Effect of GIS data quality on small watershed stream flow and sediment simulations

Simulations of total runoff and fine sediment yield in Goodwin Creek watershed, which covers 21·3 km2 in Mississippi, were carried out using a hydrological model-GIS system. The system includes the recently released Soil and Water Assessment Tool (SWAT) model version 2000 and AVSWAT version 1·0, the supporting interface with ArcView GIS. Among the required GIS input, some are commonly available in the United States with multiple options and characteristics. In our study, two available digital elevation models, three land use-land cover maps and two soil maps were grouped in all possible ways to obtain 12 applied input combinations. The objective of this study was to assess the impact of GIS input variation on the uncalibrated water runoff and sediment yield outputs and compare them with the respective observed data. The implicated issues are significant wherever multiple choices of GIS input are available. In the United States, agencies are developing TMDL (total maximum daily load) programmes at the watershed scale and are also using supporting tools along with the available GIS data. In addition, the involved water quality appraisals often include assessment of limited size watersheds, i.e. draining into a specific stream segment. This watershed, operated by the United States Department of Agriculture, Agriculture Research Service, is highly instrumented, thereby representing a severe test and a primary verification of the new system. The GIS data had a varying impact on model results. DEM choice was critical for a realistic definition of the watershed and subwatershed boundaries and topographic input, and consequently simulated outputs. Land use-land cover maps had a significant effect on both runoff and sediment yield prediction. Soil maps showed a limited influence on model results. While evidences and basic justifications of the results are provided, further investigations are needed to determine the influence of the input GIS data distribution on watersheds with various sizes, geomorphological and spatial settings.