Resumo:
Minas Gerais plays a major role in Brazil’s water resources due to its extensive river basins, which are essential for water supply and electricity generation. On the other hand, its extensive river network also makes the state susceptible to frequent extreme events, such as floods. Between 1991 and 2023, hydrological disasters affected millions of people in Minas Gerais, resulting in loss of life and significant economic damage. In this context, public centers specializing in the monitoring of extreme events were established, responsible for issuing hydroclimatic forecasts and alerts designed to support disaster prevention and response efforts. However, these operations generally focus on specific watersheds and areas previously defined as priorities, failing to comprehensively cover the entire state territory. Given this scenario, this study aims to contribute to expanding the capacity for regional-scale discharge forecasting through the integration of a regional atmospheric model and a hydrological model, applied to the main river basins of Minas Gerais. To this end, daily precipitation forecasts with a lead time of up to eight days, produced by the ETA regional atmospheric model operated by CPTEC/INPE, were used for the period from 2001 to 2024. These forecasts were employed as input data for the MGB-IPH distributed hydrological model to generate daily discharge forecasts up to eight days in advance. The accuracy of the forecasts was evaluated through visual analyses and the statistical indicators Mean Relative Error (MRE) and Mean Absolute Error (MAE). The results demonstrated that the ETA/MGB system was able to represent the observed discharge values and can be useful in supporting water resources management in the main basins of Minas Gerais. However, limitations related to the quality of hydrological data and uncertainties associated with precipitation estimates and forecasts still compromise part of the forecasts’ accuracy. Therefore, it is recommended that future studies incorporate precipitation correction techniques, more comprehensive statistical analyses, and comparisons between different atmospheric models, with the aim of improving the forecasting system’s performance.
