A synthetic seismicity model for the Xianshuihe fault, southwestern China: simulation using a rate- and state-dependent friction law

We performed a numerical simulation of seismic activity along the Xianshuihe fault, a highly active strike-slip fault approximately 350 km long, located along the eastern margin area of the Tibetan plateau, southwestern China. Historical earthquake data over the last 300 yr indicate repeated periods of seismic activity, and migration of large earthquakes along the fault during active seismic periods. To understand the characteristics of historical seismicity and to obtain insight into seismic potential, we performed a numerical simulation of slip behaviour along the fault using a 2-D elastic plate model. The friction on the model fault obeys a laboratory-derived rate- and state-dependent friction law, while the long-term slip rates along the fault are assumed to be consistent with geologically and geodetically estimated slip rates. To simulate segmented rupture behaviour we introduced non-uniformity into friction parameters along the fault. The simulation results indicate that seismic rupture is arrested in regions with frictional properties that are highly velocity-strengthening or large values of characteristic slip distance over which frictional stress evolves. In the regions where seismic rupture is arrested, post-seismic sliding occurs, causing time-dependent stress transfer along the fault. Post-seismic slip histories in highly velocity-strengthening regions are well approximated by logarithmic time functions, while those in regions with large characteristic slip distances increase more rapidly in the initial stages of post-seismic sliding. This difference in post-seismic sliding produces the difference in time-dependent stress transfer and, in consequence, the statistical characteristics of earthquake recurrence. Fault interactions near a fault branch point produce complexity in the simulated earthquake sequence. Simulated rupture commonly triggers additional rupture in a neighbouring segment with a time delay of several years. The simulated seismic activity shows good agreement with observational data with respect to the following statistical features: (1) successive seismically active periods are separated by periods of relative quiescence of 100-200 yr duration (2) and the duration of active period is a few tens of years. Strictly speaking, neither the time- nor slip-predictable models are able to describe the simulated local slip history for each fault segment. The ratio of the standard deviation to the average value of the recurrence intervals of simulated earthquakes ranges from 0.05 to 0.22, depending on the fault segment. The cumulative distribution function of the recurrence intervals of simulated earthquakes at each segment approximately obeys a Brownian passage time distribution or a lognormal distribution, which is commonly used in the statistical evaluation of earthquake occurrence.