Coupled hydrogeophysical inversion and machine learning for improved hydrological parameter estimation

• Applicant: Professor Dr. Sebastian Uhlemann
• Subject Area: Geophysics
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 544878839

Groundwater provides freshwater to more than 1500 million people, and is an important resource for agricultural and eco-systems. Yet, increasing groundwater abstraction and climate change-induced changes in precipitation patterns, with more frequent severe and prolonged drought, are stressing these critical water resources. There is a critical need for novel methods that can optimize aquifer recharge methodologies, yet estimating recharge remains highly uncertain. Here, we propose a hydrogeophysical approach that is based on geoelectrical monitoring linked with hydrological modelling and machine learning. Geoelectrical methods have been shown to provide detailed information about groundwater related processes, and linked with hydrological modelling quantitative estimates of hydrological parameters can be obtained. We will make use of these existing developments and develop a hydrogeophysical inversion approach that combines deterministic and geostatistic approaches to provide estimates of the hydraulic conductivity distribution and eventually spatially resolved estimates of groundwater recharge rates. Since these approaches are known to be computationally expensive, we will develop and test two machine learning based approaches. The first approach will make use of classification schemes to track the infiltration front and based on this information will derive hydraulic conductivities, and the second approach will use physics-based models to estimate hydraulic conductivities from the measured electrical resistances. We will apply optimized survey design techniques to test whether the increased resolution that can be obtained using such approaches can provide more accurate recharge rate estimates. To determine the most effective solution, we will compare the accuracy and computational cost of the various approaches. We will first develop these techniques using synthetic models, which resemble already aquired field data, to which the approaches will be applied at a later stage. To test and assess the accuracy of the various approaches, we will design and conduct laboratory experiments. With synthetic and laboratory testing having validated the developed hydrogeophysical approaches, we will apply them to field data that has been acquired in Southern California in previous projects, as well as to a new site. Eventually, this project will develop novel hydrogeophysical parameter estimation techniques that will be tuned to provide spatially resolved estimates of groundwater recharge rates. Such information is critical for managing and optimizing managed aquifer recharge facilities.

• DFG Programme: Research Grants