Occam’s Inversion of Magnetic Resonance Sounding Data Incorporating Groundwater Directional Information

Abstract

Magnetic resonance sounding (MRS) is a promising geophysical method for the direct detection and quantification of groundwater. However, the two-dimensional MRS inversion is usually seriously ill-posed because the amount of observational data is far less than the number of subdivisions in the inversion area. To solve this problem, it is necessary to reasonably mesh the underground space and to impose proper constraints on the inversion model according to the available prior information. Influenced by gravity and the underground structure, the distribution of groundwater often has a certain directionality. In this paper, we theoretically deduce an application framework for the directional gradient constraint in Occam’s inversion on two-dimensional unstructured triangular meshes. A two-layer slip-band model with a complex terrain was designed. Local gradient constraints that gradually change with the dip angle of the slip-band were then set according to the prior directional information of the slip-band. The results show that the accuracies of the location identification and the water content of the inverted sliding-zone water can be effectively improved using the introduced gradient constraint. Finally, a two-dimensional MRS detection field study of Quaternary sedimentary strata is presented. The local gradient constraint significantly improves the accuracy with respect to identifying the water table, soil interfaces with different particle sizes, and the location of abnormal groundwater. In summary, the introduced application framework can significantly improve the accuracy of aquifer boundary and water content determination without increasing the data volume or workload, which will undoubtedly expand the application scope of MRS.