Sensitivity of a Distributed Watershed Model to Detailed Soil Data

With modern modeling tools and datasets, it is possible to create a watershed model with explicit treatment of unsaturated-zone processes and spatially varying soil properties at a fine scale. Do the results justify the effort to prepare the required input parameters? This research used the Gridded Surface Subsurface Hydrologic Analysis (GSSHA) modeling environment to create a 9-m resolution model of the Jones Falls watershed (65.3 km2) in the Maryland Piedmont. GSSHA is a distributed, physics-based hydrologic model that considers several conditions consisting of spatial variability of soil texture. The model applies GSSHA's Richards Equation option to simulate the unsaturated zone. Soil data were obtained from the SSURGO database. The watershed contains 71 different soil units, with depths ranging from 100 to 292 cm, and with up to 10 identified horizons. A soil index map was created, assigning an index value to each 9 m × 9 m pixel; soil component and horizon data were organized into a 3-horizon soil profile for each soil index. Two versions of the model were tested and run: one with fully distributed spatially varying soil properties, and another with a single soil profile applied to the entire watershed. Differences between the output of the two versions are examined, including: partitioning of arriving rainfall into direct runoff and infiltration; peak discharge; hydrograph shape and components (i.e., surface flow vs baseflow). These results will help hydrologic modelers determine an appropriate level of detail pertaining to the soil data in watershed modeling, depending upon the purpose of their analysis.