Constraining Dynamic Rupture Properties of the 2015 Mw 7.8 Gorkha Earthquake by Integrated Inversion and Modeling Techniques

Near field recordings of the 2015 Mw 7.8 Gorkha earthquake enable us to probe the rupture process of a big inland thrust earthquake. Linear multi-time-window and nonlinear Bayesian inversion of geodetic and seismic data resolve a depth dependent rupture process with higher slip roughness, higher rupture velocity, longer rise time and higher complexity of subfault source time functions (SSTF) in the deeper portions of the rupture. Here we use strong motions recorded on top of the rupture area to develop constraints on dynamic rupture parameters. We developed a deconvolution technique to estimate average properties of slip rate functions. After interpolating the inferred coseismic slip pattern into fine scaled subfaults (1 km2), adding stochastic slip roughness and assuming all SSTFs are delta functions, we compute a set of reference strong motion seismograms. For given slip and rupture speed (Vr), if the SSTF has the same temporal shape over the whole rupture, it can be estimated by deconvolving observed seismograms by reference ones. If these assumptions are only satisfied approximately, deconvolution gives a spatial average of the SSTF near the stations weighted by certain sensitivity kernels. Applying this deconvolution technique we find that the average SSTF resembles a Yoffe function with rise time Tr 6 s and acceleration time (time to reach peak-slip-velocity) Tacc 1 s. Assuming a Yoffe function SSTF, we calculate sensitivity kernels relating the waveform correlation coefficient to small perturbations of the rupture parameters Tr, Tacc and Vr of each subfault. For a single station, the sensitivity is larger in an elliptical area of the fault whose long axis connects the hypocenter and the station. We then use dynamic rupture modeling to constrain fault friction properties. We constrain the initial stresses by the stress drop computed from the final slip obtained by finite fault inversion, and add random stress perturbations to enrich the high frequency content of the source. We constrain the slip weakening distance and strength excess using their theoretical relations to Tr and Tacc. Dynamic rupture is then simulated with the spectral element software SPECFEM3D. By fitting the spectra of the observed seismograms, we develop estimates of dynamic parameters within a portion of the fault with adequate sensitivity.