Satellite–Reanalysis Contrasts in Extreme Rainfall around the Itaipu Reservoir, Brazil

Abstract Understanding extreme rainfall and long-term hydroclimatic variability is critical for hydrological modelling, climate diagnostics, and risk assessment in subtropical basins influenced by complex mesoscale and synoptic processes. This study evaluates the ability of satellite-derived and reanalysis precipitation products to reproduce observed mean rainfall, ETCCDI precipitation extremes, and long-term trends in the Paraná III Watershed surrounding the Itaipu Reservoir, southern Brazil, a strategically important hydropower region where reliable precipitation information is critical. Daily observations from 15 rain gauges spanning 1983 to 2024 provide an independent benchmark for assessing CHIRPS, PERSIANN-CDR, ERA5, ERA5-Land, and AgERA5 at annual and seasonal scales. Performance is quantified using Willmott’s Index of agreement, percent bias (Pbias), mean absolute error, and root mean square error, while extremes and trends are analyzed through ETCCDI indices and the Mann–Kendall test with Sen’s slope estimators. CHIRPS shows the highest overall accuracy for mean precipitation and multi-day extremes (RX1Day, RX3Day, RX5Day), whereas PERSIANN-CDR exhibits intermediate performance with persistent wet bias. Short-duration extremes are generally underestimated across datasets, and ERA5-based products systematically underestimate precipitation totals, although they preserve spatial coherence. Observations reveal a marked drying signal in PRCPTOT, primarily driven by austral autumn, together with a widespread reduction in wet-spell persistence. These low-frequency signals are consistently captured by ERA5, ERA5-Land, and AgERA5 but are largely absent in CHIRPS and PERSIANN-CDR. Overall, the results demonstrate a strong dependence of hydroclimatic interpretation on dataset selection and support the complementary use of CHIRPS for magnitude-focused analyses and ERA5-based products for the assessment of long-term variability and trends in hydropower-relevant regions. Graphical Abstract This graphical abstract provides a concise and visually organized summary of the study “Satellite–Reanalysis Contrasts in Extreme Rainfall around the Itaipu Reservoir, Brazil”. The figure summarizes the main methodological steps and key findings, illustrating precipitation datasets, ETCCDI indices, and long-term hydroclimatic signals in a simplified visual format. The objective, placed at the top, highlights the evaluation of satellite-derived and reanalysis precipitation products in capturing mean and extreme rainfall over the Paraná III Watershed (P3W) surrounding the Itaipu Reservoir, using daily observations from 15 rain gauges between 1983 and 2024. In the Methodology section, data sources are depicted by a rain-gauge icon for surface records and a satellite/globe icon for gridded products (CHIRPS, ERA5, ERA5-Land, AgERA5, and PERSIANN-CDR), supported by accuracy metrics (Willmott’s d, Pbias, MAE, RMSE). Analytical methods including ETCCDI indices, the Mann–Kendall test, Sen’s slope, and the linear model slope are represented by a trend graph symbol, emphasizing multi-scale annual and seasonal assessment of extremes and trends. The Key Results section illustrates that CHIRPS provides superior accuracy for precipitation magnitudes and multi-day extremes, whereas the ERA5 family exhibits systematic underestimation, but more consistently captures the direction and spatial structure of long-term drying. Observations show significant annual and autumn (MAM) PRCPTOT declines and widespread CWD contraction, while satellite products tend toward stationarity in trend fields. The Conclusion conveys that CHIRPS is best suited for magnitude-focused applications, whereas ERA5-based datasets better diagnose hydroclimatic change, supporting dataset selection for hydrological modelling and risk assessment in this strategic hydropower basin.