Revealing the spatiotemporal precipitation patterns of ECMWF fifth-generation reanalyses since the mid-20th century in West-Africa

Reanalysis datasets are a viable alternative to ensure the continuity of hydrological studies and to assess climate variability in regions where the availability and/or quality of observational data is limited. This study revealed the performance of the fifth generation of the European Centre for Medium-Range Weather Forecasts (ECMWF) reanalyses, namely ERA5 (0.25°x0.25°) and ERA5-Land (0.1°x0.1°), over West Africa. They were analyzed on spatial and temporal scales (annual, monthly, daily, and hourly). The reanalysis time series were obtained by Selecting the Nearest Pixel (SNP) closest to the point station or by the Inverse Distance Weighted (IDW) method and compared with the observational data using Pearson correlation (c) and Relative Mean Absolute Error (RMAE). The results showed on average similar performance between ERA5 and ERA5-Land reanalyses, and also between SNP and IDW methods. However, a significant difference can be observed at some stations/areas due to the influence of topography and wind. Both reanalyses performed well on a spatial scale, with a clear distinction between the wet and dry areas. They also performed well on an annual (c = 0.60 and RMAE=22 %) and monthly (c = 0.80 and RMAE=40 %) scales. For 713 point stations, ERA5 and ERA5-Land showed negative trends in interannual variability for 82 % and 94 %, respectively. They agreed with the trends derived from observations (negative trends for 96 % of stations). However, the reanalyses performed poorly on both daily (c = 0.21 and RMAE=128 %) and hourly (c = 0.06 and RMAE=170 %) scales. Maximum daily precipitation (RX1day) is also less well represented, sometimes with negative correlations. The ECMWF fifth generation reanalyses need to be adjusted to improve their performance in describing precipitation.