Analyzing the ability of BEL1D for inverting TEM data.

Understanding the subsurface is of prime importance for many geological and hydrogeological applications. Geophysical methods offer an economical alternative for investigating the subsurface compared to costly boreholes investigation methods, but results are often obtained through an inversion problem whose solution is non-unique. There are two types of inversion approaches: deterministic and stochastic. Deterministic inversion provides a unique solution with no way to efficiently and accurately assess uncertainty  but is relatively fast to compute. Stochastic inversions investigate the full range of solutions which make them computationally very expensive. In this research, we assess the robustness of the recently introduced BEL1D method for the stochastic inversion of the time domain electromagnetic data (TDEM). We analyze the effect of the accuracy of the forward model (through the open-source SimPEG code) on the estimation of the posterior space using a synthetic case and discuss the importance of prior selection. We also apply the algorithm on field data collected in Vietnam to assess saltwater intrusions. We observed that the proper selection of timesteps and space discretization is essential to limit the computation cost while maintaining the accuracy of the posterior estimation. Secondly, the selection of the prior distribution has a direct impact on fitting the observed data and is crucial to a realistic uncertainty quantification. Furthermore, in contrast to previous studies, we suggest rejecting models not fitting the data at an early stage for reducing computational costs. Lastly, the application of BEL1D together with SimPEG for stochastic TDEM inversion is a very efficient approach as it allows us to estimate the uncertainty at a limited cost.

Keyword: Saltwater intrusion, uncertainty, TDEM, BEL1D, SimPEG