Assessing the Robustness of Bel1D for Inverting Tem Data

Summary The prime issue in the field of geophysics is the imaging of the subsurface of the earth. Geophysical methods offer an economical alternative for investigating the subsurface compared to costly boreholes investigating methods. Every geophysical method relies on physical forward models which themselves related to physical properties, such as (density, seismic velocity, or electrical conductivity) which can further transformed into the desired subsurface properties such as (permeabilities and porosity, clay content or salinity) using petrophysical relationship. Inversion approaches can be adopted to process the geophysical data. Deterministic inversion uses a regularization approach to stabilizes the inversion and resolve the non-unicity of the solution. the solution of deterministic inversion is quite certain ( Aster et al., 2013 ; kemna et al., 2007 ). Stochastic inversion method uses the Bayesian framework estimates the range of possible uncertainties. In this work, we use Bayesian evidential learning for 1D inversion of TEM fast data. BEL1D is a new approach for 1D geophysical imaging proposed by ( Michel et al, 2020 ). Here we combined the SimPEG (an open-source python package) as a forward solver to stochastically solve the inverse problem, we apply this approach on TEM data collected in Vietnam for saltwater intrusion Characterization.