Ground motion modelling in Tehran based on the stochastic method

Abstract

This paper presents the development of a seismological model for the Tehran area. This modelling approach, which was originally developed in Eastern North America, has been used successfully in other parts of the world including Australia and China for simulating accelerograms and elastic response spectra. Parameters required for input into the model were inferred from seismological and geological information obtained locally. The attenuation properties of the earth crust were derived from the analysis of accelerogram records that had been collated from within the region in a previous study. In modelling local modifications of seismic waves in the upper crust, shear-wave velocity profiles have been constructed in accordance with the power law. Information inferred from micro-zonation studies (for near-surface conditions) and from measurements of teleseismic P-waves reflected from the deeper crusts as reported in the literature has been used to constrain parameters in the power-law relationships. This method of obtaining amplification factors for the upper crust distinguishes this study from earlier studies in the Tehran area (in which site amplification factors were inferred from the H/V ratio of the recorded ground motions). The regional specific seismological model so constructed from the study enabled accelerograms to be simulated and elastic response spectra calculated for a series of magnitude–distance combinations. Importantly, elastic response spectra calculated from the simulated accelerograms have been compared with those calculated from accelerograms recorded from earthquakes with magnitudes ranging between M6.3 and M7.4. The peak ground velocity values calculated from the simulated accelerograms have also been correlated with values inferred from macro-seismic intensity data of 17 historical earthquakes with magnitudes varying between 5.4 and 7.7 and with distances varying between 40 and 230 km. This paper forms part of the long-term strategy of the authors of applying modern techniques for modelling the attenuation behaviour of earthquakes in countries which are lacking in instrumental data of earthquakes.

Introduction

Probabilistic seismic hazard analysis (PSHA) requires a well-established ground motion–attenuation relationship which can be used for predicting ground motion parameters as function of magnitude, distance, site classification (i.e. rock, stiff soil, soft soil, etc.), and styles of faulting (which has been incorporated into some but not all of the attenuation models). Different relationships have been developed for different tectonic regimes. Strictly speaking, more than one relationship is required within a tectonic regime if there are significant variations in the geological formations within the regime. However, developing an attenuation relationship with the conventional empirical modelling approach which is based on the use of instrumental strong motion data presents both technical and financial challenges given that the modelling relies mainly on recordings taken within kilometers and tens of kilometers from the epicenter of an earthquake. Consequently, empirical attenuation models that have been developed are based broadly on a tectonic regime even in well studied, and instrumented, regions like California and Japan. Thus, local variations within the regime are typically not parameterized.

The objective of this study was to develop a response spectrum–attenuation model that is representative of the area surrounding Tehran in Northern Iran. The Tehran region was chosen for the study in view of its high level of seismic activities and the recent experience of destructive earthquakes. The occurrence of more than 10 earthquakes with magnitude around 5 during the 20th century in the Tehran region indicates the intense activity of the faults (refer Section 2). The adopted modelling methodology should be suited to regions which are without adequate strong motion data for undertaking empirical modelling.

Countries which do not have the resources for capturing a significant volume of strong motion data often fall back on adopting the less reliable approach of inferring ground motion parameters such as peak ground velocities or peak ground accelerations from macro-seismic (Intensity) data taken from the study area. Note, Intensity data is better correlated with peak ground velocity data in conditions where there was damage to the infrastructure (i.e. Intensity >VI). With lower intensity values which represent the “felt” effects of the earthquake, Intensity data is better correlated with peak ground acceleration data than peak ground velocity data. One or more of such parameters so obtained from the modelling could then be used for scaling a pre-determined (imported) response spectrum model. The main shortcoming of an attenuation model which is primarily derived from Intensity data is uncertainties with the spectral contents of the ground motion given that no relevant instrumental data collected from within the study area have been incorporated into the modelling. Furthermore, the magnitude range of the attenuation relationship is limited to the distribution of magnitudes in the historical events database. Thus, the methodology is not entirely satisfactory.

Another approach of attenuation modelling is by means of stochastic simulations of the seismological model which is characterized by the separation of the ground motion model into the “source”, “path” and “local” components. Each of these components has their seismological parameters calibrated in accordance with instrumental data. Thus, the stochastic modelling approach in the original form as developed in Central and Eastern North America (CENA) is based primarily on instrumental data collated from within the region of interests. In contrasts, the macro-seismic data modelling approach is based on the use of a pre-determined (imported) spectral model and is without incorporating instrumental data at all into the modelling.

There have been attempts to undertake stochastic modelling approach in low and moderate seismicity regions like Australia and South China as was done originally in CENA but the lack of seismic data means that a seismological model similar to what have been developed in CENA could not be constructed in their entirety using solely regional/local data. The authors and collaborators have been employing an approach wherein the source component of the seismological model is assumed to be generic in nature and hence the source factor of the model is no different to that developed originally for CENA. The basic construct of the seismological model and details of the source model that has been used in the simulations are described in the later sections of the paper. The path and local components required to complete the seismological model could then be inferred from instrumental data collated from the region of interests [1], [2], [3], [4], [5], [6]. In this study, the “path” components are representative of geometrical and whole path anelastic attenuation, and are based on conditions in the area surrounding Tehran which spans a distance of about 10° from east to west as shown in Fig. 1. The derivation of these path component factors are described in Section 4. The “local” components are representative of conditions in the upper 4 km or so of the rock crust (and are not to be confused with the “site” components which deal with the effects of the soil sediments overlying bedrock at the site). Local effects that are being modelled are the upper-crustal amplification effects (which are dependent on the shear-wave velocity profile of the rock crusts [8]) and the upper-crustal attenuation effects (which can be characterized by the kappa parameter) [7]. The derivation of the local component factors are described in a later section of the paper. All filter functions representing path and local effects specific to the study area were then combined with the generic source model for constructing the seismological model in the Fourier spectrum format. Synthetic accelerograms were then generated by stochastic simulations based on random phase angles for a set of magnitude–distance combinations. A response spectrum–attenuation relationship could then be constructed from the calculated elastic response spectra of the simulated accelerograms. Finally, the simulated accelerograms were evaluated by comparing its associated response spectra with those calculated from (real) accelerograms recorded from recent events occurring within the region and by correlating the calculated PGV values with those inferred from historical Intensity data.