[S-EM12] Electric, magnetic and electromagnetic survey technologies and scientific achievements

[SEM12-P01] Structurally constrained magnetotelluric inversion using a modified regularization constraint: An alternative to the cross-gradient

*Dieno Diba¹, Han SONG¹, Makoto Uyeshima¹, Yoshiya Usui¹ (1.Earthquake Research Insititute, the University of Tokyo)

The integration of one or more independently derived physical Earth models in the magnetotelluric (MT) inversion can result in a narrower solution space of resistivity models, leading to higher confidence for geological interpretations (e.g., Moorkamp, 2017). It is called constrained MT inversion. The resulting resistivity structure is required to explain the observation and satisfy a certain relationship to the reference models. Here, we have developed a structurally constrained three-dimensional MT inversion scheme where the spatial resemblance between the inverted resistivity structure and a reference model is enforced, without necessitating a direct correlation between resistivity and the physical parameter of the reference model. As MT inversion results highly depend on the form of the regularization constraint, we imposed the structural information into the resistivity structure via the implementation of regularization (as in, e.g., Zhou et al., 2014). Typically, the regularization constraint of MT inversion involves a measure of model roughness through simple finite difference operators, thereby minimizing it will deter spurious structures in the resistivity model. In this study, we developed a weighted difference operator based on structural continuities and discontinuities in the reference model. The weights are relatively small across discontinuities, allowing for sharp changes of resistivity values within the same area, and relatively large along coherent structures, enforcing smooth changes of resistivity within the same area. This approach is implemented in the FEMTIC inversion code (Usui, 2015, 2021; Usui et al. 2017). In this presentation, we will showcase the benefits as well as possible limitations of employing this inversion scheme. Through synthetic and field data inversion experiments, we will comprehensively demonstrate to what extent our inversion scheme is usable, in comparison with the widely used cross-gradient inversion method (Gallardo and Meju, 2004).