Journalof Structural Geology, Vol. 12, No. 8, pp. 973 to 986, 1990 0191-8141/90 $03.00+0.

00
Printed in Great Britain © 1990 Pergamon Press plc

Kinematic analysis of fault-slip data

RANDALL N~iARRETTa n d RICHARD W . ALLMENDINGER

Department of Geological Sciences and Institute for the Study of the Continents, Cornell University, Ithaca,
NY 14853-1504, U.S.A.

(Received 13 February 1989; accepted in revised form 21 February 1990)

Abstract--An array of graphical and numerical techniques facilitate qualitative and quantitative kinematic
analysis of fault-slip data. Graphical contouring and Bingham statistics of the shortening and extension axes for
kinematically scale-invariant faults characterize the distributions and orientations of the principal axes of average
incremental strain. Numerical analysis by means of moment tensor summation yields the orientations and
magnitudes of the principal strain axes as well as rotational information. Field data can be weighted for moment
tensor summation using measurements of fault gouge thickness and/or fault plane width, from which average
displacement and fault area can be estimated. The greatest uncertainties of kinematic analysis derive from
assumptions about the weighting of the data, the effects of post-faulting rotation on the data, the degree to which
sampling is representative of the entire fault population, and the spatial homogeneity of strain. These
assumptions can be evaluated for a specific data set. Geometric criteria can distinguish the kinematic
heterogeneities produced by triaxial deformation, anisotropy reactivation, strain compatibility constraints
and/or multiple deformations. Strain compatibility, material anisotropy and heterogeneity may be characterized
by integrating the results of kinematic and dynamic fault-slip analyses.

INTRODUCTION drowski 1985, Lisle 1987). All dynamic methods assume

that faults slip in the direction of shear stress resolved on
MANY fundamental tectonic processes can only be stud- the fault plane (Wallace 1951, Bott 1959), which re-
ied directly in regions of active tectonics. Evidence of quires that stress is homogeneous and that faults do not
neotectonic deformation is largely limited to brittle interact mechanically. Additionally, the dynamic
deformation in the upper crust, because rocks deform- methods assume that sampling is representative and that
ing by crystal-plastic mechanisms are unexposed. Fault- there has been no post-faulting reorientation of the
ing provides the most useful constraints on brittle defor- fault-slip data. These assumptions commonly are not
mation in modern orogens for several reasons: (1) faults evaluated.
commonly accommodate the largest magnitude of de- Kinematic analyses of faults are far fewer, perhaps
formation among brittle mechanisms; (2) faults are because of the difficulties of applying continuum mech-
effectively penetrative at the Earth's surface on a re- anics to an inherently discontinuous phenomenon. Woj-
gional scale; (3) seismicity describes some fault activity tal (1989) has proposed a two-dimensional graphical
at depth; and (4) faults commonly can be well dated, approximation for finite strain. Here, we describe a
The observations that characterize the kinematics of a three-dimensional incremental strain analysis via a seis-
fault are the orientation of the fault plane, the orien- mological approach (Kostrov 1974, Molnar 1983)which
tation of the slip direction, and the sense-of-slip; these is more suitable for regional analysis in which the orien-
data collectively form a fault-slip datum. Fault-slip data tations of the principal axes, and their variation with
can be measured in the field at exposures of faults and fault magnitude, are of greater interest than the magni-
can also be gleaned in a somewhat ambiguous form from tudes of the axes. All quantitative kinematic methods
fault-plane solutions of earthquakes. Additional data present practical problems for field-based studies be-
are necessary to specify the deformation magnitude of a cause they require weighting the fault-slip data with the
fault, as discussed in detail below: the average displace- displacement and fault surface area. We consider sev-
ment and the fault surface area (for an exposed fault) or eral empirical scaling relationships for faults, including
the seismic moment (for an earthquake), fault gouge thickness and width, that allow estimation of
The classic work of Anderson (1951) was among the these parameters with field data. A simpler graphical
first attempts to relate fault geometry and kinematics to kinematic method can be applied if fault systems can be
driving stresses. More recently, considerable effort has shown to be scale-invariant.
been directed toward formulating inverse methods of Several of the assumptions made in kinematic analysis
determining stress--here referred to as dynamic (e.g. rotation, sampling and spatial homogeneity) are
methods--from fault-slip data using both numerical analogous to those made in dynamic analyses, however
techniques (e.g. Carey & Brunier 1974, Etchecopar et specific tests of each assumption are simple to carry out,
al. 1981, Armijo et al. 1982, Angelier 1984, Gephart & as demonstrated here. Techniques for analyzing kine-
Forsyth 1984, Michael 1984, Reches 1987, Gephart matically heterogeneous setsof fault-slip data produced
1988, Huang 1988) and graphical techniques (Arthaud by triaxial deformation, anisotropy reactivation, strain
1969, Angelier & Mechler 1977, Reches 1983, Aleksan- compatibility constraints and/or multiple deformations
973
974 R. MARRETrand R. W. ALLMENDINGER

are also presented. Finally, we compare the results of describing its kinematics by its scalar Mg and dividing by
fault-slip analyses of the 1971 San Fernando, California, the volume (V) of region of interest:
earthquake sequence using graphical and numerical Mg
kinematic methods as well as graphical and numerical Vuij -'- ~ (/~iti/). (3)
dynamic methods, and evaluate the strengths and weak-
nesses of each method. Single subscripts in i andj indicate vector quantities and
double subscripts in i and j indicate tensor quantities, all
in Cartesian co-ordinates using the Einstein summation
THEORETICAL BACKGROUND notation. A convention using the sense-of-slip is necess-
ary to uniquely express tl and ti. The Vu for all observed
The calculation of the incremental strain of a region faults are then added, yielding the average incremental
due to faults within it has already been solved for displacement gradient due to all of the faults studied
seismological problems (Kostrov 1974, Molnar 1983). (Vut):
The seismic moment (Mo; see Table 1 for symbols) of Vu~i = ~" Vuij.
(4)
earthquakes is simply related to average displacement
(Uave) and fault surface area (n) (Aki 1966): faults
Because of the small strain assumption, Vu t can be
Mo = I~Uaven, (1) decomposed into symmetric and antisymmetric parts,
where/~ is the elastic shear modulus. Closely related is yielding the incremental strain and rotation tensors,
the geometric moment (Sammis et al. 1987), a purely respectively. The eigenvectors of the symmetric part of
kinematic measure of deformation magnitude which Vut give the orientations of the principal incremental
simply omits the shear modulus: strain axes and the eigenvalues give their magnitudes.
Jackson & McKenzie (1988) argue that the asymmetry
Mg = Uaven. (2) of the above tensor is an artifact of Molnar's (1983)
If slip on the three-dimensional array of faults is much implicit assumption that the co-ordinate system is
smaller than the dimensions of the region, the average attached to the fault; thus the region, not the fault,
incremental displacement gradient (Vu) accommodated rotates. This assumption cannot be evaluated with fault-
by each fault can be calculated (Molnar 1983, who slip data alone and therefore they prefer Kostrov's
referred to this quantity as the "asymmetric moment (1974) symmetric tensor. This distinction is not as im-
tensor"). The tensor describing the kinematics of a fault portant as it first appears, because Kostrov's tensor and
is asymmetric because of the inherent rotation of simple the symmetric part of Molnar's tensor are identical.
shear deformation, and is calculated by forming the diad Thus, the asymmetric tensor potentially contains more
product of the unit average displacement vector (ti) and information, because its antisymmetric component de-
the unit normal vector to the fault plane (t~). Vu is scribes either the rotation of the region or the rotation of
determined for each fault by multiplying the tensor the faults.

Table 1. Symbols used in text

Symbol Parameter

u local displacement (varies with position on fault surface)

Umax maximum displacement of fault
uave displacement averaged over fault surface
tl unit average displacement vector
n surface area of fault plane
t~ unit normal vector to fault plane
t local fault gouge thickness (varies with position on fault surface)
w maximum fault width
e fault surface ellipticity
/~ elastic shear modulus
ct proportionality constant between u and t
c2 proportionality constant between Ureax and wZ/u2
c3 proportionality constant between uavc and tdmax
Mo seismic moment
M~ geometric moment
Ms geometric moment of largest fault in study area
MgZ geometric moment of second largest fault in study area
V volume of study region
Vu average displacement gradient tensor for a fault
Vut average displacement gradient tensor for all faults studied
N number of faults having M e ~ some specified value
A log N having log M e ~ 0
B negative slope of log N vs log M e relation for faults
 Kinematic analysis of fault-slip data 975

vector and the normal vector to the fault plane, and

make angles of 45° with each of the vectors (Fig. 1). The
sense-of-slip is necessary to distinguish between the two
axes. We emphasize that no interpretation is involved in
the process of determining the kinematic axes of a fault
from field measurements, because essentially one has
only converted the measurements into a fault plane
solution. Thus, the kinematic axes of a fault are merely
an alternative representation of the original data, a
representation which is particularly convenient for the
various geometric tests described below.
Contouring the shortening and extension axes of an
array of faults effectively averages them and provides
Fig. 1. Geometryof fault-slipkinematicsin lower-hemisphere,equal- descriptions of their directional distributions; we use the
area stereographicprojection.The kinematicaxes, slip directionand contouring procedure of Kamb (1959). A potential
pole to faultlie in a commonplane, with each of the kinematicaxes
forminganglesof 45"withboth the slipdirectionand the poleto fault, problem with contouring is that it treats the shortening
and extension axes as distinct rather than linked entities,
however in our analyses of over 1500 fault-slip data this
With this background, one can define explicitly the
has not proved important (Marrett et al. 1989).
nature of the so-called P- and T-axes familiar to seismol-
ogists. The names of the P- and T-axes are unfortunate, Bingham distribution statistics for axial data provide
because they suggest the stress terms 'pressure' and objective directional maxima of the shortening and ex-
'tension'. They are constructed by bisecting the ortho- tension axes of a fault array (Mardia 1972). The maxima
gonal nodal planes of a fault plane solution, and thus lie can be calculated separately or in a linked fashion by
at 45° to those planes. Analytically, however, the P- and counting one kind of kinematic axis as positive and the
T-axes are equivalent to eigenvectors of the symmetric other negative, as has been done for the analyses here.
part of the displacement gradient tensor for that fault. This procedure is identical to moment tensor summation
Thus, the axes are fundamentally kinematic in nature, with uniform weighting of the data. Because the results
representing the principal axes of the incremental strain of contouring and Bingham statistical analysis of an
tensor for the fault. We will refer to them below simply array of faults do not account for the magnitudes of
deformation, they will be related to the kinematics of the
as the shortening and extension axes. fault array only if the fault kinematics are scale-
invariant. This assumption must be tested (see "Weight-
GRAPHICAL AND NUMERICAL KINEMATIC ing test" below) for each data set.
METHODS
Moment tensor summation and fault-slip magnitude data
Ideally, collection of field data for fault-slip analysis
would include: fault plane orientation, slip direction, Measurements of average displacement and fault sur-
sense-of-slip, local bedding orientation, average dis- face area are needed to determine the magnitude of
placement and fault surface area. The fault-slip datum deformation accommodated by fault movement, yet
should be measured at a relatively planar part of the these quantities usually can be measured only where
fault which is at least subparallel to the megascopic mining (e.g. Gauthier & Angelier 1985)or seismic data
orientation of the fault. Local bedding orientation, used provide three-dimensional control. Several empirical
to perform a fold test (see below), should be measured relations make it possible to estimate the magnitude of
beyond the effects of any drag folding. Numerous fea- deformation accommodated by a fault for which Uav¢
tures indicative of sense-of-shear have already been and n are uncertain. These estimates are based on field
described in the literature (e.g. Arthaud & Mattauer measurements of fault gouge thickness and/or maximum
1972, Chester & Logan 1987, Gamond 1987, Means fault width.
1987). Because secondary fractures and fault plane sur- For our measurements of cataclastic faults, we de-
face features commonly are ambiguous indicators of fined fault gouge to be material displaced from its initial
sense-of-shear, as many indicators as possible are desir- location with respect to wall rock by more than its
able. Observations of sense-of-shear should be rated longest dimension. As such, this definition deviates from
according to relative reliability, in a manner concep- more classical definitions in that it has neither a mineral-
tually similar to quality ratings of earthquake locations, ogical nor a particle size connotation. Recent studies
have shown that particles in gouge have a fractal size
Graphicalmethod distribution (e.g. Sammis et al. 1987) and thus large
lenses of clasts meters in diameter can conceivably
The first step in a kinematic analysis is the graphical represent a coarse fraction of fault gouge.
construction of the principal incremental shortening and Measurement of fault gouge thickness (t) in the field
extension axes for a given population of faults. Each pair presents several problems. Given the presence of asperi-
of axes lies in the 'movement plane' containing the slip ties, t clearly varies from some maximum amount down
IG 1218-~
976 R. ]V[ARRE'Frand R. W. ALLMENDINGER

to zero as a function of position along a fault. However, Fault surface trace length usually is measured from air
consistent estimates can be made by choosing a tabular photographs or maps rather than measured directly in
part of each fault zone for measurement of gouge thick- the field. Because the complicated regions near fault tip
hess. The possible presence of unidentified horses pres- lines are commonly small compared with the length of
ents another problem, particularly for large faults in the fault, the uncertainty in trace length is not severe.
incompletely exposed regions. Drag folding and atten- More difficult is the assessment of the fault geometry at
dant bedding-parallel slip pose an additional problem depth and in the rock now eroded away, which is
for large faults, because they can obscure the boundaries necessary to relate fault surface trace length (which is
of the gouge zone by deforming adjacent wall rock. generally a chord in a simple elliptical fault model) to w.
Models of fault growth (Sammis et al. 1987, Cox & For many faults there is no alternative to assuming that
Scholz 1988, Power et al. 1988) predict a linear increase they are the same, which if incorrect will always lead to
of local fault gouge thickness with local displacement an underestimation of w and therefore of umax.
(u). Data from cataclastic faults with u ranging from To use one of the empirical relationships above in
10 -2 to 104 m fit these models to within an order of estimating the deformation magnitude of a fault (prefer-
magnitude (Scholz 1987, Hull 1988, this paper): ably using locally determined constants), one must first
relate t/ave with u and/or Umax, and also somehow evalu-
u = cit, (5)
ate n. The fractal nature of faulting (e.g. King 1983,
where cl is an empirical constant. We determined an Scholz & Aviles 1986, Turcotte 1986) suggests that the
average value of cl = 70 for Tertiary red beds of north- displacement functions of faults (u as a function of
western Argentina in an environment of horizontal position on a fault surface) might be scale-invariant.
shortening (Fig. 2); we will use this value below. Hull Detailed studies show that this is generally true (Mur-
(1988) independently determined a value of 63 for a aoka & Kamata 1983, Higgs & Williams 1987, Walsh &
wide variety of rock types including Mesozoic sand- Watterson 1987), although no data have been evaluated
stones deformedin horizontal extension, from faults with kilometers of displacement. This
Elliott (1976) suggested that, empirically, the surface implies a simple linear relationship between Umax and
trace length of a fault in plan view is linearly pro- t/ave:
portional to maximum displacement (Umax). Walsh &
Watterson (1988) argue that maximum displacement is //ave C3/'/max,
= (7)
proportional to the square of maximum fault plane where c3 is a constant which depends on the shape of the
width (w), defined as the maximum dimension of a fault displacement function. For example, c3 = 2/3 for an
plane normal to its slip direction. They show that elliptical displacement function and c3 = 1/3 for a tri-
Elliott's data and newly collected data are empirically angular displacement function (Fig. 3). Faults tend to
consistent, at least for fault widths under 100 km, with have displacement functions intermediate between
the following relationship: elliptical and triangular (Muraoka & Kamata 1983,
C2 -~
Uma x = ~ W-, (6) (a) u

where c2 is a variable related to the stress drop of

earthquakes averaging 2 x 10 -4 GPa 2 m -1 for faults in a . oum=,
variety of rock types with displacements ranging from
10° to 105 m. ~ii~:~~!~:i!~~:"~~
)!i~ili~i~i!i~ r
!

3 !, Oda. Carachi(N=23) I J rmax

2. = NW Argentina (N = 371
Y
-1. ** u 70 • t :::'"'":::::::::..

-2,

, r
I"max
-3 . , - , , • , -

-4 -3 -2 -1 0 Fig. 3. (a) Triangular and (b) elliptical displacement functions, show-

Io0 t {m} ing distance from the center of an elliptical fault (r) against displace-
ment (u). rrnax occurs at fault tip. Shaded regions indicate where u is
Fig. 2. Log-log plot of fault gouge thickness (t) against displacement within u~ve - ½uave. For triangular displacement function u~= = ]u~a x,
(u) for brittle fault zones in northwestern Argentine Andes. The line is and 44% of fault surface falls in shaded region. For elliptical displace-
a linear regression to the data from Quebrada Carachi: u = 707)~; ment function u~= = ~,~. . . . and 89% of fault surface falls in shaded
R 2 -- 0.99. region.
 Kinematic analysis of fault-slip data 977

Higgs & Williams 1987, Walsh & Watterson 1987), so we use of estimates from the scaling relationships is unjusti-
will use c3 = 1/2 below, fled.
Characterizing the relationship between u and uave is
less trivial. If one were to measure u at many points on a
fault in a random way and average them, the result APPLICATION OF KINEMATIC METHODS AND
would be a good approximation of uav¢. In fact, if one TESTS OF ASSUMPTIONS
were to measure u at only one randomly chosen point on
a fault with an elliptical displacement function, the The accuracy of analytical results from kinematic
probability is 89% that u is within 50% of uav~ (Fig. 3). fault-slip analysis is subject to uncertainties which prob-
When the error is greater than 50%, Ua~e is always ably depend less on the accuracy of field measurements
underestimated. Because the displacements observed than they do on the validity of assumptions: the weight-
for faults in typical arrays vary by several orders of ing of fault-slip data, possible reorientation of the fault-
magnitude, errors associated with assuming that tt is slip data, the degree to which sampling is representative
statistically the same as Ua~e should be relatively small, of the entire fault population, and, in a limited sense, the
Thus, we assume: spatial homogeneity of strain. We compare the methods
and illustrate some tests of assumptions with a data set
Uave ~ U. (8) collected in a trial study area, known as Quebrada
Kanamori & Anderson (1975) successfully explained Carachi (Fig. 4) in the Andes of northwestern Argentina
several empirically determined scaling laws of earth- (Marrett 1990). Fifty-nine measurements were made of
quakes using a model in which the surface area of slip is 46 faults (large faults were measured in several different
proportional to the square of average slip. Earthquakes places). Local displacement was measured for 23 of the
and faults are not identical phenomena, because a large faults and local fault gouge thickness was measured for
fault is the product of many earthquakes which have all faults studied. The results of moment tensor sum-
occurred in approximately the same place. The results of mation for the data set from the study area, as well as for
Walsh & Watterson (1988) imply that n is linearly subgroups of the data defined by ranges of u, are shown
proportional to/.lave, as seen by expressing n in terms of in Table 2. The graphical method shows simple point
w and substituting equations (6) and (7): maxima of shortening and extension (Figs. 5a & b)
~w 2 ~/A2 which are similar to the results of moment tensor sum-
n. . . . Ua~e, (9) mation (Table 2).
4e 4ec2c 3

where e is the ellipticity of the fault surface. Although Weighting test

few data sets are available, data from both normal and
thrust faults suggest that e varies between 2 and 3 (Walsh Weighting of fault-slip data is done in moment tensor
summation with the geometric moment. Although the
& Watterson 1987); we will use e = 2 below.
graphical kinematic method assumes that fault kinema-
Substituting equations (5)-(9) into equation (2) yields tics are scale-invariant, weighting can be qualitatively
Mg in terms of t and w: assessed by separating a data set into subgroups of faults
-- n/'/'2Cl t 2 (10) of different sizes and comparing their kinematics.
Mg 4e--ec2c ~ The data from the study area, with displacements
spanning five orders of magnitude, were separated into
~c2c3 w 4. (11) five subgroups and the kinematics of each subgroup was
Mg - --~e/.~ analyzed using both Bingham distribution statistics and
These relationships are sufficient for relative weighting moment tensor summation (Figs. 5c & d and Table 2).
of the geometric moments among observed faults, how- The results of both methods show tight clustering of
ever they are insufficient for determining the absolute shortening axes and less tight clustering of extension
axes; furthermore, the results of graphical analysisare
deformation magnitude for each fault. Using the values consistent with the results of moment tensor summation.
of c 1, c2, c3 and e cited above and kt = 12 GPa (Walsh & Fault kinematics appear to be scale-invariant f o r these
Watterson 1988), approximate relationships for a hard data, as well as for most other data sets that we and
sandstone are: others have collected. If faults generally have scale-
Mg ~ (3 x 109m)t2 (12) invariant kinematics, this represents another fractal
Mg ~ (3 x 10 -7 m - l ) w 4. (13) characteristic of the faulting process (e.g. King 1983,
Scholz & Aviles 1986, Turcotte 1986, Power et al. 1987,
We emphasize that these relationships are not as Sammis etal. 1987, Barton etal. 1988).
precise as they might appear due to the cumulative
uncertainties of the empirical constants, which can be as Fold test
large as an order of magnitude. They are primarily
intended for making order of magnitude estimates of Post-faulting reorientation of a fault-slip datum
geometric moment where the parameters of faults in a changes the orientations determined for the kinematic
typical array span many orders of magnitude. Subtler axes. The significance of differential rotation about
978 R. MARRE'IWand R. W. ALLMENDINGER

~ 65 °
' 45'W

,
•i:i ::::::::::::::
' '.'.'.'.'.'.'.'."
/ 24°

\ E •: • **~ •: * :o: ~-'~ll~j. Bernardo .~.:~:::~"

I b'..o....*..o*...-.t~. / j.-~.Z-?H
study I * : ] ' . - ~ . % ~ - ~ " ~ . f " / r. ..... -¢.x...
area. ~.°°:°~*:°***.~ ,~ / /}~!:?

5 km

Alluvium, landslides,
~--~ stream terraces,
take sediments
Continental dastic
strata (upper unit)

Continental ctastic
strata (lower unit)

Salta Group

Mesbn Group,
Santa Victoria Group

Tastil Granite

Puncoviscana
Formation

Fig. 4. G e o l o g y of Q u e b r a d a del T o r o , n o r t h w e s t e r n A r g e n t i n a ( m o d i f i e d from M a r r e t t 1990). Inset s h o w s political b o r d e r s

in solid lines and a r e a s a b o v e 3 k m a v e r a g e e l e v a t i o n in d a r k s h a d i n g .

horizontal axes can be characterized for a given data set slip directions can be inconclusive because many data
by using a fold test similar to those used in paleomagne- sets with highly variable fault and/or slip orientations
tic studies. In the fold test, the kinematic axes are have coherent kinematics. If the unfolding produces
rotated by the amount necessary to return local bedding kinematics which are more coherent than the kinematics
to horizontal. Unfolding only the fault planes or only the of the faults in present geographic orientation, then

T a b l e 2. C o m p a r i s o n of results f r o m m o m e n t t e n s o r s u m m a t i o n ( w e i g h t e d by fault g o u g e t h i c k n e s s )
a n d l i n k e d B i n g h a m d i s t r i b u t i o n statistical analysis of fault-slip d a t a from Q u e b r a d a Carachi.
D i r e c t i o n s given in t r e n d a n d p l u n g e

Displacement Moment tensor Linked Bingham

range No. of No. of summation distibution maxima
(m) measurements faults Shortening Extension Shortening Extension

all faults 59 46 316 °, 17° 102 °. 70 ° 316 °. 17° 89 °, 67 °

100-1000 8 2 316 °. 17 ° 103 °, 70 ° 316 °, 15 ° 100 °, 72 °
10-100 10 5 333 °, 13 ° 67 °, 17° 314 °. 20 ° 106 °, 67 °
1-10 10 8 322 ° , 16 ° 99 ° , 69 ° 323 °. 22 ° 125 ° , 67"
0. i - 1 23 23 309 °. 04 ° 46 °, 63 ° 314 ~, 16 ° 68 °. 56:
0.01-4). 1 8 8 310 °, 13° 178 °, 70 ° 315 °, 10° 15&, 79 °
 Kinematic analysis of fault-slip data 979

not occur. Paleomagnetic data have not been collected

in this area, so the importance of vertical axis rotation is
uncertain.

Sampling test

Due to practical limitations inherent to field-based

Shortening Axes ExtensiAxes

on studies, only a small portion of the faults that exist in a
given area are typically sampled. For the same reasons,
it is difficult to determine whether the observed faults
~ are representative of the entire fault population. Be-
../~ff'~~(d) cause many characteristics of faults and earthquakes are
. = fractal, we conjecture that faults follow a power-law
number-geometric moment relation, analogous to the
frequency-magnitude relation for earthquakes:
log N = A - B log Mg, (14)
where N is the number of faults having geometric mo-
Bingham Maxima Moment Tensor Sums ments greater than or equal to Mg, B is analogous to the

( e ) ~ ~ b-value of earthquakes (because faults and earthquakes

are different phenomena, it is unnecessary that the two
be identical) and A is a measure of the total number of
faults. One may determine A and B for a specific area
from the geometric moments of the largest two faults
(Mg1 and M2), however the use of just two faults cannot
test the assumption of a power-law number-geometric
moment relation. Alternatively, B might be determined
Unfolded Unfolded by analyzing outcrops at which all faults (above a certain
shortening Axes Extension Axes size threshold) can be identified and measured. Prefer-
Fig. 5. (a) & (b) Fault-slipkinematicsof data from Quebrada Carachi, ably, the size range of such outcrops would be as great as
northwestern Argentina (59 measurements of 46 faults). Solid dots possible and the largest faults in the study area would
represent kinematicaxes, open boxes represent Binghamdistribution also be analyzed to constrain m. This would allow one to
maxima, and shades represent distributions of 2a contour intervals evaluate how well the fault population actually follows a
(e.g. white areas indicate fewer axes at countinggrid points than that
found in a uniform distribution minus la; adjacent low density grey power-law number-geometric moment relation. Using
areas indicategrid points havingnumbersof axeswithin _loofthatin A and B, one can predict the number of faults that exist
a uniform distribution; adjacent slightly higher density grey areas at any specific size range and quantify the percent of
indicategridpoints havingnumbersof axeswith 1-3trmore than that in
a uniform distribution; all contour diagrams conform). (c) & (d) those faults actually observed.
Weightingtest of fault-slip data from Quebrada Carachi. Largest to Another measure of sampling is the portion of the
smallest boxes represent analyses of fault-slip data subgroups with total geometric moment sampled. The total geometric
displacements of 100-1000, 10--100, 1-10, 0.1-1 and 0.01-0.1 m,
respectively. Solid boxes represent shortening directions and open moment due to all faults can be written as a function of B
boxes represent extensiondirections. (e) & (f) Fold test of fault-slip and MgI by assuming that the fault array is truly self-
data from Quebrada Carachi. Contour diagrams of kinematic axes similar:
which have been independentlyrotated by the amount necessaryto
return local beddingto horizontal. 2 Mg=Mgt(l+~_~/a+ 1 1 ]
.... (15)

folding probably post-dates faulting; otherwise folding f~,lts

probably pre-dates faulting. Regional tilting, domino- Equation (15) converges only for B < 1; this is reason-
style block rotation, and vertical axis rotation cannot be able because B = 1 corresponds to the situation in which
detected with this technique, but can be addressed using the sum of geometric moments of faults in each order of
regional geologic and paleomagnetic data. magnitude range are about the same. For example,
Several NE-verging closed folds cross the study area faults with displacements of 1-10 mm would, by virtue of
(Fig. 4). Contour diagrams of the kinematic axes from their tremendous numbers, have as much total geo-
the study area are much less coherent in the unfolded metric moment as faults with displacements of 1-10 kin.
configuration (Figs. 5e & f) than in the folded configur- Were B = 1 for faults, construction of balanced cross-
ation (Figs. 5a & b), suggesting that most faulting post- sections based only on the largest faults would be a
dates folding. Approximately horizontal bedding useless exercise.
characterizes large areas in and around the study area Due to the excellent exposure in the study area, we
and suggests that regional tilting has not been important, believe that the two biggest faults were identified (Fig.
Major thrust faults flanking the study area dip both NW 6). Based on this assumption, B = 0.38. The distance on
and SE and suggest that domino-style block rotation did the number-geometric moment plot between the model
980 R. MARRE'Vrand R. W. ALLMENDINGER

2 ::i::ii!
...... o 60
:::::::::::::::::::::::::::::::::::::::::::......
:::: o
log N " " '~ • ~, ~ ; ~ i ~i~i!i!~!i!!!ii!i!i::!i~ii::.. ~=
o 40

O
"'":':'::::~ii~ u " ~ 20
1 8
0 o =
0,.
4 6 8 10 0
10gMg kn3} g ~° o~' c~- °o.
Fig. 6. Sampling test of fault-slip data from Quebrada Carachi (46 ~o_ ~
faults measured). Log-log plot of geometric moment (Mg) against the Displacement {m}
number of faults (N) having a geometric moment greater than ob-
served values of Mg. The line represents a model assuming that the two Fig. 7. Bar graph of percent of total geometric moment sampled for
largest faults observed are in fact the largest two faults in the study area faults with displacements of different orders of magnitude at Que-
and that the entire fault population fits power-law number-geometric brada Carachi, assuming that the two largest faults observed are in fact
moment relationship. Height of shaded area represents degree of the largest two faults in the study area and that the entire fault
undersampling, population fits a power-law number-geometric moment relationship.

line and the points representing the observed faults is a measurements from a region of northwestern Argentina
measure of the faults that should exist but which were about 400 km long and 200 km wide show remarkably
not measured. For example, there should be 13 faults little variation during the Tertiary and a more complex
having Mg > 10s m 3 but only five were measured in the but coherent pattern during the Quaternary (Marrett et
field, so about 38% of those faults were sampled, al. 1989).
For B = 0.38,

M g = 1.3M~. (16) INTERPRETATION OF KINEMATIC

taults HETEROGENEITY
The largest fault zone alone accounts for nearly 80% of
the total geometric moment. This suggests that the Kinematically heterogeneous faulting, represented by
prospectsofconstructing a useful balanced cross-section girdle or multi-modal patterns of shortening and/or
of the study area are good. The faults measured in the extension axes, can be produced by several mechanisms:
field represent 93% of the total geometric moment (Fig.
7), indicating that sampling is indeed representative of
the entire fault population for this data set. However, (a) ~ NE Domain ,....-a-...........,~ (b)
these estimates are based on the untested assumption ~;~J( \ / ..----r---~\
that the faults follow a power-law number-geometric
moment relation.

Spatial homogeneity test

In principle, the kinematic methods do not assume

that fault-slip kinematics are spatially homogeneous.
However, group analysis of spatially heterogeneous Shortening Axes Extension Axes

data can obscure meaningful variation in the data by

averaging statistically distinct subgroups. Testing for
spatial homogeneity assures that such variation is recog- ~ .(~: \ \c ~ ) s w Domain (d)
nized. The degree of spatial homogeneity can be evalu- ~ ~ ] ) ~ X
ated for a given data set by analyzing subgroups of faults ~
from different domains and comparing their kinematics.
Comparing the fault kinematics at the different ~J
measurement sites in the study area indicates that kine-
matics are spatially homogeneous; for the purpose of
illustration, the data were separated into two subsets
representing fault measurements from the NE and the Shortening Axes Extension Axes
SW parts of the study area (Fig. 8). The kinematics of Fig. 8. Spatial homogeneity test of fault-slip data from Quebrada
the two subsets are statistically indistinguishable. Com- Carachi. Contour diagrams of (a) shortening axes and (b) extension
axes of faults in NE part of study area (N = 36). Contour diagrams of
monly, fault kinematics are statistically homogeneous (c) shortening axes and (d) extension axes of faults in SW part of study
over large areas. For example, more than 1500 fault area (N = 23).
 Kinematic analysis of fault-slip data 981

triaxial deformation, anisotropy reactivation, strain served in the contour diagram of the data. (Fig. 10d).
compatibility constraints and/or multiple deformations. Poles to faults and slip directions commonly scatter
Geologic evidence independent of fault-slip data pro- much more than the kinematic axes determined from
vides the clearest indications of these mechanisms, them. The fault sets and slip directions have relatively
although the absence of such evidence for a specific poor orthorhombic symmetry (Fig. 10e)which is much
mechanism does not demonstrate its inactivity. Graphi- closer to that of two conjugate pairs, one with dip-slip
cal analysis of fault-slip data may allow more complete faults striking NW-SE and the other with dip-slip faults
interpretation of kinematically complex faulting be- striking NE-SW. Thus, triaxial deformation apparently
cause each mechanism produces distinct patterns of does not explain the kinematic heterogeneity in this data
poles-to-faults and slip directions (Fig. 9) which may be set.
used to qualitatively assess the importance of each
mechanism in a specific case. Anisotropy reactivation
The data previously discussed were kinematically
homogeneous (Figs. 5a & b). However, data from a Simple regional deformations can reactivate local,
larger area in NW Argentina (Fig. 4) are kinematically pre-existing anisotropies that are not ideally oriented for
heterogeneous because there is a bimodal pattern of accommodating the overall deformation, thus produc-
shortening axes (Figs. 10a & b). This area, known as ing locally heterogeneous triaxial deformation. Aniso-
Quebrada del Toro (Marrett 1990), includes the study tropies which exist only in units beneath those of interest
area previously discussed, can nevertheless control the kinematics in the overlying,
previously unfractured rock. Because the orientation of
Triaxial deformation a pre-existing anisotropy is arbitrary with respect to the
reactivating deformation, slip on the plane of anisotropy
Reches (1983) showed that triaxial deformation pro-
ducesthreeorfoursetsoffauhsarrangedwithortho- ~ ~
rhombic symmetry and an equal number of distinct slip
directions, also having orthorhombic symmetry (Fig.
9a). Such patterns have been observed in the field and in
experiments (Donath 1962, Aydin & Reches 1982,
Reches & Dieterich 1983, Krantz 1988). The degree to
which the model of Reches (1983) fits a given data set
provides an indication of how important triaxial defor-
mation was in the development of heterogeneous kine- Shortening Axes Extension Axes

matics. The presence of mutually cross-cutting fault sets

supports the interpretation of triaxial deformation. ~ ~
The study area is dominated by four sets of major
faults, although they are defined rather indistinctly by
maxima in the contour diagram of poles to faults (Fig.
10c). Four distinct maxima of slip directions are ob-

(a) (b) Poles to Faults Slip Directions

Triaxial Deformation Anisotropy Reactivation

(e) (d)

Fig. I0. (a) & (b) Fault-slip kinematics of data from Quebrada del
Toro, northwestern Argentina (N = 222). (c) Poles to faults and (d)
slip directions for data from Quebrada del Toro. Note the greater
coherence of kinematic axes than of poles to faults or slip directions.
(e) Analyses of triaxial deformation and strain compatibility for fault-
slip data from Quebrada del Toro. Maxima of poles to faults in open
Strain Compatibility Multiple Deformations dots, fault sets in great circles (different conjugate sets represented by
different width lines), and maxima of slip directions in solid dots. (f)
Fig. 9. Models of mechanisms that can produce kinematic heterogen- Analysisof anisotropy reactivation for fault-slip data from Quebrada
eity. Great circles represent sets of fault planes, open dots represent del Toro. Contour diagram of poles to bedding (N = 248) of green-
poles to fault sets, and solid dots represent dominant slip directions of schist basement underlying faulted Tertiary strata at Quebrada del
fault sets. Toro (Omarini 1983).
982 R. MARRETrand R. W. ALLMENDINGER

may be unsystematically oblique compared with slip on completely explain their slip directions and hence can-
newly forming faults. Thus, reactivation of a basement not completely explain their kinematic heterogeneity.
anisotropy should produce a fault set and slip directions
which are not related to the other fault sets by a conju- Multiple deformations
gate or orthorhombic symmetry (Fig. 9b). Anisotropy
reactivation may produce a positive correlation between Multiple deformations commonly produce hetero-
the orientations of a fault set and basement anisotropy geneous fault-slip kinematics. A special kind of aniso-
or between spatial changes in fault-slip kinematics and tropy reactivation can result when pre-existing faults are
changes in basement anisotropy orientations, reactivated, generating a second set of striae. Thus,
In the study area, a sequence of complexly folded, individual faults may show evidence for slip in two or
greenschist facies flysch, in which bedding planes define more different directions and a single set of faults may
the most important anisotropies (Fig. 10f), lies un- have widely varying slip directions (Fig. 9d). The fault-
conformably beneath the faulted strata (Omarini 1983). slip kinematics of one deformation might also be incom-
Comparison of basement bedding poles with poles to patible with the kinematics of another deformation.
faults (Fig. 10c) shows that the E-plungingpoles to faults Independent evidence for multiple deformations in-
match basement anisotropy somewhat, but the west- clude systematic cross-cutting relations between fault
plunging poles to faults are unrelated to basement aniso- sets and mutually exclusive chronologic constraints on
tropy. The two fault sets with poles plunging to the west fault sets.
are not conjugates because their slip directions are not The slip directions for NW-dipping faults in the study
appropriately oriented (Fig. 10e). Therefore anisotropy area have highly variable slip directions, while the slip
reactivation alone cannot produce the observed kinema- directions for SW-dipping faults are mostly coherent
tic heterogeneity. (Fig. 11). This indicates that the NW-dipping faults have
been reactivated whereas the SW-dipping faults have
Strain compatibility not been. Some of the larger NW-dipping faults have
multiple sets of fault striae indicating both dip- and
Strain compatibility between two differently oriented strike-slip movement, confirming the inference based on
faults or fault segments requires that they both must slip kinematic analysis. Furthermore, it appears that some of
parallel to their line of intersection if no additional the NW-dipping faults were reactivated during the
structures form. A larger difference in the orientations movement of the SW-dipping faults and thus suggests
of the faults or fault segments produces a larger differ- that the former are older than the latter. Strain compati-
ence in their kinematics. This constraint will result in bility may have controlled the slip directions of both
multiple sets of faults which have similar slip directions fault sets during the later phase of deformation.
(Fig. 9c). Geometric analysis of the fault orientations and slip
Two slip direction maxima for the data from the study directions for faults in the study area suggest that mul-
area correspond well with intersections of fault sets (Fig. tiple deformations are mostly responsible for producing
10e). A detailed comparison of slip directions for the the kinematic heterogeneity observed there. This result
faults which most tightly define the NW- and SW- is consistent with field relations: the younger NW-and
dipping fault sets (Fig. 11) shows that the SW-dipping SE-striking faults consistently cut the older NE- and
faults have mostly coherent slip which is subparallel to SW-striking faults wherever both are present (e.g. Fig.
the intersection of the fault sets. The NW-dipping faults 4). Based on this criterion, the data can be separated
have quite variable slip, some subparallel with the fault objectively into two kinematically homogeneous sub-
set intersection but most faults have more nearly dip-slip groups representing the two phases of deformation (Fig.
movement. Strain compatibility may have controlled 12).
some of the slip on these fault sets, but it cannot

t,.
COMPARISON OF DYNAMIC AND KINEMATIC

t The 1971 San Fernando, California, earthquake se-

quence is well-studied because of its proximity to a
.[ ~ \ _ + ] . major metropolitan area. Several attempts have been
made to understand the dynamics and kinematics of the
" \"~.~ / faulting based on fault plane solution and seismic mo-
ment data (e.g. Whitcomb et al. 1973, Hadley & Kana-
mori 1978, Langston 1978, Gephart & Forsyth 1984,
Julien & Cornet 1987). The same data are analyzed here
Fig. 11. Analysisof multiple deformationsfor fault-slip data from using the graphical kinematic method, moment tensor
Quebrada del Toro. Individualfaults and slip directionsrepresented summation and the P-T dihedra stress inversion (Ange-
by great circlesand solid dots. respectively.(a) Faults striking 190-
220* and dipping 30-55* (N = 13). (b) Faults striking 120-150° and iier & Mechler 1977) to complement published analyti-
dipping 30-55 ° (N = 7). cal results from grid search stress inversion (Gephart &
 Kinematic analysis of fault-slip data 983

Grid Search P-T Dlhedra

Younger Shortening Axes Younger Extension Axes

• Shortening Axes Extension Axes

Older Shortening Axes Older Extension Axes

Fig. 12. Kinematically homogeneous subsets of fault-slip data from

Quebrada del Toro. Data separated according to whether the kinema-
tics of each fault measurement most closely resemble the younger or , ,f~ \ ( e ) / ~ ~ (f)
the older kinematic pattern. (a) & (b) Younger deformation (N = 96). / \
(c) & (d) Older deformation (N = 126).
o3 .s

Forsyth 1984) and iterative numerical stress inversion

(Julien & Cornet 1987). Integrating these results, we can
understand the differences among the analytical results " - 45 o21~ 3 6

in terms of the geometry of the fault system. The results Analytical Models Poles to Nodal Planes
of the various analyses are shown in Table 3 and Fig. 13.
The results of grid search stress inversion (Fig. 13a
a n d T a b l e 3) (Gephart & Forsyth 1984) a n d t h e i t e r a t i v e Fig. 13. Analyses of fault-slip data from 1971 San Fernando, Califor-
n u m e r i c a l s t r e s s i n v e r s i o n (Table 3) (Julien & C o r n e t nia, earthquake sequence (N = 76). Best models indicated with open
1987) are virtually identical. The results of P-T dihedra boxes for cq, open dots for 03, solid boxes for shortening directions,
and solid dots for extension directions. (a) Results of grid search stress
stress inversion are similar to the results of the other inversion with 99% confidence area in light shadin~ruling and 95%
stress inversions, both in terms of the directional distri- confidence area in dark shading/ruling (analysis provided by J.
butions of good s t r e s s m o d e l s (Fig. 13b) and the best Gephart). (b) Results of P-T dihedra stress inversion with overlap of
m o d e l (Table 3). The m o m e n t t e n s o r summation (Table 56 dihedra in light shading/ruling and overlap of 66 dihedra in dark
shading/ruling. (c) Shortening and (d) extension axes analyses of
3) is dominated by the main shock, which was signifi- graphical kinematic method. (e) Comparison of best models. 1: grid
cantly different than many of the aftershocks. Because search stress inversion (Gephart & Forsyth 1984); 2: iterative numeri-
cal stress inversion (Julien & Cornet 1987); 3: P-T dihedra stress
of this, the kinematic scale-invariance assumption of the inversion; 4: moment tensor summation; 5 and 6: graphical kinematic
graphical analysis is not rigorously met. Contours of the analysis. (f) Poles to faults and slip directions for 1971 San Fernando
shortening and extension axes (Figs. 13c & d) are domi- earthquake sequence data. Contour diagram of poles to nodal planes
of fault-plane solutions (N = 152; each fault-plane solution has two
nated by the numerous aftershocks and appear similar to nodal planes). Open box represents the pole to fault of the main event
the stress analysis methods. The graphical analysis does, and the solid box represents the slip direction of the main event.

Table 3. Comparison of best models from analyses of fault-slip data from 1971 San
Fernando earthquake sequence (N = 76). Results of grid search stress inversion and
numerical iterative stress inversion from Gephart & Forsyth (1984) and Julien & Cornet
(1987), respectively. Directions given in trend and plunge

Maximum principal Minimum principal

compression or compression or
Method of analysis shortening direction extension direction

Grid search stress inversion 187°, 07° 84° , 62°

Numerical iteration stress inversion 189° , 07° 89°, 59°
P-T dihedra stress inversion 173°, 05° 75°, 57°
Moment tensor summation 31° , 05° 139°, 75°
Graphical kinematic method 195°, 01 ° 94°, 85°
168°, 05° 75°, 37°
984 R. MARRETTand R. W. ALLMENDINGER

Thrust Shortening Thrust Extension planes of the fault plane solutions show a moderately N-
~ ~ plunging point maximum and a distinct girdle with
moderately SSW- and shallowly SE-plunging maxima
(Fig. 130. The SSW- and SE-plunging maxima indicate
the orientations of the main fault and the lateral ramp,
respectively, and the N-plunging maximum indicates
their common slip direction. Two kinematically homo-
geneous subsets of the data can be obtained by separat-
J,~_____...,"'-.--r-- . ' ~ ing the faults with SSW-plunging poles from faults with
~~Z'---~_~~~ SE-plunging ones (Fig. 14 and Table 3).
The differences between dynamic and kinematic re-
~ ~ suits(Fig. 13e a n d T a b l e 3 ) might b e i n t e r p r e t e d i n
several ways: (1) the dynamic and/or kinematic methods
~,. s ~ . . - - ~ / / / ~ ~ may have failed; (2) their differences may be statistically
insignificant; or (3) strain might have been non-coaxial
~- with stress. The differences are not so great that one
(c' _.,..~ .=. must consider the analyses failures. Indeed, there are
considerable overlaps of acceptable models among the
results of grid search inversion, P-T dihedra inversion
and graphical kinematic analysis. Yet the consistency of
+ the dynamic analyses, both internally and with results
from adjacent areas in the Transverse Ranges and along
the San Andreas Fault (Jones 1988), may suggest that
the best models are more accurate than the statistics
indicate. If so, the San Fernando earthquake sequence
Ramp Shortening Ramp Extension
data represent strain which is non-coaxial with stress.
Fig. 14. Kinematically homogeneous subsets of fault-slip data from The simplest explanation of the non-coaxiality is that the
1971 San Fernando earthquake sequence. Data separated according to
whether they have a pole to nodal plane closer to the moderately SSW- strain compatibility constraint, imposed by the lateral
plunging maximum (N = 38) or the shallowly SE-plunging maximum ramp in the main fault, prevents slip from occurring in
(N = 38). (a) & (b) Kinematics of fault-plane solutions with SSW- the ideal direction.
plunging (thrust) nodal planes. (c) & (d) Kinematics of fault-plane
solutions with SE-plunging (lateral ramp) nodal planes. Bingham The kinematics of subsets of the aftershock popu-
distribution maxima of shortening and extension indicated with solid lation defined by local magnitude (M) ranges show that
boxes and dots, respectively. In center is a schematic structural both main fault andlateral ramp fault plane solutions are
contour map showing simplified contours, in kin, on the main fault
plane (modified from Whitcomb et al. 1973). Dark 0 contour rep- represented in all ranges of M < 5.0 (Fig. 15) but not
resents surface break, above. The main shock is the only event with M >5.0 for
which a fault plane solution was determined. The
hypothesis of scale-invariant fault kinematics seems to
however, display a bimodal distribution of extension break down at larger magnitudes, unless there was a
axes, hinting at an explanation, large lateral ramp event for which a fault plane solution
The differences between the results of the two kine- has not been determined. An unlocated coseismic event
matic analyses are a product of kinematic heterogeneity of M = 5.8 occurred 30 s after the main shock began
within the data set. Whitcomb et al. (1973) noted that (Hileman et al. 1975). This event may mark the begin-
two distinct kinds of fault plane solutions are common in ning of activity on the lateral ramp of the main fault once
the aftershock population: thrust events similar to the the dislocation tip propagated to it. If this is the case,
main shock and strike-slip events along a steeply NW- then the hypothesis of scale-invariant fault kinematics
dipping fault or segment of the main fault. The extension would be consistent with the data.
axes of the two kinds of fault plane solutions are suf- The various dynamic and kinematic methods of fault-
ficiently different that they produce the distinct maxima slip analysis differ in terms of assumptions, computa-
(Fig. 13d) whereas the shortening axes are similar (Fig. tional intensity, and usefulness. While it may be argued
13c). Because the results of moment tensor summation that the assumptions of kinematic analysis are not obvi-
are dominated by the main shock (thrust), they mostly ously better than those of dynamic analysis, it may be
reflect the kinematics of the thrust events, that the former are more readily testable. The P-T
Whitcomb et al. (1973) suggested that the presence of dihedra stress inversion and the graphical kinematic
a large lateral ramp in the main fault (Fig. 14) caused the method have the advantages that they are based on
kinematic heterogeneity of the San Fernando earth- simple graphical constructions. As such they are more
quake sequence data, invoking the mechanism referred suitable for preliminary analysis in the field than the
to here as strain compatibility (Fig. 9c). Aftershock other methods, which are to varying degrees computa-
locations and surface geologic data indicate that the tionally intensive. Methods which produce results in
main fault dips moderately to the NNE and that the graphical form (the grid search stress inversion, the P-T
lateral ramp dips steeply to the NW. Poles to the nodal dihedra stress inversion and the graphical kinematic
 Kinematic analysis of fault-slip data 985

(a) ~ 4.0 < M < 4 . 9 ~ (b) A largely unexplored facet of fault-slip analysis is the
potential to evaluate strain compatibility, material
anisotropy, and material heterogeneity by integrating
+ dynamic and kinematic results, such as done in a quali-
tative way here for the San Fernando earthquake se-
quence. This emphasizes that dynamic and kinematic
analyses are complementary and that both types of
(c) , 3.5 < M < . 3 ~ . . . f ~ d ) analysis should be employed routinely in studies of fault-
slip data.

Acknowledgements--We thank Trent Cladouhos, John Gephart,

Martha Grier, Geoff King and Bob Krantz for their criticism and
suggestions during the development of these ideas; nonetheless they
may not agree with all of the conclusions. Reviews by Jacques Angelier
and two anonymous reviewers were very useful. This work was
supported by a Fulbright grant, a Harold T. Stearns Fellowship
(e) 3.0 < M < 3 . 4 ~ (f) Award, an A A P G Grant-in-Aid and two Sigma Xi Grants-in-Aid to
R. Marrett and by National Science Foundation Grant EAR-8519037
to R. W. Allmendinger. Institute for the Study of the Continents
contribution No. 129.
+

REFERENCES

Shortening Axes Extension Axes Aki, K. 1966. Generation and propagation of G waves from the
Fig. 15. Fault-slip kinematics of data from 1971 San Fernando earth- Niigata earthquake of June 16, 1964, 2, Estimation of earthquake
moment, released energy, and stress-strain drop from G wave
quake sequence representing different ranges of local magnitude (M). spectrum. Bull. Earthquake Res. Inst. Tokyo 44, 73--88.
Open boxes represent kinematics of the main event (M -- 6.4). (a) & Aleksandrowski, P. 1985. Graphical determination of principal stress
(b) Kinematics of events with 4.0 -< M --- 4.9 (N = 17). (c) & (d) directions for slickenside lineation populations: an attempt to
Kinematics of events with 3.5 -< M <- 3.9 (N = 30). (e) & (f) modify Arthaud's method. J. Struct. Geol. 7, 73-82.
Kinematics of events with 3.0 <- M <- 3.4 (N = 28). Anderson, E. M. 1951. The Dynamics o f Faulting and Dyke Formation
with Applications to Britain. Oliver & Boyd, Edinburgh, U.K.
method) are more useful than those which provide only Angelier, J. 1984. Tectonic analysis of fault slip data sets. J. geophys.
Res. 89, 5835-5848.
a best model and no indication of the directional u n c e r - Angelier, J. & Mechler, P. 1977. Sur une methode graphique de
tainty. Graphical kinematic analysis may provide recherche descontraintesprincipalesegalment utilisable en tecton-
greater directional resolution than the other methods ique et en seismologie: la methode des diedres droits. Bull. Soc.
g~ol. Fr. 19, 1309-1318.
which did not discern the heterogeneity of the San Armijo, R., Carey, E. & Cisternas, A. 1982. The inverse problem in
Fernando earthquake sequence data. microtectonics and the separation of tectonic phases. Tectonophy-
sics 82, 145-160.
Arthaud, F. 1969. Mrthode de d6termination graphique des directions
de raccourcissement, d'allongement et intermrdiare d'une popu-
CONCLUSIONS lation de failles. Bull. Soc. g#ol. Fr. 11,729-737.
Arthaud, F. & Mattauer, M. 1972. Sur l'origine tectonique de certains
Kinematic analysis of fault-slip data using the tech- joints stylolitiques parall~les ala stratification; leur relation avec une
niques outlined here provides quantitative and qualitat- phase de distension (exemple du Languedoc). Bull. Soc. g~ol. Fr.
14, 12-17.
ive tools for understanding deformation in both modern Aydin, A. & Reches, Z. 1982. Number and orientation of fault sets in
and ancient orogens. The accuracy of constants relating the field and in experiments. Geology 10, 107-112.
fault gouge thickness and fault width with displacement Barton, C. C., Samuel, J. K. & Page, W. R. 1988. Fractal scaling of
fracture networks, trace lengths, and apertures (abs.). Geol. Soc.
may be improved by systematic collection of more data. Am. Abs. w. Prog. 20, A299.
But even without explicit evaluation of the constants, Bott, M. H. P. 1959. The mechanisms of oblique slip faulting. Geol.
the scaling relationships are sufficient for the relative Mag. 96,109-117.
Carey, E. & Brunier, B. 1974. Analyse throrique et numrrique d'un
weighting of data and moment tensor summation. The mod~le m~canique 616mentaire appliqu6 h l'&ude d'une population
graphical kinematic method is well suited for testing the de failles. C. r. Acad. Sci., Paris 279D, 891-894.
assumptions of kinematic analysis. For the field data Chester, F. M. & Logan, J. M. 1987. Composite planar fabric of gouge
from the Punchbowl Fault, California. J. Stuct. Geol. 9,621-634.
analyzed here, the assumptions s e e m t o be reasonable. Cox, s. J. D. & Scholz, C. H. 1988. On the formation and growth of
It might be that the assumptions generally hold for fault faults: an experimental study. J. Struct. Geol. 10,413-430.
populations, which would give insight to the nature of Donath,F. A. 1962.Analysis of Basin-Range structure, south-central
Oregon. Bull. geol. Soc. Am. 73, 1-16.
the faulting process. For example, if fault-slip kinema- Elliott, D. 1976. The energy balance and deformation mechanisms of
tics generally prove to be scale-invariant, then the fractal thrust sheets. Phil. Trans. R. Astr. Soc. A283, 289-312.
character of faulting is more profound than previously Etchecopar, A., Vasseur, G. & Daignieres, M. 1981. An inverse
problem in microtectonics for the determination of stress tensors
demonstrated. The graphical kinematic method is also from fault striation analysis. J. Struct. Geol. 3, 51-65.
useful for the analysis of kinematic heterogeneity. These Gamond, J. F. 1987. Bridge structures as sense of displacement
techniques c a n b e especially helpful in t h e field, w h e r e criteria in brittle fault zones. J. Struct. Geol. 9,609-620.
Gauthier, B. & Angelier, A. 1985. Fault tectonics and deformation: a
they may direct work toward key areas to resolve method ofquantificationusingfielddata. Earth Planet Sci. Lett. 74,
unforeseen problems. 137-148.
986 R. MARRE'ITand R. W. ALLMENDINGER

Gephart, J. W. 1988. On the use of stress inversion of fault-slip data to Marrett, R., Allmendinger, R. W. & Grier, M. E. 1989. Kinematic
infer the frictional strength of rocks (abs.). EOS Trans. Am. changes during late Cenozoic deformation of the southern Puna
geophys. Un. 69, 1462. plateau: Argentine Andes, 23°S-27"S Latitude. Proc. 28th Int.
Gephart, J. W. & Forsyth, D. W. 1984. An improved method for Geol. Congr. 2, 372-373.
determining the regional stress tensor using earthquake focal mech- Means, W. D. 1987. A newly recognized type of slickenside striation.
anism data: application to the San Fernando earthquake sequence. J. Struct. Geol. 9,585-590.
J. geophys. Res. 89, 9305--9320. Michael, A. J. 1984. Determination of stress from slip data: faults and
Hadley, D. & Kanamori, H. 1978. Recent seismicity in the San folds. J. geophys. Res. 89, 11,517-11,526.
Fernando region and tectonics in the west-central Transverse Molnar, P. 1983. Average regional strain due to slip on numerous
Ranges, California. Bull. seism. Soc. Am. 68, 1449-1457. faults of different orientations. J. geophys. Res. 88, 6430--6432.
Higgs, W. G. & Williams, G. D. 1987. Displacement efficiency of Muraoka, H. & Kamata, H. 1983. Displacement distribution along
faults and fractures. J. Struct. Geol. 9,371-374. minor fault traces. J. Struct. Geol. 5,483--495.
Hileman, J., Allen, C. & Nordquist, J. 1975. Seismicity of the Omarini, R. H. 1983. Caracterizacirn litolrgica, diferenciacirn y
Southern California region I January 1932 to 31 December 1 9 7 2 . grnesis de la Formacirn Puncoviscana entre el Valle de Lerma y la
Seismological Laboratory, California Institute of Technology, Pasa- Faja Eruptiva de la Puna. Unpublished Ph.D. dissertation Universi-
dena, California. dad Nacional de Salta, Salta, Argentina.
Huang, Q. 1988. Computer-based method to separate heterogeneous Petit, J.-P. 1987. Criteria for the sense of movement on fault surfaces
sets of fault-slip data into sub-sets. J. Struct. Geol. 10, 297-299. in brittle rocks. J. Struct. Geol. 9, 597-608.
Hull, J. 1988. Thickness-displacement relationships for deformation Power, W. L., Tullis, T. E., Brown, S. R., Boitnott, G. N. & Scholz,
zones. J. Struct. Geol. 10,431--435. C.H. 1987. Roughness of natural fault surfaces. Geophys. Res. Lett.
Jackson, J. & McKenzie, D. 1988. The relationship between plate 14, 29-32.
motions and seismic moment tensors, and the rates of active defor- Power, W. L., Tullis, T. E. & Weeks, J. D. 1988. Roughness and wear
mation in the Mediterranean and Middle East. Geophys. J. 93, 45- during brittle faulting. J. geophys. Res. 9"~, 15,268-15,278.
73. Reches, Z. 1983. Faulting of rocks in three-dimensional strain fields--
Jones, L. 1988. Focal mechanisms and the state of stress on the San II. Theoretical analysis. Tectonophysics 95,133-156.
Andreas fault in southern California. J. geophys. Res. 93, 8869- Reches, Z. 1987. Determination of the tectonic stress tensor from slip
8891. along faults that obey the Coulomb yield condition. Tectonics 6,
Julien, Ph. & Cornet, F. 1987. Stress determination from aftershocks 849-861.
of the Campania-Lucania earthquake of November 23, 1980. Reches, Z. & Dieterich, J. 1983. Faulting of rocks in three-
Annales Geophysicae 5B, 289--300. dimensional strain fields: I. Failure of rocks in polyaxial, servo-
Kamb, W. B. 1959. Ice petrofabric observations from Blue Glacier, control experiments. Tectonophysics 95, 111-132.
Washington in relation to theory and experiment. J. geophys. Res. Sammis, C., King, G. C. P. & Biegel, R. 1987. Kinematics of gouge
64, 1891-1909. formation. Pure & Appl. Geophys. 125,777-812.
Kanamori, H. & Anderson, D. 1975. Theoretical basis of some Scholz, C. H. 1987. Wear and gouge formation in brittle faulting.
empirical relations in seismology. Bull. seism. Soc. Am. 65, 1075- Geology 15,493-495.
1095. Scholz, C. H. & Aviles, C. A. 1986. The fractal geometry of faults and
King, G. C. P. 1983. The accommodation of large strains in the upper faulting. In: Earthquakes Source Mechanics (edited by Das, S.,
lithosphere of the Earth and other solids by self-similar fault Boatwright, J. & Scholz, C.). Am. Geophys. Un. Geophys.
systems: the geometrical origin of b-value. Pure & Appl. Geophys. Monogr. 37, 147-156.
121,761-815. Turcotte, D. L. 1986. A fractal model for crustal deformation. Tec-
Kostrov, B. V. 1974. Seismic moment and energy of earthquakes, and tonophysics 132,261-269.
seismic flow of rock. Izv. Acad. Sci. USSR Phys. Solid Earth 1,23-- Wallace, R. E. 1951. Geometry of shearing stress and relation to
44. faulting. J. Geol. 59, 118-130.
Krantz, R. W. 1988. Multiple fault sets and three-dimensional strain: Walsh, J. J. & Watterson, J. 1987. Distributions of cumulative dis-
theory and application. J. Struct. Geol. 10,225-237. placement and seismic slip on a single normal fault surface. J. Struct.
Langston, C. A. 1978. The February 9, 1971 San Fernando earth- Geol. 9, 1039-1046.
quake: a study of source finiteness in teleseismic body waves. Bull. Walsh, J. J. & Watterson, J. 1988. Analysis of the relationship
seism. Soc. Am. 68, 1-29. between displacements and dimensions of faults. J. Struct. Geol. 10,
Lisle, R. J. 1987. Principal stress orientations from faults: an addi- 239-247.
tional constraint. Annales Tectonicae 1,155-158. Whitcomb, J. H., Allen, C. R., Garmany, J. D. & Hileman, J. A.
Mardia, K. V. 1972. Statistics of Directional Data. Academic Press, 1973. San Fernando earthquake series, 1971: focal mechanisms and
London, U.K. tectonics. Rev. Geophys. & Space Phys. 11,693-730.
Marrett, R. 1990. The late Cenozoic tectonic evolution of the Puna Wojtal, S. 1989. Measuring displacement gradients and strains in
plateau and adjacent foreland, northwestern Argentine Andes. faulted rocks. J. Struct. Geol. 11,669-678.
Unpublished Ph.D. dissertation, Cornell University, Ithaca, New
York.