3D Stochastic Modeling of the Roussillon Aquifer System
This repository presents the main steps of the 3D multiple-point statistics (MPS) modeling workflow developed during the first part of the PhD thesis work:
PhD Thesis
https://libra.unine.ch/entities/publication/8d3e5151-252c-481a-9f97-a3416f9d89ea
This work has led to the publication of two articles:
-
Hydrogeological modeling of the Roussillon coastal aquifer (France): stochastic inversion and analysis of future stresses
DOI: https://doi.org/10.1016/j.jhydrol.2023.129693 -
3D multiple-point statistics simulations of the Roussillon Continental Pliocene aquifer using DeeSse
DOI: https://doi.org/10.1016/j.cageo.2020.104651
3D multivariate facies simulations are performed by stratigraphic intervals using multiple-point statistics (MPS).
Simulations are conditioned to borehole data and training images to reproduce realistic sedimentary architectures.
Within each simulated facies, permeability fields are generated using Gaussian Random Functions (GRF) to represent intra-facies heterogeneity and spatial variability.
The high-resolution geological grid is upscaled to the hydrogeological model grid while preserving heterogeneity structures.
An ensemble-based Monte Carlo approach is then used to calibrate permeability parameters against observed mean piezometric heads.
Multiple realizations are evaluated, and parameter sets are retained based on their ability to reproduce hydrogeological observations.
Ensemble statistics are computed over all stochastic realizations to quantify uncertainty and variability.
This includes:
- Facies proportion statistics
- Spatial variability metrics
- Connectivity analysis
- Hydraulic parameter distributions
These statistics allow evaluation of model consistency with training data and assessment of predictive uncertainty.
All steps are performed stochastically to generate multiple realizations, enabling uncertainty quantification and ensemble-based groundwater modeling.
