Evaluation of reanalysis estimates of precipitation, radiation and temperature over Benin (West Africa)

Abstract

In West Africa, climatic data issues, especially availability and quality, remain a significant constraint to the development and application of distributed hydrological modeling. As alternatives to ground-based observations, reanalysis products have received increasing attention in recent years. This study aims to evaluate three reanalysis products, namely, ERA5, WFDE5, and MERRA2, from 1981 to 2019 to determine their ability to represent four hydrological climates variables over a range of space and timescales in Benin. The variables from the reanalysis products are compared to point station data-based metrics Kling Gupta Efficiency (KGE), Mean Absolute Error (MAE), correlation, and Relative Error in Precipitation Annual (REPA). The results show that ERA5 presents a better correlation for annual mean temperature (between 0.74-0.90) compared to WFDE5 (0.63-0.78) and MERRA2 (0.25-0.65). Both ERA5 and WFDE5 are able to reproduce the observed upward trend of temperature (0.2°C/decade) in the region. We noted a systematic cold bias of ~1.3°C in all reanalyses except WFDE5 (~0.1°C). On the monthly timescale, the temperature of the region is better reproduced by ERA5 and WFDE5 (KGE ≥ 0.80) compared to MERRA2 (KGE < 0.5). At all timescales, WFDE5 produces the best MAE scores for longwave (LW) and shortwave (SW) radiation followed by ERA5. WFDE5 also provides the best estimates for the annual precipitation (REPA ∈ ]-25, 25[ and KGE ≥ 50% at most stations). ERA5 produces similar results, but MERRA2 performs poorly in all the metrics. Additionally, ERA5 and WFDE5 reproduce the bimodal rainfall regime in southern Benin, unlike MERRA2, but all products have too many small rainfall events.