Research Fellow: Jorge Lopez Alvis
hosted at Applied Geophysics, University of Liege
Frédéric Nguyen, Applied Geophysics, University of Liege
Thomas Hermans, Departement of Geology, Ghent University
CSIC Barcelona – 6 months (2019,2020)
Aquale – 2 months (2020)
Ghent University – 2 months (2020)
Integrate time-lapse geophysical data sets, at an hourly or daily temporal resolution, into a geostatistical inverse framework to gain insight on transport processes and hydrogeological parameters distribution in the subsurface.
Tasks and methodology
Bayesian evidential learning (BEL) of data-prediction relationships will be applied using geophysical data and subsurface transport models. Within the BEL framework, probabilistic falsification will be used to check for consistency of plausible geological scenarios. A combination of 3D image analysis on geophysical data (ERT, SIP and GPR), temporal information, multiple-point geostatistical simulations and dimension reduction techniques will be integrated in the framework.
The code used for the publication “A cross-validation framework to extract data features for reducing structural uncertainty in subsurface heterogeneity” is to be found on GitHub: https://github.com/jlalvis/CV_AKDE
Dissemination and communication
Publication in Water Resources Vol 133: “A cross-validation framework to extract data features for reducing structural uncertainty in subsurface heterogeneity”
Presentation at AGU Fall Meeting (Washington DC, USA): “Updating structural uncertainty through dimension reduction of geophysical data“
June 2018: Oral presentation at the 2018 Computational Methods in Water Resources (CMWR) conference: “Updating prior geologic uncertainty with GPR traveltime tomographic data”.
Poster presented during the 4th Cargèse Summer School:“Updating uncertainty in hierarchical subsurface model using geophysical data: synthetic case for crossborehole-hole GPR”.
November 2018: Geology Research seminar in University of Ghent: “Using geophysical data to update uncertainty in structural parameters of subsurface models”.
- BEL – bayesian approach where subsurface models are used to learn a statistical model that relates data and prediction variables.
- ERT – electrical resistivity tomography.
- SIP – spectral induced polarization.
- GPR – ground penetrating radar.