Refining process representation in high-resolution models of headwater catchments using internal catchment diagnostics

Abstract:

As the complexity of the problems we seek to address with process-based models continues to increase, our approaches to improving confidence in our predictions must keep pace. Process-based, distributed models have been applied in headwater catchments to address many different objectives, all of which are linked by their reliance on the selection of a catchment-representative parameter set or sets. While these parameter sets are typically obtained through calibration to the streamflow hydrograph, it is widely acknowledged that there is often insufficient information in the hydrograph to effectively address parameter equifinality. Here, we suggest that optimal parameter sets can be obtained with an additional step in the calibration process that considers the spatial representation of internal catchment behavior (e.g. space-time distributions of evapotranspiration, water table depth, presence of overland flow, soil water). Modeled internal catchment behavior is an under-utilized but valuable source of information for separating plausible from unlikely model scenarios. We demonstrate how spatial patterns of hydrologic states and fluxes across annual, seasonal, and event time scales can improve the calibration process and reduce likely parameter sets. Our approach is applied to an extensively monitored headwater catchment in Tenderfoot Creek Experimental Forest in central Montana, simulated using the Distributed Hydrology-Soil-Vegetation Model. Consideration of spatial diagnostics in the calibration process has great potential to ensure a holistic representation of catchment dynamics as well as to increase confidence in conclusions from these types of modeling applications.