Evaluation of ERA5 and ERA5L Reanalysis Data in Alagoas, Brazil

The accuracy of global reanalysis datasets, such as ERA5 and ERA5L, is crucial for reliable numerical weather prediction and climate studies. This study evaluates the performance of these reanalyses by comparing their outputs with observational data from five automatic surface stations in Alagoas, Brazil, a tropical coastal region characterized by significant altitude variation. First, we correct for the altitude differences between the model grid and the stations altitude, which varied from -153 m to +87.4 m. The ERA5 and ERA5L temperatures are corrected based on the lapse-rate, while the pressure correction is based on the hydrostatic equation. Results reveal that altitude discrepancies in ERA5 and ERA5L models led to notable biases. The pressure bias in ERA5 was reduced from -17.7 hPa to -0.4 hPa, while temperature biases varied from -0.3°C to +0.7°C, and were more effectively corrected in ERA5L. The correlation between corrected model data and observations was high, with Pearson coefficients exceeding 0.99 for pressure and ranging from 0.88 to 0.96 for temperature. The study highlights that despite high correlations, residual systematic and random errors persist. Anomalous temperature jumps were observed in ERA5’s diurnal cycle, at times when new data assimilation cycles start. These could cause problems, for instance, for using this dataset as boundary conditions for a regional model in a weather forecast system. However, these were mitigated in ERA5L by the more detailed land scheme. Seasonal cycles of both pressure and temperature are both accurately captured by ERA5 and ERA5L. The findings underscore the significance of altitude adjustments in improving reanalysis accuracy and emphasize the need for continued validation of reanalysis products. Our results contribute to enhancing the representation of atmospheric variables in reanalysis models, which is critical for accurate climate modeling and weather prediction.