ) log (DE, H, M ), = (DE2 + H 2 + S 2 ) 2 , H is focal

depth, p is the condence level, s is from site classication (details not given in paper)
and v is component direction (details not given in paper although probably v = 0 for
horizontal direction and v = 1 for vertical direction).
Response parameter is pseudo-acceleration for unknown damping level.
Use same data and weighting method as Monguilner et al. (2000a) (see Section 2.163).
Find A0 (T ) by regression of the Fourier amplitude spectra of the strong-motion records.
Estimate fault area, S , using log S = Ms + 8.13 0.6667 log(/).
Equation only valid for Mmin M Mmax where Mmin = b2 /(2b5 (T )) and Mmax =
(1 + b2 (T ))/(2b5 (T )). For M < Mmin use M for second term and M = Mmin
elsewhere. For M > Mmax use M = Mmax everywhere.
0 (T ) where S 0 (T ) is the observed pseudo Examine residuals, (T ) = log SA (T )log SA
A
acceleration
and
t
to
the
normal
probability
distribution,
p(, T ) = exp[(x (T ))/(T )]2 /((T ) 2), to nd (T ) and (T ). Find that the
residuals t the theoretical probably distribution at the 5% level using the 2 and KS7
tests.
Do not give coefcients, only graphs of coefcients.
4.97 Shabestari & Yamazaki (2000)

log y(T ) = b0 (T ) + b1 (T )M + b2 (T ) log r + b4 (T )h + ci (T )

where ci (T ) is the station coefcient, reecting the relative site effect for each period,
assuming zero mean for all stations.
Response parameters are acceleration and velocity for 5% damping.
Depths between 1 (includes earthquakes with depths reported as 0 km) and 158 km.
Exclude earthquakes with focal depths greater than 200 km.
Exclude records with vectorial composition of PGA less than 0.01 m/s2 .
Exclude data from stations which have recorded less than two records, because the
station coefcient could not be determined adequately. Use records from 823 stations.
Most records from distances between 50 and 300 km.
Use three-stage iterative partial regression method.
For T 5 s constrain horizontal anelastic coefcient to zero because get positive coefcient.
7
Probably this is Kolmogorov-Smirnov.
340
4.98
Smit et al. (2000)
See Section 2.165.

4.99
Takahashi et al. (2000)
See Section 2.166.

For periods 1 s long period noise in records leads to reduction in number of records.
Set b and e to zero at long periods because estimates not statistically signicant.
Find that soft soil site correction terms may be affected by different processing procedures for data from different sources.
4.100
See Section 2.170.

4.101
Das et al. (2002)
log[P SV (T )] = c1 (T ) + c2 (T )M + c3 (T )h + c4 (T ) log(
R2 + h2 ) + c5 (T )v
where v = 0 for horizontal and 1 for vertical.

Response spectral parameter is pseudo-velocity for 5% damping.
Use records from stiff soil/rock sites.
Use square-root-of-sum-of-squares (SRSS) to combine horizontal components to reduce strong azimuthal dependence of ground motions. Note that dividing predicted
spectra by 1.41 gives spectrum for each component separately.
Do not derive equations for T > 1 s because of baseline problems and noise in accelerograms at longer periods.
Try more complex functional forms but not enough data to constrain all parameters to
physically-realistic values.
Smooth coefcients using unspecied technique.
Report residual spectra for different probability levels not .
341
4.102 Glkan & Kalkan (2002)

See Section 2.175.
4.103 Khademi (2002)

See Section 2.176.
4.104 Manic (2002)

log PSV(T ) = c1 (T ) + c2 (T )M + c3 (T ) log(R) + c4 (T )SA

where R = (d2 + d20 )1/2
Response parameter is pseudo-velocity for 5% damping,
SA = 0 Rock, Vs,30 > 750 m/s.

SA = 1 Stiff soil, 360 < Vs,30 750 m/s.
Soft soil sites (Vs 360 m/s) do not exist in set of records.
Use technique of Ambraseys et al. (1996) to nd the site coefcient c4 (T ), i.e. use
residuals from regression without considering site classication.
Derives separate equations for Ms and ML and for rjb and repi .
4.105 Schwarz et al. (2002)

See Section 2.179.
4.106 Zonno & Montaldo (2002)

See Section 2.182.
342
4.107
Alarcn (2003)
See Section 2.183.

Response parameter is acceleration for 0, 5 and 10% damping but only report coefcients for 5% damping.
Derive equations for 848 periods but only reports coefcients for 11 periods.
4.108
See Section 2.185.

4.109
Berge-Thierry et al. (2003)
log10 PSA(f ) = a(f )M + b(f )d log10 d + c1 (f ) + c2 (f )

where c1 (f ) is for rock sites and c2 (f ) is for alluvium sites.
Use two site categories based on Vs where Vs is the average shear-wave velocity in top
30 m:
1. Rock, Vs > 800 m/s.
2. Alluvium, 300 < Vs < 800 m/s.
Note that some uncertainty in site classication due to lack of Vs values at many stations.
Response parameter is spectral acceleration for 5%, 7%, 10% and 20% damping.
Note that not enough data to derive an equation using only French data so had to use
European and US data.
Use only records from earthquakes with focal depth 30 km so as to be consistent with
shallow crustal earthquakes in France.
Predominately use corrected data from Ambraseys et al. (2000).
Supplement European data with some data from western USA to improve the magnitude
and distance distribution.
Exclude records from Ambraseys et al. (2000) from earthquakes with Ms < 4.
Exclude records from Ambraseys et al. (2000) with record lengths < 10 s.
Exclude records from Ambraseys et al. (2000) with poor visual quality.
Exclude records from Ambraseys et al. (2000) from non-free-eld stations or those
inside a building on the third oor or higher.
8
On page 8 of paper it says 88 periods.
343
Exclude records from Ambraseys et al. (2000) from stations with unknown or very soft
soil site conditions.
Processing procedure of records from Ambraseys et al. (2000) is: baseline correct
uncorrected record, re-sample record to 0.01 s time-step and bandpass ltered using a
elliptical lter with cut-offs of 0.25 and 25 Hz because most instruments were SMA-1s
with natural frequency of 25 Hz and damping of 60%. No instrument correction was
applied because instrument characteristics are not known.
Only use US records from earthquakes with M > 6.
Use the already corrected records from USGS and CDMG.
Most data from rock sites is from earthquakes with M < 6.
49.7% of data is from Italy and 16.9% is from USA. All other countries contribute less
than 10% each.
Use hypocentral distance because believe it accounts for both point and extended sources.
Use uniformly calculated Ms for data from Ambraseys et al. (2000) and Mw for data
from W. USA, which believe is equivalent for Ms for Mw > 6.
Coefcients only reported for horizontal spectral acceleration for 5% damping.
Note that recent data, e.g. Chi-Chi, shows saturation of ground motions at short distances but data used only contains a few records at close distances so data not sufcient
to model such phenomenon.
Obtain positive b(f ) coefcients for periods > 1s which believe is due to low frequency
noise and surface waves.
Believe that small difference between estimated rock and alluvium motions could be due
to incorrect site classication at some stations.
Repeat regression using a randomly selected half of the data. Find very small differences between predicted ground motions using half or complete data set so believe
equation is stable.
Repeat regression excluding data from W. USA and nd very small differences between
predicted ground motions so believe equation is not inuenced by data from W. USA.
Repeat regression using Mw rather than Ms if available and nd that predicted ground
motions are different but that the predictions using Ms are higher than those using Mw
so note that equation using Ms is conservative hence it is useful in a nuclear safety
assessment.
Repeat regression using rrup rather than rhypo and nd that predicted ground motions
using rhypo are higher than when rrup is used because using rhypo places source further
from source of energy.
Plot residuals for 0.03 and 2 s and nd not systematic bias in residuals.
344
4.110
See Section 2.187.

4.111
Campbell & Bozorgnia (2003d,a,b,c) & Bozorgnia & Campbell (2004b)
See Section 2.188.

To make regression analysis more stable set c2 equal to value from better-constrained
regression of uncorrected PGAs.
Do limited amount of smoothing of regression coefcients to reduce the considerable
amount of period-to-period variability in the regression coefcients that caused variability
in predicted pseudo-acceleration especially for small distances and large magnitudes.
4.112
Fukushima et al. (2003)
log Sa(f ) = a(f )M log(R + d(f )10e(f )M ) + b(f )R + c1 1 + c2 2

1. Rock sites with Vs > 800 m/s. 1 = 1 and 2 = 0.
2. Soil sites with Vs < 800 m/s. 2 = 1 and 1 = 0.
Note that some data (Turkish and Japanese) are associated with liquefaction phenomena and so probably Vs < 300 m/s.
Choose functional form to include effect of amplitude saturation close to source.
Note that negative Q values obtained in some ground motion estimation equations may
be due to the lack of amplitude saturation terms.
Do not investigate effect of rupture mechanism, directivity, and the hanging wall effect
because of a lack of data.
Use same set of data as Berge-Thierry et al. (2003) but with the addition of records
from the 1995 Hyogo-ken Nanbu and 1999 Kocaeli earthquakes, which are used to help
constrain the near-source characteristics. In total use 399 records from west Eurasia,
162 from USA, 154 from Hyogo-ken Nanbu and 25 from Kocaeli.
Remove records from distances greater than the distance at which the predicted PGA
is less than 10 cm/s2 (the average trigger level plus the standard error of observation)
as predicted by a previously derived ground motion prediction equation that agrees well
with the 1995 Hyogo-ken Nanbu and 1999 Kocaeli earthquakes although they note the
process should be iterative.
345
Use only records from earthquakes with M 5.5 so as to allow the use of a linear
magnitude dependence.
Due to the nonlinear functional form adopt a iterative method to nd d(f ) and e(f ).
However, due to the lack of near-source data an accurate value of e(f ) cannot be found
therefore set e(f ) to 0.42, which gives accelerations that agree with the observed peak
accelerations in the 1995 Hyogo-ken Nanbu and 1999 Kocaeli earthquakes.
Bandpass lter records with cut-offs of 0.25 and 25 Hz. Note that due to the presence of
many records from analogue instruments the results for frequencies higher than 10 Hz
are less reliable than those for lower frequencies.
Find that for frequencies > 0.4 Hz the b(f ) coefcient corresponds to positive Q values. For lower frequencies the value of b(f ) correspond to negative Q values, which
note could be due to instrumental noise or the effect of surface waves that are not well
represented by the functional form adopted.
Note that the small difference between predicted rock and soil motions may be due to
intrinsic rock amplication due to rock weathering or inappropriate site classication for
some records (e.g. those from the 1999 Kocaeli earthquake, which are all considered to
be on soil).
Plot residuals with respect to regional origin (Hyogo-ken Nanbu, USA, western Eurasian
and Kocaeli) and nd no clear bias or trend.
Note that most of the used near-fault records come from strike-slip earthquakes and so
the equation may be only should be used for prediction of strike-slip motions.
Note that the site classication scheme adopted is very basic but lack information for
more sophisticated method.
4.113 Kalkan & Glkan (2004a)

See Section 2.197.
4.114 Kalkan & Glkan (2004b) and Kalkan & Glkan (2005)
See Section 2.198.
4.115 Matsumoto et al. (2004)

Ground-motion model is (for rrup ):
log SA(T ) = Cm (T )M + Ch (T )Hc Cd (T ) log[R + 0.334 exp(0.653M )] + Co (T )

Ground-motion model is (for rq ):
log SA(T ) = Cm (T )M + Ch (T )Hc Cd (T )Xeq log Xeq + Co (T )

Hc = h for h < 100 km and Hc = 100 km for h > 100 km.
346

Data from at 91 dam sites with rock foundations. Most instruments in inspection gallery
at lowest elevation (for concrete dams) and in bottom inspection gallery (for embankment
dams). Note that 1.8 Vp 4.5 km/s for bedrock of many concrete dams and 1.5
Vp 3.0 km/s for bedrock of embankment dams, which convert to 0.7 Vs 1.5 km/s.
Select data from M > 5, de < 200 km and focal depth h < 130 km.
Most records from h < 60 km.
Most records from d < 100 km.
Classify earthquakes into three types:
Shallow crustal Epicentres located inland at shallow depths. 175 records9 .
Inter-plate Epicentres located in ocean with h < 60 km. 55 records.
Deep intra-slab Epicentres located inland with h > 60 km. 63 records.
Know fault source mechanism for 12 earthquakes.
Adopt 0.334 exp(0.653M ) from earlier Japanese study.
Derive coefcients regardless of earthquake type. Then derive correction factors for
each earthquake type.
Do not report coefcients only graphs of coefcients against period.
Find good agreement between predicted spectra and observed spectra for two stations
that recorded the magnitude 8.0 Tokati-oki 2003 earthquake.
4.116
zbey et al. (2004)
See Section 2.202.

4.117
Pankow & Pechmann (2004) and Pankow & Pechmann (2006)
See Section 2.203.

4.118
Sunuwar et al. (2004)
See Section 2.204.

Developed equations up to 5 s but do not think results for 4 and 5 s are satisfactory.
9
The authors also give number of sets as 81 for shallow crustal, 29 for inter-plate and 29 for deep intra-slab
347
4.119 Takahashi et al. (2004)

log[y(T )] = aM bx log r + e(h hc )h + SR + SI + SS + Sk

r = x + c exp(dM )
Use SR only for crustal reverse events, SI only for interface events, SS only for subduction slab events and Sk for each of the site classes (k = 1, . . . , 4). h = 0 for h < hc
and 1 otherwise. For h > 125 km use h = 125 km.
SC I Rock, natural period T < 0.2 s, Vs,30 > 600 m/s, approximately NEHRP classes A
and B. 1381 records.
SC II Hard soil, natural period 0.2 T < 0.4 s, 300 < Vs,30 600 m/s, approximately
NEHRP class C. 1425 records.
SC III Medium soil, natural period 0.4 T < 0.6 s, 200 < Vs,30 300 m/s, approximately
NEHRP class D. 594 records.
SC IV Soft soil, natural period T 0.6 s, Vs,30 200 m/s, approximately NEHRP classes
E and F. 938 records.
Site classication unknown for 62 records. Prefer using site classes rather than individual coefcients for each station because avoids possibility of source effects being
shifted into site terms and can be used when there are only a few records per station.
Classify earthquakes into three types:
Crustal Focal depths 25 km. 81 earthquakes, 1497 records.
Interface 88 earthquakes, 1188 records.
Slab 101 earthquakes. 1715 records.
Classify earthquakes into four mechanisms:
Reverse 160 earthquakes (28 crustal), 1969 records (373 crustal).
Strike-slip 82 earthquakes (39 crustal), 1674 records (1100 crustal).
Normal 26 earthquakes (4 crustal), 749 records (24 crustal).
Unknown 2 earthquakes (0 crustal), 8 records (0 crustal).
Consider differences between reverse and strike-slip motions for crustal earthquakes
because enough data but note there is not enough data to consider normal earthquakes
as a separate group.
Focal depths, h, between about 0 and 162 km with most < 60 km.
Exclude data from distances greater than a specied limit for a given magnitude in order to eliminate bias due to untriggered instruments. For subduction slab events, x
maximum distance as 300 km.
348
Note that there is little near-source data from Japan from within 30 km. All Japanese
data from within 10 km is from two earthquakes (Kobe 1995 and Tottori 2000). Add data
from with 40 km from earthquakes in western USA (h < 20 km) and from the Tabas
1978 (Iran) earthquake to help constrain near-source behaviour of derived equations.
Use data from: Japan (61 crustal earthquakes, 1301 records; 87 interface earthquakes,
1176 records; 101 slab earthquakes, 1715 records) and Iran and western USA (20
crustal earthquakes; 196 records; 1 interface earthquake, 12 records).
Note that reasonably good distribution of data for all magnitudes and focal depths.
Note strong correlation between focal depth and distance.
Use ISC relocations rather than JMA locations because nd that they are more reliable.
Use Mw values from Harvard CMT unless value from special study is available.
Prefer the one-stage maximum-likelihood method to the two-stage method because
when there many events with only a small number of records and many individual site
terms, the coefcients must be determined using an iterative method and hence their
reliability is questionable.
Find that, by residual analysis (not shown), that equations predict unbiased ground motions for crustal and interface events but biased ground motions for slab events with bias
that depends on distance. Apply this magnitude-independent
path modication factor
2
2
SF for slab events: log(SF) = SSL [log( x + Ra ) log(Rc )] where Ra = 90.0 km and
Rc = 125.0 km.
Find that, because of lack of near-source data, it is not possible to nd reliable estimates
of c and d so use a iterative method to nd d by xing c.
Estimate site coefcient, SH , for hard rock sites (Vs,30 = 1500 m/s) from 10 stations with
1020 Vs,30 2200 m/s with 1436 records, based on residuals.
Examine residuals w.r.t. magnitude, distance and focal depth for all three source types
and nd no signicant bias. Find that PGAs from two events on east coast of Hokkaido
are under-estimated and note that investigation needed to see if it is a regional anomaly.
Also nd that ground motions from 2003 Miyagi (Mw 7.0) event are under-estimated,
which note is due to a known regional anomaly.
Believe model more robust than other models for subduction events due to lower prediction errors.
Note that predictions for near-source ground motion for subduction events are largely
constrained by data from shallow crustal events from western USA hence adding subduction records from < 50 km could result in improvements.
4.120
Yu & Hu (2004)
log Y = c1 + c2 M + c3 log(R + c4 ec5 M )

349
Use data from 377 sites with Vs,30 > 500 m/s.
Use data from the Trinet broadband high and low gain channels (BH and HL). BH are
STS-1 and STS-2 instruments and HL are mainly FBA-23 instruments. Use BH data
when not clipped and otherwise HL data.
Eliminate DC offset for each record. Convert ground motions into acceleration while
applying a high-pass lter with cut-off of 40 s. Display recovered acceleration, velocity
and displacement time-histories from a ML 5.1 earthquake from the BH and HL data.
Note that they are similar and hence that reliable ground motion can be recovered from
these data.
Display the signal and noise Fourier amplitude spectra for one record and nd that the
signal-to-noise ratio is higher in the BH channel than in the HL channel. State that the
signal-to-noise ratio is still > 1 for periods of 20 s for both types of data.
Compute acceleration and relative displacement response spectra for both channels.
Find that for periods > 0.3 s the response spectra from the two channels are very close.
State that the difference for short periods is due to the low sampling rate (20 sps) for the
BH channel and the higher (80 or 100 sps) sampling rate for HL channel.
Conclude that reliable ground motions up to 20 s can be recovered from these data.
Use a two-stage regression method where rst determine c4 and c5 and then the other
coefcients.
Most data from digital instruments from M 5.5 and R < 300 km. Most data from
analogue instruments from 6.0 M 7.0 and 10 < R < 100 km.
Use data from analogue instruments for short-period range (0.043 s) and data from
Trinet instruments for long-period range (120 s). Connect the two sets of coefcients at
1.5 s after conrming that the predictions match at this period.
Do not give coefcients only predictions.
4.121 Ambraseys et al. (2005a)

See Section 2.207.
Only use spectral accelerations within passband of lter (1.25fl and fh ) where fl is the
low cut-off frequency and fh is the high roll-off frequency.
Note that after 0.8 s the number of records available for regression analysis starts to decrease rapidly and that after 4 s there are few records available. Only conduct regression
analysis up to 2.5 s because for longer periods there are too few records to obtain stable
results. Note that larger amplitude ground motions are better represented in the set for
long-periods (> 1 s).
Find that logarithmic transformation may not be justied for nine periods (0.26, 0.28 and
0.440.65 s) by using pure error analysis but use logarithmic transformation since it is
justied for neighbouring periods.
350
By using pure error analysis, nd that for periods > 0.95 s the null hypothesis of a
magnitude-independent standard deviation cannot be rejected so assume magnitudeindependent . Note that could be because magnitude-dependent standard deviations
are a short-period characteristic of ground motions or because the distribution of data
w.r.t. magnitude changes at long periods due to ltering.
Find that different coefcients are signicant at different periods so try changing the
functional form to exclude insignicant coefcients and then applying regression again.
Find that predicted spectra show considerable variation between neighbouring periods
therefore retained all coefcients for all periods even when not signicant.
Note that smoothing could improve the reliability of long-period ground-motion estimates
because they were based on less data but that smoothing is not undertaken since the
change of weighted to unweighted regression at 0.95 s means a simple function cannot
t both short- and long-period coefcients.
4.122
Ambraseys et al. (2005b)
See Section 2.208.

By using pure error analysis, nd that for periods 0.150.40, 0.600.65, 0.75 and 0.85 s
the null hypothesis of a magnitude-independent standard deviation be rejected so use
weighted regression for these periods.
4.123
See Section 2.210.

4.124
Garca et al. (2005)
See Section 2.212.

No coefcient smoothing performed because coefcients w.r.t. frequency show acceptable behaviour.
4.125
McGarr & Fletcher (2005)
See Section 2.214.

Constrain k to 0 for T 0.5 s because otherwise positive.
351
4.126 Pousse et al. (2005)

log10 (PSA(f )) = a(f )M + b(f )X log10 (X) + Sk

Select this form to compare results with Berge-Thierry et al. (2003).
Use ve Eurocode 8 categories:
A Vs,30 > 800 m/s, use S1
B 360 < Vs,30 < 800 m/s, use S2
C 180 < Vs,30 < 360 m/s, use S3
D Vs,30 < 180 m/s, use S4
E Soil D or C underlain in rst 20 m by a layer of Vs,30 > 800 m/s, use S5
where Vs,30 is average shear-wave velocity in upper 30 m. Since soil proles only available up to 20 m, use method of Atkinson & Boore (2003) to assign sites to categories
using Kik-Net proles to dene probability curves. Generate ve redistributions to test
stability of results. Find coefcients and relative stable (changes less than 10%) except
for site class A (changes up to 50%.
Use data from the K-Net and Kik-Net networks.
Process records using non-causal 4 pole Butterworth lter with cut-offs of 0.25 and 25 Hz
for consistency with earlier studies.
Select records from events with Mw > 4 and with focal depth < 25 km to exclude
records of subduction events and to remain close to tectonic conditions in France.
Exclude records from distances greater than a the distance predicted by a magnitudedependent equation predicting the location of a PGA threshold of 10 cm/s2 (corresponding to trigger of older Japanese sensors) to prevent possible underestimation of attenuation rate.
Visually inspect records to check for glitches and to use only main shock if multiple
events present.
Convert MJMA to Mw to compare results with other studies.
For 10 large earthquakes for which source dimensions are known use rrup .
Note good distribution w.r.t. Mw and rrup except between 6.1 and 7.3 where only two
events.
Find that pseudo-acceleration at 0.01 s equals PGA.
Also compute coefcients using geometric mean and nd identical coefcients and standard deviations lower by 0.02.
Find lower when use ve site classes than when no site information is used.
352
Find peak in at about 1 s. Peak also present when unltered data used. Also present
when data from different magnitude ranges (4.04.5, 4.05.0, 4.05.5 and 4.06.0) are
used.
Note that results for site class E are uncertain due to limited number of records.
Examine residuals w.r.t. distance and magnitude and nd no signicant bias.
Examine quartile plots of residuals and nd that residuals are normally distributed up to
24 s. All pass Kolmogorov-Smirnov test at 5% signicance level for normality except
at 0.01 s.
Conducted sensitivity analysis by changing minimum magnitude, geographical area and
minimum number of events recorded at each station. Find dependence of on period
was similar as were site coefcients. b shows some variations.
Coefcients not reported.
4.127
Takahashi et al. (2005), Zhao et al. (2006) and Fukushima

et al. (2006)
See Section 2.217.

4.128
Wald et al. (2005)
See Section 2.218.

4.129
Atkinson (2006)
See Section 2.219.

Compares predictions to observations grouped into 1-unit magnitude bins at 0.3 and
1.0 s and nds equations are reasonable description of data. Also compares predictions
to observations from large magnitudes events and from close distances and nds that
equations would overestimate short-period motions from large events at close distances.
Compares overall distribution of residuals for 0.3 s with normal distribution. Finds that
residuals generally follow normal distribution but data shows greater number of largeresidual observations that predicted by normal distribution, most of which come from a
single event (22/02/2000 M3.24) recorded at > 100 km. Finds no evidence for truncation of residuals up to three standard deviations.
353
For analysis of Landers events, regresses 0.3 s data for 10 stations with more than 50
records using same functional form without distance terms (since distances are almost
constant) to get site-specic equations. Find on average = 0.190.04. Therefore concludes single station-single source standard deviations much lower (60%) than standard
s.
Notes that decreasing with increasing period could be due to dominance of small
events for which long-period motions are at the moment end of the spectrum, which
should be correlated with M and independent of stress drop.
4.130 Beyer & Bommer (2006)

See Section 2.220.
Use records only up to maximum usable period specied in NGA database.
4.131 Bindi et al. (2006)

See Section 2.221.
Only use records from within passband of lter. For T > 2 s only use digital records.
4.132 Campbell & Bozorgnia (2006a) and Campbell & Bozorgnia

(2006b)
See Section 2.222.
4.133 Hernandez et al. (2006)

See Section 2.225.
4.134 Kanno et al. (2006)

See Section 2.226.
Note the poorer correlation between residuals and Vs,30 for short periods could be due
to higher modal effects or to nonlinear effects (although note that few records where
nonlinear effects are likely).
354
4.135
See Section 2.230.

4.136
Pousse et al. (2006)
See Section 2.232.

Coefcients not reported.
4.137
Sakamoto et al. (2006)
log SA(T ) = a(T )Mw + b(T )X + g + d(T )D + c(T )

where g = log(X + e)
for D 30 km
g = 0.4 log(1.7D + e) 1.4 log(X + e) for D > 30 km

e = 0.006100.5Mw
Soil characteristics known to bedrock for 571 (out of 1013) stations. Classify stations
using NEHRP classication using Vs,30 or converted N -values:
A Vs,30 > 1500 m/s, 0 stations
B 760 < Vs,30 1500 m/s, 0 stations
C1 460 < Vs,30 760 m/s, 174 stations
C2 360 < Vs,30 460 m/s, 193 stations
D1 250 < Vs,30 360 m/s, 300 stations
D2 180 < Vs,30 250 m/s, 230 stations
E Vs,30 180 m/s, 116 stations
Dene nonlinear (based on PGA at bedrock) soil amplication model using nonlinear
analyses of sampled soil conditions for each class of soils. Use this model to convert
observed ground motion to motion at a C1 site.
Focal depths, D, between 3 and 122 km.
Distribution with respective to earthquake type (based on mechanism, location and
depth) is: crustal (3 D . 25 km), 13; interplate (10 . D . 70 km), 23; and intraplate,
16 (30 . D 122 km).
PGA from 2 to 1114 cm/s2 .
355
Try including different constant terms to model effect of earthquake type but nd lower
statistical condences of results. Therefore remove these coefcients. Believe that
modelling of focal-depth dependency may already include effect of earthquake type due
to high correlation between depth and type.
Fit fourth-degree polynomials (in log(T )) through derived coefcients to generate smooth
spectra.
Compare inter- and intra-event residuals to normal distribution using Kolmogorov-Smirnov
test and nd that the intra-event residuals have a normal distribution and that the interevent residuals almost have.
Examine magnitude-dependence of the standard deviations using residuals binned within
different magnitude ranges (Mw < 6.0, 6.0 Mw < 6.5, 6.5 Mw < 7.0 and
Mw 7.0) and do not nd a clear trend for either inter- or intra-event residuals.
Examine distance-dependence of the intra-event standard deviations and nd that for
some periods the standard deviations show some depth-dependence for short and long
distances.
Examine amplitude-dependence of the intra-event standard deviations and nd some
positive dependence ( increases for higher amplitude motions) for T 0.4 s. Note that
this may be due to a lack of small amplitude motions due to nontriggering of instruments.
4.138 Sharma & Bungum (2006)

ln(A) = c2 M b ln(X + exp(c3 M ))

Response parameter is acceleration for an unspecied damping (but assumed to be
5%).
R Rock. Generally granite/quartzite/sandstone.
S Soil. Sites with exposed soil cover with different levels of consolidation.
Data from three strong-motion (SMA-1) arrays: Kangra, Uttar Pradesh and Shillong, in
the Himalayas.
Instruments generally from ground oors of buildings.
Rotate components into NS and EW directions.
Note that distribution of records is uneven. Five events have less than 9 records and
one earthquake has 43.
Note that Mw avoids magnitude saturation problems.
356
Note that lack of near-eld data (all but one record from > 20 km) means that results
are not stable. Therefore introduce nine European records from seven reverse-faulting
earthquakes for M 6.0 and de 20 km.
Use method of Campbell (1981) to avoid problems due to correlation between magnitude and distance. Divide data into a number of subsets based on distance. For each
interval, each earthquake is given equal weight by assigning a relative weight of 1/nj,l
to the record where nj,l is the total number of records from the j th earthquake within
ith distance bin. Normalise weights so that they sum to total number of records. Use
distance bins of 5 km wide up to 10 km and then bins of equal width w.r.t. logarithmic
distance.
Use rhypo rather than rrup because: a) large depth of some events and b) poorly known
fault geometries. Note that rhypo has a reasonable seismological basis and can be
reliably and easily determined for most signicant (including hypothetical design) earthquakes.
Regress all data using: ln(A) = c b ln(X) and nd b = 1.22 0.69. Next regress
using: ln(A) = aM b ln(X) + c and nd b = 0.515 0.081. Conclude that this is
due to correlation between magnitude and distance and hence conduct the rst step of
a two-step regression with dummy variables for each earthquake. Find a decay rate of
1.20 0.036. Use this xed decay rate for rest of analysis.
Try to regress on rock and soil data simultaneously by including a linear site term c4 SSR
but nd that there are problems during the regression process. Hence regress separately on rock and soil data.
4.139
Zare & Sabzali (2006)
See Section 2.234.

Response parameter is not given but assumed to be acceleration for 5% damping.
4.140
Akkar & Bommer (2007b)
See Section 2.235.

Response parameter is displacement for 2, 5, 10, 20 and 30% damping. Choose displacement because of aimed use of equations for displacement-based design.
Only use records within their usable range, dened as a fraction of the cut-off frequency
used and depending on instrument type (digital or analogue), magnitude and site class.
Note that drop-off in available records from analogue instruments is much more rapid
(starting around 1 s) than for records from digital instruments (starting around 3 s). Due
to lack of data for longer periods limit regression to periods 4 s.
Due to jagged appearance of predicted response spectra, particularly at long periods
where different data was used for each period, apply negative exponential smoothing.
Try smoothing using low-order polynomials, to achieve very smooth spectra, but complex
functional form means results are sensitive to trade-offs between smoothed coefcients.
357
Find that for periods > 3 s spectra predicted from the raw and smoothed coefcients
show differences, especially for low damping ratios.
Find that coefcients b7 -b10 weakly dependent on damping ratio so present these coefcients for 2 and 5% damping (combined), 10% and 20 and 30% damping (combined).
4.141 Bindi et al. (2007)

See Section 2.238.
Display graphs of inter-, intra-event and total standard deviations against period when
using Mw or ML .
4.142 Bommer et al. (2007)

See Section 2.239.
Derive equations only up to 0.5 s because thought that ground motions reliable up to this
limit and since equations developed only for comparative purposes. Note that usable
period range of data could be extended to 2 s but since study is for exploring inuence
of lower magnitude limit short-period motions are the most important.
4.143 Boore & Atkinson (2007) & Boore & Atkinson (2008)
See Section 2.240.
Do not use pseudo-accelerations at periods > TM AX , the inverse of the lowest useable
frequency in the NGA Flatle.
Constant number of records to 1 s, slight decrease at 2 s and a rapid fall off in number of
records for periods > 2 s.
For long periods very few records for small earthquakes (M < 6.5) at any distance so
magnitude scaling at long periods poorly determined for small events.
Choi & Stewart (2005) do not provide coefcients for site amplication for periods >
5 s so linearly extrapolate blin in terms of log period by assuming relative linear site
amplication to decrease.
To assign c3 for entire period range t quadratic to c3 s from four-event analysis with
constraints for short and long periods.
No data from normal-faulting events for 10 s so assume ratio of motions for normal and
unspecied faults is same as for 7.5 s.
Possible underprediction of long-period motions at large distances in deep basins.
358
Chi-Chi data major controlling factor for predictions for periods > 5 s even for small
events.
4.144
Campbell & Bozorgnia (2007), Campbell & Bozorgnia (2008b)

& Campbell & Bozorgnia (2008a)
See Section 2.241.

Response parameter is pseudo-acceleration (PSA) for 5% damping.
If PSA < PGA for T 0.25 s then set PSA equal to PGA, to be consistent with denition
of PSA (occurs for large distances and small magnitudes).
Due to cut-off frequencies used number of records available for periods > 45 s falls off
signicantly. Majority of earthquakes at long periods are for 6.5 M 7.9 and 70%
are from 1999 Chi-Chi earthquake.
To extend model to longer periods and small magnitudes constrain the magnitudescaling term using empirical observations and simple seismological theory.
4.145
Danciu & Tselentis (2007a) & Danciu & Tselentis (2007b)
See Section 2.242.

4.146
Fukushima et al. (2007b) & Fukushima et al. (2007a)
Ground-motion model is [same as Fukushima et al. (2003)]:
log10 (Sa(f )) = a(f )M log10 (R + d(f ) 10e(f )M ) + b(f )R + cj (f )j

j = 1 for j th site class and 0 otherwise.
Use ve site categories:
SC-1 Site natural period TG < 0.2 s, Vs,30 > 600 m/s, NEHRP classes A+B. 23 sites.
SC-2 Site natural period 0.2 TG < 0.6 s, 200 Vs,30 < 600 m/s, NEHRP classes
C+D. 100 sites.
SC-3 Site natural period TG 0.6 s, Vs,30 200 m/s, NEHRP class E. 95 sites.
SC-4 Unknown site natural period, Vs,30 > 800 m/s, NEHRP classes A+B. 44 sites.
SC-5 Unknown site natural period, 300 Vs,30 < 800 m/s, NEHRP class C. 79 sites.
Manually classify stations using the predominant period computed using average horizontalto-vertical (H/V) response spectral ratios using similar approach to Zhao et al. (2006)
and also mean residuals w.r.t. equations of Fukushima et al. (2003). Reclassify stations
of Fukushima et al. (2003), who used rock/soil classes. Some (36%) stations cannot be
classied (due to, e.g., broadband amplication) using this approach so retain rock/soil
classes for these records. Use this approach since limited geotechnical data is available
359
for most sites in their dataset. Only roughly 30% of stations have multiple records so the
average H/V ratios are not statistically robust so do not use automatic classication approach. Each co-author independently classied stations. About 90% of classications
agreed. After discussion the stations were reclassied. Originally used same categories
as Zhao et al. (2006) but nd their class SC-III too narrow so combine it with their SCII to form SC-2. Find similar average ratios for the different categories as Zhao et al.
(2006).
Use data and regression method of Fukushima et al. (2003). Eliminate data from two
stations of Fukushima et al. (2003) because of suspected soil-structure interaction.
Coefcients not reported since focus of article is the site classication procedure and
its impact on predicted response spectra and not to propose a new model for seismic
hazard assessment.
Records ltered with cut-offs at 0.25 and 25 Hz therefore present results up to 3 s to avoid
lter effects.
Find roughly 2% reduction in standard deviation using classication scheme compared
to rock/soil scheme.
4.147 Hong & Goda (2007) & Goda & Hong (2008)
See Section 2.244.
Select the period range of usable PSA values based on cut-off frequencies of the highpass lters used to correct records.
Develop an orientation-dependent ground-motion measure based on maximum resultant response and ratio between response of an (arbitrarily) oriented SDOF system and
maximum resultant response.
Derive equations for the probability of exceedance for SDOF systems designed for different ways of combining the two horizontal components subjected to ground motions
from an unknown direction.
Investigate record-to-record variability of response and implied exceedance probability
using a set of 108 records used by Boore et al. (1997) for 0.2 and 1.0 s. Conclude
that when using common methods for combining two horizontal components (such as
geometric mean) that meaning of the return period of uniform hazard spectra is not clear
because the major and minor axes of shaking are unknown before an event.
Investigate SA resolved for different directions normalized by SA along the major axis
for all selected records. Conclude that knowing SA along the major axis and the normalized SA for different direction completely denes the response in any direction. Derive
empirical equation for the normalized SA w.r.t. angle and its probability distribution.
Only report coefcients for 0.2, 0.3, 1, 2 and 3 s in article. Provide coefcients for other
periods as electronic supplement.
360
4.148
Massa et al. (2007)
See Section 2.246.

4.149
Tejeda-Jcome & Chvez-Garca (2007)
See Section 2.250.

Signal-to-noise ratios mean analysis limited to 1 s for horizontal and 0.8 s for vertical.
4.150
Abrahamson & Silva (2008) & Abrahamson & Silva (2009)
See Section 2.251.

Records only used for spectral frequencies 1.25 times the high-pass corner frequency
used in the record processing. Therefore, number of records and earthquakes available
for regression decreases with increasing period.
Fix a2 , a12 , a13 , a16 and a18 at their values for 24 s for T > 5 s because they could not
be constrained by data.
Smooth coefcients in several steps.
4.151
Aghabarati & Tehranizadeh (2008)
See Section 2.253.

4.152
Cauzzi & Faccioli (2008), Cauzzi (2008) & Cauzzi et al. (2008)
See Section 2.254.

Response parameter is displacement for 5, 10, 20 and 30% damping.
Coefcients reported as Electronic Supplementary Material.
Try replacing site terms: aB , aC and aD by b4 10b5 Mw , b6 10b7 Mw and b8 10b9 Mw but do
not report coefcients since did not lead to reduction in standard deviation.
Compare predictions and observations for Parkeld 2004 earthquake. Find good match.
Study residuals for site classes B, C and D w.r.t. predicted ground motion to check for
nonlinear site response. Find some evidence for moderate nonlinear effects in limited
period ranges.
361
4.153 Chen & Yu (2008b)

log Sa = C1 + C2 M + C3 M 2 + C4 log[R + C5 exp(C6 M )]

Use records from sites with Vs,30 500 m/s.
Use the NGA Flatle.
Data divided into magnitude intervals of: 5.05.4, 5.55.9, 6.06.4, 6.56.9 and 7.07.5
and distance intervals of: 02.9 km, 3.09.9 km, 1029.9 km, 30 59.9 km, 60-99.9 km,
100200 km and > 200 km. Use weighted regression with weights given by inverse
of number of records in each magnitude-distance bin since most data from moderate
earthquakes at intermediate distances.
Compute C5 and C6 using data from six earthquakes: 1979 Imperial Valley (M 6.53),
1980 Livermore (M 5.42), 1989 Loma Prieta (M 6.93), 1992 Landers (M 7.28), 1999
Hector Mine (M 7.13) and 2004 Parkeld (M 5.9).
4.154 Chen & Yu (2008a)

Response parameter is acceleration for 0.5, 2, 7, 10 and 20% damping.
Continuation of Chen & Yu (2008b) (Section 4.153) for other damping levels.
4.155 Chiou & Youngs (2008)

See Section 2.255.
Coefcients developed through iterative process of performing regressions for entire
spectral period range with some parts of model xed, developing smoothing models for
these coefcients with period, and then repeating analysis to examine variation of remaining coefcients. Note noticeable steps in c1 at 0.8, 1.1, 1.6, 4.0 and 8.0 s, where
there is large reduction in usable data. Suggest that this could indicate bias due to systematic removal of weaker motions from data set. To correct this bias and to smooth c1
impose smooth variation in slope of c1 w.r.t. period. Also examine shape of displacement
spectra for M 6.5 to verify that constant displacement reached at periods expected
by design spectra.
4.156 Cotton et al. (2008)

See Section 2.256.
362
4.157
Dhakal et al. (2008)
log10 Y (T ) = c + aMw + hD log10 R b1 R1 b2 R2

Use R1 , distance from hypocentre to volcanic front, and R2 , distance from volcanic front
to site, to model anelastic attenuation.
Use data from K-Net. Select earthquakes that: 1) have Mw > 5 and 2) have more
than 50 available records. To remove bias due to large number of records from forearc site compared to back-arc, select only those earthquakes with 40% of the available
records within 300 km are from back-arc region. Use both interplate and intraslab events
occurring in fore-arc region so that effect of low Q zone is clearly seen. Only use records
up to 300 km so that peaks are due to S-wave motions. Exclude records from Mw 8
earthquakes because these events radiate strong surface waves so assumption of Swave peaks may not be valid.
Focal depths, D, of intraslab earthquakes between 59 and 126 km and for interface10
earthquakes between 21 and 51 km.
Also derive model using: log10 Y (T ) = c + aMw + hD log10 R bR. Find lower s
for functional form using R1 and R2 for periods < 1 s. Examine residuals w.r.t. rhypo for
0.1 and 1.0 s with grey scale indicating ratio R1 /(R1 + R2 ) for this functional form. Note
that fore-arc sites have positive residuals and back-arc sites negative residuals. Also
plot residuals for selected functional form and nd that residuals do not show difference
between fore-arc and back-arc sites.
Regress separately for intraslab and interface earthquakes because source characteristics signicantly different.
Find that the coefcients for anelastic attenuation for fore-arc and back-arc different for
periods < 2 s.
Convert computed anelastic coefcients to Q models and nd that can relate observations to different Q models for fore-arc and back-arc regions.
4.158
Idriss (2008)
See Section 2.258.

Uses all records (including those from Chi-Chi) to constrain coefcients for 1.5 T 5 s
because inuence of Chi-Chi records decreases with increasing period.
Uses smoothed plots to obtain coefcients for T > 5 s because of lack of records.
10
Authors call them interplate.
363
4.159 Lin & Lee (2008)

See Section 2.259.
4.160 Massa et al. (2008)

See Section 2.260.
Response parameters are acceleration and pseudo-velocity for 5% damping.
4.161 Morasca et al. (2008)

See Section 2.262.
4.162 Yuzawa & Kudo (2008)

log S(T ) = a(T )M [log Xeq + b(T )Xeq ] + c(T )

Use data from KiK-Net at hard rock sites with shear-wave velocity Vs 2.0 km/s at
surface and/or in borehole. Select records from 161 sites (out of 670 sites of KiK-Net)
where spectral ratio between surface and borehole records 2 at periods > 1 s. Note
that preferable to use higher velocity (3.0 km/s) but as velocity increases number of
available sites rapidly decreases: 43 sites with Vs = 2.02.2 km/s, 33 with Vs = 2.22.4,
27 with Vs = 2.42.6, 31 with Vs = 2.62.8, 16 with Vs = 2.83.0, 8 with Vs = 3.03.2
and 3 with Vs > 3.2 km/s.
Select earthquakes based on their magnitudes, horizontal locations and depths and
types (crustal, interface and intraslab). Note that geographical distribution is not homogeneous but it covers whole of Japan.
Also derive model using Mw . Find predictions similar so use prefer MJM A for convenience of application in Japan.
Only graphs of coefcients presented.

See Section 2.265.
364
4.164
Akyol & Karagz (2009)
See Section 2.266.

Observe nonlinear site effects in residuals for periods 0.27 s, which model using site
coefcient correction terms.
4.165
Bindi et al. (2009a)
See Section 2.267.

4.166
Bindi et al. (2009b)
See Section 2.268.

4.167
Bragato (2009)
See Section 2.269.

Coefcients not reported, only s.
4.168
Ghasemi et al. (2009)
log10 Sa = a1 + a2 M + a3 log10 (R + a4 10a5 M ) + a6 S1 + a7 S2

after trying various other functional forms. Fix a5 to 0.42 from previous study due to lack
of near-eld data and unstable regression results.
Rock Vs,30 760 m/s. S1 = 1, S2 = 0.
Soil Vs,30 < 760 m/s. S2 = 1, S1 = 0.
Classify station using Vs,30 and surface geology data, if available. Otherwise use empirical H/V classication scheme.
365
Investigate differences in ground motions between Alborz-Central Iran and Zagros regions using analysis of variance (ANOVA) (Douglas, 2004b) to check whether data can
be combined into one dataset. Find that for only one magnitude-distance interval out of
30 is there a signicant difference in ground motions between the two regions. Hence,
combine two datasets.
Check that data from West Eurasia and Kobe from Fukushima et al. (2003) can be
combined with data from Iran using ANOVA. Find that for only one magnitude-distance
interval is there a signicant difference in ground motions and, therefore, the datasets
are combined.
Only retain data from R < 100 km to avoid bias due to non-triggered instruments and
because data from greater distances is of low engineering signicance.
Process uncorrected records by tting quadratic to velocity data and then ltering acceleration using a fourth-order acausal Butterworth lter after zero padding. Choose lter
cut-offs by using the signal-to-noise ratio using the pre-event noise for digital records
and the shape of the Fourier amplitude spectra for analogue records. Only use records
for periods within the passband of the lters applied.
Exclude data from earthquakes with Mw < 5 because of risk of misallocating records to
the wrong small events and because small events can be poorly located. Also records
from earthquakes with Mw < 5 are unlikely to be of engineering signicance.
Cannot nd negative anelastic coefcients for periods > 1 s and therefore exclude this
term for all periods.
Try including a M 2 term but nd that it is not statistically signicant so remove it.
Examine residuals (display graphs for 0.1 and 1 s) w.r.t. M and R. Find no signicant
(at 5% level) trends.
Examine histograms of residuals for 0.1 and 1 s and nd that expected normal distribution
ts the histograms closely.
4.169 Hong et al. (2009b)

See Section 2.270.
4.170 Hong et al. (2009a)

See Section 2.271.
Only report coefcients for three periods (0.3, 1 and 3 s).
366
4.171
Kuehn et al. (2009)
See Section 2.272.

Only use data up to highest usable period.
Note that could choose different functional form for each period separate but believe
effect would be small so use the same for all periods.
4.172
Moss (2009)
See Section 2.274.

Finds maximum decrease in is 9% at 3 s.
4.173
Rodriguez-Marek & Montalva (2010)
Ground-motion model is a simplied version of Boore & Atkinson (2008), because it is

the simplest NGA functional form11 :
ln(
y ) = Fm + Fd + Fsite (Ssurf ace ) + [F100 (S100 ) + F200 (S200 )](1 Ssurf ace )
Fd = [c1 + c2 (M Mref )] ln(R/Rref ) + c3 (R Rref )
R =
R 2 + h2
Fm = e1 + e5 (M Mh ) + e6 (M Mh )2 for M < Mh
Fm = e1 + e7 (M Mh ) for M > Mh
Fsite = blin ln(Vs,30 /Vref )
F100 = a100 + b100 ln(Vs,30 /Vref ) + c100 ln(Vs,hole /3000)
F200 = a200 + b200 ln(Vs,30 /Vref ) + c200 ln(Vs,hole /3000)
Sites characterized by Vs,30 , Vs,hole (shear-wave velocity at depth of instrument), Ssurf ace
(1 for surface record, 0 otherwise), S100 (1 for borehole record from < 150 m depth, 0
otherwise) and S200 (1 for borehole record from > 150 m depth, 0 otherwise).
Use the same data as Cotton et al. (2008) (see Section 2.256).
Develop GMPEs for use in the estimation of single-station .
Note that functional form assumes that magnitude and distance dependency are the
same for both surface and borehole records. Also assume that site amplication is
linear, which note appears to be true for most records but not all but insufcient data to
constrain nonlinearity using purely empirical method so ignore it.
11
Number of typographic errors in report so this may not be correct functional form.
367
For regression: use only surface data to constrain blin , use both surface and borehole
records to compute inter-event s and assume intra-event s independent of magnitude.
Note that nal assumption is somewhat limiting but use residual analysis to examine
dependency of intra-event terms on depth, Vs,30 and magnitude.
Compute single-station s based on residuals from the 44 stations that recorded 15
earthquakes. Averaged these 44 s to obtain a single estimate of single-station . Note
that more work on these s is being undertaken. Find single-station s are on average
25% lower than total . Find that total s obtained for borehole stations lower than those
at surface but the single-station s are not considerable different on the surface and in
boreholes.
4.174 Rupakhety & Sigbjrnsson (2009)

See Section 2.276.
Also provide coefcients for constant-ductility inelastic spectral ordinates and structural
behaviour factors for application within Eurocode 8.
Coefcients only reported for 29 periods graphs for rest.
Note that coefcients are not smooth functions w.r.t. period, which is undesirable for
practical purposes. Smooth coefcients using Savitzky-Golay procedure with a span of
19 and a quadratic polynomial and then recomputed . Verify that smoothing does not
disturb inherent correlation between model parameters by comparing correlation matrix
of coefcients before and after smoothing. Find that smoothing has little effect on matrix
nor on .
4.175 Sharma et al. (2009)

log A = b1 + b2 Mw + b3 log
2 + b2 + b S + b H
RJB
5
6
4

S = 1 Rock. 69 records.
S = 0 Soil. 132 records.
Focal depths between 5 and 33 km for Iranian events and 19 and 50 km for Indian earthquakes.
Use two fault mechanisms:
H = 0 Reverse. 8 earthquakes and 123 records.

H = 1 Strike-slip. 8 earthquakes and 78 records.
368
Seek to develop model for Indian Himalayas. Due to lack of near-source data from
India include data from the Zagros region of Iran, which has comparable seismotectonics (continental compression). Note that some differences, in particular the higher dip
angles of reverse events in the Zagros compared to those in the Himalayas.
Use data from three strong-motion arrays in Indian Himalayas: Kangra array in Himachal
Pradesh, Uttar Pradesh and Shillong array in Meghalaya and Assam, and from Iran
Strong-Motion Network. Note that records from at least three signicant Himalayan
earthquakes have not yet been digitized.
Use some non-Zagros data from Iran because of similar focal mechanisms and since no
signicant difference in ground motions between these events are those in the Zagros
was observed.
Note that data seems to be adequate between Mw 5 and 7 and up to 100 km.
To exclude data from earthquakes that show anomalous behaviour, the PGAs for each
earthquake individually were plotted against distance. Find that decay rates for 6/2/1988
and 14/3/1998 earthquakes were different than rest so data from these events were
excluded.
Also exclude data from two earthquakes (6/8/1988, 10/1/1990 and 6/5/1995) due to their
great hypocentral depths (> 90 km).
Also exclude data from eight earthquakes (9/1/1990, 24/3/1995, 14/12/2005, 29/11/2006,
10/12/2006, 9/6/2007, 18/10/2007 and 25/11/2007) because no focal mechanisms published.
Prefer rjb partly because of lack of reliable depths for most Himalayan earthquakes.
Estimate rjb for some earthquakes by using reported focal mechanism and relationships
of Wells & Coppersmith (1994).
Use explicit weighting method of Campbell (1981) with equal weights given to records
falling into three ranges: 10 km, 10100 km and more than 100 km.
Note that high standard deviations partly due to low quality of site information, large
uncertainties in source-to-site distances and simple functional form.
4.176
Akkar & Bommer (2010)
See Section 2.277.

Derive equations up to 4 s but only report coefcients to 3 s because of a signicant drop
in available data at this period and because of the related issue of a sudden change in
(particularly intra-event ) at 3.2 s.
369
4.177 Akkar & agnan

(2010)
See Section 2.278.
Data become scarce for T > 2 s due to cut-off frequencies used and, therefore, do not
derive equations for longer periods. Limit of 0.03 s is based on Nyquist (sampling rates
are generally 100 Hz) and high-cut ltering used (generally > 30 Hz). Note that this
conservative choice is based on the study of Douglas & Boore (2011).
4.178 Ghodrati Amiri et al. (2010)

log(SA) = C1 + C2 Ms + C3 log(R)
Use two site classes that are consistent with Iranian design code and derive equations
for each separately:
Soil Vs < 375 m/s.
Rock Vs > 375 m/s.
Focal depths between 5 and 59 km but most 10 km.
Based on Ghodrati Amiri et al. (2007a) (see Section 2.236) but using larger and reappraised dataset.
Derive models for Zagros and Alborz-Central Iran separately.
Note the poor quality of some Iranian strong-motion data. Selected data based on accuracy of independent parameters.
State that faulting mechanism is known for only a small proportion of data. Therefore, it
is not considered.
Use Ms because it is the most common scale for Iranian earthquakes.
Most data from Ms < 6.5 and 5 < rhypo < 200 km. Note lack of near-source data from
Ms > 6.
Because of small and moderate size of most earthquakes used and since causative
faults are not known for many earthquakes use rhypo , which compute using S-P method
because of uncertainty in reported hypocentral locations.
Data from SMA-1 (about 210 records on soil and 130 on rock) and SSA-2 (about 220
records on soil and 170 on rock).
Bandpass lter records using cut-off frequencies chosen based on instrument type and
data quality. Cut-offs chosen by trial and error based on magnitude and distance of
record and obtained velocity. Generally cut-offs are: 0.150.20 Hz and 3033 Hz for
SSA-2 on rock, 0.150.25 Hz and 2023 Hz for SMA-1 on rock, 0.070.20 Hz and 30
33 Hz for SSA-2 on soil and 0.150.20 Hz and 2023 Hz for SMA-1 on soil.
370
Choose functional form after many tests (not shown) and because it is simple but physically justied.
Note that predictions show peaks and valleys since no smoothing applied.
Report that residual analysis (not shown) shows predictions are unbiased w.r.t. magnitude, distance and site conditions.
4.179
Arroyo et al. (2010)
See Section 2.279.

4.180
Bindi et al. (2010)
See Section 2.280.

4.181
Douglas & Halldrsson (2010)
See Section 2.282.

4.182
Faccioli et al. (2010)
See Section 2.283.

Response parameter is displacement for 5% damping.
Coefcients only given for a subset of periods for which analysis conducted.
Site terms particularly important for T 0.25 s, where reduction in is between 5% and
15%.
4.183
Hong & Goda (2010)
See Section 2.285.

Present correlation models between ground motions at different periods.
371
4.184 Jayaram & Baker (2010)

See Section 2.286.
Report coefcients only for 1 s.
4.185 Montalva (2010)

See Section 2.287.
Residual analysis shown for 0.03, 0.2, 0.6, 1.0 and 1.4 s.
4.186 Ornthammarath et al. (2010b), Ornthammarath (2010) &

Ornthammarath et al. (2010a)
See Section 2.288.
4.187 Sadeghi et al. (2010)

log A = a(f ) + b(f )M c1 (f ) log R k(f )R for R R1

log A = a(f ) + b(f )M c1 (f ) log R1 c2 (f ) log(R/R1 ) k(f )R for R1 < R R2
log A = a(f ) + b(f )M c1 (f ) log R1 c2 (f ) log(R2 /R1 ) c3 (f ) log(R/R2 ) k(f )R
for R > R2
Functional form chosen to enable modelling of effect of reections off Moho and surface
wave attenuation.
Soil Vs,30 < 750 m/s or, for 30 stations classied using H/V ratios, f0 < 7.5 Hz where
f0 is peak frequency. 556 records.
Rock Vs,30 > 750 m/s or , for 30 stations classied using H/V ratios, f0 > 7.5 Hz. 213
records.
Develop models for all data and only soil records.
Data from 573 different stations.
Also develop separate models for regions of Alborz (20 earthquakes and 423 records),
Zagros (27 earthquakes and 198 records), East (32 earthquakes and 262 records) and
Central South (20 earthquakes and 175 records). Note that regionalization is limited by
lack of data for other regions.
372
Use data recorded by National Strong Motion Network of Iran from 1987 to 2007.
Select data by criterion of earthquake having being recorded by 3 stations within
350 km.
Most data from Mw < 6.5 and r < 150 km.
Insufcient data to constrain model for R > R2 therefore set geometric spreading coefcient to 0.5.
Use Monte Carlo technique to nd coefcients.
Fit a and b to functional forms: a1 +a2 exp(a3 T ) and b1 +b2 T +b3 T 2 +b4 T 3 respectively.
Also present model assuming a = a1 + a2 T + a3 T + a4 T 3 .
Plot residuals against repi .
Believe model can be applied for 5 < M < 7.5 and repi < 200 km.
4.188
Saffari et al. (2010)
log A = a(T )Mw log[X + d(T )100.5Mw ] b(T )X + cRock LR + cSoil LS

Rock LR = 1, LS = 0.
Soil LS = 1, LR = 0.
Focal depths between 7 and 72 km with most between 10 and 30 km.
Use data from Iranian Strong-Motion Network run by Building and Housing Research
Centre.
Select data based on these criteria: Mw 5, record on ground surface (free-eld) and
two orthogonal horizontal components available. Apply a Mw -dependent distance lter.
After rst regression data again truncated based on the median plus one model and a
trigger level of 10 gal.
Examine data binned by Mw w.r.t. distance and remove earthquakes with irregular
distributions (due to tectonic or other reasons).
Baseline correct and bandpass lter (cut-offs of 0.2 and 20 Hz) data based on characteristics of instruments (SSA-2 and SMA-1).
Use rock data to dene all coefcients and then compute cSoil and Soil using soil data
and the coefcients dened from rock data (details not given).
Smooth coefcients using fth-degree polynomial based on logarithm of period.
Derive coefcients for central Iran and Zagros separately.
373
374
Chapter 5
General characteristics of GMPEs for

spectral ordinates
Table 2 gives the general characteristics of published attenuation relations for spectral ordinates. The columns are the same as in Table 1 with three extra columns:
T s Number of periods for which attenuation equations are derived

Tmin Minimum period for which attenuation equation is derived
Tmax Maximum period for which attenuation equation is derived
375
376
W. USA
182
182
182
264
W.
USA,
Japan,
Papua
New
Guinea,
Mexico &
Greece
W. USA
W. USA
70
182
182
34
182
W. USA
W. USA
V
-
H
41
U
Area
W. USA
Japan
46
17+*
46
11
22
46
E
23
U
3.8
4.5*
3.8
5.3
5.3
3.8
Mmin
5.3
5.4*
68 *
11*
66 *
U5
Mostly ML
Mostly ML
15
14
62 *
ML
Mostly ML
rmin
6.3
60*
M scale
mb
U
7.7
7.7
7.7
7.8
7.6
7.7
Mmax
7.7
7.9*
They state it is two dimensional response spectrum which assume to be resolved component.
3
4
Total earthquake components (does not need to be multiplied by two)
5
Idriss (1978) nds magnitudes to be mixture of ML , mb and Ms .
6
7
8
9
McGuire (1978)
Trifunac (1978)
&
Trifunac
&
Anderson
(1978a)
Trifunac & Anderson (1978b)
Reference
Johnson (1973)
Kobayashi
&
Nagahashi
(1977)
McGuire (1977)
Trifunac (1977)
&
Trifunac
&
Anderson
(1977)
Faccioli (1978)
rhypo
4009 * repi
210* rhypo
4007 * repi
342
125 rhypo
4003 * repi
rmax r scale
149.8 repi
210* rhypo
Illustration 2: Characteristics of published spectral relations
2
3
1
3
S
1
I
91
1
91
15
16
91
Ts
14
U
0.04
1
0.04
0.1
0.1
0.04
12
1
12
8
12
B
B
B
B
Tmin Tmax C
0.055 2.469 M
0.1
5
R1
U
O
U
O
R
1
O
A
A
A
A
M
A
A
Reference
Cornell et al.
(1979)
Faccioli & Agalbato (1979)
Trifunac & Lee
(1979)
Ohsaki et al.
(1980b)
Ohsaki et al.
(1980a)
Trifunac (1980)
Devillers & Mohammadioun
(1981)
Joyner & Boore
(1982a)
Joyner & Boore
(1982b)
Kobayashi
&
Midorikawa
(1982)
Joyner & Fumal
(1984), Joyner
& Fumal (1985)
& Joyner &
Boore (1988)
Kawashima
et al. (1984)
Kawashima
et al. (1985)
Trifunac & Lee
(1985)
438
W.
N.
America
45
64
Japan
64
W.
N.
America
W.
N.
America
Japan
197
U
186
W. USA
W. USA
Japan
75
Japan
95
W.
N.
America
W.
N.
America
Japan
438
119
38
Friuli, Italy
V
-
H
70
Area
W. USA
104
90*
90
12
12
U
U
29+
14
E
U
5.0*
5.0
5.0
5.1
5.3*
5.3*
U
3.3*
3.9*
3.7
Mmin
U
1
C
1
rhypo
repi
rhypo
U
5*
MJMA
MJMA
Mw (ML )
50
0.6*
Mw
U
0.6*
10
Mw
U
U
500*
280
110*
110*
U
250*
500
3
3,
C
rhypo
repi
repi
rhypo
rjb
rjb
rjb
500*
rhypo
3*
5
190 rhypo
(repi ) (repi )
U
U
repi
S
1
ML
rmax r scale
rhypo
rmin
U
M scale
ML
7.5*
7.7
7.5
7.7
7.7
U
7.7*
7.4
7.2*
6.3
Mmax
U
91
10
10
12
12
12
91
46
86
91
Ts
7
Tmin
0.04
0.1
0.1
0.1
0.1
0.1
0.1
0.04
0.04
0.02
0.02
0.04
0.17
Tmax
15
7.5
10
15
L,
R
U
U
U
C
C
U
1
R
U
A
A
M
A
377
378
20
197
389
228
438
Mostly California
154
Tokyo
Worldwide
subduction
zones
Japan
120
11
+
+
64
N. Honshu
Europe
438
24
120
494
494
W. USA +
others
Japan
S. California
Mostly California
V
-
H
228
Area
Japan

12
Consider equations valid for Mw 8
10
Kamiyama
(1989)
Trifunac & Lee
(1989)
Petrovski
&
Marcellini
(1988)
Yokota et al.
(1988)
Youngs et al.
(1988)
K.
Sadigh
(1987)11
Annaka
&
Nozawa (1988)
Crouse et al.
(1988)
Reference
Kamiyama
&
Yanagisawa
(1986)
C.B.
Crouse
(1987)10
Lee (1987) &
Lee (1993)
104
16*
(60)
75 (U)
46
45
106
E
69
4.1
5.6* (5)
4.0
5.1
Mmin
4.5
U
U
Ms
ML
for
M . 6.5,
others for
M > 6.5
Mw
(Ms ,
MJMA
Mw
mb )
MJMA
42
Mw , Ms &
MJMA for
< 7.5
U
59
(60)
U
(15*,
20*)
3
rmin
3
M scale
MJMA
7.9
8.1* (8.1,
8.2)12
6.1
8.2
Mmax
7.9
206
(100)
U
(450*,
450*)
350
200
407
I
C
repi
7.5
repi
rrup , rhypo
for Mw .
rhypo
7.5
rhypo
rE , rhypo
for M <
rrup
1
repi
S
I
rrup
rmax r scale
repi
323
12
15
26
10
91
10
Ts
45
Tmin
Tmax
4*
15
10
0.04
0.05*
14
10*
0.1
10
(0.05) (5)
0.07 4
0.02
0.1
0.04*
0.1
0.04
0.05
0.1
C
U
1W
R
1
A (B,F)
A (S, R)
M
A
al.
235
Worldwide
subduction
zones
Intraplate
(particularly
Norway)
Unknown
112
80
Taiwan
New
Zealand
572
80
<88
Worldwide
395+31
97
Japan
30
63
30*
136+11
<51
56
E
8+3
4.0
4.6
2.4*(4.1)
5.1
4.9*
7.1
2.9
Mmin
3.60
(5.16)
ML
for
M < 6,
Ms
for
M 6
ML
Ms
(ML ,MCL )
>8
Mw (Ms ,
MJMA )
14
150*
rrup
repi
5.0
rrup , rhypo
for M < 6
178.3 rhypo
100
1300 rhypo
S
1
rseis
rmax r scale
1215 rhypo
(23)
>469 rE , rhypo
for M <
7.5
20*
1200* rhypo
(9.7) (1300)
3*
Mw
MJMA
ML
for
M < 6,
Ms
for
M 6
Ms (ML ,
mb , MCL
6
rmin
8
(8)
M scale
Mw
7.1
7.4
5.2*(6.9)
8.2
7.4
7.9
7.8
Mmax
6.00
(6.84)
Total earthquake components (does not need to be multiplied by two). 79+10 records for 0.1 s equation.
Consider more than 4 natural periods but results not reported.
15
Reported in Idriss (1993).
13
Loh
et
al.
(1991)
Matuschka &
Davis (1991)
I.M.
Idriss
(1991)15
Dahle et
(1991)
87
Worldwide
intraplate
regions
V
-
H
92+1013
Dahle et al.
(1990b)
&
Dahle et al.
(1990a)
Tamura et al.
(1990)
Tsai
et
al.
(1990)
Crouse (1991)
Campbell
(1990)
Area
E.
N.
America +
10 others
Unknown
Reference
Atkinson (1990)
16
11
0.04
0.04
0.03
0.1
414
0.07
0.1
23
10
20
0.025 4
0.04
Tmax
1
Tmin
0.1
10
14
13
89
15
Ts
4
C
B
1,
O
M
R
2
T (S,O)
M
A
379
380
17
16
83
112
22201
137

Distance to centre of array
Boore et al.
(1993) & Boore
et al. (1997)
Caillot & Bard
(1993)
Italy
W.
USA
with
4
foreign
W.
N.
America
al.
Abrahamson &
Silva (1993)
Tento et
(1992)
Campano
Lucano
W.
USA
with
4
foreign
Italy
Benito et al.
(1992)
Silva & Abrahamson (1992)
U136
234
236
48916
Mainly Italy
and former
Yugoslavia
SMART-1
array,
Taiwan
Stamatovska
&
Petrovski
(1991)
Niazi & Bozorgnia (1992)
84
V
-
H
144
Area
Italy
Reference
Mohammadioun
(1991)
5.30
3.2
40
6.0
6.1
4.7
3.6
3*
Mmin
3.0
14
118
40
U12
12
78
E
46
0.6*
Mw
Mw
3.2
ML
63
109
rhypo
rjb
rjb
for
ML 5.7,
repi otherwise
100* rrup
170
100*
3*
Mw
rseis
rhypo
3.4*
142*
3.117 119.717rhypo
S
1
ML
(MD ) for
ML < 6.6,
else Ms
ML
rmax r scale
rhypo ,
1
eq.
with
rrup
500* rhypo
186
10*
rmin
6
ML
M scale
U
Ms
if 10
ML
&
Ms 6.0
else ML
6.8
7.70
7.4
6.6
7.4
6.5
7.8
8*
Mmax
6.5
25
46
10
12
10
15
23
23
Ts
81
Tmin
Tmax
0.05
0.1
0.04
0.04
0.03
0.05
1.98
20
2.75
20
10
10
0.013 1.95
L,
G
C
B
2,
1W
2M
1M
1M
2W
R
U
A (S, R)
A (S,R)
M
A
112 (70)
W. USA
250+
285
284
19
18
Considers equation valid for M 4.7.

Considers equation valid for d 300 km.
20
21
Minimum period for vertical equations is 0.04 s.
22
Maximum period for vertical equations is 3 s.
23
Coefcients given in Boore et al. (1994b).
Fukushima
et al. (1994)
& Fukushima
et al. (1995)
Lawson
&
Krawinkler
(1994)
150+1
280
132
66
Eastern
North
America
W.
USA
with
1
foreign
W.
N.
America
Central
America &
Mexico
3 vertical
arrays in
Japan
960+4
California
with
4
foreign
Sadigh et al.
(1993)
&
Sadigh et al.
(1997)
Electric Power
Research Institute (1993a)
Sun & Peng
(1993)
Boore et al.
(1994a)
&
Boore et al.
(1997)
Climent et al.
(1994)
V
-
H
U
Area
Worldwide
Reference
Campbell
(1993)
11
42
72
14 (9)
42+1
119+2
E
U
5.8
5.0
5.30
4.1
4*
3.8 (6.8)
Mmin
U18
2*
ML
for
M < 6,
else Ms
Mw
60*
MJMA
Mw
5*
0.1
(3)
rmin
U
Mw , mLg
M scale
ML
for
M < 6.0
and
Ms
otherwise
Mw
7.4
7.7
7.70
(7.40)
7.7
6.8*
7.4 (7.4)
Mmax
U
100
400*
109
150*
rjb
rhypo
rjb
repi
rhypo for
small ones
1000* rhypo (rrup
for largest)
305 rrup
(172)20some,
for
S
2
rmax r scale
rseis
U19
38
46
10
21
Ts
15
0.0521 7.522
0.1
0.05
0.05*
0.1
0.04
0.03
L,
G
10
C
M
Tmax
Tmin
0.04
1M
1,2
1M,
2M
1M
R
O
A (R,S)23
A(R,S)
M
A (T,S)
381
382
Europe &
Mid. East
1926
1926
422
157
297
72
36+4
105+1628 -
280
15+16
88*+28*26 -
4.0
1.7
3*
4.5 (7.2)
3 (3.7)
5.3
Mmin
3.75
25
Total number, does not need to be multiplied by two.

Total number, does not need to be multiplied by two.
26
There are 116 records in total.
27
Free (1996) believes it is largest horizontal component.
28
Total number of components does not need to be multiplied by two
24
Ambraseys
et al. (1996)
Cen.
America
W.
N.
America
UK + 28*
foreign
Greece+16
foreign
Musson et al.
(1994)
Theodulidis &
Papazachos
(1994)
Dahle et al.
(1995)
Lee & Trifunac
(1995)
265
53025
W. USA
23
56
10824
California
E
183
V
313
H
313
Area
Former Yugoslavia
Mohammadioun
(1994b)
Reference
Lee & Manic
(1994) & Lee
(1995)
Mohammadioun
(1994a)
Mw (Ms ,
mb , MD )
Usually
ML
for
M 6.5
and Ms for
M > 6.5
Ms
(unspecied)
Ms , Mw ,
MJMA
ML
ML
ML
M scale
U
7.9
7.7
8*
7.0 (7.5)
4.1 (6.4)
7.7
Mmax
7.0
6*
260
rjb
for
M > 6.0,
repi otherwise
9,
3
200+ rhypo
S
6
rhypo
490*
136
rmax r scale
repi
250
Often rrup ,
rhypo in far
eld
1
250 rrup , rE if
more appropriate,
rhypo in far
eld
70*
>477.4rhypo
(>1.3) (200*)
1
128 repi
(48) (236)
3
rmin
4
46
91
73
96
96
Ts
12
Tmin
Tmax
2
0.1
0.04
15
0.025 4
0.05
0.1
0.013 5
0.013 5
0.04
R
2R
1B
U27 O
C
U
M
A
El
Salvador
&
Nicaragua
Cen. & S.
California
Stable continental regions
Japan
California
Bommer et al.
(1996)
Molas & Yamazaki (1996)
Ohno et
(1996)
650*
655*
95
99118
95
248
347
477
399
410
2166
36
238
V
417
H
-
58
2729
17
17
H:
137
138, V:
126
132
387
16
20
E
157
State equations should not be used for distances > 100 km
Worldwide
extensional
regimes
California
with some
others
Spudich et al.
(1996) & Spudich
et
al.
(1997)
Abrahamson &
Silva (1997)
29
Italy
Sabetta
&
Pugliese (1996)
al.
Crouse
&
McGuire (1996)
Free
(1996)
& Free et al.
(1998)
Area
Europe &
Mid. East
Reference
Ambraseys
&
Simpson
(1996)
4.4
5.10
4.6
5.0
0.1
0
8*
7.2
1.5,
1.5
Ms
Mw
MJMA
Mw (ML )
Ms
if
ML
&
Ms 5.5
else ML
Mw
62
Ms
0.1
rmin
0
M scale
Ms
(unspecied)
7.4
6.90
6.8
7.5
7.8
6.8
1.5
4.1
7.7
7.0
Mmax
7.9
6.0
3.7
Mmin
4.0
rjb
for
some, repi
for most
rrup
220*
rrup
102.1 rjb
179,
18029
99.6
rhypo
otherwise
rq for M >
5.3, rhypo
otherwise
Both rjb &
repi
1000* rrup
for
2
earthquakes,
820
211
rmax r scale
rjb
for
M > 6.0,
repi otherwise
260 rhypo
260
S
3
28
46
14
12
52
14
10
Ts
46
0.01
0.1
0.04
0.02
0.1
0.04
0.04
0.1
14
Tmax
2
Tmin
0.1
G,
C
C
L
1M
2M
2M
1W
R
2
A (S,O,T)
NS
R,S (R,S)
M
A
383
384
31
30
W.
N.
America
University
City, Mexico City
Japan
304
3990
20+
1031
23
1020
20+
3443
164
57
H:30,
V:22
E
11+9
Mmin
5.0
5.8
5.5
5.0
3.3
4.7
4.1
0.1
MJMA
Mw
274
mb
for
M < 6,
Ms otherwise
Mw
Ms
6.1
Mw (Ms ,
mb , MD )
Mw
(Ms ,mb )
8.5
rmin
20*
Ms
for
Ms 6,
ML
for
Ms < 6
M scale
Mw
7.7
8.1
8.1
7.9
8.2
7.6
8.1
Mmax
6.7(8.2)
Typographic error in Table 3 of Campbell (1997) does not match number of recordings in Table 4
Typographical error in Figure 3b) of Perea & Sordo (1998) because it does not match their Table 1.
Shabestari
&
Yamazaki
(1998)
Chapman
(1999)
Reyes (1998)
Perea & Sordo

(1998)
Urban area
of Puebla,
Mexico
476
Worldwide
subduction
zones
Europe &
Mid. East
200
Costa Rica
121
183
173
266 30
Worldwide
Campbell
(1997), Campbell
(2000)
&
Campbell
(2001)
Schmidt et al.
(1997)
Youngs et al.
(1997)
Bommer et al.
(1998)
V
-
H
U
Area
Cascadia
with some
foreign
Reference
Atkinson (1997)
rrup
rrup
rjb
for
most, repi
otherwise
repi
189.4 rjb
663
260
550.9 rrup , rhypo

for some
182.1 rhypo
rmax r scale
rc
for
some,
rhypo for
small ones
50
rseis
580*
S
2
24
35
0.04
0.1
0.04
1.0
10
3.0
3.5
0.075 3
0.025 4
0.05
Tmax
2
Tmin
0.1
195 0.01
66
11
13
Ts
12
L,
B
G
C
B
2M
1M
IW
R
2
NT (N,T)
A (S,R,N)
M
A
357
447
107
224 (461+66)
6017
84
S California
Japan
Japan
NZ with 66
foreign
W.
Argentina
Japan
Caucasus
Kawano et al.
(2000)
Kobayashi et al.
(2000)
McVerry et al.
(2000)
Monguilner
et al. (2000b)
Shabestari
&
Yamazaki
(2000)
Smit
et
al.
(2000)
54
275
435
273
Worldwide
California
1308
Worldwide
54
107
274
434
-
1308
183
186
Ambraseys
&
Douglas
(2000),
Douglas (2001b) &
Ambraseys &
Douglas (2003)
Bozorgnia et al.
(2000)
Campbell & Bozorgnia (2000)
Chou & Uang
(2000)
Field (2000)
V
-
H
105
132
Area
Worldwide
extensional
regimes
Worldwide
Reference
Spudich et al.
(1999)
26
94
10
(51+17)
44
28
4.0
5.0
4.3
(5.08)
5.0
5.5
5.1
5.6
4.7
36
15
33
5.83
38
44
Mmin
5.1
27
0.9*
MJMA
Mw
7*
Ms
11
Ms if ML
& Ms > 6,
ML otherwise
MJMA
7.1
6.6
7.4
(0.1)
Mw
Mw
Mw
1*
0*
Ms
Mw
rmin
0
M scale
Mw
(7.23(7.41)) Mw
7.8
7.0
7.5
7.4
7.7
7.8
Mmax
7.2
rseis
60
60*
120
230
rhypo
950*
rrup
rq
C
(6)
I,
C
4
S
2
(573) (rrup
for
some, rc
for most)
350 rhypo
400*
202
148.9 rjb
rjb
rseis
rjb
15
rmax r scale
rjb
99.4
Tmin
0.01*
0.1
0.02
0.3
0.1
0.05
0.05
0.1
0.1
22
35
0.05
0.04
200 0.1
17
25
14
46
Ts
46
Tmax
10
4*
3.0
C
G
1W
1M
1M
2M
R
1M
A (N, R,
RO)
A (R, S, O)
A (S,R,T)
A (R,S,T)
M
NS
385
386
U
43*
Mmin
4.0 (5.8)
4.0
5.5
4.5
0.9*
4.0 and
4.2
3.4*
4.5
5.7*
3.7
5* (5.8*)
Mmax
33
0.1*
Mw
(m b
for Ms < 5
and
Ms
otherwise)
Ms
and
ML
49.7
11*
Ms
Mw
Ms (Mw
for
W.
USA)
2*
ML
0
and
0
0*
1.20
ML
rmax
300*
(100*)
600*
r scale
rrup , rhypo
for some
rhypo
repi
repi
330
rhypo
and
1
4
rjb , repi for

M < 5.9
rjb
repi
S
4
rjb , repi
322.4 rhypo
550* rrup
100*
110
and
150
250*
180*
150
53.51* 153.91*rhypo
rmin
1*
(0.1*)
4*
Mw
MJMA
M scale
Mw
7.9 (7.4)
6.7
8.3
5.9
7.2
6.9 and
7.0
7.4
7.4
7.2*
6.3
8.3* (8*)
This is total number of horizontal components used. They come from 47 triaxial records.
This is total number of components. Does not need to be multiplied by two.
34
Total number of records. Does not need to be multiplied by two.
35
485 records in total but do not state number of vertical records from W. USA.
32
45 or 47
1200+
15
19
28*
19
102
E
U+7*
802+16334 403+8235 130+8
161
77
15333
683
160
160
Iran
683
9332
Turkey
N.W.
Turkey
UmbriaMarche
Colombia
Subduction
zones
Europe
&
Mid.
East+163
from
W.
USA
174
NE India
Schwarz et al.
(2002)
Zonno & Montaldo (2002)
Alarcn (2003)
Atkinson
&
Boore (2003)
Berge-Thierry
et al. (2003)
3011
3011
Former Yugoslavia
V
-
1332
Area
Japan+166
foreign
Japan
Manic (2002)
Reference
Takahashi et al.
(2000)
Lussou et al.
(2001)
Das
et
al.
(2002)
Glkan
&
Kalkan (2002)
Khademi
(2002)
Tmin
0.05
0.04
0.04
0.01
0.04
0.05
0.1
0.04
0.02
0.05
143 0.03
84
7
14
11
14
13
46
20
63
Ts
20
Tmax
10
3
3
10
U
C
L,
R
L
C
G
U
1M
R
O
A
F, B
N, O
M
A

17
63
57
47
5.0
5.0*
4.0
4.2
5.5
4.7
3636
40+10
Mmin
4.0
E
157
0.5
1.2
0*
5*
Mw (Ms )
Mw (unspecied
scales)
Mw (unspecied
scales)
MJMA
Mw (ML )
1.2
2*
rmin
0
Mw
M scale
Ms
(unspecied)
7.4
7.6*
7.4
7.4
7.4
7.7
Mmax
7.9
rhypo (rrup
for 2 earthquakes)
300*
rjb
rjb ,
repi
for
small
events
250.0 rjb ,
repi
for
small
events
195* rrup & rq
250
235
rmax r scale
rjb
for
Ms > 6.0,
repi otherwise
60*
rseis
260
S
3
31
46
46
11
14
Ts
46
0.10
0.02*
0.1
0.1
0.03
0.05
37
4.0
4*
Tmax
2
Tmin
0.1
For horizontal corrected records. There are 34 for vertical corrected records.
Authors do not state reason for different number of records used for different periods.
38
The caption of their Table 2 states that reported coefcients are for mean.
39
The authors also report that they used 139 sets, which could refer to number of records rather than the 293 components that they also report.
36
195
NW Turkey
29339
Japan
Matsumoto
et al. (2004)
zbey et al.
(2004)
95
10037
112
Turkey
Kalkan
&
Glkan (2004a)
399+341
Mainly
west Eurasia+some
US
and
Japanese
Turkey
Kalkan
&
Glkan (2004a)
439
443
Worldwide
Campbell
&
Bozorgnia
(2003d),
Campbell & Bozorgnia (2003a)
&
Bozorgnia
&
Campbell
(2004b)
Fukushima
et al. (2003)
V
-
H
422
Area
Europe &
Mid. East
Reference
Bommer et al.
(2003)
1M
1M
L38
R
1M
C
L
NS
A (B, C, F)
A (S & N,
R, T)
M
A (S, R, N)
387
388
et
al.
41
40
4400
Central
Utah coalmining
areas
3172
277
1402
277
3168
207
595
522+18740 207
595
667
667
OkhotskAmur plate
boundary
Mainly
Japan+W
USA+Iran
W USA
Europe
& Middle
East
Europe
& Middle
East
E
Alps
(45.6
46.8 N &
1214 E)
Central
Mexico
V
-
H
142
Area
Worldwide
extensional
regimes
12
16
240
59
135
38+14*
59
135
270
42
E
39
0.98
5.2
2.5
5.0
5.0
5.0
4.9*
4.0
Mmin
5.1

Call it quadratic mean, which is assumed to be geometric mean.
McGarr
&
Fletcher (2005)
Garca
(2005)
Ambraseys
et al. (2005b)
Yu & Hu (2004)
Ambraseys
et al. (2005a)
Takahashi et al.
(2004)
Reference
Pankow
&
Pechmann
(2004)
and
Pankow
&
Pechmann
(2006)
Sunuwar et al.
(2004)
4*
0.5*
Mw
ML
Mw
Mw
(MCL )
0.3*
Mw
1.5*
0
>3
MJMA
Ms
Mw
rmin
0
M scale
Mw
4.2
7.4
6.3
7.6
7.5
7.6
8.3*
5.6
Mmax
7.2
10*
400*
rrup
for
Mw > 6.5,
rhypo
otherwise
rhypo
rrup
for
some,
rhypo for
rest
575* repi
99
rjb
(repi
for
small
events)
(repi
99
rjb
for
small
events)
130 rjb & repi
300
>264 rhypo
rmax r scale
rjb
99.4
1
3
S
2
15
47
61
U
61
21
19
Ts
46
Tmin
0.1
0.04
0.05
0.05
0.04
0.05
0.02
0.05
0.1
Tmax
2.0
2.0
2.5
20
2.5
5.0
3.0
2.0
1M
O
1M
1M
2M
R
1M
2M
G41 1M
B
L
C
G,
O
A (N, T, S,
O)
A
A (N, T, S,
O)
A (B, F, R,
S)
M
NS
Bindi et al.
(2006)
Campbell & Bozorgnia (2006a)
and Campbell
&
Bozorgnia
(2006b)
Hernandez
et al. (2006)
144
239
1500+
456
UmbriaMarche
Worldwide
Haulien
LSTT
(Taiwan)
461
4973
949
Los Angeles region

Shallow
crustal
(USA,
Taiwan,
Turkey and
others)
Beyer & Bommer (2006)
California
Mmin
4.1
51
60+
456
4.0
45
4.2
4.3*
103
3.1*
509+
<249+20 5.0
E
591
2763
4518+208
Japan+208
overseas
Takahashi et al.
(2005),
Zhao
et al. (2006)
and Fukushima
et al. (2006)
Wald et al.
(2005)
Atkinson (2006)
V
-
H
6812
Area
Japan
Reference
Pousse et al.
(2005)
13.7
ML
6*
Mw
Mw
5*
Mw
1*
Mw
ML
0*
rmin
5.5
Mw
M scale
Mw
(MJMA )
7.3
7.9
5.9
7.9*
7.1*
5.3*
8.3
Mmax
7.3
rjb
repi
rhypo
rrup
134.8 rhypo
200
100*
&
repi (rjb for

some)
200* rhypo
350*
rmax r scale
rhypo (rrup
for
10
events)
300* rrup
303
S
5
0.04
0.01
3.0
0.3
0.05
10
10
5.0
I, C
Tmax
4.0
Tmin
0.01
143 0.03
14
77
0.3
20
Ts
U
1M
R
2
2M
1,
1M
2,
A,
B,
C,
D50,
G,
I50,
L,
N,
P,
R
L
1M
C
B
NS
A (U)
C (R, S/N)
& F, B
M
A
389
390
al.
Europe
and Middle
East
Europe
& Middle
East
NW Turkey
997
4047
532
4047
89
175+9
Indian Himalayas+9
European
records
Iran
89
3198
Japan
939042
Japan
Area
H
V
Japan+some 3205
foreign
3392+331
377
(shallow) &
7721
8150
(deep)
New
435
Zealand
289
528
131
55*
12+7
52
49
E
70
73+10
& 101
111
0.5
5.0
2.7
4.5 (6.0)
5.5
4.1
5.08
Mmin
5.0*
(6.1) &
5.5*
5*
0
Mw
Mw
ML 43
10
Mw (mb )
Mw
Mw
5*
Mw
(Mw )
rmin
1*
(1.5*)
&
30*
M scale
Mw
(MJMA )
7.6
5.9
7.6
7.4
7.2 (7.4)
8.3
7.3
7.09
Mmax
8.2*
(7.4) &
8.0*
43

Also derive model using Mw .
44
Also derive model using repi .
45
Their Figure 2 present s up to 2 s but the coefcients of the model are not given beyond 1 s.
42
Bindi et al.
(2007)
Bommer et al.
(2007)
Akkar & Bommer (2007b)
Zare & Sabzali

(2006)
Pousse et al.
(2006)
Sakamoto et al.
(2006)
Sharma
&
Bungum (2006)
McVerry et al.
(2006)
Reference
Kanno et
(2006)
99
200*
99
167
200
300
250*
400
rhypo 44
rjb
(repi
for
small
events)
S
C
rjb
rhypo
rhypo
rhypo (rrup
for some)
rrup
rc (rrup )
rmax r scale
450* rrup (rhypo
(350*) for some)
&
450*
10
80
21
13
11
Ts
37
0.05
0.1
0.05
0.10
0.04
0.02
0.01
0.50
145
2.5
0.075 3
Tmax
5
Tmin
0.05
L,
G
C
R
C (R, OR,
S & N) & F,
B
A
M
A
A
1WM A (N, S, R)
1M
1M
A
&
2M
1WM A (N, S, R)
1W
1M
2M
1M
R
2M
399+339
Mainly
west Eurasia+some
US
and
Japanese
California
Central
northern
Italy
Colima,
Mexico
335
Greece
162
1063
484
592
506
1561
Worldwide
shallow
crustal
162
V
-
H
600*
1574
Area
Worldwide
shallow
crustal
26
243
3439
40+10
151
2164
E
18*58
Mmin
3.3
2.5
5*
5.5
4.5
4.2749
4.27
5.0046
0.07
0*
0.5
0.2*
0*
5*
Mw
Mw (Ms )
Mw
ML
ML
47
rmin
0
Mw
M scale
Mw
5.2
5.2
7.4*
7.4
6.9
7.9050
Mmax
7.9047

Believe that model can be used to 8.0.
48
Recommend that model is not used for distances 200 km.
49
50
46
Tejeda-Jcome
&
ChvezGarca (2007)
Hong & Goda

(2007) & Goda
& Hong (2008)
Massa et al.
(2007)
Reference
Boore & Atkinson (2007) &
Boore & Atkinson (2008)
Campbell & Bozorgnia (2007),
Campbell & Bozorgnia (2008b)
&
Campbell
&
Bozorgnia
(2008a)
Danciu & Tselentis (2007a) &
Danciu & Tselentis (2007b)
Fukushima
et al. (2007b)
& Fukushima
et al. (2007a)
175
300*
100*
235
136
rhypo
rhypo
rjb
rhypo (rrup
for 2 earthquakes)
repi
199.27rrup
rmax r scale
28048 rjb
S
C
Tmin
0.1
0.1
0.03
0.10
0.01
0.01
H:10,0.07
V:9
27
31
21
Ts
21
Tmax
G,
Q,
R
L
I50
C
I50
H:0.99, G
V:0.80
1.5
10
10
2M
1M
2M
1M
1M
R
2M
A (N, ST)
A (N, R, S,
HW)
M
A (N, R, S,
U)
391
392
Yu
et
al.
130
130
195053
Northern
Japan
1132
1164
772 (B),
1749
(F)
646
646
Worldwide
shallow
crustal
Worldwide
shallow
crustal
Worldwide
shallow
crustal
Worldwide
shallow
crustal
Worldwide
shallow
crustal
Japan
3894
V
-
H
500*
2754
Area
Worldwide
shallow
crustal
10 (B),
20 (F)
337
125
60
54
E
64
135
5.4 (B),
5.1 (F)
4.26554
5.0*
5.0*
5.0
5.2
Mmin
4.2751
0*
0*
Mw
Mw
Mw
Mw
(MJMA )
52
200*
200*
150*
60
rhypo
rrup
70*
300*
100
rhypo
rrup (rhypo
for small)
0.2*56 70*57 rrup
6*
Mw
Mw
rmin rmax r scale

0.06* 200* rrup
Mw
M scale
Mw
7.0 (B),
7.3 (F)
7.3
7.9055
7.5*
7.5*
7.2
7.9
Mmax
7.952

Believe that model can be reliably extrapolated to 8.5.
53
Due to ltering number of records and earthquakes depends on period.
54
55
56
57
58
For stations on surface.
59
For borehole stations.
51
Dhakal et al.
(2008)
Cotton
(2008)
Chiou
&
Youngs (2008)
Chen &
(2008a)
Reference
Abrahamson &
Silva (2008) &
Abrahamson &
Silva (2009)
Aghabarati &
Tehranizadeh
(2008)
Cauzzi & Faccioli (2008) &
Cauzzi (2008)
Chen & Yu
(2008b)
23
458
&
259
1
16
22
0.1
0.01
0.01
0.04
0.04
3.33
10
10
10
20
0.025 10
Tmax
10
Tmin
0.01
400 0.05
26
Ts
22
4
&
C
1
S
C
I50
C
I50
2M
1M
2M
1M
R
1M
B, F
A (N, R, S,
HW, AS)
A (N, S, R)
A (N, R, S)
M
A (N, R, S,
HW)
Hong et
(2009b)
al.
Aghabarati &
Tehranizadeh
(2009)
Akyol
&
Karagz (2009)
Bindi et al.
(2009a)
Bindi et al.
(2009b)
Bragato (2009)
Morasca et al.
(2008)
Yuzawa & Kudo
(2008)
Ghasemi et al.
(2009)
al.
Massa
(2008)
et
Lee
Lin
&
(2008)
Reference
Idriss (2008)
V
-
922
Italy
418,
277
235
Italy
Mexico
(interface
& inslab)
241
241
-, -
168
678
716+177
Iran+West
Eurasia
678
1988
Japan
3090
306
Worldwide
shallow
crustal
Western
Anatolia
Italy
3090
306
4244+139 -
H
942
Molise
Area
Worldwide
shallow
crustal
NE
Taiwan+10
foreign
Northern
Italy
40, 16
116
27
27
49
55
200
18
100
82
44+10
E
72
5.0, 5.2
2.7
4.6
4.8
4.03
5.2
5.0
5.9, 5.7
2.7
3.5 & 4.0
4.1 (6.0)
Mmin
4.5
rmin
0.3
15
1*
12*
U
0.5
15
0
0
6
M scale
Mw
Mw (ML )
Mw (ML )
& ML
ML
MJM A ,
Mw
Mw
Mw
Mw (Md ,
ML )
Mw
Mw (ML )
ML
Mw
8.0, 7.4
4.5
6.9
6.9
6.40
7.9
7.4
8.0, 7.9
5.7
6.3 & 6.5
7.3 (8.1)
Mmax
7.7
100
183
190
200
60
100
60*
100*
630
rrup (rhypo
for
small
events)
rrup
rrup (rhypo
for small)
repi
I,
3,
1,
C
1
rjb
rjb , repi
rhypo
S
2
rq
rhypo
repi
rhypo
rmax r scale
199.3 rrup
27
19
18
30
26
17
45
12
&
14
12
27
Ts
31
10
2 &
4
0.1
0.3
0.03
0.03
0.05
0.025 10
0.05
0.04
0.04
0.01
Tmax
10
Tmin
0.01
G,
R,
Q
L,
G
L
I50
C
I50
1M
1M
1M
2M
1M
1M
1M
1M
1W
R
1
F, S
A (N, S, R)
A (N, R, S)
M
A
(R/RO/NO,
S/N)
A (B, F)
393
394
al.
et
al.
61
60
64+29
Europe
& Middle
East
Alborz and
central
Iran60
Pacic
coast
of
Mexico
Italy
Europe
& Middle
East
Turkey
595
561
418
416
433
532
561
1950
Worldwide
shallow
crustal
South Iceland+others
Indian Hi- 58+143

malayas+Zagros
V
-
2660
H
484
592
Worldwide
Area
California
135
107
40
189
137
131
6+10
12
125
60
E
3439
Mmin
5.0
4.0
0.2*
5*
&
10*
0
0*
Mw
Mw
Mw
Mw
20
1*
0
Mw
Mw
Mw
Ms (mb )
5*
0.1*
Mw
Mw
rmin
0.2*
M scale
Mw
7.6
6.9
8.0
7.7
3.261
5.0
7.6
7.6
6.8 & 6.6
7.67
7.90
7.9*
Mmax
7.4*
5.0
5.0
5.5 & 5.9
5.02
4.265
5.61
5*
Also develop models for the Zagros region of Iran using 309 records from 190 earthquakes.
State that only use data with Ms 4 but one earthquake in their Appendix A has Ms 3.2.
Bindi et al.
(2010)
Douglas
&
Halldrsson
(2010)
Arroyo
(2010)
Akkar & agnan (2010)

Ghodrati Amiri
et al. (2010)
Akkar & Bommer (2010)
Rupakhety
&
Sigbjrnsson
(2009)
Sharma et al.
(2009)
Kuehn et al.
(2009)
Moss (2009)
Reference
Hong et
(2009a)
99
100*
400
400*
200*
190*
&
200*
99
97
70*
200*
rjb
(repi
for
small
events)
rrup (rhypo
for Mw <
6)
rjb , repi
rhypo
rjb
rjb
rjb
S
C
rjb (repi for

some)
rrup
rjb
rmax r scale
rjb
100*
61
21
56
15
14
60
13
66
39
Ts
27
Tmin
0.05
0.03
0.04
0.1
0.03
0.05
0.04
0.04
0.1
0.01
0.1
Tmax
2.5
2.5
2.5
7.5
I50
C
G,
R
A (N, S, R)
A (N, S, R)
A (S, R)
S&O
A (N, R, S,
HW, AS)
A (N, R, S)
M
A
1WM A (N, T, S,
O, AS)
1M
1M
1M
1M
R
1M,
2M,
O
1M
(O)
1M
62
3894
883
627
Japan
Iran
Central
Iran
Zagros
&
81
3894
484
592
1561
H
1499
South Iceland
Worldwide
shallow
crustal
Japan
Area
Worldwide
shallow
crustal
California62
110
79
337
337
64
5.1
4.27
5.0
60
3439
Mmin
4.5
V
-
7.4*
7.4*
7.3
6.5
7.3
7.90
7.28
Mmax
7.6
0.2*
0.07
Mw
Mw
Mw
(MJMA )
Mw
0*
5*
Mw
(MJMA )
Mw
Mw
1*
rmin
0.2*
M scale
Mw
rjb
200*
340*
100
80*
100
rrup (rhypo
for Mw <
6.5)
repi
rrup (rhypo
for small)
rrup (rhypo
for small)
rjb (repi for
Mw < 6)
199.27rrup
100*
rmax r scale
rrup (rhypo
for small)
200*
C,
I
2
S
4
&
C
C
Also derive models for inslab (273 records from 16 earthquakes) and interface (413 records from 40 earthquakes) Mexican earthquakes.
Montalva
(2010)
Ornthammarath
et al. (2010b),
Ornthammarath
(2010) & Ornthammarath
et al. (2010a)
RodriguezMarek & Montalva (2010)
Sadeghi et al.
(2010)
Saffari et al.
(2010)
Hong & Goda

(2010)
Jayaram
&
Baker (2010)
Reference
Faccioli et al.
(2010)
19
21
21
Ts
22
Tmax
10
20
1,
2
I50
C
G
0.05
0.1
0.01
0.2
1.3622 G
0.0384 1.3622 G
0.01
0.2
Tmin
0.05
2M
2M
O
(1M)
1M
1M
R
1M
A (N, R, S,
HW)
M
A (N, R, S)
395
396
Chapter 6
List of other ground-motion models

Published ground-motion models for the prediction of PGA and/or response spectral ordinates
that were derived by methods other than regression analysis on strong-motion data are listed
below in chronological order.
Illustration 3: GMPEs derived based on simulated ground motions, often the stochastic method
Herrmann & Goertz (1981)

Herrmann & Nuttli (1984)
Boore & Atkinson (1987) and Atkinson & Boore (1990)
Nuttli & Herrmann (1987)
Toro & McGuire (1987)
Electric Power Research Institute (1988)
Boore & Joyner (1991)
Bungum et al. (1992)
Electric Power Research Institute (1993b)
Savy et al. (1993)
Atkinson & Boore (1995) & Atkinson & Boore (1997a)
Frankel et al. (1996) & Electric Power Research Institute
(2004, Appendix B)
Atkinson & Boore (1997b)
Hwang & Huo (1997)
Atkinson & Silva (2000)
Somerville et al. (2001)
Toro & Silva (2001)
Gregor et al. (2002b)
Silva et al. (2002)
Toro (2002)
Megawati et al. (2003)
Electric Power Research Institute (2004) (model clusters)
Iyengar & Raghu Kanth (2004)
Zheng & Wong (2004)
Megawati et al. (2005)
Motazedian & Atkinson (2005)
397
Eastern North America

Intraplate regions
Central and eastern USA
Cascadia subduction zone
Eastern USA
Iceland
California
Central USA
Sumatran subduction zone
Peninsular India
Southern China
Puerto Rico
Nath et al. (2005)

Collins et al. (2006)
Raghu Kanth & Iyengar (2006, 2007)
Megawati (2007)
Tuluka (2007)
Carvalho (2008)
Jin et al. (2008)1
Liang et al. (2008)
Sokolov et al. (2008)
Atkinson & Macias (2009)
Kang & Jin (2009)2
Somerville et al. (2009)
Hamzehloo & Bahoosh (2010)
Megawati & Pan (2010)
Sikkim Himalaya
Intermountain West, USA
Peninsular India
Hong Kong
African Western Rift Valley
Portugal
Fujian region, China
Southwest Western Australia
Vrancea, Romania
Sichuan region, China
Austalia
Tehran region, Iran
Illustration 4: Complete (source, path and site terms) stochastic

models that could be used within the stochastic method (e.g.
Boore, 2003)
De Natale et al. (1988)

Campi Flegrei, Italy
Atkinson (1996)
Cascadia
Atkinson & Silva (1997)
California
Gusev et al. (1997)
Kamchatka
Sokolov (1997)
Northern Caucasus
Sokolov (1998)
Caucasus
Raoof et al. (1999)
Southern California
Malagnini & Herrmann (2000)
Umbria-Marche, Italy
Malagnini et al. (2000a)
Apennines, Italy
Malagnini et al. (2000b)
Central Europe
Taiwan
Akinci et al. (2001)
Erzincan, Turkey
Parvez et al. (2001)
Himalaya
Junn et al. (2002)
South Korea
Malagnini et al. (2002)
Northeastern Italy
Bay et al. (2003)
Switzerland
Bodin et al. (2004)
Kachchh basin, India
Jeon & Herrmann (2004)
Utah and Yellowstone, USA
Halldorsson & Papageorgiou (2005) Intraplate and interplate
Scognamiglio et al. (2005)
Eastern Sicily, Italy
Vrancea, Romania
1
This may be an empirical GMPE because it is based on broadband velocity records from which acceleration
time-histories are generated by real-time simulation. This could just mean differentiation.
2
This may be an empirical GMPE because it is based on broadband velocity records from which acceleration
time-histories are generated by real-time simulation. This could just mean differentiation.
398
Akinci et al. (2006)

Allen et al. (2006)
Chung (2006)
Morasca et al. (2006)
Malagnini et al. (2007)
Meirova et al. (2008)
Zafarani et al. (2008)
Hao & Gaull (2009)
Marmara, Turkey
Southwest Western Austalia
Southwestern Taiwan
Western Alps
San Francisco, USA
Israel
Iran
Perth, Australia
Illustration 5: GMPEs derived using the hybrid stochasticempirical method (e.g. Campbell, 2003b)
Atkinson (2001)
Abrahamson & Silva (2002)
Campbell (2003b)
Atkinson (2005)
Tavakoli & Pezeshk (2005)
Douglas et al. (2006)
Douglas et al. (2006)
Campbell (2007)

Cascadia
Southern Norway
Southern Spain
Illustration 6: GMPEs derived by converting equations for the

prediction of macroseismic intensity to the prediction of PGA
and response spectral ordinates
Battis (1981)
Hasegawa et al. (1981)
Huo et al. (1992)
Malkawi & Fahmi (1996)

Canada
China
Jordan
Illustration 7: GMPEs derived using the referenced-empirical

method (e.g. Atkinson, 2008) that adjusts coefcients of published GMPEs for one region to provide a better match to observations from another
Atkinson (2008)
Scasserra et al. (2009)
Atkinson (2009, 2010)
Gupta (2010)

Italy
Hawaii
Indo-Burmese subduction zone
399
Illustration 8: Studies where one or more coefcients of previously published GMPEs are altered following additional analysis
(completely new GMPEs are not derived in these studies)
Eberhart-Phillips & McVerry (2003)

Wang & Takada (2009)
Chiou et al. (2010)
Zhao (2010)
New terms for McVerry et al. (2000)

Adjustment of Si & Midorikawa (1999, 2000) for
stations HKD100 and CHB022
New terms for Chiou & Youngs (2008)
New terms for Zhao et al. (2006)
New terms for Boore & Atkinson (2008), Atkinson
& Boore (2006) and Atkinson (2008)
Illustration 9: Non-parametric ground-motion models, i.e. models without an associated close-form equation, which are more
difcult to use within seismic hazard assessments
Schnabel & Seed (1973)

Katayama (1982)
Anderson & Lei (1994)
Lee et al. (1995)
Anderson (1997)
Fajfar & Perus (1997)
Garcia & Romo (2006)
Pathak et al. (2006)
Gll & Erelebi (2007)
Ahmad et al. (2008)
Gnaydn & Gnaydn (2008)
Cabalar & Cevik (2009)
Perus & Fajfar (2009, 2010)
400
Western North America

Japan
Guerrero, Mexico
California
Guerrero, Mexico
Europe & Middle East
Subduction zones
India
Turkey
Europe & Middle East
Northwestern Turkey
Turkey
Worldwide
Bibliography
Abdalla, J. A., Mohamedzein, Y. E.-A., & Wahab, A. A. 2001. Probabilistic seismic hazard
assessment of Sudan and its vicinity. Earthquake Spectra, 17(3).
Abrahamson, N., & Silva, W. 2008. Summary of the Abrahamson & Silva NGA ground-motion
relations. Earthquake Spectra, 24(1), 6797.
Abrahamson, N., & Silva, W. 2009 (Aug). Errata for Summary of the Abrahamson and Silva
NGA ground-motion relations by Abrahamson, N. A. and W. J. Silva. Published on PEER
NGA website.
Abrahamson, N. A., & Litehiser, J. J. 1989. Attenuation of vertical peak acceleration. Bulletin
of the Seismological Society of America, 79(3), 549580.
Abrahamson, N. A., & Shedlock, K. M. 1997. Overview. Seismological Research Letters,
68(1), 923.
Abrahamson, N. A., & Silva, W. J. 1993. Attenuation of long period strong ground motions. In:
Proceedings of Conference of American Society of Mechanical Engineers.
Abrahamson, N. A., & Silva, W. J. 1997. Empirical response spectral attenuation relations for
shallow crustal earthquakes. Seismological Research Letters, 68(1), 94127.
Abrahamson, N. A., & Silva, W. J. 2002 (Sep). Hybrid model empirical attenuation relations
for central and eastern U.S. hard and soft rock and deep soil site conditions. In: CEUS
Ground Motion Project Workshop. Not seen. Cited in Electric Power Research Institute
(2004). Only a presentation. Never ofcially published.
Abrahamson, N. A., & Youngs, R. R. 1992. A stable algorithm for regression analyses using
the random effects model. Bulletin of the Seismological Society of America, 82(1), 505510.
Aghabarati, H., & Tehranizadeh, M. 2008. Near-source attenuation relationship for the geometric mean horizontal component of peak ground acceleration and acceleration response
spectra. Asian Journal of Civil Engineering (Building and Housing), 9(3), 261290.
Aghabarati, H., & Tehranizadeh, M. 2009. Near-source ground motion attenuation relationship for PGA and PSA of the vertical and horizontal components. Bulletin of Earthquake
Engineering, 7(3), 609635.
gstsson, K., orbjarnardttir, B., & Vogfjr, K. 2008 (Apr). Seismic wave attenuation for
earthquakes in SW Iceland: First results. Tech. rept. 08005. Veurstofa slands (Icelandic
Meteorological Ofce).
Ahmad, I., El Naggar, M. H., & Khan, A. N. 2008. Neural network based attenuation of strong
motion peaks in Europe. Journal of Earthquake Engineering, 12(5), 663680.
401
Akinci, A., Malagnini, L., Herrmann, R. B., Pino, N. A., Scognamiglio, L., & Eyidogan, H.
2001. High-frequency ground motion in the Erzincan region, Turkey: Inferences from small
earthquakes. Bulletin of the Seismological Society of America, 91(6), 14461455.
Akinci, A., Malagnini, L., Herrmann, R. B., Gok, R., & Srensen, M. B. 2006. Ground motion
scaling in the Marmara region, Turkey. Geophysical Journal International, 166(2), 635651.
Akkar, S., & Bommer, J. J. 2006. Inuence of long-period lter cut-off on elastic spectral
displacements. Earthquake Engineering and Structural Dynamics, 35(9), 11451165.
Akkar, S., & Bommer, J. J. 2007a. Empirical prediction equations for peak ground velocity
derived from strong-motion records from Europe and the Middle East. Bulletin of the Seismological Society of America, 97(2), 511530.
Akkar, S., & Bommer, J. J. 2007b. Prediction of elastic displacement response spectra in
Europe and the Middle East. Earthquake Engineering and Structural Dynamics, 36(10),
12751301.
Akkar, S., & Bommer, J. J. 2010. Empirical equations for the prediction of PGA, PGV and spectral accelerations in Europe, the Mediterranean region and the Middle East. Seismological
Research Letters, 81(2), 195206.
Akkar, S., & agnan,

Z. 2010. A local ground-motion predictive model for Turkey and its comparison with other regional and global ground-motion models. Bulletin of the Seismological
Society of America, 100(6), 29782995.
Akkar, S., agnan,

Z., Yenier, E., Erdogan,
, Sandkkaya, A., & Glkan, P. 2010. The recently compiled Turkish strong-motion database: Preliminary investigation for seismological
parameters. Journal of Seismology, 14(3), 457479.
Akyol, N., & Karagz, . 2009. Empirical attenuation relationships for western Anatolia,
Turkey. Turkish Journal of Earth sciences, 18, 351382.
Alarcn, J. E. 2003. Relaciones de atenuacin a partir de espectros de respuesta para Colombia. In: Proceedings of the Second Colombian Conference on Earthquake Engineering. In
Spanish.
Alarcn, J. E. 2007 (Apr). Estimation of duration, number of cycles, peak ground velocity, peak
ground acceleration and spectral ordinates for engineering design. Ph.D. thesis, University
of London.
Alchalbi, A., Costa, G., & Suhadolc, P. 2003 (Aug). Strong motion records from Syria: A
preliminary analysis. In: Skopje Earthquake 40 Years of European Earthquake Engineering
(SE-40EEE).
Alfaro, C. S., Kiremidjian, A. S., & White, R. A. 1990. Seismic zoning and ground motion
parameters for El Salvador. Tech. rept. 93. The John A. Blume Earthquake Engineering
Center, Stanford University. Not seen. Reported in Bommer et al. (1996).
Algermissen, S. T., Hansen, S. L., & Thenhaus, P. C. 1988. Seismic hazard evaluation for El
Salvador. Tech. rept. Report for the US Agency for International Development. Not seen.
Reported in Bommer et al. (1996).
402
Allen, T. I., Dhu, T., Cummins, P. R., & Schneider, J. F. 2006. Empirical attenuation of groundmotion spectral amplitudes in southwestern Western Australia. Bulletin of the Seismological
Aman, A., Singh, U. K., & Singh, R. P. 1995. A new empirical relation for strong seismic ground
motion for the Himalayan region. Current Science, 69(9), 772777.
Ambraseys, N. 1975a. Ground motions in the near eld of small-magnitude earthquakes.
Pages 113136 of: Proceedings of the Commission on the Safety of Nuclear Installations,
Organisation of Economic Cooperation in Europe, vol. 1. Not seen. Reported in Ambraseys
(1978a).
Ambraseys, N., & Douglas, J. 2000 (Aug). Reappraisal of the effect of vertical ground motions on response. ESEE Report 00-4. Department of Civil and Environmental Engineering,
Imperial College, London.
Ambraseys, N., Smit, P., Berardi, R., Rinaldis, D., Cotton, F., & Berge, C. 2000. Dissemination
of European Strong-Motion Data. CD-ROM collection. European Commission, DirectorateGeneral XII, Environmental and Climate Programme, ENV4-CT97-0397, Brussels, Belgium.
Ambraseys, N. N. 1975b. Trends in engineering seismology in Europe. Pages 3952 of:
Proceedings of Fifth European Conference on Earthquake Engineering, vol. 3.
Ambraseys, N. N. 1978a. Middle East a reappraisal of seismicity. The Quarterly Journal of
Engineering Geology, 11(1), 1932.
Ambraseys, N. N. 1978b. Preliminary analysis of European strong-motion data 19651978.
Bulletin of the European Association of Earthquake Engineering, 4, 1737.
Ambraseys, N. N. 1990. Uniform magnitude re-evaluation of European earthquakes associated with strong-motion records. Earthquake Engineering and Structural Dynamics, 19(1),
120.
Ambraseys, N. N. 1995. The prediction of earthquake peak ground acceleration in Europe.
Earthquake Engineering and Structural Dynamics, 24(4), 467490.
Ambraseys, N. N., & Bommer, J. J. 1991. The attenuation of ground accelerations in Europe.
Ambraseys, N. N., & Bommer, J. J. 1992. On the attenuation of ground accelerations in
Europe. Pages 675678 of: Proceedings of Tenth World Conference on Earthquake Engineering, vol. 2.
Ambraseys, N. N., & Bommer, J. J. 1995. Attenuation relations for use in Europe: An overview.
Pages 6774 of: Elnashai, A. S. (ed), Proceedings of Fifth SECED Conference on European
Seismic Design Practice.
Ambraseys, N. N., & Douglas, J. 2003. Near-eld horizontal and vertical earthquake ground
motions. Soil Dynamics and Earthquake Engineering, 23(1), 118.
Ambraseys, N. N., & Simpson, K. A. 1996. Prediction of vertical response spectra in Europe.
Ambraseys, N. N., & Srbulov, M. 1994. Attenuation of earthquake-induced ground displacements. Earthquake Engineering and Structural Dynamics, 23(5), 467487.
403
Ambraseys, N. N., Bommer, J. J., & Sarma, S. K. 1992 (Nov). A review of seismic ground motions for UK design. ESEE Report 92-8. Department of Civil Engineering, Imperial College,
London.
Ambraseys, N. N., Simpson, K. A., & Bommer, J. J. 1996. Prediction of horizontal response
spectra in Europe. Earthquake Engineering and Structural Dynamics, 25(4), 371400.
Ambraseys, N. N., Smit, P., Douglas, J., Margaris, B., Sigbjrnsson, R., lafsson, S.,
Suhadolc, P., & Costa, G. 2004. Internet site for European strong-motion data. Bollettino di
Geosica Teorica ed Applicata, 45(3), 113129.
Ambraseys, N. N., Douglas, J., Sarma, S. K., & Smit, P. M. 2005a. Equations for the estimation
of strong ground motions from shallow crustal earthquakes using data from Europe and the
Middle East: Horizontal peak ground acceleration and spectral acceleration. Bulletin of
Earthquake Engineering, 3(1), 153.
Ambraseys, N. N., Douglas, J., Sarma, S. K., & Smit, P. M. 2005b. Equations for the estimation
of strong ground motions from shallow crustal earthquakes using data from Europe and
the Middle East: Vertical peak ground acceleration and spectral acceleration. Bulletin of
Anderson, J. G. 1997. Nonparametric description of peak acceleration above a subduction
thrust. Seismological Research Letters, 68(1), 8693.
Anderson, J. G., & Lei, Y. 1994. Nonparametric description of peak acceleration as a function
of magnitude, distance, and site in Guerrero, Mexico. Bulletin of the Seismological Society
of America, 84(4), 10031017.
Annaka, T., & Nozawa, Y. 1988. A probabilistic model for seismic hazard estimation in the
Kanto district. Pages 107112 of: Proceedings of Ninth World Conference on Earthquake
Engineering, vol. II.
Aptikaev, F., & Kopnichev, J. 1980. Correlation between seismic vibration parameters and type
of faulting. Pages 107110 of: Proceedings of Seventh World Conference on Earthquake
Engineering, vol. 1.
Arroyo, D., Garca, D., Ordaz, M., Mora, M. A., & Singh, S. K. 2010. Strong ground-motion
relations for Mexican interplate earthquakes. Journal of Seismology, 14(4), 769785.
Atkinson, G. M. 1990. A comparison of eastern North American ground motion observations
with theoretical predictions. Seismological Research Letters, 61(34), 171180.
Atkinson, G. M. 1996. The high-frequency shape of the source spectrum for earthquakes in
eastern and western Canada. Bulletin of the Seismological Society of America, 86(1A),
106112.
Atkinson, G. M. 1997. Empirical ground motion relations for earthquakes in the Cascadia
region. Canadian Journal of Civil Engineering, 24, 6477.
Atkinson, G. M. 2001. An alternative to stochastic ground-motion relations for use in seismic
hazard analysis in eastern North America. Seismological Research Letters, 72, 299306.
Atkinson, G. M. 2005. Ground motions for earthquakes in southwestern British Columbia and
northwestern Washington: Crustal, in-slab, and offshore events. Bulletin of the Seismological Society of America, 95(3), 10271044.
404
Atkinson, G. M. 2006. Single-station sigma. Bulletin of the Seismological Society of America,

96(2), 446455.
Atkinson, G. M. 2008. Ground-motion prediction equations for eastern North America from
a referenced empirical approach: Implications for epistemic uncertainty. Bulletin of the
Seismological Society of America, 98(3), 13041318.
Atkinson, G. M. 2009 (Mar). Ground motion prediction equations for Hawaii from a referenced
empirical approach. Final technical report 08HQGR0020.
Atkinson, G. M. 2010. Ground motion prediction equations for Hawaii from a referenced empirical approach. Bulletin of the Seismological Society of America, 100(2), 751761.
Atkinson, G. M., & Boore, D. M. 1990. Recent trends in ground motion and spectral response
relations for North America. Earthquake Spectra, 6(1), 1535.
Atkinson, G. M., & Boore, D. M. 1995. Ground-motion relations for eastern North America.
Bulletin of the Seismological Society of America, 85(1), 1730.
Atkinson, G. M., & Boore, D. M. 1997a. Some comparisons between recent ground-motion
relations. Seismological Research Letters, 68(1).
Atkinson, G. M., & Boore, D. M. 1997b. Stochastic point-source modeling of ground motions
in the Cascadia region. Seismological Research Letters, 68(1), 7485.
Atkinson, G. M., & Boore, D. M. 2003. Empirical ground-motion relations for subduction zone
earthquakes and their application to Cascadia and other regions. Bulletin of the Seismological Society of America, 93(4), 17031729.
Atkinson, G. M., & Boore, D. M. 2006. Earthquake ground-motion prediction equations for
eastern North America. Bulletin of the Seismological Society of America, 96(6), 21812205.
Atkinson, G. M., & Boore, D. M. 2011. Modications to existing ground-motion prediction
equations in light of new data. Bulletin of the Seismological Society of America, 101. In
press.
Atkinson, G. M., & Macias, M. 2009. Predicted ground motions for great interface earthquakes
in the Cascadia subduction zone. Bulletin of the Seismological Society of America, 99(3),
15521578.
Atkinson, G. M., & Silva, W. 1997. An empirical study of earthquake source spectra for California earthquakes. Bulletin of the Seismological Society of America, 87(1), 97113.
Atkinson, G. M., & Silva, W. 2000. Stochastic modeling of California ground motion. Bulletin
Aydan, . 2007. Inference of seismic characteristics of possible earthquakes and liquefaction
and landslide risks from active faults. Pages 563574 of: The 6th National Conference on
Earthquake Engineering of Turkey, vol. 1. Not seen. In Turkish.
Aydan, ., Sedaki, M., & Yarar, R. 1996. The seismic characteristics of Turkish earthquakes.
In: Proceedings of Eleventh World Conference on Earthquake Engineering. Paper no. 1270.
405
Baag, C.-E., Chang, S.-J., Jo, N.-D., & Shin, J.-S. 1998. Evaluation of seismic hazard in the
southern part of Korea. In: Proceedings of the second international symposium on seismic
hazards and ground motion in the region of moderate seismicity. Not seen. Reported in
Nakajima et al. (2007).
Battis, J. 1981. Regional modication of acceleration attenuation functions. Bulletin of the
Bay, F., Fh, D., Malagnini, L., & Giardini, D. 2003. Spectral shear-wave ground-motion scaling
in Switzerland. Bulletin of the Seismological Society of America, 93(1), 414429.
Beauducel, F., Bazin, S., & Bengoubou-Valerius, M. 2004 (Dec). Loi dattnuation B-cube
pour lvaluation rapide des intensits sismiques probables dans larchipel de Guadeloupe.
Internal report OVSG-IPGP-UAG. Observatoire Volcanologique et Sismologique de Guadeloupe.
Benito, B., Rinaldis, D., Gorelli, V., & Paciello, A. 1992. Inuence of the magnitude, distance
and natural period of soil in the strong ground motion. Pages 773779 of: Proceedings of
Tenth World Conference on Earthquake Engineering, vol. 2.
Benito, B., Cabaas, L., Jimnez, M. E., Cabaas, C., Lpez, M., Gmez, P., & Alvarez, S.
2000. Caracterizacin del movimiento del suelo en emplazamientos de la pennsula ibrica
y evaluacin del dao potencial en estructuras. proyecto daos. In: Consejo de Seguridad
Nuclear (ed), Monograa ref. 19.2000. In Spanish. Not seen.
Berge-Thierry, C., Cotton, F., Scotti, O., Griot-Pommera, D.-A., & Fukushima, Y. 2003. New
empirical response spectral attenuation laws for moderate European earthquakes. Journal
of Earthquake Engineering, 7(2), 193222.
Beyaz, T. 2004. Development of a new attenuation relationship of seismic energy for Turkey
using the strong motion records free of soil effect. Ph.D. thesis, Ankara University, Turkey.
Not seen. Reported in Selcuk et al. (2010).
Beyer, K., & Bommer, J. J. 2006. Relationships between median values and between aleatory
variabilities for different denitions of the horizontal component of motion. Bulletin of the
Seismological Society of America, 96(4A), 15121522.
Bindi, D., Luzi, L., Pacor, F., Franceshina, G., & Castro, R. R. 2006. Ground-motion predictions
from empirical attenuation relationships versus recorded data: The case of the 1997U-1998
Umbria-Marche, central Italy, strong-motion data set. Bulletin of the Seismological Society
of America, 96(3), 9841002.
Bindi, D., Parolai, S., Grosser, H., Milkereit, C., & Durukal, E. 2007. Empirical ground-motion
prediction equations for northwestern Turkey using the aftershocks of the 1999 Kocaeli
earthquake. Geophysical Research Letters, 34(L08305).
Bindi, D., Luzi, L., & Pacor, F. 2009a. Interevent and interstation variability computed for the
Italian Accelerometric Archive (ITACA). Bulletin of the Seismological Society of America,
99(4), 24712488.
Bindi, D., Luzi, L., Pacor, F., Sabetta, F., & Massa, M. 2009b. Towards a new reference
ground motion prediction equation for Italy: Update of the Sabetta-Pugliese (1996). Bulletin
406
Bindi, D., Luzi, L., Massa, M., & Pacor, F. 2010. Horizontal and vertical ground motion prediction equations derived from the Italian Accelerometric Archive (ITACA). Bulletin of Earthquake Engineering, 8(5), 12091230.
Blume, J. A. 1977. The SAM procedure for site-acceleration-magnitude relationships. Pages
416422 of: Proceedings of Sixth World Conference on Earthquake Engineering, vol. I.
Blume, J. A. 1980. Distance partitioning in attenuation studies. Pages 403410 of: Proceedings of Seventh World Conference on Earthquake Engineering, vol. 2.
Boatwright, J., Bundock, H., Luetgert, J., Seekins, L., Gee, L., & Lombard, P. 2003. The
dependence of PGA and PGV on distance and magnitude inferred from northern California
ShakeMap data. Bulletin of the Seismological Society of America, 93(5), 20432055.
Bodin, P., Malagnini, L., & Akinci, A. 2004. Ground-motion scaling in the Kachchh basin, India,
deduced from aftershocks of the 2001 Mw 7.6 Bhuj earthquake. Bulletin of the Seismological
Bolt, B. A., & Abrahamson, N. A. 1982. New attenuation relations for peak and expected
accelerations of strong ground motion. Bulletin of the Seismological Society of America,
72(6), 23072321.
Bommer, J. J. 2006. Empirical estimation of ground motion: Advances and issues. Pages 115
135 of: Proceedings of Third International Symposium on the Effects of Surface Geology
on Seismic Motion. Paper number: KN 8.
Bommer, J. J., & Alarcn, J. E. 2006. The prediction and use of peak ground velocity. Journal
Bommer, J. J., & Martnez-Pereira, A. 1999. The effective duration of earthquake strong
motion. Journal of Earthquake Engineering, 3(2), 127172.
Bommer, J. J., & Scherbaum, F. 2008. The use and misuse of logic trees in probabilistic
seismic hazard analysis. Earthquake Spectra, 24(4), 9971009.
Bommer, J. J., Hernndez, D. A., Navarrete, J. A., & Salazar, W. M. 1996. Seismic hazard
assessments for El Salvador. Geofsica Internacional, 35(3), 227244.
Bommer, J. J., Elnashai, A. S., Chlimintzas, G. O., & Lee, D. 1998 (Mar). Review and development of response spectra for displacement-based seismic design. ESEE Report 98-3.
Department of Civil Engineering, Imperial College, London.
Bommer, J. J., Douglas, J., & Strasser, F. O. 2003. Style-of-faulting in ground-motion prediction
equations. Bulletin of Earthquake Engineering, 1(2), 171203.
Bommer, J. J., Stafford, P. J., Alarcn, J. E., & Akkar, S. 2007. The inuence of magnitude
range on empirical ground-motion prediction. Bulletin of the Seismological Society of America, 97(6), 21522170.
Bommer, J. J., Stafford, P. J., & Alarcn, J. E. 2009. Empirical equations for the prediction
of the signicant, bracketed, and uniform duration of earthquake ground motion. Bulletin of
the Seismological Society of America, 99(6), 32173233.
407
Bommer, J. J., Douglas, J., Scherbaum, F., Cotton, F., Bungum, H., & Fh, D. 2010. On the
selection of ground-motion prediction equations for seismic hazard analysis. Seismological
Boore, D. M. 1983. Stochastic simulation of high-frequency ground motions based on seismological models of the radiated spectra. Bulletin of the Seismological Society of America,
73(6), 18651894.
Boore, D. M. 2003. Simulation of ground motion using the stochastic method. Pure and
Applied Geophysics, 160(34), 635676.
Boore, D. M. 2004. Estimating vs30 (or NEHRP site classes) from shallow velocity models
(depths 30 m). Bulletin of the Seismological Society of America, 94(2), 591U597.

Boore, D. M. 2005. Erratum: Equations for estimating horizontal response spectra and peak
acceleration from western north american earthquakes: A summary of recent work. Seismological Research Letters, 76(3), 368369.
Boore, D. M., & Atkinson, G. M. 1987. Stochastic prediction of ground motion and spectral
response parameters at hard-rock sites in eastern North America. Bulletin of the Seismological Society of America, 77(22), 440467.
Boore, D. M., & Atkinson, G. M. 2007. Boore-Atkinson NGA ground motion relations for the geometric mean horizontal component of peak and spectral ground motion parameters. PEER
Report 2007/01. Pacic Earthquake Engineering Research Center, College of Engineering,
University of California, Berkeley.
Boore, D. M., & Atkinson, G. M. 2008. Ground-motion prediction equations for the average
horizontal component of PGA, PGV, and 5%-damped PSA at spectral periods between
0.01 s and 10.0 s. Earthquake Spectra, 24(1), 99138.
Boore, D. M., & Joyner, W. B. 1982. The empirical prediction of ground motion. Bulletin of the
Seismological Society of America, 72(6), S43S60. Part B.
Boore, D. M., & Joyner, W. B. 1991. Estimation of ground motion at deep-soil sites in eastern
North America. Bulletin of the Seismological Society of America, 81(6), 21672185.
Boore, D. M., Joyner, W. B., & Fumal, T. E. 1993. Estimation of response spectra and peak accelerations from western North American earthquakes: An interim report. Open-File Report
93-509. U.S. Geological Survey. 70 pages.
Boore, D. M., Joyner, W. B., & Fumal, T. E. 1994a. Estimation of response spectra and peak
accelerations from western North American earthquakes: An interim report. Part 2. OpenFile Report 94-127. U.S. Geological Survey.
Boore, D. M., Joyner, W. B., & Fumal, T. E. 1994b. Ground motion estimates for strikeand reverse-slip faults. Provided to the Southern California Earthquake Center and widely
distributed as an insert in Boore et al. (1994a). Not seen. Reported in Boore et al. (1997).
Boore, D. M., Joyner, W. B., & Fumal, T. E. 1997. Equations for estimating horizontal response
spectra and peak acceleration from western North American earthquakes: A summary of
recent work. Seismological Research Letters, 68(1), 128153.
408
Boore, D. M., Watson-Lamprey, J., & Abrahamson, N. A. 2006. Orientation-independent measures of ground motion. Bulletin of the Seismological Society of America, 96(4A), 1502
1511.
Borcherdt, R. D. 1994. Estimates of site-dependent response spectra for design (methodology
and justication). Earthquake Spectra, 10(4), 617653.
Bouhadad, Y., Laouami, N., Bensalem, R., & Larbes, S. 1998. Seismic hazard estimation in
the central Tell Atlas of Algeria (Algiers-Kabylia). In: Proceedings of Eleventh European
Conference on Earthquake Engineering.
Bozorgnia, Y., & Campbell, K. W. 2004a. Engineering characterization of ground motion.
Chap. 5 of: Bozorgnia, Y., & Bertero, V. (eds), Earthquake Engineering: From Engineering
Seismology to Performance-Based Engineering. Boca Raton, FL: CRC Press.
Bozorgnia, Y., & Campbell, K. W. 2004b. The vertical-to-horizontal response spectral ratio and
tentative procedures for developing simplied V/H and the vertical design spectra. Journal
Bozorgnia, Y., Niazi, M., & Campbell, K. W. 1995. Characteristics of free-eld vertical ground
motion during the Northridge earthquake. Earthquake Spectra, 11(4), 515525.
Bozorgnia, Y., Campbell, K. W., & Niazi, M. 2000. Observed spectral characteristics of vertical
ground motion recorded during worldwide earthquakes from 1957 to 1995. In: Proceedings
of Twelfth World Conference on Earthquake Engineering. Paper No. 2671.
Bragato, P. L. 2004. Regression analysis with truncated samples and its application to groundmotion attenuation studies. Bulletin of the Seismological Society of America, 94(4), 1369
1378.
Bragato, P. L. 2005. Estimating an upper limit probability distribution for peak ground acceleration using the randomly clipped normal distribution. Bulletin of the Seismological Society
of America, 95(6), 20582065.
Bragato, P. L. 2009. Assessing regional and site-dependent variability of ground motions for
ShakeMap implementation in Italy. Bulletin of the Seismological Society of America, 99(5),
29502960.
Bragato, P. L., & Slejko, D. 2005. Empirical ground-motion attenuation relations for the eastern
Alps in the magnitude range 2.56.3. Bulletin of the Seismological Society of America,
95(1), 252276.
Brillinger, D. R., & Preisler, H. K. 1984. An exploratory analysis of the Joyner-Boore attenuation
data. Bulletin of the Seismological Society of America, 74(4), 14411450.
Brillinger, D. R., & Preisler, H. K. 1985. Further analysis of the Joyner-Boore attenuation data.
Bungum, H., Dahle, A., Toro, G., McGuire, R., & Gudmestad, O.T. 1992. Ground motions
from intraplate earthquakes. Pages 611616 of: Proceedings of Tenth World Conference
on Earthquake Engineering, vol. 2.
Cabalar, A. F., & Cevik, A. 2009. Genetic programming-based attenuation relationship: An
application of recent earthquakes in Turkey. Computers & Geosciences, 35, 18841896.
409
Cabaas, L., Benito, B., Cabaas, C, Lpez, M., Gmez, P., Jimnez, M. E., & Alvarez, S.
1999. Banco de datos de movimiento fuerte del suelo mfs. aplicaciones. Pages 111137
of: Complutense (ed), Fsica de la tierra, vol. 11. In Spanish with English abstract.
Cabaas, L., Lopez, M., Benito, B., & Jimnez, M. E. 2000 (Sep). Estimation of PGA attenuation laws for Spain and Mediterranean region. Comparison with other ground motion
models. In: Proceedings of the XXVII General Assembly of the European Seismological
Commission (ESC).
Caillot, V., & Bard, P. Y. 1993. Magnitude, distance and site dependent spectra from Italian
accelerometric data. European Earthquake Engineering, VII(1), 3748.
Campbell, K. W. 1981. Near-source attenuation of peak horizontal acceleration. Bulletin of
Campbell, K. W. 1985. Strong motion attenuation relations: A ten-year perspective. Earthquake Spectra, 1(4), 759804.
Campbell, K. W. 1989. The dependence of peak horizontal acceleration on magnitude, distance, and site effects for small-magnitude earthquakes in California and eastern North
America. Bulletin of the Seismological Society of America, 79(5), 13111346.
Campbell, K. W. 1990 (Sep). Empirical prediction of near-source soil and soft-rock ground
motion for the Diablo Canyon power plant site, San Luis Obispo county, California. Tech.
rept. Dames & Moore, Evergreen, Colorado. Prepared for Lawrence Livermore National
Laboratory. Not seen. Reported in Idriss (1993).
Campbell, K. W. 1993 (Jan). Empirical prediction of near-source ground motion from large
earthquakes. In: Proceedings of the International Workshop on Earthquake Hazard and
Large Dams in the Himalaya. Indian National Trust for Art and Cultural Heritage, New
Delhi, India.
Campbell, K. W. 1997. Empirical near-source attenuation relationships for horizontal and vertical components of peak ground acceleration, peak ground velocity, and pseudo-absolute
acceleration response spectra. Seismological Research Letters, 68(1), 154179.
Campbell, K. W. 2000. Erratum: Empirical near-source attenuation relationships for horizontal
and vertical components of peak ground acceleration, peak ground velocity, and pseudoabsolute acceleration response spectra. Seismological Research Letters, 71(3), 352354.
Campbell, K. W. 2001. Erratum: Empirical near-source attenuation relationships for horizontal
and vertical components of peak ground acceleration, peak ground velocity, and pseudoabsolute acceleration response spectra. Seismological Research Letters, 72(4), 474.
Campbell, K. W. 2003a. Engineering models of strong ground motion. Chap. 5 of: Chen,
W. F., & Scawthorn, C. (eds), Handbook of Earthquake Engineering. Boca Raton, FL, USA:
CRC Press.
Campbell, K. W. 2003b. Prediction of strong ground motion using the hybrid empirical method
and its use in the development of ground-motion (attenuation) relations in eastern North
America. Bulletin of the Seismological Society of America, 93(3), 10121033.
Campbell, K. W. 2003c. Strong-motion attenuation relations. Chap. 60 of: Lee, W. H. K.,
Kanamori, H., Jennings, P. C., & Kisslinger, C. (eds), International Handbook of Earthquake
and Engineering Seismology. London: Academic Press.
410
Campbell, K. W. 2007 (Sep). Validation and update of hybrid empirical ground motion (attenuation) relations for the CEUS. Final technical report. ABS Consulting, Inc. (EQECAT),
Beaverton, USA. NEHRP External Grants Program, U.S. Geological Survey Award Number:
05HQGR0032.
Campbell, K. W., & Bozorgnia, Y. 1994 (Jul). Near-source attenuation of peak horizontal
acceleration from worldwide accelerograms recorded from 1957 to 1993. Pages 283292
of: Proceedings of the Fifth U.S. National Conference on Earthquake Engineering, vol. III.
Campbell, K. W., & Bozorgnia, Y. 2000 (Nov). New empirical models for predicting near-source
horizontal, vertical, and V /H response spectra: Implications for design. In: Proceedings of
the Sixth International Conference on Seismic Zonation.
Campbell, K. W., & Bozorgnia, Y. 2003a. Erratum: Updated near-source ground-motion (attenuation) relations for the horizontal and vertical components of peak ground acceleration
and acceleration response spectra. Bulletin of the Seismological Society of America, 93(3),
1413.
Campbell, K. W., & Bozorgnia, Y. 2003b. Erratum: Updated near-source ground-motion (attenuation) relations for the horizontal and vertical components of peak ground acceleration
1872.
Campbell, K. W., & Bozorgnia, Y. 2003c. Erratum: Updated near-source ground-motion (attenuation) relations for the horizontal and vertical components of peak ground acceleration
2417.
Campbell, K. W., & Bozorgnia, Y. 2003d. Updated near-source ground-motion (attenuation)
relations for the horizontal and vertical components of peak ground acceleration and acceleration response spectra. Bulletin of the Seismological Society of America, 93(1), 314331.
Campbell, K. W., & Bozorgnia, Y. 2006a (Apr). Campbell-Bozorgnia Next Generation Attenuation (NGA) relations for PGA, PGV and spectral acceleration: A progress report. In:
Proceedings of the Eighth U.S. National Conference on Earthquake Engineering. Paper no.
906.
Campbell, K. W., & Bozorgnia, Y. 2006b (Sep). Next Generation Attenuation (NGA) empirical
ground motion models: Can they be used in Europe? In: Proceedings of First European
Conference on Earthquake Engineering and Seismology (a joint event of the 13th ECEE &
30th General Assembly of the ESC). Paper no. 458.
Campbell, K. W., & Bozorgnia, Y. 2007. Campbell-Bozorgnia NGA ground motion relations
for the geometric mean horizontal component of peak and spectral ground motion parameters. PEER Report 2007/02. Pacic Earthquake Engineering Research Center, College of
Engineering, University of California, Berkeley.
Campbell, K. W., & Bozorgnia, Y. 2008a. Empirical ground motion model for shallow crustal
earthquakes in active tectonic environments developed for the NGA project. In: Proceedings
of Fourteenth World Conference on Earthquake Engineering. Paper no. 03-02-0004.
Campbell, K. W., & Bozorgnia, Y. 2008b. NGA ground motion model for the geometric mean
horizontal component of PGA, PGV, PGD and 5% damped linear elastic response spectra
for periods ranging from 0.01 to 10 s. Earthquake Spectra, 24(1), 139171.
411
Carvalho, A. 2008. Modelao estocstica da aco ssmica em Portugal continental. Ph.D.

thesis, Instituto Superior Tcnico, Universidade Tcnica de Lisboa, Portugal. In Portuguese.
Not seen.
Cauzzi, C., & Faccioli, E. 2008. Broadband (0.05 to 20 s) prediction of displacement response
spectra based on worldwide digital records. Journal of Seismology, 12(4), 453475.
Cauzzi, C., Faccioli, E., Paolucci, R., & Villani, M. 2008. Long-period ground motion evaluation
from a large worldwide digital strong motion database. In: Proceedings of Fourteenth World
Conference on Earthquake Engineering. Paper no. S10-047.
Cauzzi, C. V. 2008 (Apr). Broadband empirical prediction of displacement response spectra
based on worldwide digital records. Ph.D. thesis, Politecnico di Milano.
Chang, T.-Y., Cotton, F., & Angelier, J. 2001. Seismic attenuation and peak ground acceleration in Taiwan. Bulletin of the Seismological Society of America, 91(5), 12291246.
Chapman, M. C. 1999. On the use of elastic input energy for seismic hazard analysis. Earthquake Spectra, 15(4), 607635.
Chen, Y., & Yu, Y.-X. 2008a. The development of attenuation relations in the rock sites for
periods (T = 0.04 10 s, = 0.005, 0.02, 0.07, 0.1 & 0.2) based on NGA database.
In: Proceedings of Fourteenth World Conference on Earthquake Engineering. Paper no.
03-02-0029.
Chen, Y., & Yu, Y.-X. 2008b. The development of attenuation relations in the rock sites for periods (T = 0.04 10 s, = 0.05) based on NGA database. In: Proceedings of Fourteenth
World Conference on Earthquake Engineering. Paper no. 03-02-0017.
Chen, Y.-H., & Tsai, C.-C. P. 2002. A new method for estimation of the attenuation relationship
with variance components. Bulletin of the Seismological Society of America, 92(5), 1984
1991.
Chiaruttini, C., & Siro, L. 1981. The correlation of peak ground horizontal acceleration with
magnitude, distance, and seismic intensity for Friuli and Ancona, Italy, and the Alpide belt.
Chiou, B., Youngs, R., Abrahamson, N., & Addo, K. 2010. Ground-motion attenuation model
for small-to-moderate shallow crustal earthquakes in California and its implications on regionalization of ground-motion prediction models. Earthquake Spectra, 26(4), 907926.
Chiou, B. S.-J., & Youngs, R. R. 2008. An NGA model for the average horizontal component
of peak ground motion and response spectra. Earthquake Spectra, 24(1), 173215.
Choi, Y., & Stewart, J. P. 2005. Nonlinear site amplication as function of 30 m shear wave
velocity. Earthquake Spectra, 21(1), 130.
Chou, C.-C., & Uang, C.-M. 2000. Establishing absorbed energy spectra an attenuation
approach. Earthquake Engineering and Structural Dynamics, 29(10), 14411455.
Chung, J.-K. 2006. Prediction of peak ground acceleration in southwestern Taiwan as revealed
by analysis of CHY array data. Terrestrial Atmospheric and Oceanic Science, 17(1), 139
167.
412
Climent, A., Taylor, W., Ciudad Real, M., Strauch, W., Villagrn, M., Dahle, A., & Bungum, H.
1994. Spectral strong motion attenuation in Central America. Tech. rept. 2-17. NORSAR.
Cole, S. W., Xu, Y., & Burton, P. W. 2008. Seismic hazard and risk in Shanghai and estimation
of expected building damage. Soil Dynamics and Earthquake Engineering, 28(1011), 778
794.
Collins, N., Graves, R., Ichinose, G., & Somerville, P. 2006. Ground motion attenuation relations for the Intermountain West. Final report. U.S. Geological Survey. Award
05HQGR0031.
Cornell, C. A., Banon, H., & Shakal, A. F. 1979. Seismic motion and response prediction
alternatives. Earthquake Engineering and Structural Dynamics, 7(4), 295315.
Costa, G., Suhadolc, P., & Panza, G. F. 1998. The Friuli (NE Italy) Accelerometric Network:
Analysis of low-magnitude high-quality digital accelerometric data for seismological and engineering applications. In: Proceedings of the Sixth U.S. National Conference on Earthquake Engineering. Oakland, USA: Earthquake Engineering Research Institute. Seattle,
USA. 31 May4 June.
Costa, G., Suhadolc, P., Delise, A., Moratto, L., Furlanetto, E., & Fitzko, F. 2006. Estimation of site effects at some stations of the Friuli (NE Italy) accelerometric network (RAF).
Pages 729739 of: Proceedings of Third International Symposium on the Effects of Surface
Geology on Seismic Motion, vol. 1. Paper number 089.
Cotton, F., Scherbaum, F., Bommer, J. J., & Bungum, H. 2006. Criteria for selecting and
adjusting ground-motion models for specic target regions: Application to central Europe
and rock sites. Journal of Seismology, 10(2), 137156.
Cotton, F., Pousse, G., Bonilla, F., & Scherbaum, F. 2008. On the discrepancy of recent
European ground-motion observations and predictions from empirical models: Analysis of
KiK-net accelerometric data and point-sources stochastic simulations. Bulletin of the Seismological Society of America, 98(5), 22442261.
Cousins, W. J., Zhao, J. X., & Perrin, N. D. 1999. A model for the attenuation of peak ground
acceleration in New Zealand earthquakes based on seismograph and accelerograph data.
Bulletin of the New Zealand Society for Earthquake Engineering, 32(4), 193220.
Crouse, C. B. 1991. Ground-motion attenuation equations for earthquakes on the Cascadia
subduction zones. Earthquake Spectra, 7(2), 201236.
Crouse, C. B., & McGuire, J. W. 1996. Site response studies for purpose of revising NEHRP
seismic provisions. Earthquake Spectra, 12(3), 407439.
Crouse, C. B., Vyas, Y. K., & Schell, B. A. 1988. Ground motion from subduction-zone earthquakes. Bulletin of the Seismological Society of America, 78(1), 125.
Cua, G., & Heaton, T. H. 2010. Characterizing average properties of southern California
ground motion amplitudes and envelopes. Bulletin of the Seismological Society of America.
Submitted.
Dahle, A., Bugum, H., & Kvamme, L. B. 1990a. Attenuation modelling based on intraplate
earthquake recordings. Pages 121129 of: Proceedings of Ninth European Conference on
Earthquake Engineering, vol. 4-A.
413
Dahle, A., Bungum, H., & Kvamme, L. B. 1990b. Attenuation models inferred from intraplate
earthquake recordings. Earthquake Engineering and Structural Dynamics, 19(8), 1125
1141.
Dahle, A., Bungum, H., & Kvamme, L. B. 1991. Empirically derived PSV spectral attenuation
models for intraplate conditions. European Earthquake Engineering, 3, 4252.
Dahle, A., Climent, A., Taylor, W., Bungum, H., Santos, P., Ciudad Real, M., Linholm, C.,
Strauch, W., & Segura, F. 1995. New spectral strong motion attenuation models for Central America. Pages 10051012 of: Proceedings of the Fifth International Conference on
Seismic Zonation, vol. II.
Danciu, L., & Tselentis, G.-A. 2007a. Engineering ground-motion parameters attenuation relationships for Greece. Bulletin of the Seismological Society of America, 97(1B), 162183.
Danciu, L., & Tselentis, G.-A. 2007b (Apr). Engineering ground-motion parameters attenuation
relationships for Greece. Pages 327334 of: Proceedings of the International Symposium
on Seismic Risk Reduction: The JICA Technical Cooperation Project in Romania. Paper ID
26.
Das, S., Gupta, I. D., & Gupta, V. K. 2002. A new attenuation model for north-east India. Pages
151158 of: Proceedings of the Twelfth Symposium of Earthquake Engineering, Roorkee,
India.
Davenport, A. J. 1972. A statistical relationship between shock amplitude, magnitude, and
epicentral distance and its appplication to seismic zoning. Engineering Science Research
Report BLWT-4-72. Western Ontario University. Not seen. Cited in Hays (1980).
De Natale, G., Faccioli, E., & Zollo, A. 1988. Scaling of peak ground motions from digital
recordings of small earthquakes at Campi Flegrei, southern Italy. Pure and Applied Geophysics, 126(1), 3753.
Denham, D., & Small, G. R. 1971. Strong motion data centre: Bureau of mineral resources,
Canada. Bulletin of the New Zealand Society for Earthquake Engineering, 4(1), 1530.
Denham, D., Small, G. R., & Everingham, I.B. 1973. Some strong-motion results from Papua
New Guinea 19671972. Pages 23242327 of: Proceedings of Fifth World Conference on
Earthquake Engineering, vol. 2.
Devillers, C., & Mohammadioun, B. 1981. French methodology for determining site-adapted
SMS (Sisme Major de Scurit) spectra. In: Transactions of the 6th International Conference on Structural Mechanics in Reactor Technology, vol. K(a). K 1/9.
Dhakal, Y. P., Takai, N., & Sasatani, T. 2008. Path effects on prediction equations of pseudovelocity response spectra in northern Japan. In: Proceedings of Fourteenth World Conference on Earthquake Engineering. Paper no. 03-02-0023.
Donovan, N. C. 1973. A statistical evaluation of strong motion data including the February 9,
1971 San Fernando earthquake. Pages 12521261 of: Proceedings of Fifth World Conference on Earthquake Engineering, vol. 1.
Donovan, N. C., & Bornstein, A. E. 1978. Uncertainties in seismic risk analysis. Journal of the
Geotechnical Engineering Division, ASCE, 104(GT7), 869887.
414
Douglas, J. 2001a (Jan). A comprehensive worldwide summary of strong-motion attenuation

relationships for peak ground acceleration and spectral ordinates (1969 to 2000). ESEE
Report 01-1. Department of Civil and Environmental Engineering, Imperial College, London.
Douglas, J. 2001b (Oct). A critical reappraisal of some problems in engineering seismology.
Ph.D. thesis, University of London.
Douglas, J. 2002 (Oct). Errata of and additions to ESEE Report No. 01-1: A comprehensive
worldwide summary of strong-motion attenuation relationships for peak ground acceleration
and spectral ordinates (1969 to 2000). Dept. report. Department of Civil and Environmental
Engineering, Imperial College, London.
Douglas, J. 2003. Earthquake ground motion estimation using strong-motion records: A review
of equations for the estimation of peak ground acceleration and response spectral ordinates.
Earth-Science Reviews, 61(12), 43104.
Douglas, J. 2004a (Jan.). Ground motion estimation equations 19642003: Reissue of ESEE
Report No. 01-1: A comprehensive worldwide summary of strong-motion attenuation relationships for peak ground acceleration and spectral ordinates (1969 to 2000) with corrections and additions. Tech. rept. 04-001-SM. Department of Civil and Environmental Engineering; Imperial College of Science, Technology and Medicine; London; U.K.
Douglas, J. 2004b. An investigation of analysis of variance as a tool for exploring regional
differences in strong ground motions. Journal of Seismology, 8(4), 485496.
Douglas, J. 2006 (Dec). Errata of and additions to Ground motion estimation equations 1964
2003. Intermediary report RP-54603-FR. BRGM, Orlans, France.
Douglas, J. 2007. On the regional dependence of earthquake response spectra. ISET Journal
of Earthquake Technology, 44(1), 7199.
Douglas, J. 2008 (Dec). Further errata of and additions to Ground motion estimation equations
19642003. Final report RP-56187-FR. BRGM, Orlans, France.
Douglas, J. 2010a. Assessing the epistemic uncertainty of ground-motion predictions. In:
Proceedings of the Ninth U.S. National and 10th Canadian Conference on Earthquake Engineering: Reaching Beyond Borders. Paper no. 219.
Douglas, J. 2010b. Consistency of ground-motion predictions from the past four decades.
Bulletin of Earthquake Engineering, 8(6), 15151526.
Douglas, J., & Aochi, H. 2008. A survey of techniques for predicting earthquake ground
motions for engineering purposes. Surveys in Geophysics, 29(3), 187220.
Douglas, J., & Boore, D. M. 2011. High-frequency ltering of strong-motion records. Bulletin
of Earthquake Engineering, 9. In press.
Douglas, J., & Halldrsson, H. 2010. On the use of aftershocks when deriving ground-motion
prediction equations. In: Proceedings of the Ninth U.S. National and 10th Canadian Conference on Earthquake Engineering: Reaching Beyond Borders. Paper no. 220.
Douglas, J., & Smit, P. M. 2001. How accurate can strong ground motion attenuation relations
be? Bulletin of the Seismological Society of America, 91(6), 19171923.
415
Douglas, J., Bungum, H., & Scherbaum, F. 2006. Ground-motion prediction equations for
southern Spain and southern Norway obtained using the composite model perspective.
Journal of Earthquake Engineering, 10(1), 3372.
Dowrick, D. J., & Sritharan, S. 1993. Attenuation of peak ground accelerations in some recent
New Zealand earthquakes. Bulletin of the New Zealand National Society for Earthquake
Engineering, 26(1), 313. Not seen. Reported in Stafford (2006).
Eberhart-Phillips, D., & McVerry, G. 2003. Estimating slab earthquake response spectra from
a 3D Q model. Bulletin of the Seismological Society of America, 93(6), 26492663.
El Hassan, Y. M. 1994. Structural response to earthquake ground motion in Sudan. M.Phil.
thesis, University of Khartoum, Sudan. Not seen. Reported in Abdalla et al. (2001).
Electric Power Research Institute. 1988. Engineering model of earthquake ground motion for
eastern North America. Final report NP-6074. Research project 2556-16. Investigators:
McGuire, R. K., Toro, G. R., W. J. Silva.
Electric Power Research Institute. 1993a. Empirical ground motion data in eastern North
America. In: Schneider, J. F. (ed), Guidelines for determining design basis ground motions,
vol. EPRI TR-102293.
Electric Power Research Institute. 1993b. Engineering model of strong ground motions from
earthquakes in the central and eastern United States. In: Schneider, J. F. (ed), Guidelines
for determining design basis ground motions, vol. EPRI TR-102293.
Electric Power Research Institute. 2004 (Dec). CEUS ground motion project nal report. Tech.
rept. 1009684. EPRI, Palo Alto, CA, Dominion Energy, Glen Allen, VA, Entergy Nuclear,
Jackson, MS, and Exelon Generation Company, Kennett Square, PA.
Espinosa, A. F. 1980. Attenuation of strong horizontal ground accelerations in the western
United States and their relation to ML . Bulletin of the Seismological Society of America,
70(2), 583616.
Esteva, L. 1970. Seismic risk and seismic design. Pages 142182 of: Hansen, R.J. (ed),
Seismic Design for Nuclear Power Plants. The M.I.T. Press.
Esteva, L. 1974. Geology and probability in the assessment of seismic risk. In: Proceedings
of the 2nd International Conference of the Association of Engineering Geology. Not seen.
Reported in Ambraseys (1978a).
Esteva, L., & Rosenblueth, E. 1964. Espectros de temblores a distancias moderadas y
grandes. Boletin Sociedad Mexicana de Ingenieria Sesmica, 2, 118. In Spanish.
Esteva, L., & Villaverde, R. 1973. Seismic risk, design spectra and structural reliability. Pages
25862596 of: Proceedings of Fifth World Conference on Earthquake Engineering, vol. 2.
Faccioli, E. 1978 (Jun). Response spectra for soft soil sites. Pages 441456 of: Proceedings of the ASCE Geotechnical Engineering Division Speciality Conference: Earthquake
Engineering and Soil Dynamics, vol. I.
Faccioli, E. 1979. Engineering seismic risk analysis of the Friuli region. Bollettino di Geosica
Teorica ed Applicata, XXI(83), 173190.
416
Faccioli, E., & Agalbato, D. 1979. Attenuation of strong-motion parameters in the 1976 Friuli,
Italy, earthquakes. Pages 233242 of: Proceedings of the Second U.S. National Conference
on Earthquake Engineering.
Faccioli, E., Bianchini, A., & Villani, M. 2010 (Mar). New ground motion prediction equations for
T > 1 s and their inuence on seismic hazard assessment. In: Proceedings of the University
of Tokyo Symposium on Long-Period Ground Motion and Urban Disaster Mitigation.
Fajfar, P., & Perus, I. 1997. A non-parametric approach to attenuation relations. Journal of
Field, E. H. 2000. A modied ground-motion attenuation relationship for southern California
that accounts for detailed site classication and a basin-depth effect. Bulletin of the Seismological Society of America, 90(6B), S209S221.
Frankel, A., Mueller, C., Barnhard, T., Perkins, D., Leyendecker, E. V., Dickman, N., Hanson,
S., & Hopper, M. 1996. National Seismic-Hazard Maps: Documentation June 1996. OpenFile Report 96-532. U.S. Department of the Interior, U.S. Geological Survey.
Free, M. W. 1996. The attenuation of earthquake strong-motion in intraplate regions. Ph.D.
thesis, University of London.
Free, M. W., Ambraseys, N. N., & Sarma, S. K. 1998 (Feb). Earthquake ground-motion attenuation relations for stable continental intraplate regions. ESEE Report 98-1. Department of
Civil Engineering, Imperial College, London.
Frisenda, M., Massa, M., Spallarossa, D., Ferretti, G., & Eva, C. 2005. Attenuation relationships for low magnitude earthquakes using standard seismometric records. Journal of
Frohlich, C., & Apperson, K. D. 1992. Earthquake focal mechanisms, moment tensors, and
the consistency of seismic activity near plate boundaries. Tectonics, 11(2), 279296.
Fukushima, Y., & Tanaka, T. 1990. A new attenuation relation for peak horizontal acceleration of strong earthquake ground motion in Japan. Bulletin of the Seismological Society of
America, 80(4), 757783.
Fukushima, Y., Tanaka, T., & Kataoka, S. 1988. A new attenuation relationship for peak ground
acceleration derived from strong-motion accelerograms. Pages 343348 of: Proceedings
of Ninth World Conference on Earthquake Engineering, vol. II.
Fukushima, Y., Gariel, J.-C., & Tanaka, R. 1994. Prediction relations of seismic motion parameters at depth using borehole data. Pages 417422 of: Proceedings of Tenth European
Conference on Earthquake Engineering, vol. 1.
Fukushima, Y., Gariel, J.-C., & Tanaka, R. 1995. Site-dependent attenuation relations of
seismic motion parameters at depth using borehole data. Bulletin of the Seismological
Fukushima, Y., Berge-Thierry, C., Volant, P., Griot-Pommera, D.-A., & Cotton, F. 2003. Attenuation relation for western Eurasia determined with recent near-fault records from California,
Japan and Turkey. Journal of Earthquake Engineering, 7(4), 573598.
417
Fukushima, Y., Zhao, J. X., Zhang, J., Asano, A., Ohno, Y., Oouchi, T., Takahashi, T., Ogawa,
H., Irikura, K., Fukushima, Y., Thio, H. K., & Somerville, P. G. 2006. Attenuation relations
of strong ground motion in Japan using site classication based on predominant period. In:
Proceedings of First European Conference on Earthquake Engineering and Seismology (a
joint event of the 13th ECEE & 30th General Assembly of the ESC). Paper number 683.
Fukushima, Y., Bonilla, F., Scotti, O., & Douglas, J. 2007a (Jul). Impact of site classication
on deriving empirical ground-motion prediction equations: Application to the west Eurasia
dataset. In: 7me Colloque National AFPS 2007.
Fukushima, Y., Bonilla, L. F., Scotti, O., & Douglas, J. 2007b. Site classication using
horizontal-to-vertical response spectral ratios and its impact when deriving empirical ground
motion prediction equations. Journal of Earthquake Engineering, 11(5), 712724.
Garca, D., Singh, S. K., Herriz, M., Ordaz, M., & Pacheco, J. F. 2005. Inslab earthquakes
of central Mexico: Peak ground-motion parameters and response spectra. Bulletin of the
Garcia, S., & Romo, M. 2006 (Sep). Machine learning for ground-motion relations. In: Proceedings of First European Conference on Earthquake Engineering and Seismology (a joint
event of the 13th ECEE & 30th General Assembly of the ESC). Paper no. 1438.
Garca-Fernndez, M., & Canas, J. A. 1991. Estimation of regional values of peak ground
acceleration from short-period seismograms. Pages 533539 of: Proceedings of the Fourth
International Conference on Seismic Zonation, vol. II.
Garcia-Fernandez, M., & Canas, J.A. 1992. Regional Lg-wave attenuation and estimates of
peak ground acceleration in the Iberian peninsula. Pages 439443 of: Proceedings of Tenth
World Conference on Earthquake Engineering, vol. 1.
Garcia-Fernandez, M., & Canas, J.A. 1995. Regional peak ground acceleration estimates in
the Iberian peninsula. Pages 10291034 of: Proceedings of the Fifth International Conference on Seismic Zonation, vol. II.
Gaull, B. A. 1988. Attenuation of strong ground motion in space and time in southwest Western Australia. Pages 361366 of: Proceedings of Ninth World Conference on Earthquake
Geomatrix Consultants. 1991 (Mar). Seismic ground motion study for West San Francisco
Bay Bridge. Report for caltrans, division of structures, sacramento, california. Not seen.
Cited in Idriss (1993).
Ghasemi, H., Zare, M., Fukushima, Y., & Koketsu, K. 2009. An empirical spectral groundmotion model for Iran. Journal of Seismology, 13(4), 499515.
Ghodrati Amiri, G., Mahdavian, A., & Dana, F. M. 2007a. Attenuation relationships for Iran.
Journal of Earthquake Engineering, 11(4), 469U492.

Ghodrati Amiri, G., Mahdavian, A., & Dana, F. M. 2007b. Response on the Discussion of
Attenuation relationships for Iran. Journal of Earthquake Engineering, 11(6), 1036U1037.

Ghodrati Amiri, G., Khorasani, M., Mirza Hessabi, M., & Razavian Amrei, S. A. 2010. Groundmotion prediction equations of spectral ordinates and Arias intensity for Iran. Journal of
418
Gitterman, Y., Zaslavsky, Y., & Shapira, A. 1993 (Sep). Analysis of strong records in Israel. In:
Proceedings of XVIIth regional European seminar on earthquake engineering, Haifa, Israel.
Not seen.
Goda, K., & Hong, H. P. 2008. Spatial correlation of peak ground motions and response
spectra. Bulletin of the Seismological Society of America, 98(1), 354365.
Gmez-Sobern, C., Tena-Colunga, A., & Ordaz, M. 2006 (Apr). Updated attenuation laws
in displacement and acceleration for the Mexican Pacic coast as the rst step to improve
current design spectra for base-isolated structures in Mexico. In: Proceedings of the Eighth
U.S. National Conference on Earthquake Engineering. Paper no. 1010.
Graizer, V., & Kalkan, E. 2007. Ground motion attenuation model for peak horizontal acceleration from shallow crustal earthquakes. Earthquake Spectra, 23(3), 585613.
Graizer, V., & Kalkan, E. 2008. A novel approach to strong ground motion attenuation modeling. In: Proceedings of Fourteenth World Conference on Earthquake Engineering. Paper
no. 02-0022.
Graizer, V., Kalkan, E., & Lin, K.-W. 2010. Extending and testing Graizer-Kalkan ground motion
attenuation model based on Atlas database of shallow crustal events. In: Proceedings of the
Ninth U.S. National and 10th Canadian Conference on Earthquake Engineering: Reaching
Beyond Borders. Paper no. 568.
Gregor, N., Silva, W., & Darragh, B. 2002a (Jun.). Development of attenuation relations for
peak particle velocity and displacement. A pearl report to pg&e/cec/caltrans. Pacic Engineering and Analysis, El Cerrito, U.S.A.
Gregor, N. J., Silva, W. J., Wong, I. G., & Youngs, R. R. 2002b. Ground-motion attenuation
relationships for Cascadia subduction zone megathrust earthquakes based on a stochastic
nite-fault model. Bulletin of the Seismological Society of America, 92(5), 19231932.
Glkan, P., & Kalkan, E. 2002. Attenuation modeling of recent earthquakes in Turkey. Journal
of Seismology, 6(3), 397409.
Gll, H., & Erelebi, E. 2007. A neural network approach for attenuation relationships: An
application using strong ground motion data from Turkey. Engineering Geology, 93(34),
6581.
Gnaydn, K., & Gnaydn, A. 2008. Peak ground acceleration prediction by artical neural networks for northwestern Turkey. Mathematical Problems in Engineering. Article ID
919420.
Gupta, I. D. 2010. Response spectral attenuation relations for in-slab earthquakes in IndoBurmese subduction zone. Soil Dynamics and Earthquake Engineering, 30(5), 368377.
Gupta, S., & Gupta, I. D. 2004. The prediction of earthquake peak ground acceleration in
Koyna region, India. In: Proceedings of Thirteenth World Conference on Earthquake Engineering. Paper no. 1437.
Gusev, A. A., Gordeev, E. I., Guseva, E. M., Petukhin, A. G., & Chebrov, V. N. 1997. The
rst version of the Amax (mw , r) relationship for Kamchatka. Pure and Applied Geophysics,
149(2), 299312.
419
Halldorsson, B., & Papageorgiou, A. S. 2005. Calibration of the specic barrier model to
earthquakes of different tectonic regions. Bulletin of the Seismological Society of America,
95(4), 12761300.
Halldrsson, P., & Sveinsson, B. I. 2003 (Aug). Dvnun hrunar slandi. Tech. rept. 03025.
Veurstofa slands (Icelandic Meteorological Ofce).
Hamzehloo, H., & Bahoosh, H. R. 2010. Theoretical spectral attenuation relationship for
Tehran region, Iran. In: Proceedings of Fourteenth European Conference on Earthquake
Engineering. Paper no. 821.
Hancock, J., & Bommer, J. J. 2005. The effective number of cycles of earthquake ground
motion. Earthquake Engineering and Structural Dynamics, 34, 637664.
Hao, H., & Gaull, B. A. 2009. Estimation of strong seismic ground motion for engineering use
in Perth Western Australia. Soil Dynamics and Earthquake Engineering, 29(5), 909924.
Hasegawa, H. S., Basham, P. W., & Berry, M. J. 1981. Attenuation relations for strong seismic
ground motion in Canada. Bulletin of the Seismological Society of America, 71(6), 1943
1962.
Hays, W. W. 1980. Procedures for estimating earthquake ground motions. Geological Survey
Professional Paper 1114. US Geological Survey.
Herak, M., Markusic, S., & Ivancic, I. 2001. Attenuation of peak horizontal and vertical acceleration in the Dinarides area. Studia Geophysica et Geodaetica, 45(4), 383394.
Hernandez, B., Fukushima, Y., Bossu, R., & Albaric, J. 2006. Seismic attenuation relation for
Hualien (Taiwan) at the free surface and down to 52.6 m deep. Pages 145154 of: Proceedings of Third International Symposium on the Effects of Surface Geology on Seismic
Motion, vol. 1. Paper number 008.
Herrmann, R. B., & Goertz, M. J. 1981. A numerical study of peak ground motion scaling.
Herrmann, R. B., & Nuttli, O. W. 1984. Scaling and attenuation relations for strong ground motion in eastern North America. Pages 305309 of: Proceedings of Eighth World Conference
on Earthquake Engineering, vol. II.
Hong, H. P., & Goda, K. 2007. Orientation-dependent ground-motion measure for seismichazard assessment. Bulletin of the Seismological Society of America, 97(5), 15251538.
Hong, H. P., & Goda, K. 2010. Characteristics of horizontal ground motion measures along
principal directions. Earthquake Engineering and Engineering Vibration, 9(1), 922.
Hong, H. P., Zhang, Y., & Goda, K. 2009a. Effect of spatial correlation on estimated groundmotion prediction equations. Bulletin of the Seismological Society of America, 99(2A), 928
934.
Hong, H. P., Pozos-Estrada, A., & Gomez, R. 2009b. Orientation effect on ground motion measurements for Mexican subduction earthquakes. Earthquake Engineering and Engineering
Vibration, 8(1), 116.
420
Humbert, N., & Viallet, E. 2008. An evaluation of epistemic and random uncertainties included
in attenuation relationship parameters. In: Proceedings of Fourteenth World Conference on
Earthquake Engineering. Paper no. 07-0117.
Huo, J., & Hu, Y. 1991. Attenuation laws considering the randomness of magnitude and
distance. Earthquake Research in China, 5(1), 1736.
Huo, J., Hu, Y., & Feng, Q. 1992. Study on estimation of ground motion from seismic intensity.
Earthquake Engineering and Engineering Vibration, 12(3), 115. In Chinese. Not seen.
Hwang, H., & Huo, J.-R. 1997. Attenuation relations of ground motion for rock and soil sites in
eastern United States. Soil Dynamics and Earthquake Engineering, 16(6), 363372.
Idriss, I. M. 1978 (Jun). Characteristics of earthquake ground motions. Pages 11511265
of: Proceedings of the ASCE Geotechnical Engineering Division Speciality Conference:
Earthquake Engineering and Soil Dynamics, vol. III.
Idriss, I. M. 1993. Procedures for selecting earthquake ground motions at rock sites. Tech.
rept. NIST GCR 93-625. National Institute of Standards and Technology.
Idriss, I. M. 2008. An NGA empirical model for estimating the horizontal spectral values generated by shallow crustal earthquakes. Earthquake Spectra, 24(1), 217242.
Inan, E., Colakoglu, Z., Koc, N., Baylke, N., & Coruh, E. 1996. Earthquake catalogs with
acceleration records from 1976 to 1996. Tech. rept. General Directorate of Disaster Affairs, Earthquake Research Department, Ankara, Turkey. In Turkish. Not seen. Reported in
Ulusay et al. (2004).
Iwasaki, T., Kawashima, K., & Saeki, M. 1980. Effects of seismic and geotechnical conditions
on maximum ground accelerations and response spectra. Pages 183190 of: Proceedings
of Seventh World Conference on Earthquake Engineering, vol. 2.
Iyengar, R. N., & Raghu Kanth, S. T. G. 2004. Attenuation of strong ground motion in peninsular India. Seismological Research Letters, 75(4), 530540.
Jacob, K. H., Gariel, J.-C., Armbruster, J., Hough, S., Friberg, P., & Tuttle, M. 1990 (May).
Site-specic ground motion estimates for New York City. Pages 587596 of: Proceedings
of the Fourth U.S. National Conference on Earthquake Engineering, vol. 1.
Jain, S. K., Roshan, A. D., Arlekar, J. N., & Basu, P. C. 2000 (Nov). Empirical attenuation
relationships for the Himalayan earthquakes based on Indian strong motion data. In: Proceedings of the Sixth International Conference on Seismic Zonation.
Jara-Guerrero, J. M., Jara-Diaz, M., & Hernndez, H. 2007 (Apr). Estimation of the pseudoacceleration response spectra in sites of Mexico. Pages 343350 of: Proceedings of
the International Symposium on Seismic Risk Reduction: The JICA Technical Cooperation
Project in Romania. Paper ID 13.
Jayaram, N., & Baker, J. W. 2010. Considering spatial correlation in mixed-effects regression
and the impact on ground-motion models. Bulletin of the Seismological Society of America,
100(6), 32953303.
Jeon, Y.-S., & Herrmann, R. B. 2004. High-frequency earthquake ground-motion scaling in
Utah and Yellowstone. Bulletin of the Seismological Society of America, 94(5), 16441657.
421
Jin, X., Kang, L.-C., & Ou, Y.-P. 2008. Ground motion attenuation relation for small to moderate
earthquakes in Fujian region, China. Acta Seismologica Sinica, 21(3), 283295.
Johnson, R. A. 1973. An earthquake spectrum prediction technique. Bulletin of the Seismological Society of America, 63(4), 12551274.
Johnston, A., et al. . 1994. The earthquakes of stable continental regions, Vol. 1: Assessment
of large earthquake potential. EPRI Report TR-102261. Electrical Power Research Institute,
Palo Alto.
Joyner, W. B., & Boore, D. M. 1981. Peak horizontal acceleration and velocity from strongmotion records including records from the 1979 Imperial Valley, California, earthquake. Bulletin of the Seismological Society of America, 71(6), 20112038.
Joyner, W. B., & Boore, D. M. 1982a. Estimation of response-spectral values as functions of
magnitude, distance, and site conditions. Open-File Report 82-881. U.S. Geological Survey.
Joyner, W. B., & Boore, D. M. 1982b. Prediction of earthquake response spectra. Open-File
Report 82-977. U.S. Geological Survey.
Joyner, W. B., & Boore, D. M. 1988. Measurement, characterization, and prediction of strong
ground motion. Pages 43102 of: Proceedings of Earthquake Engineering & Soil Dynamics
II. Geotechnical Division, ASCE.
Joyner, W. B., & Boore, D. M. 1993. Methods for regression analysis of strong-motion data.
Joyner, W. B., & Boore, D. M. 1996 (Aug). Recent developments in strong motion attenuation
relationships. Pages 101116 of: Proceedings of the 28th Joint Meeting of the U.S.-Japan
Cooperative Program in Natural Resource Panel on Wind and Seismic Effects.
Joyner, W. B., & Fumal, T. E. 1984. Use of measured shear-wave velocity for predicting
geologic site effects on strong ground motion. Pages 777783 of: Proceedings of Eighth
World Conference on Earthquake Engineering, vol. II.
Joyner, W. B., & Fumal, T. E. 1985. Predictive mapping of earthquake ground motion. Pages
203220 of: Evaluating Earthquake Hazards in the Los Angeles Region An Earth Science
Perspective. U.S. Geological Survey Professional Paper, no. 1360. Washington: United
States Government Printing Ofce.
Junn, J.-G., Jo, N.-D., & Baag, C.-E. 2002. Stochastic prediction of ground motions in southern
Korea. Geosciences Journal, 6(3), 203214.
Kalkan, E., & Glkan, P. 2004a. Empirical attenuation equations for vertical ground motion in
Turkey. Earthquake Spectra, 20(3), 853882.
Kalkan, E., & Glkan, P. 2004b. Site-dependent spectra derived from ground motion records
in Turkey. Earthquake Spectra, 20(4), 11111138.
Kalkan, E., & Glkan, P. 2005. Erratum: Site-dependent spectra derived from ground motion
records in Turkey. Earthquake Spectra, 21(1), 283.
Kamiyama, M. 1989 (Oct). Regression analyses of strong-motion spectra in terms of a simplied faulting source model. Pages 113126 of: Proceedings of the Fourth International
Conference on Soil Dynamics and Earthquake Engineering.
422
Kamiyama, M. 1995. An attenuation model for the peak values of strong ground motions with
emphasis on local soil effects. Pages 579585 of: Proceedings of the First International
Conference on Earthquake Geotechnical Engineering, vol. 1.
Kamiyama, M., & Yanagisawa, E. 1986. A statisical model for estimating response spectra
of strong earthquake ground motions with emphasis on local soil conditions. Soils and
Foundations, 26(2), 1632.
Kamiyama, M., ORourke, M.J., & Flores-Berrones, R. 1992 (Sep). A semi-empirical analysis
of strong-motion peaks in terms of seismic source, propagation path and local site conditions. Tech. rept. NCEER-92-0023. National Center for Earthquake Engineering Research.
Kanai, K. 1966. Improved empirical formula for characteristics of stray [sic] earthquake motions. Pages 14 of: Proceedings of the Japanese Earthquake Symposium. Not seen.
Reported in Trifunac & Brady (1975).
Kang, L., & Jin, X. 2009. Ground motion attenuation relations of small and moderate earthquakes in Sichuan region. Earthquake Science, 22, 277282.
Kanno, T., Narita, A., Morikawa, N., Fujiwara, H., & Fukushima, Y. 2006. A new attenuation
relation for strong ground motion in Japan based on recorded data. Bulletin of the Seismological Society of America, 96(3), 879897.
Katayama, T. 1982. An engineering prediction model of acceleration response spectra and its
application to seismic hazard mapping. Earthquake Engineering and Structural Dynamics,
10(1), 149163.
Kawano, H., Takahashi, K., Takemura, M., Tohdo, M., Watanabe, T., & Noda, S. 2000. Empirical response spectral attenuations on the rocks with VS = 0.5 to 3.0 km/s in Japan. In:
Proceedings of Twelfth World Conference on Earthquake Engineering. Paper No. 0953.
Kawashima, K., Aizawa, K., & Takahashi, K. 1984. Attenuation of peak ground motion and
absolute acceleration response spectra. Pages 257264 of: Proceedings of Eighth World
Conference on Earthquake Engineering, vol. II.
Kawashima, K., Aizawa, K., & Takahashi, K. 1985. Attenuation of peak ground motions and
absolute acceleration response spectra of vertical earthquake ground motion. Proceedings
of JSCE Structural Engineering/Earthquake Engineering, 2(2), 415422.
Kawashima, K., Aizawa, K., & Takahashi, K. 1986. Attenuation of peak ground acceleration,
velocity and displacement based on multiple regression analysis of Japanese strong motion
records. Earthquake Engineering and Structural Dynamics, 14(2), 199215.
Khademi, M. H. 2002 (Sep). Attenuation of peak and spectral accelerations in the Persian
plateau. In: Proceedings of Twelfth European Conference on Earthquake Engineering.
Paper reference 330.
Kobayashi, H., & Midorikawa, S. 1982. A semi-empirical method for estimating response spectra of near-eld ground motions with regard to fault rupture. Pages 161168 of: Proceedings
of Seventh European Conference on Earthquake Engineering, vol. 2.
Kobayashi, H., & Nagahashi, S. 1977. Response spectra on seismic bedrock during earthquake. Pages 516522 of: Proceedings of Sixth World Conference on Earthquake Engineering, vol. I.
423
Kobayashi, S., Takahashi, T., Matsuzaki, S., Mori, M., Fukushima, Y., Zhao, J. X., &
Somerville, P. G. 2000. A spectral attenuation model for Japan using digital strong motion records of JMA87 type. In: Proceedings of Twelfth World Conference on Earthquake
Engineering. Paper No. 2786.
Krinitzsky, E. L., Chang, F. K., & Nuttli, O. W. 1987 (Sep). State-of-the-art for assessing earthquake hazards in the United States; report 26, Parameters for specifying magnitude-related
earthquake ground motions. Tech. rept. U.S. Army Engineer Waterways Experimental Station. Miscellaneous paper S-73-1.
Krinitzsky, E. L., Chang, F. K., & Nuttli, O. W. 1988. Magnitude-related earthquake ground
motions. Bulletin of the Association of Engineering Geologists, XXV(4), 399423.
Kuehn, N. M., Scherbaum, F., & Riggelsen, C. 2009. Deriving empirical ground-motion models: Balancing data constraints and physical assumptions to optimize prediction capability.
Laouami, N., Slimani, A., Bouhadad, Y., Chatelain, J.-L., & Nour, A. 2006. Evidence for faultrelated directionality and localized site effects from strong motion recordings of the 2003
Boumerdes (Algeria) earthquake: Consequences on damage distribution and the Algerian
seismic code. Soil Dynamics and Earthquake Engineering, 26(11), 9911003.
Lawson, R. S., & Krawinkler, H. 1994. Cumulative damage potential of seismic ground motion. Pages 10791086 of: Proceedings of Tenth European Conference on Earthquake
Lee, V. W. 1987 (Jul). Inuence of local soil and geological site conditions on pseudo relative
velocity spectrum amplitudes of recorded strong motion accelerations. Tech. rept. CE 87-06.
Department of Civil Engineering, University of Southern California, Los Angeles, California,
U.S.A.
Lee, V. W. 1993. Scaling PSV from earthquake magnitude, local soil, and geologic depth of
sediments. Journal of the Geotechnical Engineering Division, ASCE, 119(1), 108126.
Lee, V. W. 1995. Pseudo relative velocity spectra in former Yugoslavia. European Earthquake
Engineering, IX(1), 1222.
Lee, V. W., & Manic, M. 1994. Empirical scaling of response spectra in former Yugoslavia.
Pages 25672572 of: Proceedings of Tenth European Conference on Earthquake Engineering, vol. 4.
Lee, V. W., & Trifunac, M. D. 1995 (May). Pseudo relative velocity spectra of strong earthquake ground motion in California. Tech. rept. CE 95-04. Department of Civil Engineering,
University of Southern California, Los Angeles, California, U.S.A.
Lee, V. W., Trifunac, M. D., Todorovska, M. I., & Novikova, E. I. 1995 (Apr). Empirical equations
describing attenuation of peak of strong ground motion, in terms of magnitude, distance,
path effects and site conditions. Tech. rept. CE 95-02. Department of Civil Engineering,
Liang, J., Hao, H., Gaull, B. A., & Sinadinovski, C. 2008. Estimation of strong ground motions
in southwest Western Australia with a combined Greens function and stochastic approach.
Journal of Earthquake Engineering, 12(3), 382405.
424
Lin, P.-S., & Lee, C.-T. 2008. Ground-motion attenuation relationships for subduction-zone
earthquakes in northeastern Taiwan. Bulletin of the Seismological Society of America, 98(1),
220240.
Liu, K.-S., & Tsai, Y.-B. 2005. Attenuation relationships of peak ground acceleration and
velocity for crustal earthquakes in Taiwan. Bulletin of the Seismological Society of America,
95(3), 10451058.
Loh, C.-H., Yeh, Y. T., Jean, W.-Y., & Yeh, Y.-H. 1991. Probabilistic seismic risk analysis in
the Taiwan area based on PGA and spectral amplitude attenuation formulas. Engineering
Geology, 30(34), 277304.
Lubkowski, Z., Bommer, J., Baptie, B., Bird, J., Douglas, J., Free, M., Hancock, J., Sargeant,
S., Sartain, N., & Strasser, F. 2004. An evaluation of attenuation relationships for seismic
hazard assessment in the UK. In: Proceedings of Thirteenth World Conference on Earthquake Engineering. Paper no. 1422.
Lungu, D., Demetriu, S., Radu, C., & Coman, O. 1994. Uniform hazard response spectra
for Vrancea earthquakes in Romania. Pages 365370 of: Proceedings of Tenth European
Conference on Earthquake Engineering, vol. 1.
Lungu, D., Coman, O., & Moldoveanu, T. 1995a (Aug). Hazard analysis for Vrancea earthquakes. Application to Cernavoda NPP site in Romania. In: Proceedings of the 13th International Conference on Structural Mechanics in Reactor Technology. Division K, Paper No.
538.
Lungu, D., Cornea, T., Craifaleanu, I., & Aldea, A. 1995b. Seismic zonation of Romania
based on uniform hazard response ordinates. Pages 445452 of: Proceedings of the Fifth
International Conference on Seismic Zonation, vol. I.
Lungu, D., Cornea, T., Craifaleanu, I., & Demetriu, S. 1996 (Jun). Probabilistic seismic hazard
analysis for inelastic structures on soft soils. In: Proceedings of Eleventh World Conference
on Earthquake Engineering.
Lussou, P., Bard, P. Y., Cotton, F., & Fukushima, Y. 2001. Seismic design regulation codes:
Contribution of K-Net data to site effect evaluation. Journal of Earthquake Engineering, 5(1),
1333.
Luzi, L., Morasca, P., Zolezzi, F., Bindi, D., Pacor, F., Spallarossa, D., & Franceschina, G.
2006. Ground motion models for Molise region (southern Italy). In: Proceedings of First
European Conference on Earthquake Engineering and Seismology (a joint event of the
13th ECEE & 30th General Assembly of the ESC). Paper number 938.
Lyubushin, A. A., & Parvez, I. A. 2010. Map of seismic hazard of India using Bayesian approach. Natural Hazards, 55(2), 543556.
Mahdavian, A. 2006. Empirical evaluation of attenuation relations of peak ground acceleration
in the Zagros and central Iran. In: Proceedings of First European Conference on Earthquake
Engineering and Seismology (a joint event of the 13th ECEE & 30th General Assembly of
the ESC). Paper number 558.
Malagnini, L., & Herrmann, R. B. 2000. Ground-motion scaling in the region of the 1997
Umbria-Marche earthquake (Italy). Bulletin of the Seismological Society of America, 90(4),
10411051.
425
Malagnini, L., Herrmann, R. B., & Di Bona, M. 2000a. Ground-motion scaling in the Apennines
(Italy). Bulletin of the Seismological Society of America, 90(4), 10621081.
Malagnini, L., Herrmann, R. B., & Koch, K. 2000b. Regional ground-motion scaling in central
Europe. Bulletin of the Seismological Society of America, 90(4), 10521061.
Malagnini, L., Akinci, A., Herrmann, R. B., Pino, N. A., & Scognamiglio, L. 2002. Characteristics of the ground motion in northeastern Italy. Bulletin of the Seismological Society of
America, 92(6), 21862204.
Malagnini, L., Mayeda, K., Uhrhammer, R., Akinci, A., & Herrmann, R. B. 2007. A regional
ground-motion excitation/attenuation model for the San Francisco region. Bulletin of the
Malkawi, A. I. H., & Fahmi, K. J. 1996. Locally derived earthquake ground motion attenuation
relations for Jordan and conterminous areas. Quarterly Journal of Engineering Geology and
Hydrogeology, 29(4), 309319.
Mandal, P., Kumar, N., Satyamurthy, C., & Raju, I. P. 2009. Ground-motion attenuation relation
from strong-motion records of the 2001 Mw 7.7 Bhuj earthquake sequence (20012006),
Gujarat, India. Pure and Applied Geophysics, 166(3), 451469.
Manic, M. I. 1998. A new site dependent attenuation model for prediction of peak horizontal
acceleration in Northwestern Balkan. In: Proceedings of Eleventh European Conference on
Earthquake Engineering.
Manic, M. I. 2002 (Sep). Empirical scaling of response spectra for the territory of north-western
Balkan. In: Proceedings of Twelfth European Conference on Earthquake Engineering. Paper reference 650.
Margaris, B., Papazachos, C., Papaioannou, C., Theodulidis, N., Kalogeras, I., & Skarlatoudis,
A. 2002a (Sep). Ground motion attenuation relations for shallow earthquakes in Greece. In:
Proceedings of the XXVIII General Assembly of the European Seismological Commission
(ESC).
Margaris, B., Papazachos, C., Papaioannou, C., Theodulidis, N., Kalogeras, I., & Skarlatoudis,
A. 2002b (Sep). Ground motion attenuation relations for shallow earthquakes in Greece. In:
Proceedings of Twelfth European Conference on Earthquake Engineering. Paper reference
385.
Marin, S., Avouac, J.-P., Nicolas, M., & Schlupp, A. 2004. A probabilistic approach to seismic
hazard in metropolitan France. Bulletin of the Seismological Society of America, 94(6),
21372163.
Massa, M., Marzorati, S., DAlema, E., Di Giacomo, D., & Augliera, P. 2007. Site classication assessment for estimating empirical attenuation relationships for central-northern Italy
earthquakes. Journal of Earthquake Engineering, 11(6), 943U967.

Massa, M., Morasca, P., Moratto, L., Marzorati, S., Costa, G., & Spallarossa, D. 2008. Empirical ground-motion prediction equations for northern Italy using weak- and strong-motion
amplitudes, frequency content, and duration parameters. Bulletin of the Seismological Society of America, 98(3), 13191342.
426
Matsumoto, N., Sasaki, T., Inagaki, K., & Annaka, T. 2004. Attenuation relations of acceleration response spectra at dam foundations in Japan. In: Proceedings of Thirteenth World
Conference on Earthquake Engineering. Paper no. 689.
Matuschka, T. 1980. Assessment of seismic hazards in New Zealand. Tech. rept. 222. Department of Civil Engineering, School of Engineering, University of Auckland. Not seen.
Reported in Stafford (2006).
Matuschka, T., & Davis, B. K. 1991. Derivation of an attenuation model in terms of spectral
acceleration for New Zealand. In: Pacic conference on earthquake engineering. Not seen.
Reported in Stafford (2006).
McCann Jr., M. W., & Echezwia, H. 1984. Investigating the uncertainty in ground motion
prediction. Pages 297304 of: Proceedings of Eighth World Conference on Earthquake
McCue, K. 1986. Strong motion attenuation in eastern Australia. In: Earthquake Engineering
Symposium. National Conference Publication 86/15. Institution of Engineers Australia. Not
seen. Reported in Free (1996).
McCue, K., Gibson, G., & Wesson, V. 1988. Intraplate recording of strong motion in southeastern Australia. Pages 355360 of: Proceedings of Ninth World Conference on Earthquake
McGarr, A., & Fletcher, J. B. 2005. Development of ground-motion prediction equations relevant to shallow mining-induced seismicity in the Trail Mountain area, Emery County, Utah.
McGuire, R. K. 1974. Seismic structural response risk analysis, incorporating peak response regressions on earthquake magnitude and distance. Research Report R74-51.
Massachusetts Institute of Technology, Department of Civil Engineering, Cambridge, USA.
Not seen.
McGuire, R. K. 1976. FORTRAN computer program for seismic risk analysis. Open-File Report
76-67. United States Department of the Interior Geological Survey.
McGuire, R. K. 1977. Seismic design spectra and mapping procedures using hazard analysis
based directly on oscillator response. Earthquake Engineering and Structural Dynamics, 5,
211234.
McGuire, R. K. 1978. Seismic ground motion parameter relations. Journal of the Geotechnical
Engineering Division, ASCE, 104(GT4), 481490.
McVerry, G. H., Dowrick, D. J., Sritharan, S., Cousins, W. J., & Porritt, T. E. 1993. Attenuation of peak ground accelerations in New Zealand. Pages 2338 of: Proceedings of the
International Workshop on Strong Motion Data, vol. 2. Not seen. Cited in McVerry et al.
(1995).
McVerry, G. H., Dowrick, D. J., & Zhao, J. X. 1995 (November). Attenuation of peak ground
accelerations in New Zealand. Pages 287292 of: Pacic Conference on Earthquake Engineering, vol. 3.
McVerry, G. H., Zhao, J. X., Abrahamson, N. A., & Somerville, P. G. 2000. Crustal and
subduction zone attenuation relations for New Zealand earthquakes. In: Proceedings of
Twelfth World Conference on Earthquake Engineering. Paper No. 1834.
427
McVerry, G. H., Zhao, J. X., Abrahamson, N. A., & Somerville, P. G. 2006. New Zealand acceleration response spectrum attenuation relations for crustal and subduction zone earthquakes. Bulletin of the New Zealand Society for Earthquake Engineering, 39(4), 158.
Megawati, K. 2007. Hybrid simulations of ground motions from local earthquakes affecting
Hong Kong. Bulletin of the Seismological Society of America, 97(4), 12931307.
Megawati, K., & Pan, T.-C. 2010. Ground-motion attenuation relationship for the Sumatran
megathrust earthquakes. Earthquake Engineering and Structural Dynamics, 39, 827845.
Megawati, K., Pan, T.-C., & Koketsu, K. 2003. Response spectral attenuation relationships for
Singapore and the Malay peninsula due to distant Sumatran-fault earthquakes. Earthquake
Engineering and Structural Dynamics, 32(14), 22412265.
Megawati, K., Pan, T.-C., & Koketsu, K. 2005. Response spectral attenuation relationships for
Sumatran-subduction earthquakes and the seismic hazard implications to Singapore and
Kuala Lumpur. Soil Dynamics and Earthquake Engineering, 25(1), 1125.
Meirova, T., Hofstetter, R., Ben-Avraham, Z., Steinberg, D. M., Malagnini, L., & Akinci, A. 2008.
Weak-motion-based attenuation relationships for Israel. Geophysical Journal International,
175, 11271140.
Mezcua, J., Garca Blanco, R. M., & Rueda, J. 2008. On the strong ground motion attenuation
in Spain. Bulletin of the Seismological Society of America, 98(3), 13431353.
Midorikawa, S., & Ohtake, Y. 2004. Variance of peak ground acceleration and velocity in
attenuation relationships. In: Proceedings of Thirteenth World Conference on Earthquake
Milne, W. G. 1977. Seismic risk maps for Canada. Page 930 of: Proceedings of Sixth World
Conference on Earthquake Engineering, vol. I. 2-508.
Milne, W. G., & Davenport, A. G. 1969. Distribution of earthquake risk in Canada. Bulletin of
Mohammadioun, B. 1991. The prediction of response spectra for the anti-seismic design
of structures specicity of data from intracontinential environments. European Earthquake
Engineering, V(2), 817.
Mohammadioun, B. 1994a. Prediction of seismic motion at the bedrock from the strong-motion
data currently available. Pages 241245 of: Proceedings of Tenth European Conference
Mohammadioun, G. 1994b. Calculation of site-adapted reference spectra from the statistical
analysis of an extensive strong-motion data bank. Pages 177181 of: Proceedings of Tenth
European Conference on Earthquake Engineering, vol. 1.
Molas, G. L., & Yamazaki, F. 1995. Attenuation of earthquake ground motion in Japan including deep focus events. Bulletin of the Seismological Society of America, 85(5), 13431358.
Molas, G. L., & Yamazaki, F. 1996. Attenuation of response spectra in Japan using new JMA
records. Bulletin of Earthquake Resistant Structure Research Center, 29(Mar), 115128.
428
Monguilner, C. A., Ponti, N., & Pavoni, S. B. 2000a. Relationships between basic ground
motion parameters for earthquakes of the Argentine western region. In: Proceedings of
Twelfth World Conference on Earthquake Engineering. Paper no. 1195.
Monguilner, C. A., Ponti, N., Pavoni, S. B., & Richarte, D. 2000b. Statistical characterization of the response spectra in the Argentine Republic. In: Proceedings of Twelfth World
Montalva, G. A. 2010 (Aug). Site-specic seismic hazard analyses. Ph.D. thesis, Washington
State University.
Morasca, P., Malagnini, L., Akinci, A., Spallarossa, D., & Herrmann, R. B. 2006. Groundmotion scaling in the western Alps. Journal of Seismology, 10(3), 315333.
Morasca, P., Zolezzi, F., Spallarossa, D., & Luzi, L. 2008. Ground motion models for the Molise
region (southern Italy). Soil Dynamics and Earthquake Engineering, 28(3), 198211.
Moss, R. E. S. 2009 (Nov). Reduced uncertainty of ground motion prediction equations
through Bayesian variance analysis. PEER Report 2009/105. Pacic Earthquake Engineering Research Center, College of Engineering, University of California, Berkeley, USA.
Moss, R. E. S., & Der Kiureghian, A. 2006 (Apr). Incorporating parameter uncertainty into
attenuation relationships. In: Proceedings of the Eighth U.S. National Conference on Earthquake Engineering. Paper no. 2010.
Motazedian, D., & Atkinson, G. 2005. Ground-motion relations for Puerto Rico. Pages 6180
of: Mann, P. (ed), Special paper 385: Active tectonics and seismic hazards of Puerto Rico,
the Virgin Islands, and offshore areas. The Geological Society of America.
Munson, C. G., & Thurber, C. H. 1997. Analysis of the attenuation of strong ground motion on
the island of Hawaii. Bulletin of the Seismological Society of America, 87(4), 945960.
Musson, R. M. W., Marrow, P. C., & Winter, P. W. 1994 (May). Attenuation of earthquake
ground motion in the UK. Tech. rept. AEA/CS/16422000/ZJ745/004. AEA Technology Consultancy Services (SRD) and British Geological Survey.
Nakajima, M., Choi, I.-K., Ohtori, Y., & Choun, Y.-S. 2007. Evaluation of seismic hazard curves
and scenario earthquakes for Korean sites based on probabilistic seismic hazard analysis.
Nuclear Engineering and Design, 237(3), 277288.
Nath, S. K., Vyas, M., Pal, I., Singh, A. K., Mukherjee, S., & Sengupta, P. 2005. Spectral
attenuation models in the Sikkim Himalaya from the observed and simulated strong motion
events in the region. Current Science, 88(2), 295303.
Niazi, M., & Bozorgnia, Y. 1991. Behaviour of near-source peak horizontal and vertical ground
motions over SMART-1 array, Taiwan. Bulletin of the Seismological Society of America,
81(3), 715732.
Niazi, M., & Bozorgnia, Y. 1992. Behaviour of near-source vertical and horizontal response
spectra at SMART-1 array, Taiwan. Earthquake Engineering and Structural Dynamics, 21,
3750.
Nowroozi, A. A. 2005. Attenuation relations for peak horizontal and vertical accelerations
of earthquake ground motion in Iran: A preliminary analysis. Journal of Seismology and
429
Nuttli, O. W., & Herrmann, R. B. 1987. Ground motion relations for eastern North American
earthquakes. Pages 231241 of: Proceedings of the Third International Conference on Soil
Dynamics & Earthquake Engineering, vol. II.
Ohno, S., Ohta, T., Ikeura, T., & Takemura, M. 1993. Revision of attenuation formula considering the effect of fault size to evaluate strong motion spectra in near eld. Tectonophysics,
218, 6981.
Ohno, S., Takemura, M., Niwa, M., & Takahashi, K. 1996. Intensity of strong ground motion on
pre-quaternary stratum and surface soil amplications during the 1995 Hyogo-ken Nanbu
earthquake, Japan. Journal of Physics of the Earth, 44(5), 623648.
Ohsaki, Y., Watabe, M., & Tohdo, M. 1980a. Analyses on seismic ground motion parameters
including vertical components. Pages 97104 of: Proceedings of Seventh World Conference on Earthquake Engineering, vol. 2.
Ohsaki, Y., Sawada, Y., Hayashi, K., Ohmura, B., & Kumagai, C. 1980b. Spectral characteristics of hard rock motions. Pages 231238 of: Proceedings of Seventh World Conference
lafsson, S., & Sigbjrnsson, R. 1999. A theoretical attenuation model for earthquake-induced
ground motion. Journal of Earthquake Engineering, 3(3), 287315.
Olszewska, D. 2006. Attenuation relations of ground motion acceleration response spectra for
the Polkowice region. Publications of the Institute of Geophysics of the Polish Academy of
Sciences, M-29(395).
Ordaz, M., & Reyes, C. 1999. Earthquake hazard in Mexico City: Observations versus computations. Bulletin of the Seismological Society of America, 89(5), 13791383.
Ordaz, M., Jara, J. M., & Singh, S. K. 1989. Riesgo ssmico y espectros de diseo en el
estado de Guerrero. Tech. rept. 8782/9745. UNAM Instituto de Ingeniera. In Spanish. Not
seen, cited in Arroyo et al. (2010).
Ordaz, M., Singh, S. K., & Arciniega, A. 1994. Bayesian attenuation regressions: An application to Mexico City. Geophysical Journal International, 117(2), 335344.
Ornthammarath, T. 2010. Inuence of hazard modeling methods and the uncertainty of
GMPEs on the results of probabilistic seismic hazard analysis. Ph.D. thesis, ROSE School,
University of Pavia, Italy.
Ornthammarath, T., Douglas, J., Sigbjrnsson, R., & Lai, C. G. 2010a. Assessment of ground
motion variability and its effects on seismic hazard analysis: A case study for Iceland. Bulletin of Earthquake Engineering. Submitted.
Ornthammarath, T., Douglas, J., Sigbjrnsson, R., & Lai, C. G. 2010b. Assessment of strong
ground motion variability in Iceland. In: Proceedings of Fourteenth European Conference
on Earthquake Engineering. Paper no. 1263.
Orphal, D. L., & Lahoud, J. A. 1974. Prediction of peak ground motion from earthquakes.
zbey, C., Sari, A., Manuel, L., Erdik, M., & Fahjan, Y. 2004. An empirical attenuation relationship for northwestern Turkey ground motion using a random effects approach. Soil
Dynamics and Earthquake Engineering, 24(2), 115125.
430
Pancha, A., & Taber, J. J. 1997. Attenuation of weak ground motions: A report prepared for
the New Zealand Earthquake Commission. Tech. rept. School of Earth Sciences, Victoria
University of Wellington, New Zealand. Not seen. Reported in Stafford (2006).
Pankow, K. L., & Pechmann, J. C. 2004. The SEA99 ground-motion predictive relations for
extensional tectonic regimes: Revisions and a new peak ground velocity relation. Bulletin
Pankow, K. L., & Pechmann, J. C. 2006. Erratum: The SEA99 ground-motion predictive relations for extensional tectonic regimes: Revisions and a new peak ground velocity relation.
Paolucci, R., Rovelli, A., Faccioli, E., Cauzzi, C., Finazzi, D., Vanini, M., Di Alessandro, C., &
Calderoni, G. 2008. On the reliability of long period spectral ordinates from digital accelerograms. Earthquake Engineering and Structural Dynamics, 37(5), 697710.
Parvez, I. A., Gusev, A. A., Panza, G. F., & Petukhin, A. G. 2001. Preliminary determination of the interdependence among strong-motion amplitude, earthquake magnitude and
hypocentral distance for the Himalayan region. Geophysical Journal International, 144(3),
577596.
Pathak, J., Paul, D. K., & Godbole, P. N. 2006 (Sep). ANN based attenuation relationship
for estimation of PGA using Indian strong-motion data. In: Proceedings of First European
Conference on Earthquake Engineering and Seismology (a joint event of the 13th ECEE &
30th General Assembly of the ESC). Paper no. 1132.
Peng, K., Xie, L., Li, S., Boore, D. M., Iwan, W. D., & Teng, T. L. 1985a. The near-source
strong-motion accelerograms recorded by an experimental array in Tangshan, China.
Physics of the Earth and Planetary Interiors, 38, 92109.
Peng, K.-Z., Wu, F. T., & Song, L. 1985b. Attenuation characteristics of peak horizontal acceleration in northeast and southwest China. Earthquake Engineering and Structural Dynamics, 13(3), 337350.
Perea, T., & Sordo, E. 1998. Direct response spectrum prediction including local site effects.
In: Proceedings of Eleventh European Conference on Earthquake Engineering.
Perus, I., & Fajfar, P. 2009 (Aug). How reliable are the ground motion prediction equations? In:
20th International Conference on Structural Mechanics in Reactor Technology (SMiRT 20).
SMiRT 20-Division IV, Paper 1662.
Perus, I., & Fajfar, P. 2010. Ground-motion prediction by a non-parametric approach. Earthquake Engineering and Structural Dynamics, 39(12), 13951416.
Petrovski, D., & Marcellini, A. 1988. Prediction of seismic movement of a site: Statisical approach. In: Proc. UN Sem. on Predict. of Earthquakes. Lisbon, Portugal, 1418 November.
Ptursson, G. G., & Vogfjrd, K. S. 2009. Attenuation relations for near- and far-eld peak
ground motion (PGV, PGA) and new magnitude estimates for large earthquakes in SWIceland. Tech. rept. V 2009-012. Icelandic Meteorological Ofce, Reykjavik, Iceland.
PML. 1982. British earthquakes. Tech. rept. 115/82. Principia Mechanica Ltd., London. Not
seen. Reported in Ambraseys et al. (1992).
431
PML. 1985. Seismological studies for UK hazard analysis. Tech. rept. 346/85. Principia Mechanica Ltd., London. Not seen. Reported in Ambraseys et al. (1992).
Popescu, E., Cioan, C. O., Radulian, M., Placinta, A. O., & Moldovan, I. A. 2007 (Oct). Attenuation relations for the seismic ground motion induced by Vrancea intermediate-depth
earthquakes. In: International Symposium on Strong Vrancea Earthquakes and Risk Mitigation.
Pousse, G., Berge-Thierry, C., Bonilla, L. F., & Bard, P.-Y. 2005. Eurocode 8 design response
spectra evaluation using the K-Net Japanese database. Journal of Earthquake Engineering,
9(4), 547574.
Pousse, G., Bonilla, L. F., Cotton, F., & Margerin, L. 2006. Non stationary stochastic simulation
of strong ground motion time histories including natural variability: Application to the K-net
Japanese database. Bulletin of the Seismological Society of America, 96(6), 2103U2117.

Radu, C., Lungu, D., Demetriu, S., & Coman, O. 1994 (Sep). Recurrence, attenuation and
dynamic amplication for intermediate depth Vrancea earthquakes. Pages 17361745 of:
Proceedings of the XXIV General Assembly of the ESC, vol. III.
Raghu Kanth, S. T. G., & Iyengar, R. N. 2006. Seismic hazard estimation for Mumbai city.
Current Science, 91(11), 14861494.
Raghu Kanth, S. T. G., & Iyengar, R. N. 2007. Estimation of seismic spectral acceleration in
peninsular India. Journal of Earth System Science, 116(3), 199214.
Ramazi, H. R., & Schenk, V. 1994 (Sep). Preliminary results obtained from strong ground
motion analzses [sic] of Iranian earthquakes. Pages 17621770 of: Proceedings of the
XXIV General Assembly of the ESC, vol. III.
Raoof, M., Herrmann, R. B., & Malagnini, L. 1999. Attenuation and excitation of threecomponent ground motion in southern California. Bulletin of the Seismological Society of
America, 89(4), 888902.
Rathje, E. M., Faraj, F., Russell, S., & Bray, J. D. 2004. Empirical relationships for frequency
content parameters of earthquake ground motions. Earthquake Spectra, 20(1), 119144.
Reyes, C. 1998. El estado limite de servicio en el diseo sismico de edicios. Ph.D. thesis,
School of Engineering, National Autonomous University of Mexico (UNAM). In Spanish. Not
seen, reported in Ordaz & Reyes (1999).
Rhoades, D. A. 1997. Estimation of attenuation relations for strong-motion data allowing
for individual earthquake magnitude uncertainties. Bulletin of the Seismological Society of
America, 87(6), 16741678.
Richter, C. F. 1958. Elementary Seismology. San Francisco, USA: Freeman and Co.
Rinaldis, D., Berardi, R., Theodulidis, N., & Margaris, B. 1998. Empirical predictive models
based on a joint Italian & Greek strong-motion database: I, peak ground acceleration and
velocity. In: Proceedings of Eleventh European Conference on Earthquake Engineering.
Rodriguez-Marek, A., & Montalva, G. 2010 (Feb). Uniform hazard spectra for site-specic
applications including uncertainties in site-response. Final technical report. Washington
State University, USA. USGS award number 08HQGR0086.
432
Rogers, A. M., Perkins, D. M., Hampson, D. B., & Campbell, K. W. 1991. Investigations
of peak acceleration data for site effects. Pages 229236 of: Proceedings of the Fourth
International Conference on Seismic Zonation, vol. II.
Romeo, R. W., Tranfaglia, G., & Castenetto, S. 1996. Engineering-developed relations derived
from the strongest instrumentally-detected Italian earthquakes. In: Proceedings of Eleventh
World Conference on Earthquake Engineering. Paper no. 1466.
Ruiz, S., & Saragoni, G. R. 2005 (Nov). Formulas de atenuacin para la subduccin de Chile
considerando los dos mecanismos de sismogenesis y los efectos del suelo. In: Congreso
Chileno de Sismologa e Ingeniera Antissmica, Novenas Jornadas. No. 01-07. In Spanish.
Rupakhety, R., & Sigbjrnsson, R. 2009. Ground-motion prediction equations (GMPEs) for
inelastic response and structural behaviour factors. Bulletin of Earthquake Engineering,
7(3), 637659.
Sabetta, F., & Pugliese, A. 1987. Attenuation of peak horizontal acceleration and velocity
from Italian strong-motion records. Bulletin of the Seismological Society of America, 77(5),
14911513.
Sabetta, F., & Pugliese, A. 1996. Estimation of response spectra and simulation of nonstationary earthquake ground motions. Bulletin of the Seismological Society of America, 86(2),
337352.
Sadeghi, H., Shooshtari, A., & Jaladat, M. 2010. Prediction of horizontal response spectra of
strong ground motions in Iran and its regions. In: Proceedings of the Ninth U.S. National
and 10th Canadian Conference on Earthquake Engineering: Reaching Beyond Borders.
Paper no. 861.
Sadigh, K., Youngs, R. R., & Power, M. S. 1978. A study of attenuation of peak horizontal
accelerations for moderately strong earthquakes. Pages 243250 of: Proceedings of Sixth
European Conference on Earthquake Engineering, vol. 1.
Sadigh, K., Chang, C.-Y., Abrahamson, N. A., Chiou, S. J., & Power, M. S. 1993 (Mar).
Specication of long-period ground motions: Updated attenuation relationships for rock site
conditions and adjustment factors for near-fault effects. Pages 5970 of: Proceedings of
ATC-17-1 Seminar on Seismic Isolation, Passive Energy Dissipation, and Active Control.
Sadigh, K., Chang, C.-Y., Egan, J. A., Makdisi, F., & Youngs, R. R. 1997. Attenuation relationships for shallow crustal earthquakes based on California strong motion data. Seismological
Sadigh, R. K., & Egan, J. A. 1998. Updated relationships for horizontal peak ground velocity
and peak ground displacement for shallow crustal earthquakes. In: Proceedings of the Sixth
U.S. National Conference on Earthquake Engineering.
Saffari, H., Kuwata, Y., Takada, S., & Mahdavian, A. 2010. Spectral acceleration attenuation
for seismic hazard analysis in Iran. In: Proceedings of the Ninth U.S. National and 10th
Canadian Conference on Earthquake Engineering: Reaching Beyond Borders. Paper no.
867.
Saini, S., Sharma, M. L., & Mukhopadhyay, S. 2002 (Dec). Strong ground motion empirical
attenuation relationship for seismic hazard estimation in Himalaya region. Pages 143150
of: 12th Symposium on Earthquake Engineering, vol. I.
433
Sakamoto, S., Uchiyama, Y., & Midorikawa, S. 2006 (Apr). Variance of response spectra
in attenuation relationship. In: Proceedings of the Eighth U.S. National Conference on
Earthquake Engineering. Paper no. 471.
Sanchez, A. R., & Jara, J. M. 2001. Estimacin del peligro ssmico de Morelia. Ciencia
nicolaita, 29, 6376. Not seen. Cited in Jara-Guerrero et al. (2007). In Spanish.
Sarma, S. K., & Free, M. W. 1995 (November). The comparison of attenuation relationships
for peak horizontal acceleration in intraplate regions. Pages 175184 of: Pacic Conference
Sarma, S. K., & Srbulov, M. 1996. A simplied method for prediction of kinematic soilfoundation interaction effects on peak horizontal acceleration of a rigid foundation. Earthquake Engineering and Structural Dynamics, 25(8), 815836.
Sarma, S. K., & Srbulov, M. 1998. A uniform estimation of some basic ground motion parameters. Journal of Earthquake Engineering, 2(2), 267287.
Savy, J. B., Boussonnade, A. C., Mensing, R. W., & Short, C. M. 1993 (Jun). Eastern U.S.
seismic hazard characterization update. Tech. rept. UCRL-ID-115111. Lawrence Livermore
National Laboratory, USA.
Scasserra, G., Stewart, J. P., Bazzurro, P., Lanzo, G., & Mollaioli, F. 2009. A comparison
of NGA ground-motion prediction equations to Italian data. Bulletin of the Seismological
Schenk, V. 1982. Peak particle ground motions in earthquake near-eld. Pages 211217 of:
Proceedings of Seventh European Conference on Earthquake Engineering, vol. 2.
Schenk, V. 1984. Relations between ground motions and earthquake magnitude, focal distance and epicentral intensity. Engineering Geology, 20(1/2), 143151.
Scherbaum, F., Cotton, F., & Smit, P. 2004. On the use of response spectral-reference data for
the selection and ranking of ground-motion models for seismic-hazard analysis in regions
of moderate seismicity: The case of rock motion. Bulletin of the Seismological Society of
America, 94(6), 21642185.
Schmidt, V., Dahle, A., & Bungum, H. 1997 (Nov.). Costa Rican spectral strong motion attenuation. Tech. rept. NORSAR, Kjeller, Norway. Reduction of Natural Disasters in central
America Earthquake Preparedness and Hazard Mitigation Phase II: 19962000, Part 2.
Schnabel, P. B., & Seed, H. B. 1973. Accelerations in rock for earthquakes in the western
United States. Bulletin of the Seismological Society of America, 63(2), 501516.
Schwarz, J., Ende, C., Habenberger, J., Lang, D. H., Baumbach, M., Grosser, H., Milereit,
C., Karakisa, S., & Znbl, S. 2002 (Sep). Horizontal and vertical response spectra on the
basis of strong-motion recordings from the 1999 Turkey earthquakes. In: Proceedings of
the XXVIII General Assembly of the European Seismological Commission (ESC).
Scognamiglio, L., Malagnini, L., & Akinci, A. 2005. Ground-motion scaling in eastern Sicily,
Italy. Bulletin of the Seismological Society of America, 95(2), 568578.
Seed, H. B., Murarka, R., Lysmer, J., & Idriss, I. M. 1976. Relationships of maximum acceleration, maximum velocity, distance from source, and local site conditions for moderately
strong earthquakes. Bulletin of the Seismological Society of America, 66(4), 13231342.
434
Selcuk, L., Selcuk, A. S., & Beyaz, T. 2010. Probabilistic seismic hazard assessment for Lake
Van basin, Turkey. Natural Hazards, 54(3), 949965.
Sen, M. K. 1990 (May). Deep structural complexity and site response in Los Angeles basin.
Pages 545553 of: Proceedings of the Fourth U.S. National Conference on Earthquake
Shabestari, K. T., & Yamazaki, F. 2000. Attenuation relation of response spectra in Japan
considering site-specic term. In: Proceedings of Twelfth World Conference on Earthquake
Engineering. Paper No. 1432.
Shabestari, T. K., & Yamazaki, F. 1998. Attenuation of JMA seismic intensity using recent
JMA records. Pages 529534 of: Proceedings of the 10th Japan Earthquake Engineering
Symposium, vol. 1. Not seen. Reported in Shabestari & Yamazaki (2000).
Sharma, M., & Bungum, H. 2006. New strong ground-motion spectral acceleration relations
for the Himalayan region. In: Proceedings of First European Conference on Earthquake
Engineering and Seismology (a joint event of the 13th ECEE & 30th General Assembly of
the ESC). Paper number 1459.
Sharma, M. L. 1998. Attenuation relationship for estimation of peak ground horizontal acceleration using data from strong-motion arrays in India. Bulletin of the Seismological Society
of America, 88(4), 10631069.
Sharma, M. L. 2000. Attenuation relationship for estimation of peak ground vertical acceleration using data from strong motion arrays in India. In: Proceedings of Twelfth World
Conference on Earthquake Engineering. Paper No. 1964.
Sharma, M. L., Douglas, J., Bungum, H., & Kotadia, J. 2009. Ground-motion prediction equations based on data from the Himalayan and Zagros regions. Journal of Earthquake Engineering, 13(8), 11911210.
Shi, S., & Shen, J. 2003. A study on attenuation relations of strong earth movements in
Shanghai and its adjacent area. Earthquake Research in China, 19, 315323. In Chinese.
Not seen. Cited in Cole et al. (2008).
Si, H., & Midorikawa, S. 1999. New attenuation relationships for peak ground acceleration
and velocity considering effects of fault type and site condition. Journal of structural and
construction engineering, aij, 523, 6370. In Japanese with English abstract. Not seen.
Si, H., & Midorikawa, S. 2000. New attenuation relations for peak ground acceleration and
velocity considering effects of fault type and site condition. In: Proceedings of Twelfth
World Conference on Earthquake Engineering. Paper No. 0532.
Sigbjrnsson, R. 1990. Strong motion measurements in Iceland and seismic risk assessment.
Pages 215222 of: Proceedings of Ninth European Conference on Earthquake Engineering,
vol. 10-A.
Sigbjrnsson, R., & Ambraseys, N. N. 2003. Uncertainty analysis of strong ground motion.
Bulletin of Earthquake Engineering, 1(3), 321347.
Sigbjrnsson, R., & Baldvinsson, G. I. 1992. Seismic hazard and recordings of strong ground
motion in Iceland. Pages 419424 of: Proceedings of Tenth World Conference on Earthquake Engineering, vol. 1. Rotterdam, The Netherlands: A. A. Balkema. Madrid, Spain.
1924 July.
435
Silva, W., & Abrahamson, N. A. 1992. Quantication of long period strong ground motion
attenuation for engineering design. In: Huang, M. J. (ed), Proceedings of (strong Motion
Instrumentation Program) smip92 seminar on seismological and engineering implications of
recent strong-motion data. California Division of Mines and Geology, Sacramento, USA.
Silva, W., Gregor, N., & Darragh, R. 2002 (Nov). Development of regional hard rock attenuation relations for central and eastern North America. Tech. rept. Pacic Engineering and
Analysis.
Simpson, K. A. 1996. Attenuation of strong ground-motion incorporating near-surface foundation conditions. Ph.D. thesis, University of London.
Singh, R. P., Aman, A., & Prasad, Y. J. J. 1996. Attenuation relations for strong seismic ground
motion in the Himalayan region. Pure and Applied Geophysics, 147(1), 161180.
Singh, S. K., Mena, E., Castro, R., & Carmona, C. 1987. Empirical prediction of ground motion
in Mexico City from coastal earthquakes. Bulletin of the Seismological Society of America,
77(5), 18621867.
Singh, S. K., Gutierreez, C., Arboleda, J., & Ordaz, M. 1993. Peligro ssmico en El Salvador.
Tech. rept. Universidad Nacional Autnoma de Mxico, Mxico. Not seen. Reported in
Bommer et al. (1996).
Skarlatoudis, A., Theodulidis, N., Papaioannou, C., & Roumelioti, Z. 2004. The dependence of
peak horizontal acceleration on magnitude and distance for small magnitude earthquakes
in Greece. In: Proceedings of Thirteenth World Conference on Earthquake Engineering.
Paper no. 1857.
Skarlatoudis, A. A., Papazachos, C. B., Margaris, B. N., Theodulidis, N., Papaioannou, C.,
Kalogeras, I., Scordilis, E. M., & Karakostas, V. 2003. Empirical peak ground-motion predictive relations for shallow earthquake in Greece. Bulletin of the Seismological Society of
America, 93(6), 25912603.
Slejko, D., Javakhishvili, Z., Rebez, A., Santulin, M., Elashvili, M., Bragato, P. L., Godoladze,
T., & Garcia, J. 2008. Seismic hazard assessment for the Tbilisi test area (eastern Georgia).
Bollettino di Geosica Teorica ed Applicata, 49(1), 3757.
Smit, P. 1998 (Sep). Strong motion attenuation model for central Europe. In: Proceedings of
Eleventh European Conference on Earthquake Engineering. smisma.
Smit, P. 2000 (Dec). Personal communication 4/12/2000.
Smit, P., Arzoumanian, V., Javakhishvili, Z., Areev, S., Mayer-Rosa, D., Balassanian, S.,
& Chelidze, T. 2000. The digital accelerograph network in the Caucasus. In: Balassanian, S. (ed), Earthquake Hazard and Seismic Risk Reduction Advances in Natural and
Technological Hazards Research. Kluwer Academic Publishers. Presented at 2nd International Conference on Earthquake Hazard and Seismic Risk Reduction, Yerevan, Armenia,
15/9/199821/9/1998.
Sobhaninejad, G., Noorzad, A., & Ansari, A. 2007 (Jun). Genetic algorithm (GA): A new
approach in estimating strong ground motion attenuation relations. In: Proceedings of the
Fourth International Conference on Earthquake Geotechnical Engineering. Paper no. 1313.
436
Sokolov, V. 1997. Empirical models for estimating Fourier-amplitude spectra of ground acceleration in the northern Caucasus (Racha seismogenic zone). Bulletin of the Seismological
Sokolov, V., Loh, C.-H., & Wen, K.-L. 2000. Empirical model for estimating Fourier amplitude
spectra of ground acceleration in Taiwan region. Earthquake Engineering and Structural
Dynamics, 29(3), 339357.
Sokolov, V., Bonjer, K.-P., Oncescu, M., & Rizescu, M. 2005. Hard rock spectral models for
intermediate-depth Vrancea, Romania, earthquakes. Bulletin of the Seismological Society
of America, 95(5), 17491765.
Sokolov, V., Bonjer, K.-P., Wenzel, F., Grecu, B., & Radulian, M. 2008. Ground-motion prediction equations for the intermediate depth Vrancea (Romania) earthquakes. Bulletin of
Earthquake Engineering, 6, 367388.
Sokolov, V. Y. 1998. Spectral parameters of the ground motions in Caucasian seismogenic
zones. Bulletin of the Seismological Society of America, 88(6), 14381444.
Somerville, P., Collins, N., Abrahamson, N., Graves, R., & Saikia, C. 2001 (Jun). Ground
motion attenuation relations for the central and eastern United States. Tech. rept. Research
supported by the U.S. Geological Survey, under award number 99HQGR0098.
Somerville, P., Graves, R., Collins, N., Song, S. G., Ni, S., & Cummins, P. 2009 (Dec). Source
and ground motion models of Australian earthquakes. In: Proceedings of the 2009 Annual
Conference of the Australian Earthquake Engineering Society.
Souriau, A. 2006. Quantifying felt events: A joint analysis of intensities, accelerations and
dominant frequencies. Journal of Seismology, 10(1), 2338.
Spudich, P., & Boore, D. M. 2005. Erratum: SEA99: A revised ground-motion prediction
relation for use in extensional tectonic regimes. Bulletin of the Seismological Society of
America, 95(3), 1209.
Spudich, P., Fletcher, J., Hellweg, M., Boatwright, J., Sullivan, C., Joyner, W., Hanks, T., Boore,
D., McGarr, A., Baker, L., & Lindh, A. 1996. Earthquake ground motions in extensional
tectonic regimes. Open-File Report 96-292. U.S. Geological Survey. Not seen. Reported in
Spudich et al. (1997).
Spudich, P., Fletcher, J. B., Hellweg, M., Boatwright, J., Sullivan, C., Joyner, W. B., Hanks,
T. C., Boore, D. M., McGarr, A., Baker, L. M., & Lindh, A. G. 1997. SEA96 A new predictive relation for earthquake ground motions in extensional tectonic regimes. Seismological
Spudich, P., Joyner, W. B., Lindh, A. G., Boore, D. M., Margaris, B. M., & Fletcher, J. B. 1999.
SEA99: A revised ground motion prediction relation for use in extensional tectonic regimes.
Srinivasan, C., Sharma, M. L., Kotadia, J., & Willy, Y. A. 2008. Peak ground horizontal acceleration attenuation relationship for low magnitudes at short distances in south Indian region.
In: Proceedings of Fourteenth World Conference on Earthquake Engineering. Paper no.
02-0135.
437
Stafford, P. J. 2006 (Feb). Engineering seismological studies and seismic design criteria
for the Buller region, South Island, New Zealand. Ph.D. thesis, University of Canterbury,
Christchurch, New Zealand.
Stafford, P. J., Berrill, J. B., & Pettinga, J. R. 2009. New predictive equations for Arias intensity
from crustal earthquakes in New Zealand. Journal of Seismology, 13(1), 3152.
Stamatovska, S. 2002 (Sep). A new azimuth dependent empirical strong motion model for
Vranchea subduction zone. In: Proceedings of Twelfth European Conference on Earthquake Engineering. Paper reference 324.
Stamatovska, S., & Petrovski, D. 1991. Ground motion parameters based on data obtained from strong earthquake records. In: National Progress Report of Yugoslavia,
UNDP/UNESCO Project Rep/88/004, Task Group 3, Second Meeting. Zagreb, Yugoslavia:
Geophysical Institute.
Stamatovska, S., & Petrovski, D. 1996. Empirical attenuation acceleration laws for Vrancea
intermediate earthquakes. In: Proceedings of Eleventh World Conference on Earthquake
Steinberg, V. V., Saks, M. V., Aptikaev, F. F., et al. . 1993. Methods of seismic ground motion
estimation (handbook). Pages 597 of: Voprosy inzhenernoi seismologii. iss. 34. nauka. In
Russian. Not seen. Cited by Lyubushin & Parvez (2010).
Sun, F., & Peng, K. 1993. Attenuation of strong ground motion in western U.S.A. Earthquake
Research in China, 7(1), 119131.
Sunuwar, L., Cuadra, C., & Karkee, M. B. 2004. Strong ground motion attenuation in the
Sea of Japan (Okhotsk-Amur plates boundary) region. In: Proceedings of Thirteenth World
Takahashi, T., Kobayashi, S., Fukushima, Y., Zhao, J. X., Nakamura, H., & Somerville, P. G.
2000 (Nov). A spectral attenuation model for Japan using strong motion data base. In:
Proceedings of the Sixth International Conference on Seismic Zonation.
Takahashi, T., Asano, A., Saiki, T., Okada, H., Irikura, K., Zhao, J. X., Zhang, J., Thio, H. K.,
Somerville, P. G., Fukushima, Y., & Fukushima, Y. 2004. Attenuation models for response
spectra derived from Japanese strong-motion records accounting for tectonic source types.
In: Proceedings of Thirteenth World Conference on Earthquake Engineering. Paper no.
1271.
Takahashi, T., Asano, A., Ono, Y., Ogawa, H., Zhao, J. X., Zhang, J., Fukushima, Y., Irikura, K.,
Thio, H. K., Somerville, P. G., & Fukushima, Y. 2005 (Aug). Attenuation relations of strong
motion in Japan using site classication based on predominant period. In: 18th International
Conference on Structural Mechanics in Reactor Technology (SMiRT 18). Paper SMiRT18K02-1.
Tamura, K., Sasaki, Y., & Aizawa, K. 1990 (May). Attenuation characteristics of ground motions
in the period range of 2 to 20 seconds for application to the seismic design of long-period
structures. Pages 495504 of: Proceedings of the Fourth U.S. National Conference on
Earthquake Engineering, vol. 1.
Tavakoli, B., & Pezeshk, S. 2005. Empirical-stochastic ground-motion prediction for eastern
North America. Bulletin of the Seismological Society of America, 95(6), 22832296.
438
Tavakoli, B., & Pezeshk, S. 2007. A new approach to estimate a mixed model-based ground
motion prediction equation. Earthquake Spectra, 23(3), 665684.
Taylor Castillo, W., Santos Lopez, P., Dahle, A., & Bungum, H. 1992 (Nov.). Digitization
of strong motion data and estimation of PGA attenuation. Tech. rept. NORSAR, Kjeller,
Norway. Reduction of Natural Disasters in central America Earthquake Preparedness and
Hazard Mitigation Seismic Zonation and Earthquake Hazard Assessment.
Tejeda-Jcome, J., & Chvez-Garca, F. J. 2007. Empirical ground-motion estimation equations in Colima from weak motion records. ISET Journal of Earthquake Technology, 44(34),
409420.
Tento, A., Franceschina, L., & Marcellini, A. 1992. Expected ground motion evaluation for
Italian sites. Pages 489494 of: Proceedings of Tenth World Conference on Earthquake
Theodulidis, N. P., & Papazachos, B. C. 1992. Dependence of strong ground motion on
magnitude-distance, site geology and macroseismic intensity for shallow earthquakes in
Greece: I, peak horizontal acceleration, velocity and displacement. Soil Dynamics and
Earthquake Engineering, 11, 387402.
Theodulidis, N. P., & Papazachos, B. C. 1994. Dependence of strong ground motion on
magnitude-distance, site geology and macroseismic intensity for shallow earthquakes in
Greece: II horizontal pseudovelocity. Soil Dynamics and Earthquake Engineering, 13(5),
317343.
Tong, H., & Katayama, T. 1988. Peak acceleration attenuation by eliminating the ill-effect of the
correlation between magnitude and epicentral distance. Pages 349354 of: Proceedings of
Ninth World Conference on Earthquake Engineering, vol. II.
Toro, G. R. 2002 (Jun). Modication of the Toro et al. (1997) attenuation equations for large
magnitudes and short distances. Tech. rept. Risk Engineering.
Toro, G. R., & McGuire, R. K. 1987. An investigation into earthquake ground motion characteristics in eastern North America. Bulletin of the Seismological Society of America, 77(2),
468489.
Toro, G. R., & Silva, W. J. 2001 (Jan). Scenario earthquakes for Saint Louis, MO, and Memphis, TN, and seismic hazard maps for the central United States region including the effect
of site conditions. Tech. rept. Research supported by the U.S. Geological Survey (USGS),
under award number 1434-HQ-97-GR-02981.
Trifunac, M. D. 1976. Preliminary analysis of the peaks of strong earthquake ground motion
dependence of peaks on earthquake magnitude, epicentral distance, and recording site
conditions. Bulletin of the Seismological Society of America, 66(1), 189219.
Trifunac, M. D. 1977. Forecasting the spectral amplitudes of strong earthquake ground motion.
Pages 139152 of: Proceedings of Sixth World Conference on Earthquake Engineering, vol.
I.
Trifunac, M. D. 1978. Response spectra of earthquake ground motions. Journal of the Engineering Mechanics Division, ASCE, 104(EM5), 10811097.
439
Trifunac, M. D. 1980. Effects of site geology on amplitudes of strong motion. Pages 145152
of: Proceedings of Seventh World Conference on Earthquake Engineering, vol. 2.
Trifunac, M. D., & Anderson, J. G. 1977. Preliminary empirical models for scaling absolute acceleration spectra. Tech. rept. 77-03. Department of Civil Engineering, University of
Southern California, Los Angeles, California, U.S.A.
Trifunac, M. D., & Anderson, J. G. 1978a. Preliminary empirical models for scaling pseudo
relative velocity spectra. Tech. rept. 78-04. Department of Civil Engineering, University of
Southern California, Los Angeles, California, U.S.A.
Trifunac, M. D., & Anderson, J. G. 1978b. Preliminary empirical models for scaling relative
velocity spectra. Tech. rept. 78-05. Department of Civil Engineering, University of Southern
California, Los Angeles, California, U.S.A.
Trifunac, M. D., & Brady, A. G. 1975. On the correlation of peak acceleration of strong motion with earthquake magnitude, epicentral distance and site conditions. Pages 4352 of:
Proceedings of the U.S. National Conference on Earthquake Engineering.
Trifunac, M. D., & Brady, A. G. 1976. Correlations of peak acceleration, velocity and displacement with earthquake magnitude, distance and site conditions. Earthquake Engineering
and Structural Dynamics, 4(5), 455471.
Trifunac, M. D., & Lee, V. W. 1979. Dependence of pseudo relative velocity spectra of strong
motion acceleration on depth of sedimentary deposits. Tech. rept. 79-02. Department of
Civil Engineering, University of Southern California, Los Angeles, California, U.S.A.
Trifunac, M. D., & Lee, V. W. 1985. Preliminary empirical model for scaling pseudo relative
velocity spectra of strong earthquake acceleration in terms of magnitude, distance, site
intensity and recording site condition. Tech. rept. 85-04. Department of Civil Engineering,
Trifunac, M. D., & Lee, V. W. 1989. Empirical models for scaling pseudo relative velocity
spectra of strong earthquake accelerations in terms of magnitude, distance, site intensity
and recording site conditions. Soil Dynamics and Earthquake Engineering, 8(3), 126144.
Tromans, I. 2004 (Jan). Behaviour of buried water supply pipelines in earthquake zones. Ph.D.
thesis, University of London.
Tromans, I. J., & Bommer, J. J. 2002 (Sep). The attenuation of strong-motion peaks in Europe.
In: Proceedings of Twelfth European Conference on Earthquake Engineering. Paper no.
394.
Tsai, Y. B., Brady, F. W., & Cluff, L. S. 1990 (May). An integrated approach for characterization
of ground motions in PG&Es long term seismic program for Diablo Canyon. Pages 597606
of: Proceedings of the Fourth U.S. National Conference on Earthquake Engineering, vol. 1.
Tuluka, M. 2007. An estimate of the attenuation relationship for strong ground motion in the
Kivu Province, Western Rift Valley of Africa. Physics of the Earth and Planetary Interiors,
162, 1321.
Ulusay, R., Tuncay, E., Sonmez, H., & Gokceoglu, C. 2004. An attenuation relationship based
on Turkish strong motion data and iso-acceleration map of Turkey. Engineering Geology,
74(3-4), 265291.
440
Ulutas, E., & Ozer, M. F. 2010. Empirical attenuation relationship of peak ground acceleration
for eastern Marmara region in Turkey. The Arabian Journal for Science and Engineering,
35(1A), 187203.
Vives, V., & Canas, J.A. 1992. Anelastic attenuation and pseudoacceleration relations in
eastern Iberia. Pages 299304 of: Proceedings of Tenth World Conference on Earthquake
Wald, D. J., Worden, B. C., Quitoriano, V., & Pankow, K. L. 2005. ShakeMap manual.
Technical Manual, users guide, and software guide Version 1.0. USGS Techniques and
Methods 12-A1.
Wang, B.-Q., Wu, F. T., & Bian, Y.-J. 1999. Attenuation characteristics of peak acceleration in
north China and comparison with those in the eastern part of North America. Acta Seismologica Sinica, 12(1), 2634.
Wang, G., & Tao, X. 2000 (Nov). A new two-stage procedure for tting attenuation relationship
of strong ground motion. In: Proceedings of the Sixth International Conference on Seismic
Zonation.
Wang, M., & Takada, T. 2009. A Bayesian framework for prediction of seismic ground motion.
Weisburg, S. 1985. Applied Linear Regression. 2nd edn. John Wiley & Sons.
Wells, D. L., & Coppersmith, K. J. 1994. New empirical relationships among magnitude,
rupture length, rupture width, rupture area, and surface displacement. Bulletin of the Seismological Society of America, 84(4), 9741002.
Wiggins Jr., J. H. 1964. Construction of strong motion response spectra from magnitude and
distance data. Bulletin of the Seismological Society of America, 54(5), 12571269.
Wills, C. J., Petersen, M., Bryant, W. A., Reichle, M., Saucedo, G. J., Tan, S., Taylor, G., &
Treiman, J. 2000. A site-conditions map for California based on geology and shear-wave
velocity. Bulletin of the Seismological Society of America, 90(6B), S187S208.
Wu, Y.-M., Shin, T.-C., & Chang, C.-H. 2001. Near real-time mapping of peak ground acceleration and peak ground velocity following a strong earthquake. Bulletin of the Seismological
Xiang, J., & Gao, D. 1994. The attenuation law of horizontal peak acceleration on the rock site
in Yunnan area. Earthquake Research in China, 8(4), 509516.
Xu, Z., Shen, X., & Hong, J. 1984. Attenuation relation of ground motion in northern China.
Pages 335342 of: Proceedings of Eighth World Conference on Earthquake Engineering,
vol. II.
Yamabe, K., & Kanai, K. 1988. An empirical formula on the attenuation of the maximum acceleration of earthquake motions. Pages 337342 of: Proceedings of Ninth World Conference
on Earthquake Engineering, vol. II.
Yamazaki, F., Wakamatsu, K., Onishi, J., & Shabestari, K. T. 2000. Relationship between geomorphological land classication and site amplication ratio based on JMA strong motion
records. Soil Dynamics and Earthquake Engineering, 19(1), 4153.
441
Yokota, H., Shiba, K., & Okada, K. 1988. The characteristics of underground earthquake
motions observed in the mud stone layer in Tokyo. Pages 429434 of: Proceedings of
Ninth World Conference on Earthquake Engineering, vol. II.
Youngs, R. R., Day, S. M., & Stevens, J. L. 1988. Near eld ground motions on rock for large
subduction earthquakes. Pages 445462 of: Proceedings of Earthquake Engineering & Soil
Dynamics II. Geotechnical Division, ASCE.
Youngs, R. R., Abrahamson, N., Makdisi, F. I., & Sadigh, K. 1995. Magnitude-dependent
variance of peak ground acceleration. Bulletin of the Seismological Society of America,
85(4), 11611176.
Youngs, R. R., Chiou, S.-J., Silva, W. J., & Humphrey, J. R. 1997. Strong ground motion
attenuation relationships for subduction zone earthquakes. Seismological Research Letters,
68(1), 5873.
Yu, Y., & Hu, Y. 2004. Empirical long-period response spectral attenuation relations based
on southern California digital broad-band recordings. In: Proceedings of Thirteenth World
Yuzawa, Y., & Kudo, K. 2008. Empirical estimation of long-period (110 sec.) earthquake
ground motion on hard rocks. In: Proceedings of Fourteenth World Conference on Earthquake Engineering. Paper no. S10-057.
Zafarani, H., Mousavi, M., Noorzad, A., & Ansari, A. 2008. Calibration of the specic barrier model to Iranian plateau earthquakes and development of physically based attenuation
relationships for Iran. Soil Dynamics and Earthquake Engineering, 28(7), 550576.
Zare, M., & Sabzali, S. 2006. Spectral attenuation of strong motions in Iran. Pages 749758
of: Proceedings of Third International Symposium on the Effects of Surface Geology on
Seismic Motion, vol. 1. Paper number 146.
Zar, M., Ghafory-Ashtiany, M., & Bard, P.-Y. 1999. Attenuation law for the strong-motions in
Iran. Pages 345354 of: Proceedings of the Third International Conference on Seismology
and Earthquake Engineering, Tehran, vol. 1.
Zhao, J. X. 2010. Geometric spreading functions and modeling of volcanic zones for strongmotion attenuation models derived from records in Japan. Bulletin of the Seismological
Zhao, J. X., Dowrick, D. J., & McVerry, G. H. 1997. Attenuation of peak ground acceleration
in New Zealand earthquakes. Bulletin of the New Zealand National Society for Earthquake
Engineering, 30(2), 133158.
Zhao, J. X., Zhang, J., Asano, A., Ohno, Y., Oouchi, T., Takahashi, T., Ogawa, H., Irikura, K.,
Thio, H. K., Somerville, P. G., Fukushima, Y., & Fukushima, Y. 2006. Attenuation relations
of strong ground motion in Japan using site classication based on predominant period.
Zheng, S., & Wong, Y. L. 2004. Seismic ground motion relationships in southern China based
on stochastic nite-fault model. Earthquake Engineering and Engineering Vibration, 3(1),
1121.
Zonno, G., & Montaldo, V. 2002. Analysis of strong ground motions to evaluate regional
attenuation relationships. Annals of Geophysics, 45(34), 439454.
442
443
Scientic and Technical Center

Natural Risks and CO2 Storage Safety Division
3, avenue Claude Guillemin - BP 36009
45060 Orleans Cedex 2 - France - Tel: +33 (0)2 38 64 34 34

Douglas 2011

Uploaded by

Copyright:

Available Formats

Douglas 2011

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Douglas 2011

Uploaded by

Copyright:

Available Formats

Ground-motion prediction equations

Ground-motion prediction equations

Keywords: earthquakes, seismic hazard, ground-motion prediction equations (GMPEs),

In bibliography, this report should be cited as follows:

Ground-motion prediction equations 19642010

BRGM/RP-59356-FR Final Report

Ground-motion prediction equations 19642010

BRGM/RP-59356-FR Final Report

Ground-motion prediction equations 19642010

Stamatovska & Petrovski (1991) . . . . . . . . . . . . . . . . .

BRGM/RP-59356-FR Final Report

Ground-motion prediction equations 19642010

BRGM/RP-59356-FR Final Report

Ground-motion prediction equations 19642010

BRGM/RP-59356-FR Final Report

Ground-motion prediction equations 19642010

BRGM/RP-59356-FR Final Report

Ground-motion prediction equations 19642010

Akkar & Bommer (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . .

Akkar & agnan

3 General characteristics of GMPEs for PGA

Ground-motion prediction equations 19642010

Trifunac & Lee (1989) . . . . . . . . . . . . . . . . . . . . . . . . .

BRGM/RP-59356-FR Final Report

Ground-motion prediction equations 19642010

Bommer et al. (1998) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

BRGM/RP-59356-FR Final Report

Ground-motion prediction equations 19642010

Bindi et al. (2006) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Akkar & agnan

BRGM/RP-59356-FR Final Report

Ground-motion prediction equations 19642010

Arroyo et al. (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5 General characteristics of GMPEs for spectral ordinates

6 List of other ground-motion models

BRGM/RP-59356-FR Final Report

Ground-motion prediction equations 19642010

BRGM/RP-59356-FR Final Report

Ground-motion prediction equations 19642010

1. Model is derived for an inappropriate tectonic environment (such as subduction-zone

1.1 Other summaries and reviews of GMPEs

BRGM/RP-59356-FR Final Report

Ground-motion prediction equations 19642010

GMPEs summarised here

BRGM/RP-59356-FR Final Report

Ground-motion prediction equations 19642010

BRGM/RP-59356-FR Final Report

Ground-motion prediction equations 19642010

BRGM/RP-59356-FR Final Report

Ground-motion prediction equations 19642010

BRGM/RP-59356-FR Final Report

Summary of published GMPEs for

Esteva & Rosenblueth (1964)

Ground-motion model is:

2.2 Kanai (1966)

Milne & Davenport (1969)

Ground-motion model is:

where A is in percentage of g, a1 = 0.69, a2 = 1.64, a3 = 1.1 and a4 = 1.10 ( not

Ground-motion prediction equations 19642010

2.4 Esteva (1970)