Sensitivity of Ground Motion Simulation Results to Input Parameters Using Stochastic Finite Fault Method: A Case Study for Niavaran Fault, Tehran

In strong ground motion simulation methods, results are strongly dependent on the accuracy of input parameters; so that the resultant peak parameters, frequency content and the duration of motion could be easily affected. In the current study, for investigating this matter further, the extent in which each parameter influences the results, has been examined. For this purpose, a reference simulation scenario considering Niavaran fault has been assumed first. Then, by increasing and decreasing input parameters, 19 other scenarios have been considered and the results have been explored. 

Stochastic finite fault method [1] was used for ground motion simulation. In this method, which has its roots in stochastic point source method [2], the fault plane is divided into sub-faults, each of which acting as a point source. Resultant recordings from each sub-fault are added in the observation point considering their time delays so the main recording is obtained. Stochastic methods benefit from using the Fourier amplitude of motion that includes the effects of source, path and site. Therefore, in this study, by changing each of these parameters and parameters relating to the geometry of the fault, accurate selection of these parameters are illustrated. Niavaran fault with length and width of 44 and 20 km respectively, has the capacity of producing an earthquake with magnitude 7. By changing the dip and rupture starting point as fault geometrical parameters, moment magnitude, rupture velocity, stress drop and pulsing area as source parameters, quality factor and geometrical spreading as path parameters and kappa as site parameter, we investigated the effect of uncertainty of these parameters on the results.

At first, the simulation was done for the reference scenario with magnitude 7 and on engineering bedrock (Vs(30)=620 m/s) for distances up to 60 km from the fault plane. It should be noted that rupture starting point was decided based on the study of Chiou and Youngs [3] and other parameters are chosen based on the study of Motazedian [4]. For the site amplification Boor and Joyner [5] curves were used. For this scenario, at closest point with 3 km distance from the fault plane, PGA reaches 600 cm/c/c and PGV to about 60 cm/s. Most motions are observed in near field points, especially those above the rupture plane in the area of 1300 km2 which an average have acceleration of 330 cm/s/s. The results for PGA and PGV are compared with four ground motion prediction equations and seem to be satisfactory. Considering that the target fault is within the area of Tehran, this scenario can be a real threat to the city. After comparing the testing scenario with the reference one, the effects of uncertainty could be stated as follows: Parameters that have the most significant effects on the motions are the parameters with greater moment magnitude, greater stress drop, greater b coefficient of geometrical spreading and smaller kappa. Therefore these parameters could be called “worst case scenario” since these scenarios besides having larger PGA and PGV, have a bigger area affected with peak parameters. Effects of uncertainty of parameters have also been investigated in frequency domain. Motions with low frequency are usually controlled by rupture starting point and moment magnitude parameters, which have most effects on larger structures. Higher frequencies which are more important for smaller structures are controlled by stress drop and quality factor parameters. The most important effects of tested scenarios compared to the reference one is observed in distances under 30 km, therefore it is necessary for the near field hazard analysis that parameters with highest accuracy are considered.