Modelagem hidrológica na bacia do Rio Piquiri no estado do Paraná

Abstract

Due to the recurring scenario of droughts that have affected the central-southern region of Brazil in the last 10 years and the increasing demand for water resources, the use of environmental modeling tools through prediction software gains prominence, including SWAT+ (Soil and Water Assessment Tool). This study aimed to obtain a spatiotemporal model capable of simulating the streamflow of the Piquiri River basin, located in the central-western region of the State of Paraná. Thus, the characterization of the study area was necessary, encompassing physical data such as topography, soil, land use and land cover, as well as climatological data of precipitation, wind, humidity, temperature, and solar radiation. This study was divided into 3 stages: 1st - obtaining and preparing input data for software use; 2nd - preprocessing to delimit the basin and hydrological response units; and 3rd - processing involving calibration, validation, and simulation of monthly streamflow. For the construction of the database necessary for the development of the SWAT+ model, soil data from SiBCS of EMBRAPA were used, climatological data were extracted from the IDR-Paraná system and CFSR provided by SWAT+ itself. In the model development, 1622 sub-basins and 10828 hydrological response units were generated in an area of 20,925.34 km². Among all existing streamflow stations in the study area, the Balsa Santa Rita station was chosen as the watershed outlet and as a control point for calibration and validation. To identify the most sensitive parameters as well as the most suitable limits for model calibration, 10 different calibration attempts were made, with the best calibration result for the period between 2003 and 2006 with a Nash-Sutcliffe index of 0.441. After obtaining the most suitable limits to maximize the model's efficiency, a calibrated model for the monthly streamflow of the Piquiri River basin was achieved, reaching a Nash-Sutcliffe efficiency index equivalent to 0.55, considered satisfactory. However, the model overestimated the average streamflow by about 38% as indicated by the PBIAS index. Parameters K, AWC, CHK, CHN, and MSK_CO2 were identified as the most sensitive, and the main cause for the low model efficiency can be attributed to the limited availability of error-free climatic data with long series for the evaluated region for all variables required by SWAT+, as well as data series like evapotranspiration and sediment that could be compared to generated data to serve as calibration parameters.