You are here:
Comparison of Varied Precipitation and Soil Data Types for Use in Watershed Modeling.
Citation:
PRICE, K., S. T. PURUCKER, S. R. KRAEMER, AND J. E. BABENDREIER. Comparison of Varied Precipitation and Soil Data Types for Use in Watershed Modeling. Presented at AGU Fall Meeting 2011, San Francisco, CA, December 05 - 09, 2011.
Impact/Purpose:
Presentation at American Geophysical Union, Fall Meeting 2011
Description:
The accuracy of water quality and quantity models depends on calibration to ensure reliable simulations of streamflow, which in turn requires accurate climatic forcing data. Precipitation is widely acknowledged to be the largest source of uncertainty in watershed modeling, and soil properties strongly influence hydrologic routing, in large part determining whether precipitation is moved as overland flow or infiltrates to subsurface storage. Uncertainty introduced by these input data propagates through all subsequent stages of water quantity, quality, and ecosystem service modeling and analysis. Most watershed models are designed to easily incorporate publicly-available precipitation data from rain gauges (e.g., data provided by the National Climatic Data Center), but several additional data products from ground-based radar are available and provide more spatially-explicit precipitation estimates. However, these datasets are associated with greater computational requirements and expertise for use in watershed models. The USDA-NRCS produces two digital soil data products, STATSGO and SSURGO, with SSURGO characterized by much higher spatial resolution and greater computational requirements. In addition to easier use, STATSGO and gauge precipitation data are more representative of data availability outside the U.S. than the higher resolution products. Here, we investigate whether the use of higher-resolution Multisensor Precipitation Estimator (MPE) and SSURGO soil data improve the accuracy of daily streamflow simulations using the Soil and Water Assessment Tool (SWAT) watershed hydrology model. SWAT-simulated daily streamflows are compared with USGS-observed streamflows for four nested subwatersheds of the Neuse River basin in North Carolina (21, 203, 2960, and 10100 km^2 watershed area), for a 9-year simulation period (2002-2010). Standard metrics for streamflow calibration, such as Nash-Sutcliffe Efficiency (NSE) and R^2, are heavily influenced by high flows. Calibration and accuracy assessment based on these objective functions capture minimal information on accuracy of low flows or flow variability, which are as important as high flows to habitat suitability and water availability. In addition to standard approaches, we used an alternative calibration strategy, using a combination of objective functions representing high flows (NSE), low flows (modified NSE), and flow variability (standard deviation ratio), and focused on likelihood distributions rather than individual "winning" simulations. Streamflow simulation accuracy improved with MPE data for the two smaller watersheds (10s to 100s km^2) but not in the larger watersheds (1000s of km^2). This nested approach needs to be repeated in other study areas, but it appears that semi-distributed modeling in watersheds large enough to include multiple rain gauges within the watershed boundary may not benefit from use of MPE data. SSURGO data did not lead to significantly improved streamflow simulations at any watershed scale. These results suggest that for many watershed modeling tasks, rain gauge and STATSGO data are of sufficient resolution, particularly for semi-distributed and lumped models. These results also serve as a reminder that increased spatial resolution does not necessarily equal increased data accuracy.
URLs/Downloads:
Comparison of Varied Precipitation and Soil Data Types for Use in Watershed Modeling.