Determining trustworthy boundaries for 4D watershed models

While surface watersheds are easily delineated using topographic divides, the delineation of groundwater catchments is not nearly as straightforward. A groundwater catchment is not completely controlled by its topographically-defined surface watershed, but rather can be influenced by neighboring surface watersheds. In addition, groundwater divides, unlike surface water divides, can change over short periods of time, even seasonally. However, when groundwater data is scarce, as is the case in most mountainous areas, the surface water divide is still assumed to be a proxy for the groundwater divide. Traditional watershed models either simplify the groundwater portion or model surface water and groundwater as completely separate systems. Equating the groundwater catchment to the surface watershed in these traditional scenarios is not necessarily problematic. However, the new generation of 4D integrated distributed parameter models, where groundwater is represented in three spatial dimensions (GSFLOW, for example), allow for a detailed representation of groundwater and its connection to and effects on surface water. Using the same watershed boundary assumptions in this type of model can lead to fundamental errors when modeling coupled surface and groundwater behavior, particularly in mountainous or semi-arid regions. Instead, the boundaries of a groundwater model should be extended beyond the surface watershed, but by how much? By testing cross-sections through a 3D model of regional groundwater flow at multiple nested scales, this paper tries to determine by what amount the boundaries should be extended in order to produce more trustworthy results. Using COMSOL Multiphysics Earth Science Module, we are able to explore the effects of hydraulic conductivity, depth of circulation, recharge, and topography on parameters like groundwater discharge to streams and its residence time. Rather than arbitrarily assigning boundaries to groundwater models, this work allows us to determine which model boundaries provide the best estimates under different physical and environmental conditions.