Combining Multi-temporal Electric Resistivity Tomography and Predictive Algorithms for supporting aquifer monitoring and management

Abstract

Summary This work presents the results of geophysical data prediction by applying statistical and predictive algorithms to a multi-temporal Electric Resistivity Tomography dataset. A cross-hole time-lapse resistivity survey was carried out during an experiment addressed to monitor a tracer diffusion in a real aquifer. In order to retrieve a number of “predicted” pseudo sections of apparent resistivity values, we applied the Vector Autoregressive (VAR) and Recurrent Neural Network (RNN) algorithms over a sequence of 18 ERT surveys. Real and predicted dataset allow to delineate plume evolution under 30 m depth, describing a complex transport pathway influenced by hydraulic properties of the studied aquifer.