Influência de entradas pedológicas e pluviométricas em modelagens de vazão com o modelo SWAT : Bacia do Rio São Lourenço-MT

Abstract

This research is an assessment of the influence of spatial and attribute detailing of pedological information on the performance of streamflow simulations using the SWAT model (Soil and Water Assessment Tool). In a case study in the São Lourenço river basin, comparing daily and monthly resolution, fed with three different spatial discretizations of pedological information from secondary sources, available for most Brazilian river basins from their respective Ecological-Economic Zoning projects and the EMBRAPA national soil database. To ensure comparability between the models, the SWAT was initially fed with a single set of input data, leaving only the discretization level of the mapping units and the way of aggregating the pedological information of profiles available for each unit variable. Four sources of rainfall data were used: the GPM and ERA5 Land global grids, and the physical gauges of ANA without and after a spatial interpolation to the parameterized sub-basins. Daily and monthly stream flow simulations were validated using statistical metrics such as coefficient of determination (R2 ), residual mass coefficient (CMR) and Nash-Sutcliffe Efficiency (NSE). The results showed that none of the two more detailed soil parameterizations improves the modeling performance when compared with the more generalized pedological inputs. The non-calibrated models using ANA's interpolated precipitation data showed the best agreement with the observed flow. Achieving a similar accuracy, the ERA5 Land outperformed the GPM dataset. The simulations of all input data sets resulted in a better performance in a monthly than a daily temporal resolution.