ALBERT

Description: [1]  We develop a three-step Maximum-A-Posteriori probability (MAP) method for coseismic rupture inversion, which aims at maximizing the a posterior probability density function (PDF) of elastic deformation solutions of earthquake rupture. The method originates from the Fully Bayesian Inversion (FBI) and Mixed linear-nonlinear Bayesian inversion (MBI) methods, shares the same posterior PDF with them, while overcoming difficulties with convergence when large numbers of low-quality data are used and greatly improving the convergence rate using optimization procedures. A highly-efficient global optimization algorithm, Adaptive Simulated Annealing (ASA), is used to search for the maximum of a posterior PDF (" mode " in statistics) in the first step. The second step inversion approaches the " true" solution further using the Monte Carlo Inversion (MCI) technique with positivity constraints, with all parameters obtained from step one as the initial solution. Then slip artifacts are eliminated from slip models in the third step using the same procedure of the second step, with fixed fault geometry parameters. [2]  We first design a fault model with 45°-dip angle and oblique slip, and produce corresponding synthetic InSAR datasets to validate the reliability and efficiency of the new method. We then apply this method to InSAR data inversion for the coseismic slip-distribution of the April 14, 2010 Mw 6.9 Yushu, China earthquake. Our preferred slip model is composed of three segments with most of the slip occurring within 15 km depth and the maximum slip reaches 1.38 m at the surface. The seismic moment released is estimated to be 2.32e + 19 Nm, consistent with the seismic estimate of 2.50e + 19 Nm.