EGU24-8304, updated on 08 Mar 2024
https://doi.org/10.5194/egusphere-egu24-8304
EGU General Assembly 2024
© Author(s) 2024. This work is distributed under
the Creative Commons Attribution 4.0 License.

Water table height maps prediction from passive surface-wave dispersion using deep learning

José Cunha Teixeira1,2, Ludovic Bodet1, Agnès Rivière3, Marine Dangeard2, Amélie Hallier2, Alexandrine Gesret3, Amine Dhemaied2, and Joséphine Boisson Gaboriau2
José Cunha Teixeira et al.
  • 1Sorbonne Universite, CNRS, EPHE, UMR 7619 METIS, 4 place Jussieu, 75252 Paris 05, France (jose.teixeira@sorbonne-universite.fr)
  • 2SNCF Réseau, 6 avenue François Mitterrand,7 93210 Saint-Denis, France
  • 3Geosciences Department, Mines Paris - PSL, PSL University, Paris, France

Monitoring underground water reservoirs is challenging due to limited spatial and temporal observations. This study presents an innovative approach utilizing supervised deep learning (DL), specifically a multilayer perceptron (MLP), and continuous passive-Multichannel Analysis of Surface Waves (passive-MASW) for constructing 2D water table height maps. The study site, geologically well-constrained, features two 20-meter-deep piezometers and a permanent 2D geophone array capturing train-induced surface waves. For each point of the 2D array, dispersion curves (DCs), displaying Rayleigh-wave phase velocities (VR) across a frequency range of 5 to 50 Hz, have been computed each day between December 2022 and September 2023. In the present study, these DCs are sampled in wavelengths ranging from 4.5 to 10.5 m in order to focus the monitoring on the expected water table depths. All VR data around one of the two piezometers is used to train the MLP model. Water table heights are then predicted across the entire geophone array, generating daily 2D piezometric maps. Model's performance is tested through cross-validation and comparisons with water table data at the second piezometer. Model’s efficiency is quantified with the root-mean-square error (RMSE) and the coefficient of determination (R²). A R² is estimated above 80 % for data surrounding the training piezometer and above 55 % for data surrounding the test piezometer. Additionally, the RMSE is impressively low at 0.03 m at both piezometers. Results showcase the effectiveness of DL in generating predictions of water table heights from passive-MASW data. This research contributes to advancing our understanding of subsurface hydrological dynamics, providing a valuable tool for water resource management and environmental monitoring. The ability to predict 2D piezometric maps from a single piezometer is particularly noteworthy, offering a practical and efficient solution for monitoring water table variations across broader spatial extents.

How to cite: Cunha Teixeira, J., Bodet, L., Rivière, A., Dangeard, M., Hallier, A., Gesret, A., Dhemaied, A., and Boisson Gaboriau, J.: Water table height maps prediction from passive surface-wave dispersion using deep learning, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-8304, https://doi.org/10.5194/egusphere-egu24-8304, 2024.

Comments on the supplementary material

AC: Author Comment | CC: Community Comment | Report abuse

supplementary materials version 1 – uploaded on 13 Apr 2024, no comments