Performance evaluation of ERA5, MERRA 2 and PERSIANN-CDR reanalysis grids in the Tambo basin.

Grid-based meteorological data products, such as reanalysis data, are essential for overcoming data scarcity in sparse and uneven hydrometeorological networks, especially in developing countries. However, these products require detailed validation to ensure their accuracy and reliability. This research evaluates the performance of ERA5, MERRA-2, and PERSIANN-CDR products in the Tambo basin, using monthly precipitation data from meteorological stations for the period 1985-2019 and applying two approaches: point values from stations and areal averages. The performance metrics considered are the Pearson correlation coefficient (CC) and percentage bias (PBIAS). In the station-based evaluation, ERA5 demonstrated excellent temporal correlation (with a CC between 0.83 and 0.91), but there was significant overestimation of precipitation (PBIAS between 72.0% and 217.5%). PERSIANN-CDR also overestimated precipitation (with a PBIAS between 8.0% and 86.9%) and showed low temporal correlation (with a CC between -0.14 and 0.15). In contrast, MERRA-2, despite underestimating precipitation (with a PBIAS between -30.8% and -77.9%), showed good temporal correlation (CC between 0.72 and 0.82). In the areal average evaluation, ERA5 had the best correlation (with a CC of 0.94) and a high bias (106.8%). MERRA-2 exhibited a low bias (a PBIAS of -53.5%) and a slightly lower correlation (0.88). PERSIANN-CDR had moderate bias (24.5%) and an intermediate correlation (0.70). These findings highlight the need to improve accuracy and reduce bias in products like ERA5 to optimize their use in hydrological applications and water resource management, thereby contributing to better forecasting and planning in water-scarce regions and helping to mitigate the impacts of climate change, especially in vulnerable areas.