Bayesian evidential learning: An alternative to hydrogeophysical coupled inversion

Thomas Hermans (1) , Nicolas Compaire (2, 1, 3) , Robin Thibaut (1) , Nolwenn Lesparre (3)

1 UGENT - Universiteit Gent = Ghent University
2 ISAE-SUPAERO - Institut Supérieur de l'Aéronautique et de l'Espace
3 ITES - Institut Terre Environnement Strasbourg

Deterministic geophysical inversion suffers from a lack of realism because of the regularization, while stochastic inversion allowing for uncertainty quantification is computationally expensive. In this contribution, we propose to use Bayesian Evidential Learning as an alternative to hydrogeophysical coupled inversion. We demonstrate the ability of the approach to successfully predict a hydrogeological target from time-lapse ERT data in the context of a heat injection and storage experiment.

Domaines: Géophysique [physics.geo-ph]; Hydrologie

Thomas Hermans, Nicolas Compaire, Robin Thibaut, Nolwenn Lesparre. Bayesian evidential learning: An alternative to hydrogeophysical coupled inversion. First International Meeting for Applied Geoscience & Energy, Sep 2021, Denver, United States. pp.3125-3129, ⟨10.1190/segam2021-3580979.1⟩. ⟨hal-03426940⟩