Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
DOI 10.1007/s10518-006-9021-2
O R I G I NA L R E S E A R C H PA P E R
Ezio Faccioli
Abstract The main features of the Risk-UE project approach to assessing the
ground-shaking (and related hazards) distribution within urban areas are described,
as a basis for developing seismic damage scenarios for European cities. Emphasis was
placed in the project on adoption of homogeneous criteria in the quantitative treat-
ment of seismicity and in constructing the ground-shaking scenarios, despite wide
differences in amount and quality of data available for the cities involved. The ini-
tial steps of the approach include treatment of the regional seismotectonic setting
and the geotechnical zonation of the urban area, while the hazard assessment itself
takes the form of both a deterministic analysis, and of a probabilistic, constant-hazard
spectra analysis. Systematic 1D site response analyses were used, mostly in the softer
soil zones, to modify (when needed) the obtained ground motion maps. Earthquake
induced hazard effects, such as liquefaction and landsliding, are also briefly dealt with
at the end.
1 Introduction
This article describes the Risk-UE approach to estimating the earthquake ground-
shaking hazard (and induced effects) in urban areas, and to representing it in ways
suitable for the derivation of damage scenarios. The methods adopted for tackling
the different steps involved are presented in the following sections, illustrating in few
cases applications to some of the cities involved in the project, namely Barcelona
E. Faccioli (B)
Department of Structural Engineering, Politecnico di Milano
Piazza Leonardo da Vinci 32, 20133 Milano, Italy
e-mail: faccioli@stru.polimi.it
342 Bull Earthquake Eng (2006) 4:341–364
actually taken in the project, i.e. developing both a deterministic and constant-hazard
scenarios for all the cities, and introducing comparisons between the two (for a
discussion on methods preferable for hazard scenarios see e.g. Bommer 2002).
Also pointed out in the work programme was the need for a “. . . cross-validation
of the local ground-shaking representations obtained through the use of different
methods (. . .)like (H/V) spectral analysis of microtremors recorded on various soil
types, and numerical models in 1D (Shake, CyberQuake) or 2D ”.
To put the last statement into focus, it is anticipated that the basic (determin-
istic) ground-shaking representation for a city was to be produced through a GIS,
by combining the features of the scenario seismic source(s), appropriate attenuation
relations, and a geological or geotechnical zonation of the zone of interest. As shown
in a later section the validation of the GIS generated map could be fully implemented
only for the aspect of 1D numerical simulations in city zones susceptible of high soil
amplification.
The quoted statements from the Risk-UE work programme outline the contents
of the following sections, that is:
– Aspects related with the seismo-tectonic setting and its treatment.
– Geologic and geotechnical zonation of city areas.
– Deterministic and constant-hazard evaluation methods.
– Validation of ground motion estimations using 1D site response analyses.
– Induced hazard effects.
– Conclusions.
Highlighted at the beginning of each section are the main guidelines adopted for the
topic at hand and the salient aspects of the method used; a flowchart showing the
main ingredients of the overall approach proposed and their mutual relationships is
illustrated in Fig. 1.
Fig. 1 Flowchart showing the main ingredients of the approach proposed for earthquake hazard
assessment in urban areas
344 Bull Earthquake Eng (2006) 4:341–364
Rupture length
Fig. 2 (left) Polygons (numbered 71, 73,..) showing seismic source zones (SSZ) for probabilistic,
constant-hazard analysis in Eastern Sicily, taken from a SSZ model for Italy (Meletti et al. 2000),
together with epicentral locations, year of occurrence and MS magnitudes of earthquakes that caused
damage in the city of Catania. (right) Sketch model of main tectonic features and earthquake sources
in SE Sicily, with estimated rupture length and year of occurrence. From Azzaro and Barbano (1999)
For seismically active faults, useful data are in the European Catalogue of Seismo-
genic Sources FAUST (http://www.faust.ingv.it/current_2.htm) and in the DISS
Database of Individual Seismogenic Sources in Italy (http://www.ingv.it/DISS/), that
allowed to check the current state of knowledge on active faults in the vicinity of
Catania. In cases where a recognised active fault lies inside an extended source zone
with more diffuse seismicity, it was agreed that the seismic activity of the latter should
be reckoned separately from that of the fault.
Overall, since the type and detail of seismotectonic information available for the
different European regions vary widely, the difficulty encountered in translating the
contents of a seismotectonic map into a SSZ map for developing earthquake scenarios
differed appreciably from city to city.
As discussed in more detail below, one basic assumption in Risk-UE was to take
the maximum historically felt intensity (or recorded ground motions) in a city, and
the generating earthquake, as the basic deterministic scenario. This appears rather
natural for regions where strong earthquakes are well documented over a long his-
torical record, as for Bucharest or Catania. In regions where deformation rates are
low and earthquake return periods are very long, historical seismicity data may not
suffice to identify a plausible scenario earthquake, be in location, magnitude or source
geometry, but can be used in support of recent seismicity data and tectonic evidence.
Table 1 displays the main characteristics of the deterministic earthquake scenarios
for each city, together with representative acceleration response spectral values on
rock (see Sect. 4).
346 Bull Earthquake Eng (2006) 4:341–364
Table 1 Deterministic scenario earthquakes for each city; acceleration response spectrum ordinates
SA(0) and SA(1) at 0 and 1 s period, respectively, according to Ambraseys et al. (1996)
Notes: For Bucharest, a different attenuation relation was used, with coefficients calibrated for the
Vrancea source
To construct site dependent, scenario ground motion maps and to identify zones
prone to seismically induced slope failure and liquefaction, a suitable geotechnical
zonation of the urban area is needed. While the importance of ground motion maps
giving a realistic picture of seismic site effects hardly needs emphasizing, it is less
obvious to identify what precisely should be represented in a geotechnical zonation
map to achieve that picture. A crucial simplification is that an earthquake scenario
aims at producing areal estimations of damage, and not at predicting ground (and
structural) response at specific sites. Hence, the representation of ground conditions
could be simpler than that used by current seismic codes to determine the design
elastic spectrum at a site.
Assuming a geological map of the city area is available, a general approach to the
geotechnical zonation issue was identified in Risk-UE as follows:
1. Gather from public and private sources as much ground data as possible from
borings, water wells, geophysical investigations and geotechnical laboratory tests;
most important are the deep geotechnical borings, i.e. reaching formations assim-
ilated to “seismic bedrock”, with shear wave propagation velocity VS of the order
of 800 m/s.
2. After careful processing of the material of step 1, classify available ground profile
data by means of suitable geotechnical units (GUs). In addition to being geologi-
cally consistent, such units should use as main mechanical parameters supporting
the classification CPT or SPT resistances, undrained strength (cu ) and, especially,
VS values. If in-situ measurements of VS are absent, in-hole velocity surveys are
strongly recommended at a few (e.g. two) sites in order to assign VS values to
the most representative GUs. Literature correlations may be used to estimate VS
Bull Earthquake Eng (2006) 4:341–364 347
from SPT resistance or other geotechnical parameters for units with no velocity
measurements.
3. Create a GIS-supported, geotechnical database of all borehole data in which the
original ground profile description has been re-cast in terms of the GUs intro-
duced in step 2, with the VS profile.
4. Draw geotechnical cross-sections across the urban area, depicting the GUs, high-
lighting the VS values, and extended in depth to reach bedrock wherever feasible.
5. Compute along each cross-section the distribution of the weighted-average value,
VS30 , in the uppermost 30 m of the ground, following the Eurocode 8 (CEN 2004,
Table 3.1) formula
30
VS30 = . (1)
hi
i=1,N Vi
6. Arrange the VS30 values by ground classes as in Eurocode 8; the first four classes
(labelled A, B, C, and D) are in particular retained as they are consistent with the
simplified site classification of state-of-art attenuation relations used for ground
motion mapping (see Sect. 4).
7. Guided by the geological map (possibly at a scale not less than 1:10,000), draw the
VS30 contours throughout the zone at study using the results of step 6 as a basis.
3 Experience has shown that, provided a sufficient velocity contrast occurs between near-surface and
underlying materials, the frequency of the fundamental H/V peak at a site is closely related to the
thickness of the surface layers and the S wave propagation velocity and, hence, it is a good indicator
of local seismic response.
348 Bull Earthquake Eng (2006) 4:341–364
vibration periods of common building structures. Bedrock depths from several tens to
hundreds of m, especially in valley- or basin-like configurations will affect the surface
ground response at longer periods (typically exceeding 1 s), which are of interest only
for the taller and more flexible structures (more than, say, 10–12 stories in height), that
are in fact present in significant number in some of the studied cities. As the lesson
of the 1995 Kobe earthquake has shown, in the presence of basin-like configurations
the deeper geological structure must be carefully considered, because it may generate
complex (“basin edge”) effects leading to strong ground motion amplification and
catastrophic damage in limited portions of the urban area (Nagakawa et al. 1996).
Only deep reflection geophysical surveys, if feasible, can in such cases resolve the
deeper geological structure.
As to VS30 as key factor, this parameter, first introduced by Joyner and Fumal
(1984), should not be regarded as a panacea for predicting local ground response
through a geotechnical zonation. VS30 is a synthetic expression of a ground profile
dynamic properties, apt for quantifying the statistical influence of site conditions on
the elastic response spectrum, but may not actually control local response in several
situations. Among these one has, on one side, ground class E of Eurocode 8, since
averaging out the VS values over 30 m would clearly conceal the crucial velocity con-
trast and, on the opposite side, the deep, basin-like configurations. In the latter the
crucial parameters affecting the response (substantial thickness of soil deposits with
soil–rock contrast at depths larger than 30 m) would not be reflected in VS30 . In cases
where ground class E is present, one could use, in some urban areas, a VS value aver-
aged over a depth smaller than 30 m, e.g. 10–20 m. In general, however, the treatment
of class E ground profiles would require more accurate characterisation and more
detailed studies.
Only for Catania and Thessaloniki the existing data allowed to thoroughly apply the
general zonation approach outlined in 3.1. For the remaining five cities of Risk-UE the
situation was very diversified, ranging from absence of a zonation (either geological
or geotechnical) tout court, as in the case of Bitola, to a zonation based only on obser-
vations from records of recent earthquakes, as in the case of Bucharest. Thus, while
all the cities were asked to produce a VS30 map in order to allow proper comparisons,
this proved possible only in a few cases.
the depth of the contact. While a map of VS30 values was not produced, VS values
were estimated for the main materials and found to range from 1,200 m/s for Tertiary
materials to 200 m/s in the deltaic deposition soils.
Where sufficient strong motion records are available and favourably distributed across
the city area, a zonation using some ground response measure is feasible, which may
just need a general geological support. The only such example in Risk-UE was that of
Bucharest: one possible use of instrumental data, with the seismic zoning proposed, is
mapped in Fig. 4. The map relies on the data of the deep Vrancea earthquake of 1986
(M 7.2), for which ground motions were recorded at some eight accelerograph stations
in the municipality; the zonation parameter used is the period (Tc , in s) bounding the
plateau of the acceleration response spectrum on the long period side. A clear ten-
dency of periods increasing from NW to SE is apparent, but comparison with records
of other earthquakes has shown that data from a number of earthquakes are needed
to achieve a sufficiently stable zonation. On the other hand, owing to the presence of
very deep (order of 1 km) unconsolidated sediments in the Bucharest area, a typical
geological/geotechnical zonation centered on descriptions of near-surface soil lay-
ers—however accurate—would fail to bring out the crucial factors controlling local
ground response.
Fig. 4 Seismic zonation of Bucharest displaying the spatial distribution of the control period Tc of
the acceleration response spectrum, based on records of the Vrancea earthquake of August 30, 1986.
The observed Tc values, in s, are shown at the accelerograph sites. Modified after Lungu et al. (1999)
ing down-hole and cross-hole VS profiles measured at some 18 different sites, in which
all the main GUs are represented. This made it possible to associate representative Vs
ranges to the main units identified. Owing to the proximity to Mt. Etna volcano, on a
large part of the city the ground consists of heterogeneous, solidified lava flows from
ancient and historical eruptions, from many tens to a few m thick. From the geotechni-
cal database and zonation, detailed cross-sections and a VS30 contour map have been
constructed, following steps 3–6 in par. 3.1. From such map it has been straightforward
to derive a simplified representation showing three basic ground types, consistent with
the Eurocode 8 ground classification.
A substantial geotechnical database was also assembled for Thessaloniki and the
geotechnical zonation map has been designed based on physical and mechanical soil
properties. The general geological configuration under the urban area is basin-like,
with the basement rock outcropping on the hills E and NE of the city; the thickness
of the overlying clay deposits increases as the coastline is approached. Following the
same reasoning just shown for Catania, a simplified zonation map consistent with
EC 8 has been created. The abundant boring data allowed to generate thickness
maps for each soil category: the map for artificial fills is shown as an example in
Fig. 5.
Bull Earthquake Eng (2006) 4:341–364 351
Fig. 5 Thessaloniki: map of the thickness (in m) of the near-surface artificial fills
In the first case, the scenario would typically show one or few EMS98 intensity value(s)
concisely describing the effects of the selected historical earthquake(s). In the sec-
ond case, the acceleration (or displacement) response spectrum ordinates at vibration
periods T = 0.0, 0.3, 0.6, 1.0 and 2.0 s4 were in principle to be computed by the
attenuation relationship of Ambraseys et al. (1996), hereinafter denoted as AMB96.
If feasible and justified, the results were to be compared with those yielded by region-
ally available relationships. Also, the spectral response scenario had to include site
effects: the standard way for doing this was through the geotechnical zonation (Sect.
3) and the attenuation relation, integrated by additional site response analysis in the
presence of unconsolidated soil deposits (as explained in Sect. 5).
A strong requirement was that the reference earthquake for the deterministic sce-
nario should be the “true” maximum historical earthquake affecting the city. This
choice differs substantially from that of the maximum credible earthquake, typically
adopted in single-site deterministic hazard evaluations of critical structures such as
nuclear power plants, dams, etc. Stated differently, the use of seismotectonic criteria
to “shift” the epicentres of historical earthquakes nearest to city to derive Maximum
Credible Earthquake effects, or the like, was not seen as recommendable for a dam-
age scenario of a European city. Although the historical record was in some cases
recognised to be insufficiently long, the maximum historical earthquake was regarded
in all cases as preferable, to avoid in the first place overly conservative scenario earth-
quakes. With average slip rates on active faults well below 1 mm/yr in most seismic
regions of Europe, maximum credible earthquakes would typically have recurrence
periods in the thousands of years (see e.g. the quoted DISS database for seismogenic
areas of Italy), and were felt to be difficult to handle by most city administrations for
civil defence purposes. Second, the lack of reliable tectonic constraints recognised for
most of the project cities would make the determination of the maximum credible
earthquake parameters extremely uncertain.
Also, if the city at study had historically been exposed to earthquakes of signifi-
cantly different intensity, e.g. both destructive and damaging events (as in the case of
Catania), two different deterministic scenarios were considered desirable. However,
for any given scenario the preferred hypothesis for the seismic source had to be clearly
indicated and justified. Finally, to provide a feeling on the return periods implied by
the deterministic scenario earthquake for a representative city site, the corresponding
response spectrum was to be compared (on the same plot) with the constant-hazard
(probabilistic) spectra for different return periods.
The maximum felt intensity for the city was chosen as the basis of the macroseis-
mic scenario, so as to reduce the arbitrariness associated with other options, such as
the maximum credible earthquake (MCE)5 and the problems arising from the vastly
different time spans of the historical records of each city.
While Barcelona had the lowest felt intensity (IS =VI–VII), other cities experi-
enced intensities as high as VIII or IX, with Catania at the highest level with X–XI
4 If necessary to better describe the plateau of the spectra, the values could be changed: e.g. spectra
of low-magnitude earthquakes are better described using T = 0.2 s.
5 The choice of the maximum credible event was considered admissible only if constrained by good
paleo-seismic data in addition to a long and reliable historical catalogue. This choice was adopted for
the city of Nice.
Bull Earthquake Eng (2006) 4:341–364 353
Fig. 6 Scenario intensity map for Sofia city, generated with the elliptical attenuation models for
Bulgaria (Glavcheva et al. 1983)
(re-assessed at IX–X in a later stage of the project). The relatively high values for
some cities, e.g. VIII for Nice or IX for Sofia, should be interpreted with caution as
scenario intensities for the present, but some of the events were well documented and
truly devastating: as an example, the 1693 Catania earthquake caused a death toll of
10,000–12,000 on a population of about 18,000. Since the source identification of the
reference historical events is difficult, the worst scenario resulting from equally likely
assumptions had to be adopted.
When, as it often occurs, the assessment of the strongest felt intensity is uncer-
tain, independent evaluations had to be sought either by attenuating the epicentral
intensity of the scenario event through suitable relationships, or by estimating local
intensity from independently generated, strong motion parameters. Regional atten-
uation relations for intensity have been developed in several countries, e.g. of the
type point source–circular attenuation [e.g. Grandori et al. (1991), used for Catania
and Nice], or point source–elliptical attenuation [e.g. Glavcheva et al. (1983), used
for Sofia]. The macroseismic scenario map obtained for Sofia is illustrated in Fig. 6 as
an example, taking the 1858 earthquake (I0 = IX–X EMS98, M 6.3) as reference, and
including the effects of soil amplification. For Sofia, as for other cities, different inten-
sity levels are predicted in the urban area, either because of distance attenuation, or of
near-surface geology contrasts. On the latter, a consensus was reached in the project
that an intensity increment of half a degree with respect to hard ground was applicable
on (sufficiently extended) areas on medium stiff clays and medium dense cohesionless
soils. Concerning topographic amplification, as much as one degree increase could be
applied for hilltop, crests and severely sloping ground (Faccioli et al. 2002).
In a few cases, rather than relying on historical sources, it was considered less
uncertain estimating the scenario intensity through correlations with strong motion
parameters such amax or vmax (maximum horizontal ground acceleration and veloc-
ity), based on regional data. The case of Catania seems worth mentioning because: (a)
despite the wealth of historical sources and recent studies on the 1693 earthquake (e.g.
Boschi and Guidoboni 2001), the X–XI (most recently X) felt intensity reported in
354 Bull Earthquake Eng (2006) 4:341–364
catalogues remains possibly too high6 for an event with estimated M 7.1–7.3, assumed
on the offshore Malta Escarpment at 12–15 km distance from the city, and (b) an
intensity X would result in a scenario of widespread destruction, unlikely for the city
of today. Hence, to obtain an independent intensity assessment, it was considered that
an empirical correlation between intensity and vmax , using mostly data from Italian
earthquakes, would be sufficiently stable and less dispersed than with respect to amax ;
other correlations, such as intensity vs. EPA (effective peak acceleration) or vs. some
spectral ordinate would likely be suitable, but were not explored within Risk-UE. For
the purpose of constructing the correlation, EMS98 was roughly assumed equivalent
to MSK 64 and MCS (used for Italian data) intensities between degrees V and IX.
The following expression was obtained:
IEMS = 8.69 log vmax (m/s) + 1.80 R2 = 0.61; σ = 0.71 (2)
applicable from IEMS IV to IX. The calibration data set included 75 observed inten-
sities (mostly MCS), and as many vmax values, from 23 earthquakes of Italy, Turkey
(only Izmit 1999) and few other European Countries. Since the 1693 ground motion
scenario gives vmax values between about 35 and 50 cm/s in Catania, Eq. (2) would give
an intensity of VIII–IX, i.e. over one degree less than the historically based assess-
ment. The value assumed in the end for the actual scenario was IX–X, intermediate
between the previous one and the historical one.
As mentioned, the strong motion attenuation relation AMB96 was adopted to gener-
ate input ground motion maps for level II damage evaluations. AMB96 is calibrated
mostly on European data and makes no distinction among types of source mechanism.
In addition, it covers large ranges in distance (up to 200 km) and magnitude (MS 4.0–
7.5), and uses a simplified ground classification i. e. four ground classes (from rock to
very soft soil) almost coincident with the eurocode 8 ground classification (CEN 2004).
AMB96 was the basic relation used for the project most important shaking scenarios
(acceleration spectral ordinates), and since it uses MS , the surface wave magnitude
was selected as the reference magnitude scale for the scenario earthquakes wherever
possible.
Some cities, like Bucharest, lie in regions where AMB96 is not applicable; for the
deep earthquakes of the Vrancea region that have historically damaged Bucharest an
ad hoc relationship was developed, based on 71 three-component or two-component
accelerograms recorded at 47 free-field stations in Romania. Elsewhere the appli-
cation of AMB96 had to be checked against local data, because the seismic regime
may not be adequately represented in the European database, as e.g. in the case of
Catalonia, in Spain, characterised by low level seismicity. In such cases, AMB96 could
be substituted by local relationships; note, however, that for Spain no strong motion
record for earthquakes bigger that M 5 exists at the moment, an this constitutes an
important limitation for deriving this type of relationships. Only a strong motion
model in terms of PGA is published (Cabañas et al. 1999).
Finally, for hazard maps in terms of vmax , or maximum ground displacement, dmax ,
significant for estimating damage to underground lifelines, the relationships proposed
6 Most likely because of cumulative damage effects: the event in question occurred 2 days after a
first, strong shock (M ∼ 6) that had already caused significant damage in the city.
Bull Earthquake Eng (2006) 4:341–364 355
Legend Legend
Ring Road
Ring Road
Determ.Assiros_1 (PSA-T=1.0sec)
Determ.Assiros_0 (PGA-T=0sec)
HIgh : 206.61gal HIgh : 187.80gal
Medium : 166.62gal Medium : 148.36gal
Low : 126.63gal Low : 108.91gal
Fig. 7 Deterministic maps of response spectral acceleration for period T = 0 s (left) and T = 1 s (right)
for the city of Thessaloniki (Assiros earthquake) including the site effects in a simplified way. Units
are cm/s2
were those by Sabetta and Pugliese (1996), using a database of Italian events with
4.5 ≤ M ≤ 6.87 , Bommer et al. (2000), based on the same European database used
in AMB96, slightly modified, and Tromans and Bommer (2002), including in the
European database also significant crustal events.
An example of spectral acceleration maps is shown in Fig. 7 for Thessaloniki at
T = 0 (maximum ground acceleration), and at T = 1 s. While in both maps accel-
erations range roughly over a factor of 2, the zones of strong response are not the
same, since at 1 s the effect of the depth of looser deposits and fills near the coastline
is considerably enhanced.
7 The Sabetta and Pugliese relationship uses magnitude M = M for M up to 5.5 and M = M for
L S
the larger magnitudes; thus, based on Table 1, for all cities where the possible occurrence of induced
effects was considered significant enough to require a hazard map in terms of vmax the magnitude of
the reference events was the same as for AMB96. Maps in terms of maximum ground displacement
were not produced in practice.
356 Bull Earthquake Eng (2006) 4:341–364
– A regional model of SSZs, see Sect. 2. Seismic activity in each SSZ was to be
quantified in terms of activity rates and maximum magnitude, based on the best
available earthquake catalogues and regional seismo-tectonic models.
– Attenuation relations for the selected hazard parameter(s): being the same as for
the deterministic scenarios in Sect. 4, these will not be further considered here.
The probabilistic hazard description had to satisfy the level II damage scenario
requirements (see Lagomarsino and Giovinazzi this issue); as for the determinis-
tic case, a separate map had to be produced for each spectral ordinate at the specified
period values. A rather uncommon requirement was that each spectral response map
was to be obtained by interpolating values computed for the target probability of
exceedence over a sufficiently dense grid of points covering the city area of interest.
The treatment of the site effects would have to follow the same lines as in the deter-
ministic maps, i.e. through the zonation and the ground classes of the attenuation
relation, supported by more refined evaluations based on 1D site response analyses
for zones on soft deposits (see Sect. 6).
Fundamental requirements for the probabilistic evaluation were the uniformity of
the method and flexibility and efficiency of the computational tool of analysis, consid-
ering that the construction of hazard maps through a sufficiently dense grid of points
would be computationally heavy.
Concerning the computational tool, a code was chosen with enough flexibility to
match the generality8 requirements of Risk-UE. After some tests, CRISIS (version
1999), a code developed by Ordaz and co-workers of Mexico City UNAM was found
satisfactory for the purpose9 . The theoretical background for the code, following the
probabilistic formulation of Esteva (1970) integrated with a description of the char-
acteristic–earthquake occurrence process, may be found in Ordaz et al. (1991). Assets
of CRISIS include: the option of treating the occurrence process as Poissonian or as a
characteristic–earthquake model for any SSZ, the freedom in describing the geome-
try of any SSZ (e.g. an inclined plane), and—in its most recent (2003) version—three
options for the source-to-site distance: hypocentral, epicentral, and fault distance.
Since the performance of CRISIS99 had not been widely documented before,
extensive tests were carried out to check its sensitivity and stability to the choice of
key computation parameters. These included:
– The minimum size of the triangles by which the SSZs are discretised, and the ratio
triangle-to-site distance/triangle size; the choice is important for the sites close to
the source. Triangle sizes of the order of 5 km and ratios of 5–10 were found to
give very stable results.
– The uncertainty in the attenuation relation; the residuals of the estimate of e.g.
amax are assumed as lognormally distributed with median value amax and standard
deviation σlog(a) . The attenuation relation is input in the code as a table, where each
row corresponds to a magnitude and each column to a hypocentral distance. In the
current version, introducing in the table the parameter N, the code performs the
integration from 0 to log(amax ) + Nσlog A , and the function will be re-normalised
8 For instance, evaluation of seismic hazard for a city like Bucharest requires introducing a SSZ
like Vrancea lying deep below the Earth surface and having the geometry of an inclined plane. This
representation is not directly feasible with the well known code SEISRISKIII.
9 To obtain the current (2006), updated version of the code (CRISIS2003, version 3.0.1), interested
readers should address Professor M. Ordaz, Instituto de Ingenieria, UNAM, Mexico City.
Bull Earthquake Eng (2006) 4:341–364 357
in order to obtain unity in the full range. The tests showed that, as |N| decreases,
the hazard curves for a given site tend to the curve free of uncertainty.
– The shortest distance required by the code: when a city area lies inside the surface
projection of a SSZ, the lowest distance used in the attenuation table plays an
important role on its seismic hazard assessment. This is because, if such distance is
not close enough to 0, the code must extrapolate values introducing errors in the
calculations. For instance, it was found for Barcelona that the smallest distance
should be ≤ 1 km.
Attenuation relations that—like AMB96—use the shortest distance between the sur-
face projection of the fault rupture and the site are not consistent with the distance
measure adopted in CRISIS99. The latter makes use of the focal distance R (see Fig. 8),
which is greater than the distance (RS ) adopted by AMB96. The ground motion com-
puted in this way is thus underestimated. This difficulty has been circumvented by
constructing attenuation tables consistent with the geometry shown in Fig. 8, where:
with Mw =moment magnitude, ranging between 4.8 and 8.1, and σ = 0.16.
From Fig. 8 one has ρ 2 = R2 − H 2 , and since RS = ρ − L/2, one easily obtains
When the site (S) lies above the seismic source, RS vanishes. Thus, attenuation tables
for a given SSZ were constructed such that, when for a given magnitude the focal
distance is R, the table yields the acceleration associated to distance RS .
Fig. 9 Catania: rock site (Cathedral), comparison of deterministic acceleration response spectra
with constant-hazard spectra: the latter labelled as “new” if computed using distances consistent
with AMB96 definition, and as “old” if not. Same for soft site (ENEL accelerograph site), but only
constant-hazard spectra are shown
The error caused by using inconsistent distance measures can be sizable. Figure 9
illustrates, together with the deterministic spectra, the constant-hazard spectra com-
puted for two representative sites in Catania (on rock and on soft ground) using:
(a) the distance consistent with the AMB96 definition and the CRISIS99 require-
ment (labelled as “new” in the figure), and (b) placing the SSZs at zero depth, so
that the distances computed by CRISIS would be the same as those from the surface
projection of the ruptured fault (labelled as “old” in the figure). Note that the remark-
able difference both in the spectral level and in shape.
No comments are given here to the data pre-processing of earthquake catalogues
for identifying the completeness time intervals for different magnitude classes, but a
strong disparity was found in the catalogues used for the different cities; the ensuing
problems will require in the medium term a solution that can only stem from a unified
European catalogue of damaging earthquakes.
Concerning regional models of SSZs, the EU countries most actively participating
in the development of Eurocode 8—notably France, Greece, Italy and Spain—have
elaborated SSZ models reflecting the state of the art in knowledge of the active tec-
tonic lineaments and seismicity models (García Mayordomo et al. 2004). For the other
countries involved in Risk-UE, such tools were not available in the same terms, and
regional maps of SSZs affecting the relevant cities were constructed ad hoc, mostly
based on the distribution of historical and present seismicity.
The simplified assumptions of a homogeneous behaviour and a random earth-
quake distribution in space and time within a single source zone suggested a separate
treatment of active fault segments located within a SSZ; such segments could gen-
erate some of the strongest events affecting the city, especially if short distances are
involved, as for Catania or Thessaloniki.
The spacing of the grid of points where seismic hazard is calculated obviously con-
trols the accuracy and resolution of the results and the computational time. Systematic
tests were carried out for most cities; for Barcelona, in particular, it was found that
Bull Earthquake Eng (2006) 4:341–364 359
good accuracy and stability in the computed amax values could be achieved with a
receiver grid spacing of 0.1◦ with a medium computational effort. The grid spacings
actually used ranged between less than 0.01◦ and 0.1◦ (about 1–10 Km), depending on
the extension of the city and other factors.
It was already mentioned that for some RISK-UE cities the earthquake sources of
the deterministic scenarios differed from those of the probabilistic analysis; typically
they were faults in the first case, and extended SSZs in the second. Thus, significantly
different spectra could result in the two cases, see e.g. Fig. 9. A more complete com-
parison is provided in Fig. 10, which includes a hard ground site for each city (except
Bucharest, which does not have one).
Note that in Fig. 10 the strongest difference between deterministic and constant-
hazard spectra occurs for Catania and Thessaloniki, and also that such difference has
opposite signs in the two cases. This depends mostly on the distance from the city at
which the governing fault is located in the deterministic scenario. For the other cities
the difference between the two scenario spectra is limited. Also worth noting is that,
if the deterministic spectrum at a site is compared with the constant-hazard ones for
Fig. 10 Comparison of deterministic (top) and constant-hazard (bottom) acceleration response spec-
tra at a reference rock site for the different cities (excluding Bucharest)
360 Bull Earthquake Eng (2006) 4:341–364
different return periods, the former tends to match the latter at long vibration peri-
ods and at the longer return periods (e.g. 1,000 years), as a result of the statistically
stronger dependence on magnitude of the long period spectral ordinates.
Due to scarcity of data in the calibration set, and the complexity of site effects, the
evaluations of seismic response yielded by the site-dependent AMB96 attenuation
relation were judged to be too crude for zones on soft soil deposits or artificial fills.
Hence, additional verifications were required on such zones, complying with the fol-
lowing criteria:
– Sufficient 1D response analyses had to be performed on soft deposits, using avail-
able ground profiles down to depths including either local bedrock or materials
with VS ≥ 700 m/s; the deposits referred to would typically include ground classes
D, E, F of Eurocode 8.
– The 1D analyses were to use as input 5 real accelerograms recorded on hard
ground, in tectonic environments and at distances comparable to those pertinent
for the city at study. In addition, for such input waveforms:
– Amplitude scaling factors were allowed, within ±1σ of the attenuation relation;
– The average response spectrum of scaled waveforms should indicatively dif-
fer by not more than 20% from the scenario spectrum on hard ground at the
prescribed vibration periods;
– In the 0.3–1 s period range, the H/V ratios of input response spectra should be
approximately constant;
– The spatially averaged surface spectra computed from 1D analyses were to provide
the basis for appropriate correction factors (if any) to be applied to the original
scenario spectra.
– Alternatively, if the seismic zonation of a city was directly based on the simula-
tion of local effects (as for Barcelona), and if the zones with similar behaviour
were characterised by a transfer function and by an amplification factor for amax ,
the previous validation procedure could involve the confirmation of the spectral
amplification factors.
While the tool more readily available for the 1D calculations was the linear equivalent
EERA code (Bardet et al. 2000), other codes were used such as CYBERQUAKE
(http://www.brgm.fr/logiciels.htm) for Nice.
The supplementary 1D analyses seemed especially appropriate for Thessaloniki,
due to steeply irregular bedrock morphology, Catania (complex geological setting),
Sofia (presence of deep Neogene and Quaternary deposits), and Nice. For Bitola
and Bucharest, instead, such analyses were not possible either due to the scarcity of
available geotechnical borings, or the very large thickness of the sediments. Where a
comprehensive 3D subsoil model was available, as in the case of Nice and Thessalo-
niki, the validation could have been performed at any site; although sophisticated 3D
analyses were carried out for Nice, their results could not be used within Risk-UE.
Where the geotechnical zonation was supported by advanced data (depth contour
lines of bedrock, thickness of each soil formation, and dense 2D cross-sections), the
grid resolution of 1D analyses could be increased in the most interesting zones: this
Bull Earthquake Eng (2006) 4:341–364 361
Fig. 11 Comparison between constant-hazard and 1D analyses results at site of the historical centre
of Thessaloniki
was done e.g. for Thessaloniki, where a total of 301 1D calculation sites were used,
with spacing of 500 and 250 m, depending on the zone.
The selection of a sufficient number of input waveforms satisfying the stated
requirements proved difficult, both for the deterministic and the probabilistic analyses.
For Catania, for instance, the consistency with the target rock spectra was achieved,
but not the constancy of H/V spectral ratios. Exceptions had to be allowed where a
few regionally recorded strong motion records were available, as for Bucharest.
For each 1D calculation site a 5% damped average spectrum (mainly of accel-
eration) was computed from the output spectra obtained from the propagation of
the selected inputs. Then, a single spectrum was obtained by spatially averaging the
spectra at the grid points of a given zone. The spatially averaged spectrum from the
1D analyses was then compared with the deterministic and the probabilistic scenario
spectra, to check whether any modifications to the latter were necessary.
1D analyses tended to provide on soft soil larger responses than those estimated by
the attenuation relationship, with some exceptions. Quite understandably, the ampli-
fication depended on the extent at which the frequency content of the input motion
matched the 1D fundamental frequency of the soil profiles. More complex situations
also emerged, typically on shallow surface deposits, showing amplification or deam-
plification, depending on the period range, as e.g. in Catania.
For Thessaloniki the predictions could be compared with the (single) instrumental
record of the damaging 1978 earthquake, as illustrated in Fig. 11. The 1D results are
in good agreement with the constant-hazard spectrum at periods >0.5 s, while, at
smaller periods, the probabilistic ordinates considerably exceed the 1D values. The
discrepancy has very likely to do with the influence of soil non-linearity affecting the
1D calculations in the considered range of periods. Also note that the spectrum of the
deterministic scenario matches reasonably well the spectra of the 1978 earthquake.
362 Bull Earthquake Eng (2006) 4:341–364
Only in few regions of Europe did historical earthquakes produce strong impact in
terms of induced permanent deformations, such as ground settlement and rupture
due to liquefaction and lateral soil spreading, or even rock- and land-slides; when the
impact was massive, as in the destructive Calabria (Southern Italy) earthquakes of
1783, the affected zones were for the most part sparsely populated. In most cases, the
paucity of ground deformation phenomena is related to the limited magnitude of the
events. The same can be said for co-seismic surface fault rupture. Indeed, also for
the Risk-UE cities, observed cases of induced hazard effects were very few. Hence,
earthquake induced permanent ground movements were not dealt with extensively in
Risk-UE; only some general guidelines and criteria were provided, aimed at balancing
the effective demand of protection against induced hazard effects and the resources
needed for carrying out the analysis. The low emphasis placed on this topic was also
motivated by the fact that in many cities the data required for proper analysis (digital
thematic maps and databases, DEMs) are lacking.
The following general criteria were formulated, mainly with reference to the Risk-
UE task dealing with vulnerability assessment of lifelines:
– Since few places are likely to suffer soil failure caused by lateral spreading, the
weight of the analysis was shifted to a realistic case-by-case assessment of local
situations;
– As a general criterion, earthquake induced landslides deserved attention from
a scenario viewpoint provided: (i) there was evidence of seismically triggered
landslides within the city area in past earthquakes, (ii) slopes are predomi-
nantly subject to frictional types of failures and, (iii) a DEM of the terrain is
available.
Bull Earthquake Eng (2006) 4:341–364 363
8 Conclusions
The main conclusions of the effort deployed in Risk-UE in the realm of seismic hazard
assessment in urban areas can be outlined as follows:
References