My research topics deal essentially with development and application of simulation tools for water, mass and heat flow in aquifer structures. I try and give some elements to answer two very basic questions: what is inside a simulation tool, and what is it good for ? The main line of this report is twofold: - On the one hand, showing the perpetual and fruitful two-way exchange between the needs stemming from modeling a practical problem and the search for the best numerical techniques, in terms of efficiency but also of parsimony (looking for the simplest possible solution) - On the other hand, illustrating the increased understanding of flow mechanisms in aquifer structure, brought by modeling. In a search for the most efficient numerical methods, I have developed and tested various approaches, with a varying degree of success. I feel that this experience can be of some use for a modeler, and I developed some aspects of this work in the first part of this report. Three problems are presented: - Simulating sharp fronts while solving the dispersion equation: three formulations are proposed, using Spectral Elements, Discontinuous Elements and a Eulerian Lagrangian (ELLAM) approach. - Simulating flow in media with strongly varying hydraulic conductivity: a multigrid method has been developed. - Merging a densely discretized porous domain, or alternatively a fractured domain represented by a Discrete Fractures Network, into a regional flow model: two solutions are proposed, one based on Boundary Elements, one on Finite Elements In a second part I discuss some of my work related to code development, among which the development of the METIS code, which simulates flow, solute and heat transfer in a porous / fractured medium. I show on a few examples how the simulation tools are tuned to give to the modeler specific services which facilitate the building of an application and allow to extract therefrom as much information as possible. The tools I developed have been used for various applications, ranging from engineering studies such as flow through a dam to more academic subjects such as the behavior of natural tracers in large scale aquifer systems. I chose to show some of these applications, with emphasis on those which rested on a strong experimental support, in order to illustrate the whole modeling work from data analysis to simulation. This constitutes the third part of the report. A first series of examples deals with the simulation of natural tracers movement (noble gases) in two large aquifer systems (Paris Basin in France and Carrizo aquifer system in the USA). These examples show how modeling natural tracers permits to improve the robustness of identification of flow parameters (hydraulic conductivity), and thus to better quantify water fluxes. Natural tracers are further used to compute water ages, with interesting applications in the field of paleoclimatic reconstruction. My second example deals with the quantification of leakage flux through geotextiles used for isolation at the bottom of surface waste disposal sites. The system to be modeled is partly made of natural and partly of artificial material. A key feature is the very fine interface between the host medium and the geotextile. This work is based on a strong experimental program developed by CEMAGREF, and on a modeling work using the METIS code.
