Joint inversion for magnetotelluric data based on cuckoo search algorithm and least squares method

Abstract

The main drawback of using traditional linear or quasi-linear methods for magnetotelluric (MT) data inversion is its strong dependence on the initial model, which leads to its tendency to fall into a local optimum. However, the lack of sufficient a priori information makes it difficult to obtain suitable initial models in most geophysical inversions. To solve this problem, a joint approach (CS-MLS) based on the cuckoo search (CS) algorithm and the moving least squares (MLS) method is proposed to perform MT data inversion. The CS algorithm has a good global search performance without being sensitive to the initial model, which is used for the initial search of the stratum model parameter space. The search results are then used as the initial model to explore the quasi-linear inversion based on MLS. The synthetic and field data experiments show that the reliable initial model provided by the CS algorithm has greatly refined the results of the MLS inversion. The joint CS-MLS algorithm has great inversion performance, with significant improvement in inversion efficiency and accuracy, and is especially suitable for MT data inversion without initial information.