STRONG EARTHQUAKE SIMULATION WITH EMPIRICAL GREEN'S FUNCTIONS; FIRST ATTEMPTS IN THE FRAMEWORK OF THE EUROPEAN PROJECT SERGISAI*.

Ph. Rosset, J.-J. Wagner, CERG-UNIGE, Geneva, Switzerland

M. Garcia-Fernandez, M. J. Jimenez, ICT-CSIC, Barcelona, Spain

Abstract

The method of empirical Green's functions (EGF) was applied to simulate realistic signals of expected earthquakes in the Barcelona region (Spain) for the hazard assessment in the framework of the European project SERGISAI*. The methodology can provide such signals when the seismic catalogue is well enough and the parameter uncertainties (e.g. fault related values, etc ..) are taken into account. First results of magnitude 6 earthquakes simulation in the Barcelona city are presented.

Introduction

The project SERGISAI* aims to develop a tool for seismic risk assessment at different scales (i.e regional, sub-regional and local). An estimate of ground-motion due to an anticipated large earthquake is a basic element of information required to produce a risk zonation map. The general lack of strong ground-motion data in Europe has led to the development of numerical simulations for predicting time-histories of hypothetical but realistic earthquakes. After a detailed compilation of the existing approaches, the Empirical Green's Functions (EGF) technique has been selected for its a priori advantages. It uses actual recordings of small earthquakes, typically with magnitude from 2 to 3, to calculate the effect of much larger ones ranging magnitude between 5 to 7.

Applications of this approach have been performed on Barcelona which is the local scale test area of the project.

Methodology

The model developed is based on one proposed by Hutchings (1988, 1991, 1994). It uses a simple kinematic rupture model to describe the source and the EGF to constrain propagation path and site response information.

A Kostrov rupture model (1964), approximated as a ramp shape, is used to calculate the slip at a point on the fault surface. The rupture velocity V_R is either considered constant or variable as 0.8 to 0.9 time the S-wave velocity-depth profile. The position of the hypocenter of the modeled earthquake within the fault plane is determined by its distances to the edges of the fault. The position where initiates the rupture and the focal mechanisms of EGFs are calibrated by comparing observed and equivalent calculated seismograms for a given seismic stations. The general relation of Hanks and Kanamori (1979) has been chosen in a first stage to convert the magnitude into seismic moment.

Application at the local scale OF Barcelona

The two principal seismic areas highlighted by the historical and recent seismicity are the south-eastern part of the Pyrenées and the Catalan coastal range (Olivera et al., 1996). Around Barcelona, the earthquakes catalogue contains only events which occurred after 1985 (Servei Geologic de Catalunya, SGC) recorded at the three stations POB, VAN and EBR. For reasons of seismic sources proximity, three swarms of events were selected which occurred in the Catalan coastal ranges; namely, December 1991, August 1987 and May 1995. The location of these events, the position of the seismic stations and the major seismo-tectonic informations are indicated on the figure 1.

The Mediterranean continental platform where all these events are generated is formed by an NE-SW extensive group of faults which contribute to the basin formation during the Neogene period.

The actual catalogue of earthquakes and the seismotectonic environment lead to think that the magnitude of the maximum probable earthquakes is around M=6. Therefore this magnitude has been chosen to be modeled for seismic risk scenarios at Barcelona.

A single application is here illustrated with the simulation of the mentionned magnitude earthquake using two aftershocks of the May 1995 main event of M_L=4.2 located at 35 km SE of Taragonna. The two replicas were relocated in order to join the NNE fault direction of the main event considering the inaccuracy of the source determination in the off-shore region. The focal mechanisms of the cluster of earthquakes individualise a sinistral strike-slip fault in agreement with the micro-tectonic survey (Fleta and Escuer, 1991). The fault model is represented as a rectangular rupture surface with length L=2W, W being the width.

The calibration stage allows to choose a constant rupture velocity on the fault area around 2.9 km/s and a rupture initiated near an edge of the fault area. Its rupture surface was discretized into 0.01 km² elemental areas in order to obtain continuous modeled rupture for frequencies up to 25 Hz. Table 1 summarises informations on aftershocks used as EGF and on the simulated M=6 earthquake.

Figure 2 presents the Fourier and response spectra for one horizontal component of the different calculations. It shows the large variability of results in maximum acceleration (from 0.3 to 6% of g at 25 Hz) and maximum displacement (from 0.01 to 0.1 cm at 0.1 Hz). Indeed, it is to deal with uncertainties of parameters values that these calculations have been done with different fault surfaces and rupture velocities.

The results give a range of slip velocity from 2.0 to 7.0 m/s, faulting duration from 4.0 to 7.0 s and displacement from 16 to 44 cm.

The synthetic seismograms of the figure 3 are considered to be representative of the M=6 scenario and correpond to the Fourier spectrum closest to the mean spectrum (the thicker line in figure 2). It shows a peak acceleration near 3% of g (0.3 m/s²) and a duration greater than 12 s.

The global prediction of vertical response spectra in Europe given by Ambraseys and Simpson (1996) is then considered to compare the simulated results. Figure 4 shows the good agreement between the value of the calculated vertical peak ground acceleration (PGA) for the epicentral distance of 71 km and the M=6 predicted curve. A factor 2 between vertical and horizontal spectra values of calculated signals is observed, it corroborates the observations of Ambraseys and Simpson (1996).

It should be point out that the calculated vertical PGA value for the magnitude 5 scenario is 4 time lower than the value given by the predictive curve. This could be due to the sensitivity of the program to the seismic moment given for the small earthquakes in regards with the low magnitude calculations. Further investigation is needed for that use.

Although the parameter uncertainties and the influence of the model have been preliminary investigated, it should also be done with a larger range of calculations and parameters values.

CONCLUSION

The aim of the CERG-UNIGE participation (Rosset and Wagner, 1998) was to elaborate a strategy of evaluation of expected strong ground motion for hypothetical earthquakes in a given area. In this sense, the technique which uses empirical Green's functions has been chosen. The calculation program, from Hutchings (1994), has been modified in order to take into account the specificities of the European seismic context and its utilisation within the user's interface of the SERGISAI* environment.

Barcelona city, the test area at local scale, has been selected to check the program in elaborating calibration and scenarios for expected earthquakes of magnitude 5 and 6. The validation of the method was not simple to obtain the best results one could hope for, because only few records on only two seismic stations (more than 100 km distant from Barcelona) were useable in the earthquake catalogue of the region. Synthetic seismograms of the figure 3 were calculated from the three component recordings of two aftershocks at the POBlet site. They represent the simulated records of a M=6 earthquake, 71 km distant from Barcelona. One can point out that these have been used to perform evaluation of site effects, with the program SHAKE91 (1992) in different part of the city.

At this stage of the development, the methodology suffers a few limitations, in terms of the number of small earthquakes needed and the number of seismic stations where to calculate synthetic signals.

To better test the validity of the method, one should include the selection of regions with higher seismic activity and an appropraite seismic network. The addition of a procedure to generate synthetic Green's functions in order to complete the low frequency range (0 to 0.5 Hz) of the simulated earthquakes would also be necessary.

In the future, the development of the European seismic networks and its instrumentation argue the larger utilisation of the EGF approach in order to simulate M=6 and larger earthquakes which have a low probability of occurrence but are potentially dangerous.

AKNOWLEDGEMENTS

This research was in part supported by the EC Environment Research programme (contract Nr ENV4-CT96-0279, Climatology and Natural Hazards) and the Swiss part by the "Office Fédéral de l'Education et de la Science" of Switzerland (contract Nr OFES95.0452). The Lawrence Livermore National Laboratory (California, USA) has provided the Empsyn program. A special thanks to L. Hutchings and S. Jarpe for regular comments and discussions and to Dr. I. Hedley for his english reviewing.

The seismic data were provide by the Institut d'estudis Catalan and the Servei Geologic de Catalunya. The good achievement of the SERGISAI project was only able to the close collaboration between colleagues within the working group.

REFERENCES