Development of the AquiferSim model of cumulative effect on groundwater of nitrate discharge from heterogeneous land use over large regions

Abstract

Regional-scale prediction of the effects of nitrate discharge from land use on the quality of the underlying groundwater requires two major model components: (1) a climate-driven model of agricultural land use that predicts nitrate discharge at the bottom of the plant root zone, for locations over horizontal space; and (2) a groundwater transport model, which predicts nitrate concentration at horizontal and vertical locations within the aquifer as well as the nitrate discharge to surface-water bodies. AquiferSim is a recently developed groundwater transport model, for which the inputs of recharge and nitrate from the land surface are received from a GIS user interface that accesses the root zone nitrate discharge model. The AquiferSim groundwater transport model is designed to address two particular requirements. The first is that model run times should allow for real-time examination of land use scenarios and assessment of uncertainty. The second is that dimensions of computational cells should allow for realistic transport dispersion in both horizontal and vertical dimensions, as well as allowing improved accuracy of flowpaths to surface-water bodies. These latter requirements imply very large numbers of computational cells for regional-scale studies, with associated costs in model run time. These issues are addressed in the AquiferSim model by: assuming steady-state groundwater flow and transport; solving 2D horizontal groundwater flow on ~10⁶ computational cells with a fast, fullmulti- grid solver; and then solving flow and transport on vertical sections of ~10⁴ cells along selected groundwater flowpaths, with a successive over- relaxation solver. The software was developed entirely in Microsoft Visual C# on the .NET framework. This enables the AquiferSim engine to run on modern Windows PCs and on Linux and clustered environments using the MONO platform. Computational time performance of the AquiferSim engine enables the horizontal 2D steady-flow groundwater problem and pathline tracking to be solved in about 5 s for a region occupying one third of the 1025 x 1025 computational grid. Solution for groundwater flow, nitrate transport, and groundwater age on the ~ 10⁴ cells of one vertical slice requires up to 20 s. Implementation of AquiferSim within a regional council for environmental planning purposes is the next phase of development. The major issues are likely to be: quantifying the predictive uncertainty caused by inadequate description of aquifer recharge and properties; and software design for interrogation of AquiferSim output to meet yet unspecified requirements for end-user information.