Long-term hydrological modeling in the Tocantins river basin upstream Serra da Mesa hydropower plant

Abstract

Hydrological models are important tools for water resource management, and their applicability is directly constrained by the availability and quality of precipitation time series. The adequate precipitation measurement is challenging due to its spatial and temporal variability, especially where the monitoring network is sparse. Moreover, precipitation time series must be reliable in areas with hydrological and ecosystem relevance, such as the Cerrado biome, one of the world's hotspots. In this context, this study aimed to evaluate the applicability of precipitation products obtained from different methodologies for hydrological modeling purposes in the Tocantins-Araguaia watershed using the Soil and Water Assessment Tool (SWAT) model. For this purpose, the following analyses were performed: i) Calibration and validation of satellite-estimated, point-by-pixel precipitation series; ii) simulation of future scenarios with calibrated parameters using interpolated precipitation data; iii) simulation of a long streamflow series, designated here as hydrological reanalysis, for the 20th century using climate reanalysis. The study site is located in the Cerrado with the control section defined at the Serra da Mesa hydroelectric power plant, comprising a drainage area of 51,237 km2. The precipitation datasets evaluated were: Satellite Precipitation Products - SPP (TMPA and IMERG); Interpolated Observed Data (IOD); and Climate Reanalysis (ERA-20CM, ERA-20C, and 20CRv3). The highest correlation and lowest errors between rain gauges and datasets were obtained at the monthly scale compared to the daily scale, in which SPP represented precipitation more adequately than the other products. The SPPs showed "very good" results for the monthly streamflow simulation, and IMERG stood out as a potential substitute for the TRMM in hydrologic modeling applications. The streamflow simulation from the IOD was the one that presented the best statistical indexes, and thus its calibrated parameters were used for the simulation of future scenarios combined with the Regional Climate Models Eta-HadGEM-ES and Eta-MIROC5. The results showed a possible reduction of up to 70% in the streamflow for the end of the XXI century, while runoff could reduce 54% in the first time slice. ERA-20CM presented lower quality results among the reanalysis, explained by the absence of the data assimilation of observations. Although ERA-20C presented superior results to ERA-20CM, o 20CRv3 was the only one capable of satisfactorily simulating the streamflow in the validation period and was used for the hydrological reanalysis of the 20th century. According to Pettitt's analysis, the main change point occurred in the 1970s, and the lowest average streamflow was observed during the first half of the 20th century. In the streamflow series derived from the hydrological reanalysis was identified extreme flood and drought events. However, some important events were not captured. Overall, SPP, IOD, and the 20CRv3 presented reliability as alternative input data for hydrological models and allowed to recreate past and future information besides overcoming the limitations of the spatial and temporal coverage of conventional monitoring.