1994). Réflexions sur la modélisation de la propagation de polluants dans les hydrosystèmes souterrains About pollutants transport modelling in groundwater

Résumé

Les modèles de simulation de propagation de polluants dans les eaux souterraines sont de plus en plus utilisés comme outils de gestion de cette ressource. La qualité des simulations dépend étroitement des connaissances que l'on a des processus et des paramètres de transport nécessaires à la mise en application des modèles. La fiabilité des résultats repose sur:

- le choix du bon modèle en fonction de l'échelle d'observation

- la mesure des paramètres représentatifs du transport liée à l'échelle de discrétisation du site

- la spatialisation de mesures locales.

Compte tenu des spécificités des hydrosystèmes souterrains (invisibilité, accès coûteux), la connaissance du milieu restera trop fragmentaire pour réaliser des simulations fines. Seules les approches stochastiques permettent alors d'intégrer ces incertitudes dans les simulations.

Mots-clés:

• Eaux souterraines,
• pollution,
• modélisation

Abstract

Groundwater quality modelling has become a tool for water management. The accuracy of the simulations closely depends on the available knowledge concerning the transport processes and the parameters used in the model. The accuracy of the results depends on the choice ofa suitable model adapted to the observation scale, the measurement of the effective parameters linked to the discretization of the field and the spatialization of the local measurements.

The mathematical model used to describe mass transport in porous media is the dispersion-convection equation. The velocity is calculated by solving the flow equation. For heterogeneous media, numerical schemes which simultaneously solve heads and velocities have to be preferred to classical finite element or finite differences techniques. The dispersion coefficient represents the velocity fluctuations around the average velocity. Therefore, it strongly depends on the dimension and the scale of the discretization.

A predictive simulation of the Twin Lake Tracer Test experiment has been done. After a very fine calibration of the flow (differences between measured and calculated heads less than 1 cm), the transport simulation did not succeed. The headgradients were not calculated with enough accuracy and the simulated plume travelled in a wrong direction.

Due to the nature of groundwater (invisible, expensive rneasurements), knowledge of the structure of the aquifer will always be too incomplete to perform very detailed simulations. Stochastic computations may be the way to take into account uncertainities in groundwater modening.

Keywords:

• Groundwater,
• pollution,
• simulation