Impact of Land Use and Land Cover Categorical Uncertainty on SWAT Hydrologic Modeling

Abstract

Abstract. The land use and land cover (LULC) map of a watershed is a critical input to the Soil and Water Assessment Tool (SWAT) model. LULC is a categorical geospatial data layer that is typically developed based on models that establish relationships between pixel-based spectral reflectance and corresponding ground-truth information. Hence, LULC maps, like other classified remote-sensing datasets, are subject to error, which varies for each LULC category. The purpose of this study was to evaluate the effect of published LULC categorical errors on SWAT model uncertainty. A new algorithm was developed within the SWAT2009_LUC tool framework [see Transactions of the ASABE 54(5): 1649-1658] to produce multiple realizations of the LULC layer based on LULC categorical errors and integrate them dynamically within SWAT simulations. The enhanced SWAT2009_LUC tool and algorithm were tested for the SWAT model of the Illinois River Drainage Area in Arkansas (IRDAA) watershed. Tool evaluation results showed that the algorithm successfully integrated the LULC realizations within SWAT runs. Uncertainty evaluation showed that LULC categorical errors produced deviations in water yield output ranging from 0% to 8% at an annual scale and 0% to 19.9% at a monthly scale for the IRDAA subwatersheds. Results from this research highlight the importance of LULC categorical accuracy in the SWAT model and provide a generic tool for examining this uncertainty in any other watershed.