A Deep Learning Approach for Aquifer Prospectivity Mapping

Aquifer exploration requires the effective integration of diverse and large-scale geoscience datasets (e.g., geophysical, hydrological, geological, geochemical) in order to locate, delineate, and/or characterize water resources. We propose a deep learning approach to demonstrate how this technology can enrich the prospectivity mapping process. The backbone of our approach is a VNet, a deep convolutional neural network that we designed to learn multi-scale features. We exemplify our approach using data from the Northern Territory in Australia.