Regularization analysis of three‐dimensional magnetotelluric inversion

Abstract

Inversion of MT data is an inherently nonunique and unstable problem due to the ill‐posedness of the electromagnetic inverse problem. A variety of models may fit the data very well. To overcome this ill‐posed nature of the inverse problem, we use Tikhonov's regularization in which the ill‐posed problem is replaced by a family of well‐posed problems. We also analyze the behavior of the Tikhonov regularization parameter to find out its optimal value for a typical model of a hydrocarbon reservoir in a marine environment. We have compared two regularization techniques: rigorous and adaptive regularizations. The results of this numerical study demonstrate that adaptive regularization provides practically the same inverse image as the rigorous regularization, while reducing the computational time dramatically.