Skip to main content
ARS Home » Research » Publications at this Location » Publication #240475

Title: Effects of the resolution of soil dataset and precipitation dataset on SWAT2005 streamflow calibration parameters and simulation accuracy

Author
item Moriasi, Daniel
item Starks, Patrick

Submitted to: Journal of Soil and Water Conservation Society
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/1/2009
Publication Date: 3/15/2010
Citation: Moriasi, D.N., Starks, P.J. 2010. Effects of the resolution of soil dataset and precipitation dataset on SWAT2005 streamflow calibration parameters and simulation accuracy. Journal of Soil and Water Conservation Society. 65(2):63-78.

Interpretive Summary: The level of uncertainty of input parameters associated with hydrologic modeling has significant impact on the model performance and the uncertainty of the resulting model outputs. The goals of this study were to: 1) investigate the effect of soils data resolution on the Soil and Water Assessment Tool model (SWAT2005) streamflow simulation performance and parameter values using four precipitation data types with varying spatial resolutions, and 2) determine the best combination of soil and precipitation data sets for three different size sub-watersheds within the Fort Cobb Reservoir Experimental Watershed located in southwestern Oklahoma. The study results revealed that soils data resolution did not have a significant effect on SWAT2005 streamflow simulation performance across sub-watersheds, irrespective of the precipitation data type used. However, the model performed better when higher resolution precipitation datasets were used and especially for the larger size sub-watershed. There were slight to large differences in the resultant calibration parameter values depending on the calibration parameter, the precipitation data used, and the sub-watershed. The differences in the calibration parameter values led to some significant differences in the simulated water balance components such as surface runoff, ground water, lateral flow, and deep aquifer recharge values due to the soil dataset resolution especially when used in combination with the coarser resolution precipitation datasets in the smaller sub-watersheds. Large differences in the simulated surface runoff and deep aquifer recharge due to soils data set resolution could lead to significant differences in the simulated water quality components such as sediments and nutrients. This is important because significant differences in simulated sediments and/or nutrients could lead to significantly different outcomes in terms of the impacts of a given conservation practice for studies like the Conservation Effects Assessment Project (CEAP). Due to the lack of measured data to validate the simulated water balance components, it was recommended to use both the fine and coarse resolution soil datasets in combination with the fine spatial resolution precipitation datasets and that the simulated water balance components of interest be reported as a range.

Technical Abstract: The resultant calibration parameter values and simulation accuracy of hydrologic models such as the Soil and Water Assessment Tool (SWAT2005) depends on how well spatial input parameters describe the characteristics of the study area. The objectives of this study were to: 1) investigate the effect of soils data resolution using different precipitation data sets on SWAT2005 streamflow calibration parameters and simulation accuracy and 2) determine the combination of soil/precipitation datasets that results in the most representative streamflow calibration parameter values in three sub-watersheds (Cobb, Lake, and Willow Creeks), within the Fort Cobb Reservoir Experimental Watershed (FCREW) located in Caddo and Washita counties, Oklahoma. SWAT2005 was calibrated and validated for streamflow for the three sub-watersheds using State Soil Geographic (STATSGO) and Soil Survey Geographic (SSURGO) soils databases for each of the four available precipitation datasets with different spatial resolutions. The four sources of rainfall data included the National Weather Service’s network of Co-operative weather stations (COOP), statewide Oklahoma Mesonet (MESONET), USDA Agricultural Research Service’s weather station network (MICRONET), and NWS Next generation radar precipitation estimates (NEXRAD). The model performance was assessed using the Nash-Sutcliffe efficiency coefficient (NSE), root mean square error (RMSE), and percent bias (PBIAS) statistics. During the calibration period, there were no significant differences in the model monthly performance statistics between the higher resolution SSURGO soil database the lower resolution STATSGO soils database across sub-watersheds, irrespective of the rainfall data set used. However, the model performed better when the NEXRAD and MICRONET precipitation datasets were used. There were slight to large differences in the resultant calibration parameter values depending on the calibration parameter, the precipitation data used, and the sub-watershed. The model did not simulate streamflow during the short (6 months) validation period as well as during the calibration period. However, the monthly statistics revealed that there was no significant difference in the simulation accuracy due to using the higher resolution SSURGO soil database compared to STATSGO soils database across sub-watersheds, irrespective of the rainfall data set used. Based on the validation period, the model generally performed better using the NEXRAD and MICRONET precipitation datasets especially for the larger size Cobb Creek sub-watershed. Depending on the precipitation data set used and the sub-watershed, the study results showed that there were some significant differences in water balance component values due the soil dataset resolution especially using the MESONET and COOP precipitation datasets. Due to the lack of measured data to validate the simulated water balance components, it is recommended to use both SSURGO and STATSGO datasets with the finer spatial resolution NEXRAD and MICRONET precipitation datasets.