Does soil data resolution matter? State Soil Geographic database versus Soil Survey Geographic database in rainfall-runoff modeling across Wisconsin

Abstract

Whether or not the use of generalized, State Soil Geographic (STATSGO) data in place of higher resolution Soil Survey Geographic (SSURGO) data reduces the accuracy of hydrologic and nonpoint source pollution models has thus far been an open question. Comparative studies have yet to reveal a systematic bias in STATSGO–based model outputs on account of their small sample sizes and differences in the models employed. In an effort to determine whether a bias exists, direct runoff was modeled for a hypothetical 24-hour rainfall event, using STATSGO and SSURGO as alternative inputs to a series of standard rainfall-runoff models in nearly 300 contiguous watersheds, spanning most of the state of Wisconsin. The Long-Term Hydrologic Impact Assessment (L-THIA) modeling tool was used for this analysis. Results indicate that there is a negative bias in STATSGO–based runoff over the large majority of the study area and that the degree of underprediction is highest for spatially disaggregated (distributed parameter) models. Runoff was also modeled for daily precipitation in six gauged watersheds and was compared to observed runoff, with SSURGO–based, distributed models typically producing the most accurate outputs. In addition, a series of regression analyses was conducted to determine whether, and in what direction, the STATSGO bias is affected by the percent coverage of land uses that discourage infiltration. The results of these analyses suggest that STATSGO–based, lumped, and partially distributed models, on average, underpredict the relative impact of increasing land-use intensity. These findings indicate that two of the most common approaches to improving the computational efficiency of watershed modeling systems: the use of lower resolution soils data and the lumping of model parameters to larger spatial units of analysis, combine to reduce the accuracy of modeled runoff under current conditions, while simultaneously underestimating the impact of potential future land-use change.