Field-Level Targeting Using SWAT: Mapping Output from HRUs to Fields and Assessing Limitations of GIS Input Data

Abstract

Soil erosion from agricultural fields is a fundamental water quality and quantity concern throughout the U.S. Watershed models can help target general areas where soil conservation measures are needed, but they have been less effective at making field-level recommendations. The objectives of this study were to demonstrate a method of field-scale targeting using ArcSWAT and to assess the impact of topography, soil, land use, and land management source data on field-scale targeting results. The study was implemented in Black Kettle Creek watershed (7,818 ha) in south-central Kansas. An ArcGIS toolbar was developed to post-process SWAT hydrologic response unit (HRU) output to generate sediment yields for individual fields. The relative impact of each input data source on field-level targeting was assessed by comparing ranked lists of fields on the basis of modeled sediment-yield density (Mg ha-1) from each data-source scenario. Baseline data of field-reconnaissance land use and management were compared to NASS and NLCD data, 10 m DEM topography were compared to 30 m, and SSURGO soil data were compared to STATSGO. Misclassification of cropland as pasture by NASS and aggregation of all cropland types to a single category by NLCD led to as much as 75% and 82% disagreement, respectively, in fields identified as having the greatest sediment-yield densities. Neither NASS nor NLCD data include land management data (such as terraces, contour farming, or no-till), but such inclusion changed targeted fields by as much as 71%. Impacts of 10 m versus 30 m DEM topographic data and STATSGO versus SSURGO soil data altered the fields targeted as having the highest sediment-yield densities to a lesser extent (about 10% to 25%). SWAT results post-processed to field boundaries were demonstrated to be useful for field-scale targeting. However, use of incorrect source data directly translated into incorrect field-level sediment-yield ranking, and thus incorrect field targeting. Sensitivity was greatest for land use data source, followed closely by inclusion of land management practices, with less sensitivity to topographic and soil data sources.

Hydrological Responses to Various Land Use, Soil and Weather Inputs in Northern Lake Erie Basin in Canada

“…Model performance criteria (statistics) recommended by Moriasi et al [43] were used to evaluate the performance of the model for compared scenarios. In this study, statistics of percent bias (PBIAS) and Nash-Sutcliffe efficiency (NSE) were used, as these two statistics were used in various hydrological studies [12,21,[44][45][46] In addition, graphical comparisons of time variable plots and flow duration curves (FDCs) were also used to compare simulated and observed flows for different scenarios. Graphical plots provided greater insights into hydrograph representation, and low flow and high flow comparisons.…”

“…Most Ontarians found that the two land use layers studied resulted in greater differences compared to the two soil layers studied. Daggupati et al [21] compared the impacts of land use (field-reconnaissance land use layer vs. National Agricultural Statistics Service (NASS) vs. National Land Cover Dataset (NLCD data, topography (10 m vs. 30 m DEM) and soils (SSURGO vs. STATSGO) on sediment yields. They found out that the sediment yield outputs were greatly sensitive towards the land use data source and was less sensitive towards topographic and soil data sources.…”

Application of Soil and Water Assessment Tool Model to Estimate Sediment Yield in Kaw Lake

“…As a physically based hydrological model, SWAT requires a great deal of input data (Daggupati et al, 2011;Hosseini et al, 2011). Major input datasets include topography, soils, land use/land cover data and management practices, weather and hydrography.…”

Predicting Ephemeral Gully Location and Length Using Topographic Index Models

“…Daggupati et al (2011) reported that the land use data source has a major impact on modeling results for field-scale targeting, so this study evaluated the impacts on topographic index model performance of masking agricultural lands using either USDA National Agricultural Statistical Service (NASS) data or field reconnaissance land use data. Field reconnaissance land use data were collected during EG field reconnaissance in 2011, described above.…”