SEISMIC MODELLING OF FAULT RUPTURE FOR FOUR IRANIAN EARTHQUAKES

Abstract

Ground motion records from the epicenteral region can be of great help in understanding the earthquake process as effects of transmission path are minimal find rupture process on the causative fault can be modelled. In the present work strong motion records of four Iranian earthquakes, the 1978 Tabas earthquake, the 1990 Rudbar earthquake, the 1994 Zanjiran earthquake and the 1997 Ardebii earthquake have been analyzed using the data generated by the Strong Motion Instruments Array installed by Building and Housing Research Center (BHRC), Iran. The observed records are compared parametrically with synthetic strong motion records generated for various models to arrive at a preferred rupture model for each earthquake. For each earthquake the most probable causative fault is identified on the combined basis of tectonic map of the region, location of aftershocks, fault plane solutions, geological cross section, isoseismal map, isoacceleration map and field observations. Estimation of peak acceleration from a preliminary simulated record is on the basis modelling parameters of rupture plane instead of empirical relations for peak acceleration. These modelling parameters are length (L), dip (5 ), strike (§) of causative fault, downward extension of rupture plane (D), number of elements (N), length of element (Le), velocity of S waves in Ihe medium (V), rupture velocity (Vr) and starting point of rupture or nucleation point. The preliminary simulated record is refined to have basic spectral shape,© model of Brune's spectrum, by using envelope waveform. The efficiency of the new hybrid method has been tested for four earthquakes in Iran. Fourteen features have been extracted from observed and simulated records in time and frequency domain. These are Pa, TD, Tat, Trt, Tdt, Rat, Tarea, Fp, F,, F2, F3, F4, F5 and F6. For an objective choice from amongst different rupture models, two criteria are used. First, it is considered which model gives maximum number of extracted features showing less than 20% difference between observed and simulated records. Secondly, which model gives maximum number of stations for which root mean square error (rmse) is minimum between observed and simulated response spectra. The efficiency of the developed method has been tested for these four earthquake in Iran. A good comparison is observed for extracted features Pa, TD and Rat compared to other extracted features. The study indicates that peak acceleration can be estimated reliably by this method within 20% error. This is especially significant for stations close to the causative fault. The closest station to the causative fault for the Tabas, Rudbar, Zanjiran and Ardebli earthquakes are Tabas (Rs =3 km), Abbar (Rs = 8 km), Zanjiran (Rh = 9 km) and Ardebil (Rh = 29.8 km), respectively. The simulation method gives synthetic records which show (i) good comparison between observed and simulated extracted features both in time as well as frequency domain for Tabas and Abbar stations and in time domain for Zanjiran and Ardebil stations and (ii) good match is observed between observed and simulated response spectra for stations close to causative fault for all earthquakes. It is observed that none of the attenuation relations, which used hypocenteral distance and epicenteral distance, could predict peak acceleration at Tabas and Abbar stations for Tabas and Rudbar earthquake, respectively. It is observed that the observed peak acceleration is more at Tabas although it is at a greater hypocenteral distance as compared to Deyhook. This feature is well replicated in the records generated by simulation method. In empirical relations the peak acceleration decreases with increasing distance. This suggests that for large earthquakes relations using hypocenteral and epicenteral distances are not suitable for stations close to causative fault. Another aspect of simulation method is that a decision can be made about the fault plane and extend of the most probable causative. In the case of Tabas and Rudbar earthquakes, surface ruptures have been observed and the causative faults well

identified. Surface rupture was not observed for Zanjiran earthquake. In the case of Ardebil earthquake opinions differ about the causative fault. Simulation and comparison of strong ground motion for different dip, strike and nucleation points indicate most probable causative faults for Zanjiran and Ardebli earthquakes are in approximately north- south direction. As a test case, a hypothetical earthquake of magnitude Ms 7.3 is assumed to take place on Lahbari fault in the vicinity of Karkheh dam in Iran. Effect of this earthquake at the dam site has been estimated using the modelling method developed in the present study. Five records are simulated at Karkheh dam site for the five nucleation points within the rupture plane. Peak acceleration obtained for the five simulated records at Karkheh varies between 361 to 482 cm/sec . This suggests that the peak acceleration at Karkheh will be at least 361 cm/sec and can go up to 482 cm/sec depending on the position of nucleation point and the direction of propagation within the rupture plane, for a Ms 7.3 earthquake on Lahbari fault.