Two Dimensional Joint Inversion Of MT And Seismic Data

We have developed a2-D synchronous joint inversion algorithm ofmagnetotelluric (MT) and seismic data on the base of cross-gradient theory proposedby Gallardo and Meju. Using the central differences to discrete the cross-gradientequation makes the constraints on the model’s boundary more reliable and improvesthe exactness of inverse results. For MT forward, the Maxwell’s equations areapproximated by finite-difference method. For MT inversion, the Gauss-Newton(GN) method of inversion and regularization are used to control the ill-posedproblem. For seismic forward, the2-D visco-acoustic wave equation in thefrequency domain is discretized in a finite-difference sense using the so-called9-point mixed-grid approach. For seismic inversion, we use the full-waveforminversion. Furthermore, the gradient method of inversion is used and the diagonalelements of the approximate Hessian are used to provide a preconditioner of thegradient that properly scales the perturbation model and make the convergence faster.In the seismic iterative process, the simultaneous inversion of multi-frequencies isused with multi-frequency data weighting which is designed to balance contributionsfrom the different frequency measurements and to prevent high-frequencycomponents from dominating the inversion.Theoretical model data are tested by our inverse algorithm described above.The test results demonstrate that our inversion algorithm with cross-gradientconstraints can cause the velocity and resistivity to constrain each other so that theresults of inversion are more reliable. It’s also proved that the joint inversion is betterthan traditional separate inversion. Therefore, the2-D synchronous joint inversionalgorithm of magnetotelluric and seismic data is verified to be validity.