Comparing observed and reanalysis data for trends in extreme temperature events in Brazil (1996–2022)

This study compares and contrasts two climate datasets, ERA5 reanalysis and the Integrated Surface Database (ISD) ground data, to assess their ability to identify extremes in temperature across Brazil from 1996 to 2022. Given the challenges posed by Brazil’s vast geographical size and uneven distribution of weather stations, the study aims to evaluate how these datasets capture climate extremes in different regions of the country. Key metrics such as Pearson correlation, Mean Absolute Error (MAE), Root Mean Square Error (RMSE), hit rates for extreme events, and the percentage of matching days at the 5th and 95th percentiles were used for comparison. The results reveal a high overall correlation ( r  = 0.89) between ERA5 and ISD, with regional variability, particularly in the northern and northeast regions. The ERA5 dataset showed a consistent increase in heat extremes across all regions, in line with global warming projections, whereas ISD data presented different trends in some regions. Cold extremes showed a general decline in both datasets, reflecting the impact of climate warming. Discrepancies between the datasets, particularly in extreme temperature predictions, highlight challenges associated with data sparsity, methodological differences, and modeling uncertainties. These findings offer critical insights into the accuracy and limitations of reanalysis and station-based data for monitoring climate extremes in Brazil, with implications for future climate research and policy.