关键词: 正则化, L曲线, 联合反演, 模拟退火, 大地电磁, 地震

Abstract: All the geophysical inversion problems are ill-posed.The application of regularization theory and joint inversion provide us ideas to solve the ill-posed inverse geophysical problems.Regularization theory has been widely used in single geophysical inversion problem which has achieved certain effects;however,what geophysicists concerned in the previous studies on joint inversion just was data misfit with little consideration about regularization theory.In this paper,based on the joint inversion method with random resistivity and velocity distributions proposed by YU Peng,the conception of regularization is introduced into synchronous joint inversion of magnetotelluric(MT) and seismic data.We use priori information to improve the stability of solution,and set stabilizer as the smallest model.How to meet the best trade-off between data misfit and model constraint,and what's intellectual choice for the regularization factor have always been the hotspots in the geophysical inversion problem.We use L-curve method to determine the quasi-optimal value of the regularization parameter,and apply it into joint inversion.To avoid immerging in local minimum of linear inversion method,nonlinear simulated annealing method is used to realize joint inversion.The forward modeling of MT data is obtained by using finite element method and the seismic travel-time computing is obtained by using improved ray-tracing method.Testing the regularized and conventional synchronous joint inversion on synthetic data sets shows that regularized joint inversion with effective model constraint has powerful availability to improve stability of solution and efficiency of computation,and to help us to obtain a stable solution which is close to the fact.

Key words: regularization, L-curve, joint inversion, simulated annealing, MT, earthquake