The Impact of LULC dynamics representation on SWAT+ modeling in the Pampa Region of South America

Global food demand has increased pressure on natural ecosystems and transform forested and grassland areas to intensive agriculture. This is the case in the Pampa region, that lost 15% of its natural grasslands to a complex and dynamic agricultural scenario, led by soybean production since the 2000s. However, agriculture is normally mapped at a low level of thematic detail, consisting in a broad class of annual agriculture, which does not consider the cropping practices and rotation systems, impairing the ability to represent these situations. To fill the gaps in LULC dynamics, the Vegetation Indexes, such as the EVI or NDVI presents as a possible solution, as they are available with a temporal resolution of around one to two weeks. The purpose of this study was to identify if the added use o NDVI could improve LULC dynamics description and provide better results for hydrologic modeling simulations. The new version of the Soil and Water Assessment Tool (SWAT+) was set up in the Ibicuí River (42.650 km²), an important regional hotspot for LULC change. LULC scenarios were: (a) SWAT+ default land use with single class for annual agriculture; and (b) seasonal LULC dynamics as captured by pairing NDVI datasets with available databases. The results for the uncalibrated model showed a better fit using the dynamic LULC description, as the single LULC class artificially created multiple peaks. Using SWAT+Toolbox to calibrate the models, both scenarios showed very good statistics or validation NSE = 0.78 and PBIAS = -6%, NSE = 0.81 and PBIAS = -1.8%, respectively. The water balance parameters were very close between the models, being the main difference in the Evapotranspiration and Revap components. Results shows that it is extremely important to correctly characterize the LULC dynamics, especially in ungauged basins, and that NDVI could help in this representation.