Application of SWAT Model to a Brazilian Semiarid Basin without Observed Discharge Data for Changing Scenarios of Land Use and Surface Cover

The increasing demand for water resources coupled with frequent droughts in the semi-arid region of Brazil, makes it necessary to understand the underlying hydrosedimentological processes for the best use of water resources. This can be done utilizing hydrological models such as SWAT. The Sucurú river basin (938.7 km²) located in the semi-arid region of the State of Paraíba (Brazil), has no observed runoff data but climatic data as well as land use and soil cover maps for three distinct periods (1990, 2002, 2013) are available. The basin is typical of many in the northeast of Brazil and hence a satisfactory modeling of this basin could serve as a reference for other basins in the region without runoff data. The precipitation data for the period 1994-2015 are available from 13 rain gauge stations. The land use/land cover data were generated from Landsat5 TM images (1990 and 2002) and Landsat7 (2013) for the following classes: Water, Bare Soil and Shrub Vegetation. The most influential model parameters on the runoff process were calibrated through trial and error. In the absence of observed flow data, the parameters were calibrated such that most of the annual runoff coefficients lay between 5 and 10 % which would correspond to the bulk of the years from lower than average and higher than average annual precipitation. The simulations were carried out for the period 1994-2015 with each of the three land use maps (1990, 2002, and 2013). The three periods were also separately simulated utilizing the soil use map applicable for the period: 1994 - 2001 (1990 map); 2002 - 2012 (2002 map) and 2013 - 2015 (2013 map). After several trials of adjustment, the parameters were adjusted satisfactorily. As a result, the simulations with the 1990 and 2002 soil use maps generated on the average, larger surface runoff than with the soil map of 2013. This was indeed to be expected as the vegetal cover in the map of 2013 is a lot more extensive with a significant reduction of bare soil areas. The Annual runoff coefficients were well within the range of 5 to 10%, and the simulations were highly consistent. The model also generated the annual soil erosion in the basin which was also qualitatively consistent. Thus the SWAT model can be considered as a very useful tool to simulate the runoff and erosion processes even in basins with no recorded flow data.