Validation of precipitation data generated by ERA5 reanalysis for the Mirim-São Gonçalo watershed, Brazil

Abstract

The transboundary Mirim-São Gonçalo watershed (MSGW) is of a paramount importance for the development of its region, mainly for the agricultural sector. Obtaining climate data is fundamental in the management of water resources, as it allows estimating missing data, making long-term forecasts and checking possible trends. The European Center for Medium-Range Weather Forecasts (ECMWF) produces numerical weather forecasts, being the fifth generation of reanalysis (ERA5) the most recent. This work aimed to evaluate the efficiency of ERA5 in representing precipitation for the Brazilian side of MSGW, comparing its grid points data with the observed ones from rain gauges. For this purpose, the ERA5 and observed monthly climatological 1981-2020 rain data of the MSGW was calculated. The stationarity of the data was evaluated by applying the Mann-Kendall’s (MK) and Sen’s Slope (SS) tests, while the data homogeneity was verified with the Pettitt (Pt) test. For data validation, the Mean Absolute Percentage Error (MAPE), Root Mean Squared Error (RMSE) and Pearson's correlation coefficient (rpearson) were calculated. MK and SS tests verified that there is no significant trend in the majority of the precipitation series, while the Pt test pointed out that there is no significant changing point (5% of significance level). The error measures (MAPE and RMSE) were low while the rpearson presented high correlation values (strong and positive), indicating the good performance and the potential of the ERA5 model in representing the precipitation data in the MSGW.