Development of a Ground Motion Model (GMM) for Western China Based on Simulation and Observation Data

Although dense strong motion observation arrays have been established nationwide in China, data from large earthquakes, particularly those at near-fault distances, remain limited, hindering the development of reliable ground-motion model (GMM). Over the past several years, we have employed a simulation scheme that utilizes stress drop, quality factor, and site transfer function, inverted from historical strong motion recordings by using the generalized inversion technique, as input parameter for the stochastic finite-fault method. Comparisons of simulated pseudo-spectral accelerations (PSAs) with observations from several historical earthquakes have demonstrated that this simulation scheme can produce reliable PSA at frequencies above 0.1 Hz. Thus, we aimed to develop a GMM specially for western China by integrating both simulated data from tens of historical earthquakes and observations. The resulting GMM (GMM1) was compared to two other GMMs: one developed solely from observation data (GMM2), and another incorporated near-fault distance data collected from the NGA-West2 dataset in addition to observation data (GMM3). The result shows that the median values of GMM1 are closely similar to those of GMM2 within a period range of 1 to 10 s. At periods below 1 s, the median values of the two GMMs are comparable only at distances greater than 100 km, whereas the median values of GMM1 and GMM3 are comparable. It is challenging to solidly judge which GMM is more reliable at this stage due to the lack of near-fault recordings of large earthquakes in the studied area. Nevertheless, unlike the published stochastic-based GMMs, which have significantly smaller standard deviations (SD) around 0.2, the total SD of GMM1 is closely match that of GMM2. However, although the SD curve shows similar shapes, the contribution of within-event and inter-event residuals to the total SD differ between GMM1 and GMM2. Although incorporating more comprehensive source models, lateral heterogenous path attenuation effects, and nonlinear site effects could potentially enhance the reliability of the simulation data, these findings indicate that the method for developing GMM based on simulation and observation data is promising.