Soil smarts: A taxonomic approach to speed up watershed modeling without sacrificing water quality insights

Abstract Hydrologic modeling studies within the United States frequently utilize the USDA‐Natural Resources Conservation Service State Soil Geographic (STATSGO, 1:250,000 resolution) or Soil Survey Geographic (SSURGO, 1:24,000 resolution) database to characterize soil properties. The finer resolution of SSURGO enables more precise spatial modeling that is beneficial in locating hydrologically vulnerable regions. However, the coarser‐resolution STATSGO results in quicker simulation runtimes, which can be computationally necessary in larger watersheds and for multi‐decadal estimation periods. Extending runtime by several minutes for a single simulation translates into multiple days during calibration and optimization modeling, for which hundreds to thousands of runs may be needed. We developed a method to aggregate SSURGO using soil taxonomy while maintaining distinctions critical to hydrologic processes. We automated this process in R for the Soil and Water Assessment Tool (SWAT; https://www.climatehubs.usda.gov/hubs/international/tools/soil-and-water-assessment-tool ) soil input data. Our method identifies hydrologically sensitive SSURGO map units, aggregates them up to the taxonomic subgroup level, and then creates the SWAT‐required inputs from the subgroup characteristics. Soil horizons are then standardized by depth‐weighting the SSURGO horizons. The taxonomic SWAT model ran twice as fast as the SSURGO model, predicted similar nutrient and sediment losses, and matched SSURGO identification of hydrologically vulnerable areas. The STATSGO model ran most quickly but also held the most water in the soil profile, likely due to much larger map units (309 ha each vs. 26 ha for SSURGO and 52 ha for taxonomic). Conversion of SSURGO into a taxonomy‐based input layer supports more exploratory research by reducing processing time while maintaining similar precision levels.