Magnetic Inversion of Adaptive Regularized Parameter Selection

Chongjin Zhao,Peng Yu,Haohao Lan,Yang Xiang

DOI: https://doi.org/10.1190/gem2015-024

2015-01-01

Abstract:Summary How to select a reasonable regularization parameter is a core problem in regularization inversion. Potential field inversion is difficult to conduct if there is no a priori information for selecting regularization design. Yet there are no clear and suitable methods for selecting the regularization parameter in regularized potential field inversions. In this paper, we compare several methods for this purpose. Based on the improved adaptive regularized parameter selection method in magnetotelluric inversion, we proposed an improved automatic selection method of adaptive regularized parameter for magnetic inversion. Firstly, we get an initial value of the regularization parameter after running Zhdanov's adaptive regularization inversion method. Then, we determine the new regularization parameter based on the initial value and the decreasing rate of data misfit, so we can automatically conduct 2nd round of adaptive inversion and obtain a good data fitting and stable inversion result by the adaptive decay of the new regularized parameter with a big attenuation factor. The results of model comparisons show that this improved automatic selection method is effective especially for potential field inversion.