Cumulative Conductance Mapping for the Direct Transformation of TDEM Data into 1D Resistivity Models

Inversion of Time-Domain Electromagnetic (TDEM) data is an inherently non-unique problem. To address this, we present a fully data-driven method that directly transforms 1D TDEM data into layered resistivity models via cumulative conductance mapping. By defining a cumulative conductance model function of the diffusion time, and by transforming the TDEM data into a conductance by using the S-transformation, we establish a mapping function that allows the conversion of the transient data into a cumulative conductance model, that is then differentiated to recover a 1D resistivity model. To generalize this for unknown data without a priori constraints, a neural network is trained to predict this mapping. Furthermore, the predicted mapping function follows the true current density maxima, highlighting how the mapping function is related to the diffusion of the EM field. To evaluate the proposed methodology on field data, we used the Southern Thomson Orogen VTEM plus Airborne Electromagnetic dataset, published by Geoscience Australia. The survey’s goal to characterize underlying geological structures makes it a good case study for assessing the robustness the methodology in a real-world scenario. Our results show that our method retrieves resistivity profiles that follow the same trend as the CDI profiles published in the repository.