Estimation of groundwater storage from seismic data using deep learning

Convolutional neural networks can provide a potential framework to characterize groundwater storage from seismic data. Estimation of key components, such as the amount of groundwater stored in an aquifer and delineate water table level, from active-source seismic data are performed in this study. The data to train, validate and test the neural networks are obtained by solving wave propagation in a coupled poroviscoelastic–elastic media. A discontinuous Galerkin method is applied to model wave propagation, whereas a deep convolutional neural network is used for the parameter estimation problem. In the numerical experiment, the primary unknowns estimated are the amount of stored groundwater and water table level, while the remaining parameters, assumed to be of less of interest, are marginalized in the convolutional neural network-based solution. Results, obtained through synthetic data, illustrate the potential of deep learning methods to extract additional aquifer information from seismic data, which otherwise would be impossible based on a set of reflection seismic sections or velocity tomograms.     

Interlobate esker architecture and related hydrogeological features derived from a combination of high-resolution reflection seismics and refraction tomography,Virttaankangas, southwest Finland

A novel high-resolution (2–4 m source and receiver spacing) reflection and refraction seismic survey was carried out for aquifer characterization and to confirm the existing depositional model of the interlobate esker of Virttaankangas, which is part of the Säkylänharju-Virttaankangas glaciofluvial esker-chain complex in southwest Finland. The interlobate esker complex hosting the managed aquifer recharge (MAR) plant is the source of the entire water supply for the city of Turku and its surrounding municipalities. An accurate delineation of the aquifer is therefore critical for long-term MAR planning and sustainable use of the esker resources. Moreover, an additional target was to resolve the poorly known stratigraphy of the 70–100-m-thick glacial deposits overlying a zone of fractured bedrock. Bedrock surface as well as fracture zones were confirmed through combined reflection seismic and refraction tomography results and further validated against existing borehole information. The high-resolution seismic data proved successful in accurately delineating the esker cores and revealing complex stratigraphy from fan lobes to kettle holes, providing valuable information for potential new pumping wells. This study illustrates the potential of geophysical methods for fast and cost-effective esker studies, in particular the digital-based landstreamer and its combination with geophone-based wireless recorders, where the cover sediments are reasonably thick.