Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

Available online at www.sciencedirect.

com

Computers and Geotechnics 35 (2008) 505–523


www.elsevier.com/locate/compgeo

Methodology for automated GIS-aided seismic microzonation studies


Achilleas G. Papadimitriou a,*, Andreas A. Antoniou b,
George D. Bouckovalas b, Pavlos G. Marinos b
a
Department of Civil Engineering, University of Thessaly, Pedion Areos 38334, Volos, Greece
b
Geotechnical Department, School of Civil Engineering, National Technical University of Athens, Greece

Received 14 September 2007; accepted 1 October 2007


Available online 19 November 2007

Abstract

This paper presents an automated methodology for performing geographic information system (GIS)-aided seismic microzonation
studies. It presupposes the existence of a geotechnical database containing data from sampling boreholes and in situ geotechnical or geo-
physical tests. The database has been related to a GIS that is being used only for graphically locating the available borehole and in situ
test sites for the study area and for the compilation of digital contour maps. The processing of data and the seismic ground response
analyses are automatically performed externally to the GIS and the geotechnical database, via a computer code (created in MS Visual
Basic) whose flow chart and characteristics are described in detail herein. This code processes data from the geotechnical database (that
has a relational structure and has been created in MS Access here), along with engineering geological and seismological information for
the study area. Then, it proceeds in performing the seismic ground response analyses via recently proposed multi-variable relations that
duplicate successfully the results of the 1D equivalent linear numerical method. As an application, the paper presents an exemplary GIS-
aided seismic microzonation study for an urban municipality of the greater Athens (Greece) area, which reveals the efficiency of the auto-
mated methodology and explores its limitations.
Ó 2007 Elsevier Ltd. All rights reserved.

Keywords: Earthquake; GIS; Geotechnical database; Seismic microzonation; Site effects; Soil amplification

1. Introduction and interpretation of only the pertinent data for establish-


ing the subsurface profile of the study area in a known
Earthquake hazard zonation for urban areas, mostly seismo-tectonic regime (Step 2). The third step is to evalu-
referred to as seismic microzonation, is the first and most ate for each location within the study area all aspects of the
important step towards a seismic risk analysis and mitiga- seismic ground response, namely (Step 3):
tion strategy planning. Practically, a seismic microzonation
study refers to a specific design earthquake (e.g. the earth- (a) the elastic response spectra (taking into account soil,
quake with a return period of 475 years in the European basin and topography effects),
Seismic Code EC-8 and the Greek Seismic Code EAK- (b) the potential for liquefaction and its related effects
2000) and should include data from various scientific fields. (e.g. lateral spreading, settlements),
As shown in Fig. 1, the first step of a seismic microzonation (c) the risk for slope instability or failure, and
study (ideally) includes the collection of seismological, tec- (d) the liability of having surface rupture of active faults
tonic, geological, geotechnical, geophysical and topo- within the study area.
graphic data (Step 1). The second step is the selection
The estimation of the above is performed by various
numerical and/or empirical techniques. The selection of
*
Corresponding author. Tel.: +30 24210 74140; fax: +30 24210 74169. the appropriate technique for each of these aspects depends
E-mail address: apapad@civ.uth.gr (A.G. Papadimitriou). on the scale of the study area (few km2 or tens or even hun-

0266-352X/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compgeo.2007.10.001
506 A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523

Fig. 1. Outline of 4-Step procedure for a GIS-aided seismic microzonation study, and range of proposed automated methodology.

dreds of km2), the detail and reliability of available data both geotechnical and geological data that are imple-
and obviously the level of accuracy of the performed seis- mented as coverages in the GIS and produces the input file
mic microzonation study (preliminary or final). The final for the use of Shake91 [15]. Nevertheless, the authors give
step (Step 4) of the study is the production of contour maps no description of the automation involved nor of the gen-
showing the variation of seismic intensity parameters in the eric nature of their methodology, i.e. whether their modu-
study area, i.e. the maps that will be used as input for a lar tool can be used without alterations in any seismic
seismic risk analysis and mitigation strategy planning that microzonation study. More recently, [19] use GIS interpo-
follows the completed seismic microzonation study. lating techniques to construct the 3D subsurface profile of
It becomes obvious that both input (data) and output the study area by incorporating a rather complex yet auto-
(maps) of a seismic microzonation study are referenced mated methodology and then apply linear visco-elastic
by their geographic location, and consequently the use of solutions for estimating the soil amplification. As a whole,
a geographic information system (GIS) in such a study their approach is promising yet complicating, and cannot
becomes very advantageous. As a minimum, a GIS can take soil non-linearity into account, a crucial drawback
be used for the presentation of the end results, namely for performing seismic microzonation studies for areas of
for presenting contour maps of seismic intensity parame- high seismicity.
ters, like the peak ground acceleration amax (Step 4 of the To the contrary, this paper presents an accurate, user-
study in Fig. 1). Yet, even if the mapping is performed with friendly and generic methodology for performing auto-
the aid of a GIS, the core of the seismic microzonation mated GIS-aided seismic microzonation studies. The pro-
study, namely the selection and the manipulation of the posed methodology is materialized via a new computer
data for creating the subsurface model (Step 2) and the exe- code named Auto-Seismo-Zonation (ASZ), which imports
cution of the seismic ground response analyses (Step 3) are geotechnical, engineering geological and seismological data
usually performed manually and independently of the GIS. to automatically construct the subsurface model at distinct
Steps towards more automated GIS-aided seismic locations (the locations of the available boreholes and/or
microzonation studies have recently started to appear in in situ tests). Then, the code estimates the seismic ground
the literature. For example, [27] describe ShakeMap an response at each of these locations, and the foregoing point
interface written in MS Visual Basic which prepares the estimates of seismic ground surface response are exported
input file for the use of Shake91 [15], a computer code exe- to the GIS, where seismic intensity contour mapping is per-
cuting 1D equivalent linear analysis for any given horizon- formed. Hence, the proposed methodology automates the
tally layered soil profile over visco-elastic bedrock. This core of the seismic microzonation study, namely Steps 2
interface automates soil profiling with depth, by making and 3, as outlined with the thick dashed line in Fig. 1.
use of an engineering geological database related to the The paper initiates with an outline of the proposed
GIS environment. Yet, it does not take into account geo- methodology (Section 2) and then it proceeds in presenting
technical data, like information from boreholes and/or the form of the data input (Section 3), the subsurface pro-
in situ tests. Hence, although it offers the potential for filing method (Section 4) and the form of the multi-variable
using an accurate numerical tool like Shake91 [15], the relations that are used for the seismic ground response
accuracy of the respective analyses is hindered by the analyses (Section 5). The paper ends with a detailed appli-
crudeness of the input data. Following the same path, cation of the proposed methodology for a GIS-aided seis-
[17,30] present an integrated modular tool, which compiles mic microzonation study of Ano Liossia, an urban
A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523 507

municipality of the greater Athens area, which suffered at this stage of development of the automated
severely in the 7 September 1999 Athens earthquake methodology.
(Ms = 5.9). To better understand the proposed methodology, Fig. 2
presents its flow chart. The remaining of this section
2. Outline of methodology explains the flow chart, always in relation to the 4-Step
procedure outlined in Fig. 1. Specifically:
The proposed methodology does not alter the 4-Step
general procedure of performing seismic microzonation 2.1. Collection of data (Step 1)
studies outlined in Fig. 1. Rather it merely automates the
actions enclosed by the thick dashed line in this figure. In Based on Fig. 2, the proposed methodology requires
this perspective, it simplifies the general procedure and input from:
focuses on the most important aspect of the seismic ground
response, namely the spatial variation of the response spec- – a geotechnical study (denoted as database (i) in Figs. 1
tra (i.e. aspect (a) of Fig. 1), as do most researchers in the and 2), with data and profiling from borehole, in situ
field of GIS-aided seismic microzonation (see and laboratory tests,
[17,19,27,30]). This is also why aspects (b), (c) and (d) of – an engineering geological study (denoted as database (ii)
the seismic ground response appear in gray rather than in Figs. 1 and 2), with profiling of the geological strata in
black color in Fig. 1. This does not mean that these other terms of elevation and depths, and
aspects are not important and hence they have also – a seismological (or seismo-tectonic) study (denoted as
attracted attention in GIS-related literature, but not neces- database (iii) in Figs. 1 and 2), with information on
sarily in relation to seismic microzonation (e.g. [20] for liq- the design earthquake parameters that apply to the
uefaction and [29] for slope stability problems). Their study area.
exclusion here is merely based on the need for simplicity
As implied from above, the foregoing studies need to be
presented in the form of respective electronic databases for
ease of data access and manipulation, thus ending Step 1 of
the seismic microzonation study (see Fig. 1).

2.2. Import and interpretation of selected data (Step 2)

In general, the three databases may refer to areas larger


than the study area itself. The selection of only the relevant
data is performed by relating the databases to a GIS envi-
ronment, whose software provides options for selecting the
relevant data in a graphical manner, namely by outlining
the geographic limits of the study area. Then, standard
query language (SQL) queries are used for exporting only
the relevant data in the form of tables, since the databases
generally include data that are not necessarily useful for a
seismic microzonation study (e.g. the compressibility con-
stants Cc and Cr estimated by oedometer testing).
The foregoing graphical procedure for databases (i) and
(ii) leads to the selection of N locations of boreholes and/or
in situ geotechnical (e.g. SPT) or geophysical tests (e.g.
cross-hole, SASW), where the value of N depends on the
wealth of the geotechnical database and of course the size
of the study area. After retrieving the pertinent geological,
geotechnical and geophysical data (in the form of tables
geological, geotechnical, geophys-c, geophys-d), the pro-
posed methodology proceeds in creating the subsurface
profile of the study area. This profile is not geographically
continuous, but is constructed separately for each one of
the N distinct locations, as depicted by the loop in terms
of j = 1–N in the flow chart of Fig. 2.
Irrespective of the wealth of the geotechnical database
Fig. 2. Flow chart of automated methodology for a GIS-aided seismic (i), accurate contouring of the seismic ground response
microzonation study. usually requires more than N points. For this reason, the
508 A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523

automated methodology incorporates M distinct locations which then proceeds to the compilation of digital contour
of geologic bedrock outcrops (in table bedrock), based on maps of any of the foregoing site and seismic motion
the surface geology information derived from database parameters. In this process, the user may make use of var-
(ii). In addition, database (ii) may also provide geological ious options for surface contouring (e.g. kriging, spline)
layering information for the N locations of known geotech- offered by the GIS software. These digital maps are used
nical conditions (in table strata). for dividing the study area in zones of practically uniform
Finally, database (iii) provides data for defining the seismic hazard, taking into account the standing seismic
design seismic excitation parameters (e.g. peak acceleration codes in the study area.
at outcropping bedrock abmax ) for the study area, which are Note that the adopted approach of performing point
assumed uniform in the proposed methodology, for rea- estimates of seismic ground response and performing con-
sons of simplicity. touring of the response is simpler than performing the seis-
Further details on the table format of the data input of mic response analysis of the 3D subsurface profile of the
the methodology from the three databases are presented in whole study area, since the latter requires a lot of assump-
Section 3 and Fig. 3, while emphasis on the interpretation tions for the interpolation of soil properties between the
of the data (subsurface profiling) is provided in Section 4 N + M locations and heavy computational effort. More-
and Fig. 4 of this paper. over, this approach may even prove more accurate, since
the numerical codes that are capable of performing 3D seis-
2.3. Seismic ground response analysis and export of results mic ground analysis usually are not equipped with consti-
(Step 3) tutive models appropriate for this kind of non-linear
analysis.
In Step 3, the methodology estimates the 1D seismic
response at the ground surface of each of the N locations 2.5. Discussion
(e.g. peak ground acceleration amax), by employing a set
of multi-variable relations [10–12] for an optimal combina- The computer code ASZ that materializes the flow
tion of simplicity and accuracy. These relations use as input chart of Fig. 2 is written and compiled in MS Visual
the site parameters for each location and the (uniform) Basic. The choice between programming languages is of
design seismic excitation parameters for the outcropping minor importance for this endeavor, since the described
seismic bedrock (all estimated in Step 2). Further details automated procedure is performed via the produced exe-
on the multi-variable relations used for the seismic ground cutable file, externally to the GIS environment and the
response analysis are presented in Section 5 and Fig. 5 of involved databases (i), (ii) and (iii). Similarly, the choice
this paper. between available software for creating these databases
Concluding the previous tasks, one proceeds in export- is again of minor importance. On the contrary, what is
ing the results of the analyses in the GIS environment in of importance is that these databases are linked to the
the form of a table. This table, named output, has N + M same GIS environment, and that one may extract selected
rows and a large number of columns that provide the X data in the form of tables via SQL queries, since these
and Y coordinates, the Z (elevation), as well as a number tables are read as input by the executable file and not
of key site and seismic motion parameters for these of these the databases themselves.
N + M locations, namely:
3. Form of data input
– the thickness of soil deposits H overlying the bedrock,
– the linear (elastic) site period TS,o, As outlined above, the automated methodology requires
– the non-linear site period TS, data from three different databases. This section focuses on
– the amplification ratio Aa of the peak ground accelera- the form of the data input from each of these three
tion, or the peak ground acceleration amax ¼ Aa abmax databases.
(where abmax is the peak acceleration at outcropping
bedrock), 3.1. Data from the geotechnical database
– the amplification ratio Av of the peak ground velocity,
or the peak ground velocity vmax ¼ Av vbmax (where vbmax The database that is indispensable for the automated
is the peak velocity at outcropping bedrock), methodology is the geotechnical database (i). In this paper,
– the spectral acceleration Sa (for 5% damping) at selected this database is entitled HelGeoRDaS [2], has been created
structural periods Tstr. in MS Access and contains information from sampling
boreholes, laboratory and in situ geotechnical or geophys-
ical tests from over 2000 locations within the greater Ath-
2.4. Seismic intensity mapping (Step 4) ens (Greece) area. The database is built on seven primary
tables, which contain different types of information that
Importing table output in the GIS environment marks are produced during geotechnical investigations. Besides
the beginning of Step 4 of the seismic microzonation study, geotechnical data, the types of information included in
A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523 509

these tables are geographic, geological and geophysical.  Table geotechnical: A table with in situ geotechnical test
Specifically, the seven primary tables are as follows: data having six columns referring to: BOREHOLE_ID,
X, Y and Z coordinates, depth of SPT test ZSPT and the
 BOREHOLE: Includes general information for every number of blow counts NSPT. This table has enough
geotechnical investigation and of interest here, the rows as to present all the SPT tests performed within
BOREHOLE_ID (index number), and the X, Y and Z the N locations in the study area.
(elevation) coordinates of the borehole head.  Table geophys-d: A table with geophysical test data
 REAL_ID: This table stores the identification number characterized by discrete measurements of the shear VS
of the investigation as recorded in the borehole log. and the pressure wave VP velocities (e.g. cross-hole test)
 WATER_TABLE: The information deals with the having seven columns referring to: BOREHOLE_ID, X,
fluctuation of the water table during drilling, or, in a Y and Z coordinates, depth ZV of velocity measurement,
case of a piezometer the depth of the water table as VS and VP values. This table has enough rows as to pres-
recorded. ent all discrete geophysical test data performed within
 LITHOLOGY: This table conveys the geological infor- the N locations in the study area.
mation of the stratum. In particular the roof and the  Table geophys-c: A table with geophysical test data char-
base depth of each geological unit are presented, as well acterized by continuous measurements of the shear VS
as the geological description and its colour. and the pressure wave VP velocities (e.g. down-hole or
 IN SITU_TESTS: Information obtained from in situ SASW test) having eight columns referring to: BORE-
tests and carried out inside boreholes is enclosed, HOLE_ID, X, Y and Z coordinates, depth of top ZVtop
namely SPT, CPT, pressuremeter tests, permeability of layer with uniform VS and VP values, depth of bot-
tests and finally geophysical (down-hole, cross-hole, tom ZVbot of the same layer, VS and VP values for
SASW) tests at the various depths. the layer. This table has enough rows as to present all
 LAB_TESTS: Includes data referring to laboratory continuous geophysical test data performed within the
results for soil and rock specimens. For soil specimens N locations in the study area.
in particular, besides the initial and final depth of the
sample, the table includes its natural properties (e.g.
Atterberg limits) as well as its mechanical properties 3.2. Data from the engineering geological database
(e.g. strength parameters c 0 and / 0 , compressibility
parameters Cc and Cr). Presenting an engineering geological study in the form
 ROCKMASS_CLASSIFICATION: This table refers to of a database related to a GIS practically entails that all
rock mass rating (RMR and GSI). information on the type and succession of the geological
strata has been implemented in the form of different cover-
HelGeoRDaS is fully relational, thus allowing rapid data ages depicting the elevation of the head of each stratum.
retrieval. Thereby the information from the BOREHOLE This information can then be used for creating a surface
table is related with data of other tables with a ‘‘one-to- geology map, but also for geological sections at selected
many’’ type of relation, because a geotechnical investiga- locations. As shown in Figs. 2 and 3, such an engineering
tion refers to many lithological units, in situ and laboratory geological database (ii) can be used for two purposes:
tests, etc. Further information on the structure and opera-
tion of this database is beyond the scope of this paper and  Identifying M locations of outcropping bedrock based
can be found in [3]. on the GIS-compiled surface geology map, and thus
What is of interest here is that the required data are exporting a table named bedrock that has M rows and
exported in table format via SQL queries. The geological, four columns referring to: index of bedrock location,
geotechnical and geophysical information that is used for X, Y and Z (elevation) coordinates.
the subsurface profiling is exported from HelGeoRDaS in  Exporting table strata with the geological section at the
the form of four tables. The type of information provided N locations of known geotechnical properties, a table
in these tables is graphically shown in Fig. 3 and is that has seven columns referring to: index number of
explained below: N location, X, Y and Z (elevation) coordinate, depth
of top Zgtop of geological stratum, depth of bottom
 Table geological: A table with geological information Zgbot of stratum, geological description of the stratum.
having seven columns referring to: BOREHOLE_ID This table has enough rows as to present the foregoing
(index number), X, Y and Z (elevation) borehole coordi- geological data for all layers found at the N borehole
nates, depth of top Ztop of layer, depth of bottom Zbot locations in the study area.
of same layer and the geotechnical characterization
according to the USCS. This table has enough rows as Note that a surface geology map is the absolute mini-
to present the foregoing geological data for all layers mum of any engineering geological study (or database)
found within the N borehole locations depicted in the and therefore table bedrock is consider a prerequisite for
study area. an accurate seismic microzonation study. On the other
510 A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523

Fig. 3. Types of data included in tables exported from the geotechnical (i) and the engineering geological databases (ii).

hand, the table strata is considered useful but not manda-


tory. The reason is that this table is expected to merely
enhance the data provided in table geological from the geo-
technical database (i), in the sense that it provides profiling
information for depths greater than the bottom of the
borehole contained in table geological, information that
may not be crucial if the boreholes have drilled through
the bedrock head.

3.3. Data from the seismological database

Presenting a seismological (or seismo-tectonic) study in


the form of a database related to a GIS practically entails
that the spatial variability of the seismic excitation param-
eters is presented in the form of different coverages. In this
paper, the term ‘seismic excitation’ refers to the accelera-
tion time history at the outcropping bedrock (which is
not the actual ground surface if soil covers the bedrock).
In addition, the foregoing ‘excitation parameters’ refer to:
(a) the peak value of the seismic excitation abmax , (b) the pre-
dominant period Te of the seismic excitation defined as the
period that corresponds to the maximum spectral accelera-
tion and (c) the number of significant cycles n defined as the
number of cycles of the seismic excitation that exceed an
effective acceleration level (e.g. 0:65abmax Þ.
As shown in Fig. 2, the excitation parameters in the pro-
posed methodology are assumed a priori uniform through-
out the study area, a simplifying assumption which is valid
for relatively small areas that are not situated directly
above the seismic sources.

4. Subsurface profiling
Fig. 4. Flow chart of subsurface profiling subroutine included in the
As shown in Fig. 2, after importing the necessary data proposed automated methodology.
from the databases (i), (ii) and (iii) in the form of the var-
ious tables, the automated methodology proceeds in esti- involves substantial engineering judgment and has been
mating the subsurface profile separately for each one of implemented in the ASZ code, as a separate subroutine
the N locations. This subsurface profiling procedure whose flow chart is presented in Fig. 4. Observe in this fig-
A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523 511

ure that the subsurface profiling procedure can be divided VS–NSPT relation is not the measured 50, but a value
into three successive parts, which are the subjects of the that is linearly extrapolated in terms of the depth of pen-
remaining paragraphs of this section. etration of the sampler, i.e. NSPT = 50 (30/20) = 75.
– If a layer has more than one SPT blow count, then this
layer is subdivided into sublayers with interfaces
4.1. Soil profiling
assigned at the mid-distance of successive SPT blow
counts (similarly to what is performed for data from
The soil profile required for performing seismic wave
table geophys-d above).
propagation analysis consists of the shear wave velocity
– If a layer has no SPT blow count and its thickness is
VS variation with depth of the soil column. Ideally, the VS
small (less than 1 m) then the layer is not taken into
profile must be based on direct measurements. More specif-
account. Yet, if its thickness is large, then its VS value
ically, as shown in Fig. 4, if a location j has VS measure-
is estimated on the basis of the average NSPT of the
ments either from table geophys-c or from table geophys-
two neighboring blow counts (above and below), the soil
d, the VS profile is assigned on the basis of these direct mea-
type of the layer and the selected empirical VS–NSPT
surements. Specifically, if the direct measurements come
relation, or on the basis of literature VS values.
from table geophys-c, then the VS profile is assigned directly
as the measurements show, i.e. for layer i with VS=VS(i) the
4.2. Depth to bedrock
upper interface is at ZV,top(i) and the lower interface at
ZV,bot(i), respectively (see also Fig. 3). On the other hand,
As shown in Fig. 4, after defining the VS profile for
if the direct measurements come from table geophys-d, then
location j, the subsurface profiling subroutine proceeds
the VS profile comprises of layers whose interfaces are
in assigning the depth of the seismic bedrock H. Ideally,
assigned at the mid-distance of successive measurements,
this is assigned at the interface between the soil and the
i.e. for layer i with VS = VS (i) the upper interface is at
geological bedrock. Yet, in practice, the seismic and the
[ZV(i  1) + ZV(i)]/2 and the lower interface at
geological bedrock do not necessarily coincide and the
(ZV(i) + ZV(i + 1))/2, respectively (see also Fig. 3).
former is sometimes shallower than the latter. For this
Yet, having VS measurements from geophysical tests is
purpose, the automated methodology takes into account
rarely the case for all N locations. Hence, if a location j
primarily the VS profile and secondarily the data from
does not have direct VS measurements, then the VS profile
table geological and the optional table strata. More specif-
is assigned indirectly from the soil stratigraphy, the NSPT
ically, the first approximation of H is based on the VS
values and an empirical relation that relates the NSPT blow
profile and especially at the depth where the VS profile
count to the shear wave velocity VS on the basis of the soil
shows an intense contrast, i.e. at an interface between lay-
type at hand. The profiling initiates from the soil layering
ers where the upper VS is smaller than half the value of
exported in table geological, i.e. the upper interface of layer
the lower VS. To avoid having a definition of H at a
i is at depth Ztop(i) and the lower interface at Zbot(i), with
locally stiff interlayer, this value of H is accepted only if
these depth values explained in Fig. 3. Then, the value of
the value of the lower VS is greater or equal to 550 m/s
VS(i) is estimated from the soil type of the layer (from table
and if the values of VS from depth H and below are con-
geological), the value of NSPT that was executed within the
sistently larger than 550 m/s. This basic procedure
layer (from table geotechnical) and the selected empirical
requires a series of secondary assumptions in order to
VS–NSPT relation. There are many such relations in the lit-
cover all potential cases in practice, the most important
erature and the ASZ code provides the user with various
of which are the following:
options, namely the use of well established relations from
the literature [1,16,21] and the possibility to implement a
– If the VS profile at location j has no interface with
set of case-specific relations for the study area itself. The
intense contrast, or there are more than one such inter-
indirect VS profiling procedure is very easy to implement,
faces, then the seismic bedrock is assigned at the depth
given the simplicity of the empirical relations, e.g. the rela-
H where the VS values becomes consistently larger than
tions of [16] have the following form:
( 0.8 Vb, where Vb is the value of the seismic wave velocity
87:8N 0:282
SPT ; non-cohesive soils in the geological bedrock, as is defined by the user based
V S ðm=sÞ ¼ ð1Þ on the data from tables geological and strata.
107N 0:274
SPT ; cohesive soils
– If the depth to seismic bedrock H has not been assigned
The outlined indirect VS profiling procedure requires a ser- by any of the foregoing criteria, this means that the
ies of secondary assumptions in order to cover all potential borehole or the in situ (geotechnical or geophysical) test
cases in practice, the most important of which are the did not reach a sufficient depth to enter the seismic bed-
following: rock. In this case, the user has the option of defining the
depth to seismic bedrock H manually, either at the end
– If a SPT blow count reached a maximum value, say 50, of the borehole or the in situ test (shallow bedrock) or at
and the sampler did not enter a full 30 cm but something the depth of the geological bedrock as it is concluded
less, say 20 cm, then the NSPT value used in the selected from the data of table strata (deep bedrock).
512 A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523

4.3. Site parameters or as the VS value that corresponds to the same travel time
of S waves from the seismic bedrock up to the ground
As shown in Fig. 4, the subsurface profiling routine surface:
concludes with the estimation of the site parameters, i.e. H
the basic characteristic values of the site, that are to be V S;o ¼ X ; T S;o ¼ 4H =V S;o ð4Þ
k hðiÞ
used in the ground response analyses that follow. The pri- i¼1 V ðiÞ
S
mary site parameter is the linear (elastic) site period TS,o.
Its estimation is performed according to the simplified In closing, the third and final site parameter corresponds to
Rayleigh method [6], which subdivides the VS profile into the elastic period of a bedrock column having the same
k layers starting enumeration from the seismic bedrock thickness H, a parameter defined as Tb = 4H/Vb, where
(as shown to Fig. 5) and estimates the TS,o according Vb is the (assumed uniform) shear wave velocity in the seis-
to, Eq. (2): mic bedrock.
T S;o ¼ 2p=x;
Pk 5. Multi-variable relations for seismic ground response
2 2
2 4 i¼1 ½½H  zðiÞ hðiÞ=V S ðiÞ  analysis
x ¼ Pk 2
;
i¼1 ½½xðiÞ þ xði þ 1Þ hðiÞ
2 As outlined in Fig. 2, after concluding the subsurface
xði þ 1Þ ¼ xðiÞ þ ½½H  zðiÞhðiÞ=V S ðiÞ  ð2Þ
profiling for the N selected locations, the proposed meth-
where h(i) and VS(i) are the thickness and the shear wave odology estimates the seismic ground response (including
velocity of layer i, while z(i) is the distance of the midpoint effects of soil non-linearity), independently for each of
of layer i from the seismic bedrock. Note that x(1) = 0 in these locations. The seismic ground response at the surface
the foregoing calculations, and that for greater accuracy of each location j is estimated via multi-variable relations
the total number of layers k should be greater or equal [10–12], which are based on a regression analysis of the
to the number of layers with different VS values. Practically input data and the results of over 700 numerical analyses
speaking layers with h(i) 6 1 m should be used in this esti- of seismic ground response and have been shown accurate
mation, by appropriately dividing the uniform layers of the enough to reproduce their results. The remainder of this
VS profile. Given the value of TS,o, the second site param- section presents briefly these relations and their application
eter, i.e. the average shear wave velocity VS,o of the soil col- in practice, while details can be found in the aforemen-
umn, is estimated as VS,o = 4H/TS,o. tioned publications.
It should be underlined, that the foregoing procedure
for estimating the TS,o is more accurate than widely used 5.1. General
empirical methodologies that estimate the VS,o first, either
as the average value of the VS profile, weighted in terms The site model used for these over 700 numerical analy-
of the h(i): ses consisted of a column of horizontal soil layers, with
Pk non-linear visco-elastic response, resting upon uniform, lin-
hðiÞV S ðiÞ ear visco-elastic bedrock. Computations followed the
V S;o ¼ i¼1 ; T S;o ¼ 4H =V S;o ð3Þ
H equivalent linear method [15,26], assuming 1D vertical
propagation of earthquake-induced shear waves between
the seismic bedrock to the ground surface. The seismic bed-
rock in all these analyses was defined within Neogene or
older geological formations. In terms of soil categories of
seismic codes, the seismic bedrock corresponded to NEH-
RP soil category B (and rarely A or stiff C), or EC-8 cate-
gory A (and rarely stiff B).
The seismic excitations used for the analyses were actual
strong motion recordings that were applied at the surface
of the outcropping seismic bedrock and had a peak value
abmax varying between 0.01 g and 0.45 g. Their predominant
period Te varied from 0.1 to 0.8 s, capturing a wide spec-
trum of potential earthquake events. Their duration was
also highly variable, with equivalent number of uniform
cycles n = 0.5–24.
The soil profiles used in the numerical analyses corre-
sponded to sites where geotechnical investigations had
been performed, including measurements (mainly cross-
Fig. 5. Form of input parameters for the estimation of the linear (elastic) hole) of shear wave velocity VS. Soil non-linearity in these
site period TS,o via the simplified Rayleigh method. analyses was introduced by shear modulus degradation and
A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523 513

hysteretic damping increase curves. Lacking actual mea- ratio TS/Te that has the most pronounced effect on Av, sim-
surements, the experimental curves of [28] for soil materials ilarly to what was observed for Aa above. Yet, notice that
with plasticity index Ip between 0% and 50% were used in the number of cycles n was found statistically independent
the majority of these analyses. of the value of Av and that the peak values of Av are ex-
The basic site parameter for a seismic ground response pected for TS/Te = 1.5, since the predominant period of
analysis is the fundamental period of the soil column over- the velocity time history is usually higher than that of the
lying the seismic bedrock. As such, the non-linear value of respective acceleration history.
the site period TS was assigned as the basic site parameter
of the proposed relations, a parameter estimated by: 5.3. Horizontal elastic response spectra
 b 1:04
a
ðT S =T S;o Þ2 ¼ 1 þ 5330ðV S;o Þ1:3 max ð5Þ To focus on the frequency content of the elastic response
g
spectra, the emphasis of [10–12] was given on the 5%
where TS,o and VS,o are the linear (elastic) values of the site damped normalized spectral acceleration, S a ¼ S a =amax .
period and the average shear wave velocity, whose estima- Thus, the corresponding soil surface-to-bedrock outcrop
tion is based on Eq. (2). Observe that the ratio TS/TS,o prac- normalized spectral amplification ratio, ASa , is approxi-
tically quantifies the effect of soil non-linearity on the site mately given by:
period, and that the relative effects of the included parame- 2
ðS  Þ 1 þ ASa;r ðT str =T S Þ
ters abmax and VS,o are opposite, but quantitatively similar (in ASa ¼  a soil ¼ vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðS a Þbedrock u "   2 #2 " ! #2
percentage of differentiation). Finally, note that the VS,o in u 1 þ ASa;r
Eq. (5) should be entered in (m/s), while g is the acceleration t 1  T str þ
T str

TS ASa;p TS
of gravity entering Eq. (5) in the dimensions of abmax .
ð10Þ
5.2. Peak seismic motion parameters
where Tstr is the fundamental period of a single degree of
This paragraph presents the relations for the soil sur- freedom structural system, and ASa;p and ASa;r are the peak
face-to-bedrock outcrop peak acceleration and velocity and residual values of the normalized spectral amplification
ratios, Aa and Av, respectively. The former is given by: ratio ASa that are expected for Tstr/TS = 1 (resonance) and
Tstr/TS ! 1, respectively. Given the form of Eq. (10), the
2
amax 1 þ C 1;a ðT S =T e Þ statistical analysis of [10–12] focused on ASa;p and ASa;r ,
Aa ¼ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð6Þ
abmax 2 2 2 which gives, in turn:
½1  ðT S =T e Þ  þ ½C 2;a ðT S =T e Þ
8  0:058
>
> TS
where amax and abmax are the peak ground acceleration val- >
> 1 þ 0:318
>
> Te
ues at soil surface and outcropping bedrock, respectively >
>  0:504  
<
and:  Tb 0:613 TS
ASa;p ¼ 1:318 þ 0:279 n 1 ð11Þ
 b 0:17 >
> TS Te
a n0:5 >
>  0:504
C 1;a ¼ 1:2 max ; >
>
1 þ n0:5 >
> T
g : 1:318 þ 0:837 b n0:613
TS
C 2;a ¼ 1:05 þ 0:57ðT b =T S Þ ð7Þ
where Eqs. (11a) and (11c) hold for TS/Te < 1 and TS/
Observe from Eqs. (6) and (7), that of the included param-
Te > 4 respectively, while Eq. (11b) holds for 1 6 TS/
eters abmax , n, Tb/TS and TS/Te, it is the normalized site per-
Te 6 4. Similarly, the ASa;r is given by a three-partite
iod ratio TS/Te that has the most pronounced effect on Aa.
equation:
For example, note that for TS/Te = 1, the Aa becomes max- 8  
imum due to resonance. > 1  0:302 T S
>
>
The relation for Av is similar to Eq. (6), namely: >
> Te
>
>
2
>
<  0:474  
vmax 1 þ C 1;v ðT S =1:5T e Þ  Tb 0:406 T S
Av ¼ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð8Þ ASa;r ¼ 0:698 þ 0:189 n 1 ð12Þ
b
vmax >
> TS Te
2 2
½1  ðT S =1:5T e Þ  þ ½C 2;v ðT S =1:5T e Þ
2 >
>  
>
> Tb
0:474
>
>
where vmax and vbmax are the peak ground velocity values at : 0:698 þ 0:945 n0:406
TS
soil surface and outcropping bedrock, respectively and
 b 0:124 where Eqs. (12a) and (12c) hold for TS/Te < 1 and TS/
a Te > 6 respectively, while Eq. (12b) holds for 1 6 TS/
C 1;v ¼ 0:88 max ;
g Te 6 6. Observe that for both the ASa;p and ASa;r , the TS/
C 2;v ¼ 1:087 þ 0:598ðT b =T S Þ ð9Þ Te is the most important parameter, while abmax proved sta-
tistically insignificant.
Observe from Eqs. (8) and (9), that of the included param- As described in detail in [10–12], the foregoing relations
eters abmax , Tb/TS and TS/Te, it is the normalized site period have an unbiased error of ±20–26% with respect to the
514 A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523

results of the 1D equivalent linear method on which they thermore, this figure shows that the study area is almost
are based. Yet, whenever the 1D equivalent linear method surrounded by outcropping rock formations, the lime-
fails to reproduce the recorded response, so do the pro- stones of Parnitha and Egaleo mountains to its north and
posed relations. In other words, these relations inherit the south, respectively, and marly formations to its east. The
accuracy as well as the limitations of the 1D equivalent lin- soil in Ano Liossia (shown in white color in Fig. 6) consists
ear method, and should be used merely as a user-friendly of pleistocenic talus cones and in some locations sandy-clay
alternative. sediments, and is underlain by clayey-marly materials or
limestone, depending on the location, which comprise the
6. Application of proposed methodology geological bedrock. In all cases, the soil is rather stiff
(NEHRP category C, EC-8 category B).
As an application of the proposed automated methodol- Step 2 continued with selecting the pertinent geotechni-
ogy for GIS-aided seismic microzonation studies, the area cal data for the Ano Liossia municipality. Specifically, after
of Ano Liossia Municipality was chosen. It is located in outlining the study area in the GIS, data from 13 boreholes
the greater Athens (Greece) area (10 km NNW of the city and/or in situ test sites were located in HelGeoRDaS [2,3].
center) and suffered severe structural damage from the 7 Hence, N = 13 in Figs. 2 and 4 for Ano Liossia, and the
September 1999 Athens earthquake (MS = 5.9) that struck locations of these 13 sites are marked by black squares
at an epicentral distance 10 km to its west. The aim of this and are denoted as G1 through G13 in Fig. 6. Further-
application is not to actually perform a seismic microzona- more, Fig. 6 includes (with hollow squares) the M sites of
tion study for this area, but to show how such a study may bedrock outcrops that were located manually on the basis
be performed with the use of the proposed automated of the surface geology map and summarized in table
methodology. Hence for the purpose of this application, bedrock.
the Athens earthquake is assumed to be one of the poten- The retrieval of geotechnical data from HelGeoRDaS
tially devastating scenarios, and not necessarily the produced tables: geological, geotechnical and geophys-c,
‘‘design’’ earthquake for the study area (an issue to be i.e. besides the stratigraphy and the SPT blow counts,
decided upon by seismologists). Yet, this assumption geophysical data were also available, namely from
enables a rough comparison of predicted seismic motion down-hole measurements that were performed next to
amplification with recorded structural damage during the each of the respective 13 borehole locations. Hence,
foregoing earthquake, but also a detailed comparison with according to the subsurface profiling subroutine of
the results of the actual seismic microzonation study of the Fig. 4, the VS profiles of these 13 locations were based
area that also used the Athens earthquake as one of the directly on the down-hole measurements and disregarded
potential devastating seismic scenarios [14]. Some prelimin- the geotechnical data. Fig. 7 presents the thus constructed
ary results from this application were presented in [22], but 13 VS profiles and the automatically assigned depth H of
this section provided full details and the final form of this the seismic bedrock. Observe that in most locations the H
application following the 4-Step procedure presented was assigned as the depth beyond which the VS measure-
earlier. ments became consistently larger than 0.8 Vb = 600 m/s,
with Vb = 750 m/s, a value selected by the user on the
6.1. Collection of data for Ano Liossia (Step 1) basis of the geological and geotechnical data. Exceptions
to this rule are locations G6 and G12, whose VS measure-
The application initiates with the collection of geotech- ments did not reach very high values and the depth H was
nical (database (i)), engineering geological (database (ii)) assigned at the bottom of the borehole and its respective
and seismological information (database (iii)) for the area. down-hole test, and locations G1, G7 and G13 that H
In particular, the geotechnical data were organized in Hel- was assigned at the depth of an intense contrast in VS val-
GeoRDaS [2,3] (see Section 3), the engineering geological ues, clearly marking the end of soil deposits and the
information was retrieved by [14] and organized in data- beginning of the underlying bedrock. Overall, the depth
base (ii), while details on the seismological data, with refer- H varies between 4 and 35m at these 13 locations, accord-
ence to the Athens earthquake, may be found in [24]. ing to the automated methodology.
Based on Fig. 4, after compiling the VS profiles and
6.2. Import and interpretation of selected data for Ano assigning depth H, the subsurface profiling subroutine
Liossia (Step 2) (Fig. 4) ends with the estimation of the site parameters:
the linear (elastic) site period TS,o its respective average
Step 2 initiated with the selection and interpretation of shear wave velocity VS,o and the linear (elastic) bedrock
the pertinent data, starting with Fig. 6 that shows a surface period Tb, for all 13 locations. This procedure led to values
geology map with elevation contours that was created in of TS,o ranging from 0.04 to 0.29 s and to values of VS,o
the GIS environment, after importing the pertinent data ranging from 388 to 518 m/s agreeing with the engineering
from database (ii). As shown in this figure, the Ano Liossia geological description of ‘‘stiff’’ soil deposits in the study
Municipality has a surface area of approximately 10 km2 area. The respective values of Tb ranged from 0.02 to
and is characterized by a mild slope to the north-west. Fur- 0.19 s.
A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523 515

Fig. 6. Geological map with surface elevation contours of study area and locations of N = 13 borehole and/or in situ tests and M selected bedrock
outcrops.

Fig. 7. Shear wave velocity VS profiles of 13 locations within the study area and estimates of depths to seismic bedrock H, according to the automated
methodology.

As shown in Fig. 2, Step 2 ends with the estimation of the ens earthquake. This was a moderate to strong earthquake
design seismic excitation parameters. As explained in the with an MS = 5.9, a focal depth 8–16 km, while the focal
introduction of this section, the seismic excitation used in mechanism analysis indicated normal faulting with ESE-
this application corresponds to the 7 September 1999 Ath- WNW direction and about 40° dip to the SW [25]. In lack
516 A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523

of recordings of the main shock in the study area, for the been performed as part of the actual seismic microzonation
scope of this paper a synthetic acceleration time history is study of Ano Liossia Municipality that was concluded in
used as an approximation of the main shock in the epicen- year 2000 [14], i.e. before conducting the application pre-
tral area [24]. More specifically, Fig. 8 presents this acceler- sented herein. Obviously these numerical analyses include
ation time history and its respective elastic response engineering judgment and case-specific assumptions that
spectrum (for 5% damping) and shows that for this seismic are not necessarily identical to those implemented in the
excitation, the procedure outlined in the Section 3.3 leads to generic automated methodology proposed here. Neverthe-
the following values for the excitation parameters of the less, the input data and results from these analyses pose as
automated methodology: abmax ¼ 0:633 g, Te = 0.13 s and an ‘objective’ approximation of how this seismic microzo-
n = 4. nation study has been performed manually by a group of
experts in this field of work.
6.3. Seismic ground response analysis for Ano Liossia and Hence, a direct one-to-one comparison of processed
export of results (Step 3) input data and results between the ‘manual’ methodology
(of reducing the data and using Shake91 [15]) and the pro-
Step 3 involves the estimation of the seismic ground posed ‘automated’ methodology for the 13 locations is pre-
response at the surface each of the 13 locations, by using sented in Fig. 9. In particular, the comparison is made in
Eqs. (5)–(12). Yet, merely presenting the pertinent results terms of H, VS,o, TS,o, TS, Aa, Av, Sa[Tstr = 0.2 s] and
would not provide an estimate of the accuracy of the pro- Sa[Tstr = 0.4 s] in plots 9a through 9 h, respectively.
posed automated methodology. Therefore, actual numeri- Regarding the last two parameters, i.e. the spectral acceler-
cal analyses using Shake91 [15] were performed for each ation Sa at the ground surface for structural periods
of the 13 locations with the acceleration time history of Tstr = 0.2 and 0.4 s, it should be mentioned that they corre-
Fig. 8 used as the seismic excitation. These analyses had spond to the peak acceleration suffered by 1–2 storey and

Fig. 8. Definition of excitation parameters abmax , n and Te based on the acceleration time history and its respective elastic response spectrum (for 5%
damping), according to the automated methodology.

Fig. 9. Comparison between the manual and the automated methodology, in terms of various site and seismic motion parameters: H, VS,o, TS,o, TS, Aa,
Av, Sa[Tstr = 0.2 s] and Sa[Tstr = 0.4 s].
A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523 517

3–4 storey buildings, respectively, i.e. the large majority of H = 0 m, VS,o = 750 m/s, TS,o = 0, TS = 0, Aa = 1,
the building stock in Ano Liossia Municipality. Specifi- Av, = 1, Sa[Tstr = 0.2 s] = 1.43 g and Sa[Tstr = 0.4 s]
cally, in every plot of Fig. 9 each symbol corresponds to = 0.76 g, based on the seismic excitation of Fig. 8.
a different location and is obtained using as coordinates,
on one hand, the value of the pertinent parameter from 6.4. Seismic intensity mapping for Ano Liossia (Step 4)
the ‘manual’ methodology and, on the other hand, the
respective value from the ‘automated’ methodology. The Importing table output (with the N + M rows) to the
solid diagonal line is the locus of points of perfect agree- GIS environment in Step 4 enables contour mapping of
ment and the comparison of Fig. 9 shows an overall satis- the foregoing site and seismic motion parameters, a proce-
factory agreement between the two methodologies. In more dure that was also performed for the results of the ‘manual’
detail, the most satisfactory comparison is obtained in methodology for comparison purposes.
terms of TS,o and TS in plots 9c and 9d, and the least sat- Hence, Fig. 10 compares the contours for the depth to
isfactory for Aa and Av in plots 9e and 9f, respectively. bedrock H(m) as is concluded from the proposed auto-
Although Fig. 9 shows an overall satisfactory agreement mated (dashed lines) and the manual (solid lines) method-
between the results of the ‘manual’ and the ‘automated’ ology described above. Observe that, in practical terms, the
methodology, the best estimate of the reliability of the lat- variability of the depth H to the seismic bedrock is very
ter is a comparison in terms of the produced digital contour well reproduced by the automated methodology. Further-
maps, i.e. the end result of any seismic microzonation more, note how important is the inclusion of the M points
study. Hence, ending Step 3, table output is constructed of bedrock outcrop in the perimeter of the soil deposits for
having as columns the X and Y coordinates, the Z (eleva- the accurate mapping of the actual depth H to the bedrock.
tion) of each location and the aforementioned site and seis- Fig. 11 presents a similar comparison of contours for the
mic motion parameters: H, VS,o, TS,o, TS, Aa, Av, linear (elastic) site period TS,o(s) within the study area.
Sa[Tstr = 0.2s] and Sa[Tstr = 0.4 s]. The first N = 13 rows Observe that the very good agreement between the two
of this table correspond to the 13 sites (G1—G13) shown methodologies is retained and this poses as an optimistic
in Fig. 6 and include the values for the foregoing site and precursor for similar agreement of the results of the seismic
seismic motion parameters that are presented in Fig. 9. ground response analyses. The first but most important
In the sequel, this table includes M rows with the respective output parameter from these analyses is the non-linear site
data for the M sites of bedrock outcrops (also shown in period TS, because this parameter is the cornerstone of the
Fig. 6) to which the following values are assigned: proposed multi-variable relations. Hence, Fig. 12 presents

Fig. 10. Comparison between the manual and the automated methodology, in terms of the digital contour maps for the depth to bedrock H(m).
518 A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523

Fig. 11. Comparison between the manual and the automated methodology, in terms of the digital contour maps for the linear (elastic) site period TS,o (s).

Fig. 12. Comparison between the manual and the automated methodology, in terms of the digital contour maps for the non-linear site period TS (s).
A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523 519

Fig. 13. Comparison between the manual and the automated methodology, in terms of the digital contour maps for the peak ground acceleration
amplification ratio Aa.

Fig. 14. Comparison between the manual and the automated methodology, in terms of the digital contour maps for the peak ground velocity
amplification ratio Av.
520 A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523

Fig. 15. Comparison between the manual and the automated methodology, in terms of the digital contour maps for the spectral acceleration Sa(g) for
Tstr = 0.2 s (5% damping).

Fig. 16. Comparison between the manual and the automated methodology, in terms of the digital contour maps for the spectral acceleration Sa(g) for
Tstr = 0.4 s (5% damping).
A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523 521

the comparison of contours for TS(s) in the known format engineering geological and seismological data in the form
of dashed and solid lines for the automated and the manual of databases related to the GIS and Step 4 being the com-
methodologies, respectively. Observe that the two method- pilation of digital contour maps of seismic intensity param-
ologies coincide to a very large degree, i.e. note how the eters in the GIS environment. Yet, the emphasis here is on
two distinct regions with TS > 0.2 s in the southern and a methodology that automates the core of this general pro-
eastern sides of Ano Liossia are zoned almost identically cedure, namely the selection of only the pertinent data and
by the two methodologies. their interpretation for establishing the subsurface profile
This very good agreement is slightly relaxed in Fig. 13 of the study area (Step 2) and the estimation of the seismic
that compares the respective contours for the peak ground surface response for a given seismic excitation sce-
ground acceleration ratio Aa. For example, the region of nario (Step 3). This automated methodology is performed
Ano Liossia with intense amplification (Aa > 1.3) is zoned externally to the GIS environment, via a computer code
a bit further to the north by the automated methodology, named Auto-Seismo-Zonation (ASZ) whose flow chart,
yet the region of generally considerable amplification characteristics and assumptions are described in detail in
(Aa > 1.15) is similarly zoned by both methodologies. this paper.
Similar, in nature, are the results for the peak ground More specifically, the automated methodology presup-
velocity ratio Av that are presented in Fig. 14. In particu- poses that, at minimum, the geotechnical data are orga-
lar, the region of Ano Liossia with very intense amplifica- nized in the form of a geotechnical database related to a
tion (Av > 1.4) is zoned mainly to the east, but the GIS. In this paper, the geotechnical database is named Hel-
automated procedure predicts it to have a relatively smal- GeoRDaS [2,3], has a relational structure and contains
ler area. Yet, the region of generally considerable amplifi- information from sampling boreholes, laboratory and
cation (Av > 1.2) is very similarly zoned by both in situ geotechnical or geophysical tests from over 2000
methodologies. locations within the greater Athens (Greece) area. In par-
The foregoing results refer to the peak seismic motion ticular, the ASZ code processes selected data from Hel-
parameters at the ground surface. Yet, of most importance GeoRDaS and along with engineering geological
for the structural integrity of buildings are the values of the information proceeds in establishing the shear wave veloc-
spectral acceleration Sa(g) at representative values of Tstr. ity VS profile with depth at selected locations within the
In particular, the comparison in terms of the Sa values study area (Step 2). Based on the seismological data for
for Tstr = 0.2 s (that corresponds to 1-2 storey buildings) the study area, the methodology proceeds in estimating
is shown in Fig. 15. This comparison is considered very sat- the seismic ground surface response via recently proposed
isfactory. For example, observe in Fig. 15 that the regions multi-variable relations [10–12] (in Step 3) that duplicate
with very high spectral acceleration (Sa > 2 g) from the two successfully the results of a more robust but not as user-
methodologies are very similar, while the coincidence of the friendly numerical methodology, that of the 1D equivalent
respective regions for Sa > 1.7 g is even greater. Finally, for linear method of Shake91 [15].
Tstr = 0.4 s (that corresponds to 3–4 storey buildings), the An exhaustive evaluation of the proposed automated
comparison of Sa values from the two methodologies methodology for the seismic microzonation study of Ano
becomes even more satisfactory (see Fig. 16). Liossia (Greece) shows satisfactory similarity to an analo-
In closure, it should be mentioned that the structural gous study that was executed ‘manually’ by a group of
damage in the Athens 1999 earthquake in Ano Liossia experts in this field of work [14] using Shake91 [15]. The
was mostly situated to the center and the east of the area depicted differentiations between the results of the auto-
[9], i.e. the regions that correspond roughly to Sa[Tstr = mated and the manual methodologies are attributed to:
0.2 s] > 2 g and Sa[Tstr = 0.4 s] > 0.9 g in Figs. 15 and 16,
respectively. The fact the seismic excitation of Fig. 8 that – The intrinsic, yet unbiased, error of ±20–26% of the
was used by both methodologies was synthesized to simu- multi-variable relations as compared to Shake91 [15],
late the seismic motion at the outcropping bedrock of the an error that has been well documented by [10–12].
area during the main shock of that earthquake leads to – The differentiations in critical decisions of the methodol-
the conclusion, that the heavy structural damage toll in ogy (e.g. the depth to bedrock H) that unavoidably
these regions can be related to, and roughly explained by, occur by using the generic and deterministic computer
the very high spectral accelerations predicted by both ver- code ASZ, as compared to the case-specific opinion of
sions of seismic microzonation studies of Ano Liossia an expert.
Municipality, a fact that adds to the reliability of the pro-
posed automated methodology. The accuracy and the user-friendliness of the proposed
methodology are mostly based on the respective qualities
7. Summary and discussion of the foregoing multi-variable relations. In particular,
these relations are more accurate than the similarly simple
In this paper, the general procedure for performing GIS- analytical solutions or pertinent empirical methods (e.g.
aided seismic microzonation studies has been divided into strong motion attenuation relations [7,18]) and much easier
four Steps, with Step 1 being the collection of geotechnical, to use than similarly accurate numerical methods for 1D
522 A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523

seismic wave propagation (e.g. Shake91 [15], Spectra [8], the methodology, if he sees it fit. Nevertheless, this paper
EERA [4], NERA [5]) that also take into account soil- shows that a generic, user-friendly automated methodology
non-linearity. The generic nature of the methodology orig- for GIS-aided microzonation studies is possible without loss
inates from the fact that it can be used with any pre-exist- of accuracy, at least for a preliminary estimation of the seis-
ing geotechnical database, given that the required data are mic risk and for depicting high-risk locations where further
exported in table format via SQL queries, i.e. via a proce- geotechnical investigations should be performed and more
dure applicable to any database. detailed (possibly numerical) analyses should be executed.
Furthermore, it has to be underlined that the proposed
automated procedure makes the following three assump- Acknowledgements
tions, which pose as the limitations of the procedure itself:
The authors wish to thank the Earthquake Protection
(a) The seismic wave propagation is considered practi- and Planning Organization of Greece (O.A.R.P.) for partly
cally one-dimensional and vertical, which means that funding this research. Dr. Andreas A. Antoniou would also
neither the seismic bedrock nor the ground surface like to acknowledge support from the State Fund Institu-
are characterized by intense topographic irregulari- tion of Greece (I.K.Y.).
ties. Although this assumption is realistic for the
study area in the presented application, in cases that
References
these conditions are not met, then neither the multi-
variable relations [10–12] nor any 1D numerical [1] Andrus RD, Stokoe II KH. Liquefaction of soils from shear
method (e.g. Shake91 [15], EERA [4]) can be consid- wave velocity. J Geotechn Geoenviron Eng, ASCE
ered as accurate and 2D (or even 3D) case-specific 2000;126(11):1015–25.
seismic ground response analysis must be performed. [2] Antoniou AA. HELGEORDAS: HELlenic GEOtechnical Relational
DAta System, in cooperation with GIS. Greece: Doctoral disserta-
Alternatively, simplified criteria for the topography
tion. Geotechnical Department, School of Civil Engineering,
aggravation of seismic ground motion may be National Technical University of Athens; 2003.
employed, as the ones proposed by [13] for the case [3] Antoniou A, Papadimitriou AG, Tsiambaos G. A geographic
of single faced slopes. information system managing geotechnical data for Athens (Greece)
(b) The soils in the study area are stable under seismic and its use for automated seismic microzonation. Natural Hazards
(under review); 2008.
motion, i.e. they are neither liquefiable nor sensitive
[4] Bardet JP, Ichii K, Lin CH. EERA: a computer program for
and do not fall under the S1 and S2 categories of equivalent-linear earthquake site response analysis of layered soil
EC-8 or the F category of NEHRP. Although this deposits. University of Southern California, Department of Civil
assumption is realistic for the study area in the pre- Engineering; 2000. p. 38.
sented application, in cases that such soils are [5] Bardet JP, Tobita T. NERA: a computer program for nonlinear
earthquake site response analysis of layered soil deposits. University
depicted on the basis of the geotechnical data, then
of Southern California, Department of Civil Engineering; 2001.
these should be presented as a different coverage in [6] Biggs JM. Introduction to structural dynamics. New York: McGraw-
the GIS environment for the potential definition of Hill; 1965.
zones where future construction works may be per- [7] Boore DM, Joyner WD, Fumal TE. Empirical near-source attenu-
formed only after soil improvement. ation relationships for horizontal and vertical peak ground acceler-
ation, peak ground velocity and pseudo-absolute acceleration
(c) There are no significant active faults traversing the
response spectra. Seismol Res Lett 1997;168(1):154–79.
study area. Although this assumption is realistic for [8] Borja RI, Chao HY, Montans FJ, Lin CH. Non-linear ground
the study area in the presented application, in cases response at Lotung LSST site. J Geotechn Eng, ASCE 1999;125(3):
that the seismo-tectonic study depicts such faults then 187–97.
these should be considered in the respective database [9] Bouckovalas GD, Kouretzis GP. Stiff soil amplification in the 7
September 1999 Athens (Greece) earthquake. Soil dynamics and
as a different coverage in the GIS environment for the
earthquake engineering 2001;21(8):671–87.
potential definition of setback limits for future con- [10] Bouckovalas GD, Papadimitriou AG. Multi-variable relations for
struction works. As an example of the width of such soil effects on seismic ground motion. Earthquake engineering and
setback limits, the reader is referred to [23], for cases structural dynamics 2003;32(10):1867–96.
of normal and reverse active faults covered by soil [11] Bouckovalas GD, Papadimitriou AG. Multi-variable relations for
soil effects on peak seismic motion parameters. Techn Chron Sci J
layers.
2005;25(1):9–22. January–April.
[12] Bouckovalas GD, Papadimitriou AG. Multi-variable relations for
In closure, it is emphasized that the authors believe that soil effects on elastic response spectra. Techn Chron Sci J
there is no deterministic code that could replace the edu- 2005;25(1):23–36. January–April.
cated opinion of an expert, and for this reason the ASZ code [13] Bouckovalas GD, Papadimitriou AG. Numerical evaluation of slope
topography effects on seismic ground motion. Soil Dynam Earth-
has options for user intervention. From another point of
quake Eng 2005;25(7–10):547–55.
view, an educated code programmer of the methodology [14] Bouckovalas GD, Papadimitriou AG, Antoniou AA, Andrianopou-
could merely retain its basic flow chart (Figs. 2 and 4) and los KI, Kouretzis G. Analysis of seismic risk of soils in Ano Liossia
its ‘heart’ (the multi-variable relations [10–12]) and poten- Municipality. Technical Report. Geotechnical Department, School of
tially change the assumptions involved in various steps of Civil Engineering, NTUA; 2000.
A.G. Papadimitriou et al. / Computers and Geotechnics 35 (2008) 505–523 523

[15] Idriss IM, Sun JI. SHAKE91 – A computer program for conducting rupture. In: Proceedings of the 4th international conference on
equivalent linear seismic response analysis of horizontally layered earthquake geotechnical engineering. Thessaloniki; June 25–28, 2007,
soils. CGM research report. UC Davis; 1992. Paper 1583.
[16] Imai T, Tonuchi K. Correlation of N value with S-wave velocity and [24] Papadimitriou P, Kaviris G, Voulgaris N, Kassaras I, Delibasis
shear modulus. In: Proceedings of the second European symposium N, Makropoulos K. The September 7, 1999 Athens earthquake
on penetration testing. Amsterdam, Netherlands; 1982. sequence recorded by the Cornet network: preliminary results of
[17] Jimenez MJ, Garcia-Fernandez M, Zonno G, Cella F. Mapping soil source parameters determination of the main shock. Annales
effects in Barcelona, Spain, through an integrated GIS environment. Gelogiques des Pays Helleniques, le serie XXXVIII; 2000. p. 29–
Soil Dynam Earthquake Eng 2000;19:289–301. 40.
[18] Kawashima K, Aizawa K, Takashi K. Attenuation of peak ground [25] Papanastassiou D, Stavrakakis G, Drakatos G, Papadopoulos G.
acceleration, velocity and displacement based on multiple regression The Athens, September 7, 1999, Ms = 5.9 earthquake: first results on
analysis of Japanese strong ground motion records. Earthquake Eng the focal properties of the main shock and the aftershock sequence,
Struct Dynam 1986;14:199–218. Annales Gelogiques des Pays Helleniques, 1e serie XXXVIII; 2000. p.
[19] Kienzle A, Hannich D, Wirth W, Ehret D, Rohn J, Ciugudean V, 73–88.
et al. A GIS-based study of earthquake hazard as a tool for the [26] Schnabel PB, Lysmer J, Seed HB. SHAKE – A computer program for
microzonation of Bucharest. Eng Geol 2006;87(1-2):13–32. earthquake response analyses of layered soils. EERC research report.
[20] O’Rourke TD, Bonneau AL, Pease JW, Ship P, Wang Y. Liquefac- UC Berkeley; 1972.
tion and ground failures in San Francisco. Earthquake Spectra [27] Slob S, Hack R, Scarpas T, van Bemmelen B, Duque A. A
2006;22(Sp.Issue 2):S91–S112. methodology for seismic microzonation using GIS and Shake – a
[21] Ohta Y, Goto N. Empirical shear wave velocity equations in terms of case study from Armenia, Colombia. In: Proceedings, engineering
characteristic soil indices. Earthquake Eng Struct Dynam. geology for developing countries, 9th congress of IAEGE, Durban,
1978;6:167–87. South Africa; September 16 – 20, 2002, p. 2843–52.
[22] Papadimitriou AG, Antoniou A, Bouckovalas GD, Marinos P. [28] Vucetic M, Dobry R. Effect of soil plasticity on cyclic response. J
Approximate relations for GIS-aided evaluation of soil effects on Geotechn Eng Div, ASCE 1991;117(1):89–107.
seismic ground motion. In: Proceedings of the 11th international [29] Xie M, Esaki T, Cai M. GIS-based implementation of three-
conference on soil dynamics and earthquake engineering – 3rd dimensional limit equilibrium approach of slope stability. J Geotechn
international conference on earthquake geotechnical engineering, vol. Geoenviron Eng, ASCE 2006;132(5):656–60.
2. Berkeley, USA; January, 2004. p. 39 – 46. [30] Zonno G, Garcia-Fernandez M, Jimenez MJ, Menoni S, Meroni F,
[23] Papadimitriou AG, Loukidis D, Bouckovalas GD, Karamitros D. Petrini V. The SERGISAI procedure for seismic risk assessment. J
Zone of excessive ground surface distortion due to dip-slip fault Seismol 2003;7:259–77.

You might also like