ON THE CHALLENGE OF INTEGRATED SURFACE-SUBSURFACE FLOW AND TRANSPORT MODELLING AT MULTIPLE CATCHMENT SCALES

Over the past several years, substantial research has been directed towards understanding flow and solute transport processes occurring at the interface between surface water and groundwater. To date, however, relatively few numerical models have been formulated that couple these processes in a holistic, physically-based framework. In this paper, we will examine these coupling strategies in the context of the HydroGeoSphere model, a surface-subsurface control-volume finite element model. HydroGeoSphere is a fully-integrated 3D model that can simulate water flow, heat flow and advective-dispersive solute transport on the 2D land surface and in the 3D subsurface under variably-saturated conditions. Full coupling of the surface and subsurface flow regimes is accomplished implicitly by simultaneously solving one system of non-linear discrete equations describing flow and transport in both flow regimes. The model capabilities and main features are demonstrated with several high-resolution 3D numerical simulations performed for catchments of various scales, ranging from the scale of an intensively-monitored rainfall-runoff tracer experiment (~ 2000 m²), to a regional-scale watershed of about 1000 km², to the continental scale that comprises the entire Canadian land mass. The simulations highlights the difficulties and challenges for representing water flow and solute flux in complex natural systems, and stresses the advantage of using a process-based model such as HydroGeoSphere for prediction of current and future water management scenarios. Among the challenges associated with such simulations is the discrepancy in spatial and temporal resolutions needed for the surface and subsurface flow domains, the forms of the needed constitutive relations and effective parameter values at various scales, and dealing with data uncertainty. New algorithms such as sub-timing and sub-gridding, together with domain decomposition for parallel processing, have been developed to alleviate some of the computational demands and are demonstrated here.