Benchmarking of Variable Density Model Codes against Henry's Problem

The modelling of groundwater flows in high-density brine systems provides a challenging problem for the groundwater modeller. The equations of variable density groundwater flow are non-linear, requiring robust, and usually numerically intensive methods to provide reliable solutions. There are a large number of modelling codes that have been designed to simulate variable density groundwater flow, none of which has distinguished itself as an 'industry standard'. The standard way to test the performance of a numerical code is to compare its numerical solution against the analytical solution for a representative problem. Due to the non-linearity of the equations governing variable density flow, a fully analytical solution cannot be achieved. However, the semi-analytical solution to Henry's Problem of salt encroachment into a coastal aquifer (Henry, 1964) is widely regarded as one of the standards against which variable density codes should be compared. In addition to comparisons against Henry's Problem, benchmarking against other numerical codes provides increased confidence in the code's performance. This methodology of applying different numerical codes to same problem and comparing the results provides an indication of how different codes perform under specific conditions. As part of a project between Water Management Consultants Ltd and Minera Escondida Limitada, a number of different variable density modelling codes have been benchmarked, including SWIFT, HST3D, 3DFEMFAT, DSTRAM, and FEFLOW. This project is ongoing. These codes have been compared against Henry's Problem, and the sensitivity of each code's solution compared, for changes in maximum density, viscosity, dispersivity, porosity, and the effects of pumping. The ability of the codes to simulate a sharp saline interface (Voss and Souza, 1987) was also investigated. Once numerical models had been constructed to simulate Henry's Problem, model parameters were changed in a step-wise fashion until conditions resembled those of an inland drainage system (salt pan or Andean salar). These models simulate a thin layer of freshwater flowing over high-density brines, exiting in evaporation zones. The effect of changes in the grid dimensions on the solutions of these models was also investigated. This paper presents the results of the initial comparisons to Henry's Problem, the sensitivity analysis, and sharp interface investigations.