Soil Dynamics and Earthquake Engineering 69 (2015) 262–269

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering

journal homepage: www.elsevier.com/locate/soildyn

Seismic vulnerability assessment based on typological characteristics.

The first level procedure “SAVE”
G. Zuccaro a,b, F. Cacace a,b,n
a
Plinivs Studies Center – Lupt, University of Naples “Federico II”, Via Toledo, 402-80134 Naples, Italy
b
Department of structures for Engineering and Architecture, University of Naples “Federico II”, Via Claudio 21, Naples, Italy

art ic l e i nf o a b s t r a c t

Article history: In this paper a procedure is described for quick seismic vulnerability assessment according to the EMS 98
Received 30 September 2012 classification. The criteria adopted by Ems 98 to assign the typological classes are mainly based on the
Received in revised form vertical structure type. The proposed methodology aims to reduce the uncertainty in the class assessment
3 November 2014
using a set of parameters, related to typological features, that are identified as modifiers of the vulnerability
Accepted 6 November 2014
level. The paper shows how the weight of each of these parameters is evaluated through the analysis of the
Available online 9 December 2014
seismic damages recorded during past earthquakes. A synthetic damage parameter is then defined in order
Keywords: to compare the seismic response of different sets of buildings under the same seismic intensity. Finally, the
Structure vulnerability assessment obtained on a set of buildings by the application of the methodology is compared
Seismic vulnerability assessment
with expert evaluations derived from the direct inspections on buildings.
Seismic damage analysis
& 2014 Elsevier Ltd. All rights reserved.
Typological features

1. Introduction: vulnerability and “scale” problems The first approach has the limitation of not being very reliable
for single buildings. On the other hand it has the undoubted
The seismic vulnerability assessment of buildings has been advantage of demanding both little information and rapid proces-
extensively studied during the last 30 years, and has been sing. For this reason it is useful for investigating a wide range of
analyzed by using different approaches and by adopting different buildings (urban scale or wider). The research in this context aims
levels of approximation. The methods proposed can be grouped in to achieve a greater reliability of results while maintaining an
two main “categories”: acceptable quick investigation.
The second approach provides assessments certainly more reliable
– Statistical/observational approach on single buildings, but it requires detailed knowledge of technical
The vulnerability is derived from the synthetic analysis of the features of the buildings and the development of time consuming
formal and structural characteristics of the building [2,3,5,6,9,10,12- structural calculations. Therefore it is difficult to implement it at large
15,17]. scale. In this context, the research is oriented to elaborating methods
A restricted number of building categories, called “vulnerability of calculation more simple in modeling and running, but without
classes”, is identified as a function of the typological and structural significant loss in terms of accuracy of the result.
characteristics. Each class is characterized by a vulnerability func- It should also be considered that, for the Civil Protection purposes,
tion that generally is calibrated by analyzing the damage observed the scale of work is wide (regional or national), and the observational
during past events. approach seems to be at the moment the best choice.
– Mechanical approach Seismic impact evaluation at regional scale either in terms of
The vulnerability evaluation is the result of accurate computations buildings or population damage (scenario analysis) is the basic
using techniques provided by the Structural Mechanics [18–21]. tool for developing, in “uneventful time” the planning of emer-
The damage evaluation is formulated on the basis of analytical gency management. Furthermore, during seismic events, in the
calculations to determine the seismic response of the building, earliest phase of the emergency, it is very useful to have in few
the stress and corresponding strain state are derived. minutes after the event a reliable estimation of the impact on the
region, to immediately assess the entity of the disaster and to
properly size the rescue operation, optimizing the use of resources
These are, clearly, two quite different approaches, particularly
available through their location on the territory.
in terms of “scale”.
Similarly, another important elaboration is the seismic risk
assessment per unit area (risk maps). It represents an essential
n
Corresponding author. tool for the optimal allocation of resources in mitigation planning.

http://dx.doi.org/10.1016/j.soildyn.2014.11.003
0267-7261/& 2014 Elsevier Ltd. All rights reserved.
 G. Zuccaro, F. Cacace / Soil Dynamics and Earthquake Engineering 69 (2015) 262–269 263

Enhancement of studies based on observational approach is A different approach has been pursued by Bernardini et al.
aimed to improve the reliability of vulnerability curves. This is [8,9], and consists in setting up DPM directly derived from the
pursued by trying to identify more accurately the typological definition of European Macroseismic Scale (EMS98) [11]. A further
factors that may affect (positively or negatively) the seismic development of this method uses the “fuzzy” technique to define
response of the building (vulnerability factors), and consequently “blur” DPM matrices, that take into account the uncertainties in
the assignment of the building to a specific vulnerability class. the definitions of relationship stress/damage “implicitly” con-
To perform a vulnerability analysis finalized to territorial tained in the EMS.
applications, the inventory of buildings in the study area has to About the aspect regarding the vulnerability classes, some
be grouped by the following steps: researchers have investigated the possibility to make more reliable
the assignment of the classes taking into account additional
a) Define a fixed small number of “typological vulnerability classes”, vulnerability factors in spite of the simple combination of vertical
establishing the assignment criteria to each building class. and horizontal structure. In this regard, there are several interest-
b) Choose a shaking parameter that allows to quantify the seismic ing works (Lagomarsino [10], Bernardini [14]) in which typological
action. characteristics (such as age, number of floors, regularity and
c) Define a graduated scale to describe the seismic damage. accuracy in elevation, construction details, position in the aggre-
d) Evaluate the seismic response and set the probabilistic vulner- gate, staggered floors, etc.) are considered “modifier” parameters
ability law for each typological class. of the vulnerability. The weight of these parameters results in a
score from 0 to 100, according to Petrini vulnerability index Iv [3]
assigned by expert judgment.
It should be clarified that “the typological class” is a theoretical In this paper a methodology is proposed for the assessment of
abstraction. In other words, the issue is not to determine the vulnerability in which the weight of the factors modifying vulner-
vulnerability of a particular type of building or a structural model, ability is calibrated using statistics on observed seismic damage.
but rather to identify the building types characterized by a similar
seismic response.
Set the problem in this term, the main tip becomes the 2. The first level procedure “SAVE”
identification of suitable vulnerability functions able to describe
with good reliability the behavior of each typological class. The first level procedure “SAVE” (Strumenti Aggiornati per la
The beginning of a modern and systematic strand of research in Vulnerabilità sismica del patrimonio Edilizio e dei sistemi urbani –
Italy on seismic risk can be dated back to the period between the Updated Tools for the Seismic Vulnerability Evaluation of the
late 70s and early 80s, after the earthquake in Friuli in 1976 and Italian Real Estate and of Urban Systems), starting from the
the earthquake in southern Italy in 1980. In that period the Italian typological classification given in the definition of the European
researchers implemented the definitions of “risk” and “vulner- Macroseismic Scale [16] (EMS'98), is aimed at reduce the implicit
ability” introduced in the late 70's (UNDRO 1979 [1]). At the time uncertainty of the EMS'98 vulnerability classification (Fig. 1). This
arose the problem of assessing the seismic risk at Regional or is pursued by identifying a set of typological features which can be
urban scale, so as the need to develop “scenario” procedures able considered as vulnerability modifier, and giving to each of these a
to estimate damage impact on the territory caused by an earth- numerical weight, calibrated using an extensive database of
quake of given magnitude. seismic damage observed (using survey forms) after the most
Soon after the Irpinia earthquake in 1980 was carried out a important earthquakes occurred in Italy since the 1980 Irpinia
systematic analysis of the damages, and developed the “first level” earthquake to today.
methodology for a rapid assessment of the seismic vulnerability The database used for all the elaborations is made by forms
[2] which defined three vulnerability classes depending on the compiled during post-earthquake survey campaigns, indicating
combination of vertical and horizontal structures, and was char- the typological features of the observed damage. It includes
acterized by vulnerability curves based on the statistical analysis of approximately 170,000 buildings in about 500 municipalities.
the observed damages. Seismic events surveyed:
In the same years was developed [3] another methodology
called “level two”. According this method a large number of – Irpinia 1980. About 38,000 buildings. Almost all the involved
building features are analyzed in detail and a numerical score is Municipalities were surveyed building by building, however
given to each one. Then summing up all the scores, weighted by a the information collected per building are not very detailed.
confidence factor, the vulnerability index Iv is derived. It is Seismic intensity range is V–X. Irpinia survey form [2,23].
expressed numerically by a value between 0 and 100. – Abruzzo 1984. About 51,100 buildings. Almost all the involved
As one can easily understand the two key points of the Municipalities were surveyed building by building, but the infor-
question are on one hand the developing of vulnerability curves, mation collected per building are not very detailed. Seismic
expressed as Damage Probability Matrix (DPM), that provide a intensity range is V–VIII. Abruzzo 1984 survey form [23].
good simulation of the real behavior of buildings, and on the other – Sicilia 1990. About 12,800 buildings. Minor event. Partial survey,
hand the appropriate definition of vulnerability classes and classi- the information collected on the buildings have a good level of
fication criteria. detail. Seismic intensity range is V–VII. GNDT survey form [23].
With regard to the first of these aspects, several studies have – Parma 1983. About 380 buildings. Minor event. Partial survey,
proposed the adoption of new DPM functions, developed by using information collected on the buildings have a good level of
a wider data-base including damage surveys conducted after detail. Seismic intensity range is V–VII. GNDT survey form [23].
earthquakes, in various Italian cities. Several studies have been – Umbria-Marche 1997. About 42,300 buildings, partial survey,
produced by the National Seismic Service in 1997 [4,5], by the (only damaged buildings were surveyed), information collected
authors in [7,22] and by working group National Group for the on the buildings have a good level of detail. Seismic intensity
defense by Earthquakes and National Seismic Service, Italian Civil range is V–VII. Aedes 09/1997 survey form [23].
Protection Department (GNDT–SSN) in 2002. All these studies – Etna 2002. About 5100 buildings. Minor event. Partial survey,
were based, with different criteria, on the interpretation of information collected on the buildings have a good level of
observed seismic damage. detail. Seismic intensity range is V–VIII.
264 G. Zuccaro, F. Cacace / Soil Dynamics and Earthquake Engineering 69 (2015) 262–269

SPD

Building percentage
Damage level

Fig. 2. Synthetic parameter of damage (SPD).

typology, the historical damage data occurred in previous seismic

events have been statistically processed. Namely, different damage
distribution (D0, D1,…, D5) for buildings with the same vertical
structure, grouped according to parameters investigated, have
been built. In order to compare different damage distributions, a
synthetic damage index has been defined.
This value, called Synthetic Parameter of Damage (SPD), repre-
sents the barycentric abscissa of the damage distribution, as
shown in Fig. 2, and allows to represent synthetically in a single
value the average level of seismic damage suffered by a set of
buildings.
5 N dI dI
SPDI ¼ ∑ ð2:1Þ
dI ¼ 1 N tot

where, I is the seismic intensity MCS, dI is the damage level

according to EMS 98, NdI is the number of buildings with damage
Fig. 1. Vulnerability classification according to the EMS 98.
level dI, and Ntot is the total number of buildings in I intensity area.
Given a set of buildings “X”, homogeneous for typological
– S. Giuliano di Puglia 2002 About 21,200 buildings. Minor event.
features, and given the distribution of the damage due to a
Partial survey, but information collected on the buildings have
specified intensity, the corresponding parameter SPD allows to
a good level of detail. Seismic intensity range is V–VII. Aedes
compare in a simplified way the behavior of the set “X” with that
05/2000 survey form [24].
of another set of buildings “Y”, for instance characterized by
– Pollino 1998. About 5750 buildings. Minor event. Complete
different typological features, in order to analyze the differences
survey in three municipalities, information collected on the
in vulnerability with changes in features.
buildings have a good level of detail. Seismic intensity VI. Aedes
From the value of SPDI, evaluated on a damage distribution due
05/2000 survey form [24].
to a shaking intensity I, is possible to derive an estimate of the


5
The most of surveyed buildings are masonry structure, and the binomial coefficient p I of the Damage Probability Matrix
main part of the fields in the data-base is chosen with regard to I
this kind of buildings, therefore the survey forms do not include (DPM) for intensity I, by the formula:


data specifically related to the vulnerability elements of reinforced 5 SPDI
concrete, such as short columns, etc., this is why the correlations pI ¼ ð2:2Þ
I 5
regarding the vulnerability of reinforced concrete buildings are
obtained using only the few data available. This result can be
significantly improved by the data collected after the earthquake 2.2. Building classification according to vertical structure
in L'Aquila in 2009. Further development of the research will
achieve this improvement. Once collected in an homogeneous Database, the data of
damages observed, occurred in various past seismic events (Irpi-
2.1. Definition of the synthetic parameter of damage SPD nia'80, Abruzzo'84, Sicilia'90, Parma'83, Umbria-Marche'97,
Etna'02, Molise-Puglia'02), have been grouped according to their
The European Macroseismic Scale (EMS 98) foresees that the vertical typology. For each of these classes the Synthetic Parameter
assignment of a building to a vulnerability class essentially of Damage (SPD) has been calculated at different intensity level, as
depends on the vertical structure typology, with some percentage shown in Fig. 3.
of uncertainty as shown in Fig. 1. This uncertainty range could be According to the trend of the SPD index, the different vertical
rather large and therefore able to heavily affect risk or scenario structures analyzed have been re-grouped in the EMS classes, as
analyses. This investigation has prompted the consideration of shown in Table 1, and for each of these EMS vertical vulnerability
other structural characteristics that could better address the classes the SPD value has been re-evaluated, varying the seismic
vulnerability classification inside the range of admissibility of the intensity level (Fig. 4).
EMS table. Therefore, in order to evaluate the influence of different For each of these EMS vertical vulnerability classes the SPD
parameters on the structural seismic behavior of a given vertical mean values in the intensity range VI–VIII have been calculated
 G. Zuccaro, F. Cacace / Soil Dynamics and Earthquake Engineering 69 (2015) 262–269 265

2.6 SPD mean value for EMS Vertical Structure

2.4 vulnerability classes
2.4 Av
2.2 CEMA
2.2 Bv
2 MFOR
2
SPD

MIST
1.8 Cv
MPIE 1.8
1.6 MREG 1.6 Dv
1.4 PIRR 1.4
1.2 PREG 1.2
1
1
5 6 7 8 Av Bv Cv Dv
MCS Int. Fig. 5. SPD medium value in the range VI–VIII (Int.) for the EMS vertical
vulnerability classes.
Fig. 3. SPD trend for different kinds of vertical structures. Reinforced concrete
(CEMA), hollow bricks (MFOR), mixed masonry-r.c. (MIST), solid bricks (MPIE),
regular masonry (MREG), irregular stones masonry (PIRR), and regular stones Table 2
(PREG). SPD ranges for the assignment of the
vulnerability class.

Table 1 A B C D
Building classification according to the typology of vertical structure.
– 2.0 1.7 1.4
Vuln. Typology of the vertical structure 2.0 1.7 1.4 1.0
classes

Av Stones, irregular masonry

increase or reduce the vulnerability. These characteristics are
Bv Stones, regular masonry, mixed (r.c. þmas. on different floors),
hollow bricks grouped as follows:
Cv Solid bricks, mixed (r.c. þ masonry, both on all the floors)
Dv Frame, reinforced concrete or steel Structural typology:
 Horizontal typology
 Roof typology
 Isolated column in masonry building
3  Tie rods
2.8  Mixed r.c.-masonry structures
2.6 Geometry and site:
2.4  Number of floors or maximum height
Av
2.2
 Plan and elevation regularity
Bv 
SPD

2 Infill panel regularity (only for r.c. buildings)

1.8
Cv  Building location in the block
Dv  Site topography
1.6
Other:
1.4
 Building age
1.2
 Pre-existent damage
1
5 6 7 8
Since the parameters listed above are not always completely
MCS Int
available, the SAVE procedure allows to separately consider the
Fig. 4. SPD trend for the EMS vulnerability classes reported in Table 1. influence of each of them for the assessment of the vulnerability
class. Therefore, such a procedure is applicable independently
(Fig. 5). These parameters represent a first indication about the from the number of known parameters.
vulnerability and the expected damage, based only on the typol- So, each of the EMS vertical vulnerability classes previously
ogy of the vertical structure. identified have been further subdivided, grouping the buildings
The correlation showed in Fig. 5 represents the reference according to the different values of the considered parameter. For
criterion assumed for the further vulnerability evaluations, based example the buildings belonging to the EMS vertical vulnerability
on the SPD value (see Table 2). class A have been split to five groups, according to the five
different horizontal structural types considered.
The numeric value of the weight of each parameter is deter-
2.3. Parameter modifying vulnerability. Numeric weight of each mined as follows:
parameter
1. A set “V” of buildings is extract from the data-base with a given
As pointed out in the previous paragraphs, the knowledge of vertical structure. The damage distribution is built from the
the only vertical structural system could be not sufficient to assign damage data-base and the SPDv (medium in the range of
accurately the vulnerability class. In fact, the seismic behavior of a intensity VI–VIII) is evaluated.
building could widely vary, depending on a number of factors 2. Several subsets “Pi” are extracted from the set “V”, each of
which could either improve or worsen its performance under the which selected according to a given parameter. In example, a
seismic actions. set for each type of horizontal structure (wood, steel, r.c., vaults,
In order to investigate the weight of typological features, the mixed). The distribution of damage is built from the damage
whole post-earthquake damage survey data-base has been ana- data-base and the mean SPDp in the intensity range VI–VIII is
lyzed, pointing out the data about features that potentially could evaluated.
266 G. Zuccaro, F. Cacace / Soil Dynamics and Earthquake Engineering 69 (2015) 262–269

3. The value of the parameter pi is given from: (Fig. 6)

SPDv pi ¼ SPDp  SPDv

SPDp
Positive spreads express a vulnerability increase, negative
spreads mean a vulnerability decrease. For instance, taking the
values shown in Fig. 7c, the parameter “horizontal structure” for
chaotic masonry buildings (A) and “deformable” slab is 0.215. This
means that the post-earthquake damage (expressed as SPD) suf-
fered by a group of buildings with chaotic masonry (A) and
“deformable” slabs was on average 21.5% higher than the post-
Fig. 6. Influence of a generic parameter on the SPD value per a fixed intensity. earthquake damage observed on the full set of buildings with

0.6
Building age Geometric regularity
0.1
0.4 0.08
0.06
0.2 0.04
0.02
-1E-15 0
-0.02 Av
-0.2 Av -0.04
Bv
Bv -0.06
-0.4 -0.08 Cv
Cv
-0.1 Dv
-0.6 Dv non-regular regular
-> 1919 1919-1945 1945-1961 1962-1971 1972-1981 1982 ->
Av 0.028779801 -0.028779801
Av 0.30210739 0.09963162 -0.1163757 -0.2277823 -0.149505 0.09192397
Bv 0.064936996 -0.064936996
Bv 0.41323659 0.24142526 0.07754763 -0.0756106 -0.2820542 -0.3745447
Cv 0.086551726 -0.086551726
Cv 0.5115788 0.41612527 -0.0804982 -0.2535889 -0.5543997 -0.0392172
Dv 0.071531974 -0.071531974
Dv 0 0.03743 0.14174537 0.02927338 -0.2093787 0.00092995

0.4 0.1 0.08

horizontal structure tie roads Roof Type
0.08 0.06
0.3 0.06 0.04
0.04 0.02
0.2 0.02
0
0
0.1 -0.02 Av -0.02
Av -0.04
-0.04
0 Bv -0.06
Bv -0.06
-0.08 Cv -0.08
-0.1 Cv not
-0.1 thrusting
no yes thrusting
-0.2
flexible rigid semirigid vaults Av 0.043697476 -0.04369748 Av -0.02603129 0.026031286
Av 0.214843631 -0.17450491 -0.08920236 0.317539066 Bv 0.085143864 -0.08514386 Bv -0.04495626 0.044956256
Bv 0.01954592 -0.12329065 -0.02019925 0.123943977 Cv 0.036670264 -0.03667026 Cv -0.06279113 0.062791131
Cv 0.150793821 -0.14782931 -0.10732117 0.157570392

0.35 0.25
building position in the block Roof Structure
0.3 0.2
0.25
0.2 0.15
0.15 0.1
0.1 0.05
0.05 Av
0 Av
0 Bv
-0.05 -0.05 Bv
Cv
-0.1 -0.1
-0.15 Cv
Dv
corner external internal insulated -0.15
Av 0.114349209 0.0503571 -0.066695362 -0.098010942 -0.2
Steel R.C. wood
Bv 0.052599981 0.012415401 -0.033010613 -0.03200477
Av -0.0615294 -0.1138506 0.17537998
Cv 0.280406088 -0.04638296 -0.105639458 -0.128383651
Bv -0.0550699 -0.0733636 0.12843351
Dv 0.053169068 0.019903928 -0.052221946 -0.02085105
Cv -0.0257076 -0.164418 0.19012557

Fig. 7. Influence of some relevant parameters on the SPD value.

 G. Zuccaro, F. Cacace / Soil Dynamics and Earthquake Engineering 69 (2015) 262–269 267

Table 3
Non-correlation matrix among the dependent parameters q.

Roof struct. Slab struct. Age Floors nr.

LIGHT HEAVY DEFO SEMIRIGI RIGI VOLT VO-SOL o 1919 1919-1946 1946-1961 1961–1971 1971–1981 1981 4 1–2 3–4 5–6 7–8

Floors nr.
1–2 0.4096 0.8469 0.7517 0.8111 0.8179 0.9264 0.9995 0.7444 0.8764 0.8118 0.9239 0.9405 0.9993
3–4 0.8482 0.9146 0.9061 0.9164 0.9344 0.9565 0.9982 0.8985 0.9633 0.9196 0.9678 0.9812 0.9981
5–6 0.9962 0.9884 0.9978 0.9893 0.9983 0.9977 0.9999 0.9978 0.9980 0.9903 0.9961 0.9985 0.9990
7–8 0.9998 0.9964 0.9999 0.9968 0.9999 0.9999 1.0000 0.9998 0.9998 0.9972 0.9991 0.9998 0.9997

Age
– 41919 0.6444 0.9676 0.8386 0.9755 0.8845 0.9343 0.9989 SYM
1919–1945 0.8568 0.9768 0.9476 0.9682 0.9357 0.9864 0.9989
1945–1961 0.8535 0.8944 0.9652 0.8359 0.9310 0.9921 0.9994
1962–1971 0.9626 0.9303 0.9903 0.9101 0.9910 0.9974 1.0000
1972–1981 0.9351 0.9830 0.9717 0.9786 0.9797 0.9930 1.0000
1982 –4 0.9997 0.9956 0.9999 0.9962 0.9999 1.0000 1.0000

Slab struct.
DEFO 0.6599 0.9793 SYM SYM
SEMIRIGI 0.8415 0.7493
RIGI 0.6340 0.9420
VOLT 0.8623 0.9743
VO-SOL 0.9972 0.9993

Roof struct.
LIGHT SYM SYM SYM
HEAVY

chaotic masonry (A), therefore the coefficient value adopted in As previously seen, the parameter “horizontal structure” for
vulnerability assessment is Pi ¼0.215. chaotic masonry buildings (Av) and deformable slabs is 0.215, the
In the following Fig. 7 some diagrams are shown as examples. same way the coefficient for buildings “Av” built before 1919 is
The procedure for correcting the vulnerability considering a p ¼0.3. The coefficient values are therefore calculated independently
single parameter “Pi” can be summarized as follows: for each parameter. In other words, the computation of the coeffi-
cient “deformable floor” does not take into account that the set of
1. Each building is assigned to a “vertical” vulnerability class EMSv buildings with deformable floor also contains buildings with age
on the basis of the vertical structural system. prior to 1919. In fact, analyzing the data-base one can find that is
2. The average value of SPDv corresponding to the assigned class observed that 25% of buildings “Av” have both “deformable floor”
EMSv is taken as base score. and age “before 1919”. This means that the two features are not
3. The base score is multiplied by the coefficient of influence independent, and the application of both coefficients in the scoring
(positive or negative) corresponding to the value of the para- of vulnerability would lead to an overestimation of vulnerability.
meter “Pi” considered. Thus the weight of the coefficients was reduced proportionally to the
4. Adding up the value obtained to the base SPDv score the statistical dependence between them.
“modified” SPD score is calculated. In order to investigate the independence between these para-
5. The vulnerability class is reallocated according to the new value meters, a matrix was defined in which the rows and columns are all
of the SPD. This may result in vulnerability class different from the potentially related parameters, and each element represents, for
that obtained on the basis of the vertical structure alone. the corresponding couple of parameters and for each possible
combination of values, the percentage of occurrence of buildings in
The SPD ranges used for the definition of the vulnerability the database for which it was observed that the two parameters do
classes have been calibrated using the statistical analysis of the not appear simultaneously.
damage data referred to the buildings grouped according to the The matrix thus defined is an array of non-correlation between the
vertical structure typology (see Table 2). dependent parameters q (Table 3).
However, usually more than one parameter is available, in this Of course this matrix has no value at crossings between self-equal
case is possible to assess the vulnerability taking into account the parameter, and is symmetrical with respect to the main diagonal.
combined influence of known features. According to the procedure described above, the percentage of
Such a procedure allows to consider the influence of each single the SPD spread has been evaluated for each of the parameters
parameter in the vulnerability assessment. It has to be pointed out that identified, in order to modify the original assignment to the vulner-
not all typological parameters can be considered as independent ability classes. For the vulnerability evaluation, the influence of the n
variables: e.g., if a masonry building has wooden floors, probably the independent parameters q and the influence of the m dependent
age of its construction will be prior to 1919, so it would not be correct, ones p, considered proportionally to the not-correlation coefficients
in assessing the vulnerability score, simply adding the variation due to cij, are combined as shown in Eq. (2.1):
the presence of wooden floors with the one due to old age. Similarly
are not independent, for example, the type of roof and type of floors, 0 m m 1
the age and the number of floors, etc. It was therefore defined a B ∑ ∑ δ i;j ðp j þ p i Þc i;j C
n
correlation law that allows to take into account the dependencies B j¼1i¼1 C
SPDI ¼ SPDvI B1 þ ∑ qs þ C ð2:3Þ
between the various parameters in the final assessment of @ s¼1 m  1 A
vulnerability.
268 G. Zuccaro, F. Cacace / Soil Dynamics and Earthquake Engineering 69 (2015) 262–269

where, δi;j is the Kronecker operator, δi;j ¼0 if i¼ j, δi;j ¼ 1 if iaj. i, j On the other hand it should be noted that a wide investigation
are the indices of non-correlation matrix, p is the non- on a large number of buildings would be extremely onerous,
independent parameter, q is the independent parameter, cij is especially for classes c and d, for which the choice of the sample
the non-correlation coefficient, m is the total number of non- buildings could strongly affect the test results. For these reasons,
independent parameters, and n is the total number of independent none of the cited similar studies, aimed to investigate the vulner-
parameters. ability of building classes and based on “poor” data has analogous
Therefore the methodology described allows to automatically checks.
assign the EMS vulnerability classes, once the vertical typology A major inspection “downstream” can be directly carried out by
and other typological characteristics of the building are known, analyzing the relationship between types and damage on the
even in the case of an inhomogeneous level of information. buildings affected by the earthquake in L'Aquila in 2009.

4. Conclusions
3. Development of a vulnerability ranking for strategic
buildings in Campania Region by the proposed methodology The methodology described allows for the rapid assessment of
the vulnerability of even large samples of buildings on the basis of
In this regard may be interesting to report the experience “poor” data, but without excluding the possibility to use more
carried out in Campania Region. In 1997, the Civil Protection detailed data if available.
Department funded a census on the seismic vulnerability of By combining the available information it is possible to dras-
strategic buildings in some Italian regions. These data (geometric tically reduce the uncertainties of assignment, inevitable if we
and typological features) were collected in a data-base. In 2005, consider only the vertical structure.
the Campania Region bureau of Civil Protection, on the basis of In this regard it may be noted that the proposed procedure
these data, has developed a vulnerability classification of all differs from similar techniques, where the vulnerability assess-
strategic buildings in the region, in order to establish priority in ment is done through complex amounts of “scores” assigned to
the assignation of the funds earmarked for strengthening. So 60 different characteristics in an “expert” way, more or less arbitra-
buildings were chosen in high-hazard areas and ranked with high rily, and whose result can lose of meaning when not all the data
vulnerability (class “A”). Since the typological features were needed for the procedure are available or when the available
surveyed eight years earlier, before delivering the funding the information are not “coded” so that is not possible to attribute
Regional bureau of Civil Protection has sent its engineers to each of them a score.
the 60 buildings, in order to verify if in the meanwhile there had The main advantages in using this procedure are:
been no changes, and also to check the accuracy of the vulner-
ability assessment. Accurate inspections on site performed in – Unlike similar methods that require knowledge to every build-
cooperation with the Municipal Technical Offices showed a very ing in a defined set of typological characteristics, the method is
good correspondence with the results of SAVE procedure. The “deformable” to be used with different levels of detail of
result of these on-site inspections was: available data. This allows one to not ignore more detailed
Eight years after the first survey, 27 buildings have not been information when available.
modified or retrofitted, (about 45%), for 26 of them (43%) the – The weight of vulnerability modifiers parameters is derived
assignment class “A” is confirmed correct, while for only one directly by the analysis of actually occurred and observed
(about 2%) the assessment from surveyed data was wrong. damage. This “empirical” approach ensures that obtained
Four buildings (7%) were demolished, not because of damage results can properly describe the actual behavior of buildings.
caused by an earthquake, but probably just because they were – The typological characteristics determining the choice of the
very old and poorly maintained, and there was no longer any vulnerability modifiers coefficients can be detected quickly
advantage to keep them up. with a simply “observation” of the building, maybe even if
The remaining 29 buildings (48%) were had already been done by a non-expert surveyor. This allows to apply the
strengthened according to National Seismic Code'96, as shown in method easily to a wide number of buildings.
the diagram. In particular 8 buildings(13%) were reinforced, 21 – The method is quite open to two levels of further improve-
buildings (35%) were retrofitted. In any case, all these interven- ments, in fact the coefficients expressing the numerical weight
tions confirm a high vulnerability at the time of the survey (Fig. 8). of each chosen parameter could be recalculated and refined by
This test is clearly not exhaustive, since is limited to 60 analyzing further available set of damage data (level 1). More-
buildings in class “A”, but can give a little contribution to validate over, if other kind of characteristics (in example short column,
the proposed methodology. wide distance between walls or between columns, or anything
that can be considered relevant) will be surveyed, they will be
RESULTS OF CHECK ON THE SET OF BUILDINGS SELECTED easily related to the damage and their coefficients could be
BY USING "SAVE" PROCEDURE
not verified evaluated in the same way is shown for the parameters
demolished 2% reported in this paper (level 2).
7%
strengthened
13% Weak points of this method are related mainly to the shaking
parameter used, i.e. the macroseismic intensity. This parameter
verified
presents in fact the known problems (discrete and non-continuous
verified scale, estimate of the shaking affected by the damage observation,
retrofitted
43%
strengthened etc.). Nevertheless it has often been preferred for the possibility of
demolished comparison with the historical catalogs of events, that implicitly
not verified encompass the local effect, that's why the vulnerability DPM
retrofitted matrices have often been expressed in terms of intensity.
35%
The robustness of the proposed methodology can be increased
Fig. 8. Percentages of confirmation of the vulnerability. analytically checking the weight of the adopted parameters, by
 G. Zuccaro, F. Cacace / Soil Dynamics and Earthquake Engineering 69 (2015) 262–269 269

using an iterative computational model. Such a procedure is [10] Lagomarsino S, Giovinazzi S. Macroseismic and mechanical models for the
currently under study. vulnerability assessment of current buildings. (Special Issue “Risk-Ue Project”).
Bull Earthq Eng 2006;4(4):415–43.
Finally, although, as it said, there is still room for improvement, [11] EMS. European Macroseismic Scale, 1998. Conseil de l'Europe. European
the methodology appears to be effective in the identification and Seismological Commission. 8 LUXEMBOURG 1998; 1998.
numerical quantification of essential factors to reduce uncertain- [12] Dolce M, Masi A, Zuccaro G, Cacace F, Samela L, Santarsiero G, et al. Esame
ties in the attribution of vulnerability classes. delle caratteristiche tipologiche e del danneggiamento del patrimonio edilizio
di San Giuliano di Puglia.
[13] Dolce M, Zuccaro G, Papa F, Masi A. Remarks on the seismic damage in the
References recent earthquakes in Europe. Proceedings of the 12th European Conference
on Earthquake Engineering. Londra: Elsevier Science Ltd.; 2002.
[14] Bernardini A. Classi macrosismiche di vulnerabilità degli edifici in area veneto-
[1] UNDRO (United Nations Disater Relief Coordinator). Natural disasters and friulana. Atti XI Congresso nazionale “L'ingegneria sismica in Italia”, Genova,
vulnerabilityanalysis in report of expert group meeting (9–12 July 1979). CD_Rom;2004a.
Geneva: UNDRO; 1979 (49 p.). [15] Zuccaro G, Cacace F. Proposta per una procedura di valutazione semplificata di
[2] Braga F, Dolce M, Liberatore D. A statistical study on damaged buildings and an rischio sismico a scala nazionale dell'edilizia scolastica 2007, Atti Convegno
ensuing review of the M.S.K.-76 scale. In: proceedings of the 7th European
ANIDIS, Pisa; 2007.
conference on earthquake engineeering, Atene; 1982.
[16] Gruntal G editor. European Macroseismic Scale 1998. Cahiers du Centre
[3] Benedetti D, Petrini V. On seismic vulnerability of masonry buildings: proposal
Européen de Géodynamique et de Séismologie, Luxembourg; 1998: vol. 15.
of an evaluation procedure. L'industria delle Costruzioni 1984;18:66–78.
[17] Zuccaro G, Cacace F. Valutazione speditiva della vulnerabilità per gli edifici
[4] Di Pasquale G, Orsini G. Proposta per la valutazione di scenari di danno
strategici della Regione Campania. Ingegneria Sismica, Anno XXIII no. 2; 2006.
conseguenti ad un evento sismico a partire dai dati ISTAT. Atti dell'81
[18] FAJFAR P. A nonlinear analysis method for performance-based seismic design.
Convegno nazionale L'Ingegneria Sismica in Italia, Taormina; 1997. p. 477–86.
Earthq Spectra 2000;16:573–91.
[5] Di Pasquale G, Orsini G, A Pugliese, Romeo RW. Damage scenario for future
[19] Fajfar P. Capacity spectrum method based on inelastic demand spectra. Earthq
earthquakes. In: proceedings of the 11th European conference on earthquake
engineering, Paris; 1998. Eng Struct Dyn 1999;28:979–93.
[20] Cosenza E, Manfredi G, Polese M, Verderame GM. A multilevel approach to the
[6] Coburn AW Pomonis A, Spence RJS. Factors determining human casualty levels
in earthquakes: mortality prediction in Building collapse. In: proceedings of capacity assessment of RC buildings. J Earthq Eng 2005;2005(9):1–22.
the Xth World Conference on Earthquake Engineering, Madrid; 1992. [21] Masi A. Seismic vulnerability assessment of gravity load designed R/C frames.
[7] Zuccaro G, Papa F, e Baratta A. Aggiornamento delle mappe a scala nazionale di Bull Earthq Eng 2003;1(3):371–95.
vulnerabilità sismica delle strutture edilizie in “la vulnerabilità degli edifici”, [22] Bernardini A editor. La vulnerabilità degli edifici: valutazione a scala nazionale
GNDT; 2001. della Vulnerabilità sismica degli edifici ordinari, CNR-Gruppo Nazionale per la
[8] Bernardini, A, Giovinazzi S, Lagomarsino, S, Parodi, S. Matrici di probabilità di Difesa dai Terremoti – Roma þ CD-ROM allegato; 2000. 175 p.
danno implicite nella scala EMS-98 per tipologie di edilizia abitativa. In: [23] Angeletti et al. Italian Civil Protection Department “Rapporto finale della
proceedings of the XII Convegno ANIDIS “L'ingegneria sismica in Italia”, Pisa; Commissione tecnico-scientifica per l'aggiornamento dell'Inventario e della
2007. vulnerabilità degli edifici residenziali e pubblici”, Roma; 2002 ( in Italian).
[9] Bernardini, A, Giovinazzi, S, Lagomarsino, S. Comparison and refinement of [24] Baggio C, Bernardini A, Colozza R, Corazza L, Della Bella M, Di Pasquale G, et al.
two EMS-98 derived vulnerability methods: towards a unique definition. In: Manuale di istruzioni per la compilazione della scheda di 11 livello di
Proceedings of the First European Conference on Earthquake Engineering and rilevamento del danno, pronto intervento ed agibilità nell'emergenza post-
Seismology, Abstract Book, Paper Number: 909, Geneva, Switzerland; 3–8 sismica (AeDES). Italian Civil Protection Department; Novembre 2000 (in
September 2006. p. 9. Italian).