Search

Gridded climate datasets such as satellite and reanalysis products have biases related to the methods used to develop them. This study aims to improve the hourly rainfall distribution of the WFDE5 reanalysis over all grid boxes in Benin, based on bias adjustment through Quantile Mapping (QM). The bias adjusted product (called NEW) was evaluated on an hourly scale, using the percentage and quantity of precipitation events per modality class ([0–1), [1–2), [2–3), [3–4), [4–5) and ≥ 5 mm/h) and the relative mean absolute error (RMAE). On a daily scale, evaluation was based on Pearson correlation and RMAE values using nine extreme precipitation indices. The mean absolute error (MAE) and Mann Kendall trend are sometimes shown. Our result showed that assimilated rainfall from WFDE5 performed well at seasonal scales (RMAE < 15%) but underperformed at the hourly scale (RMAE sometimes > 400%). NEW offers RMAE values generally < 100%, i.e. an almost four-fold reduction in bias. The QM method improves the rainfall distribution, particularly with regard to the percentage of rainfall between 0 and 1 mm/h (WFDE5 = 95.40%, Obs = 97.87% and NEW = 97.83% at the Nalohou station) and the cumulative rainfall quantity greater than 5 mm/h (WFDE5 = 131 mm/year vs. Obs = 880 mm/year and NEW = 790 mm/year at the Nalohou station). The bias adjustment also significantly improved the description of climate extremes in Benin, particularly in terms of bias. At Cotonou station, WFDE5 was associated with an average RMAE of 61% for the nine indices, compared to 33% for NEW. Finally, NEW presents mean values closer to observation data, and can be used for hydrological impact studies in Benin. [ABSTRACT FROM AUTHOR]

In West Africa, the validation of distributed models is limited by the quality and availability of point station data measured in-situ. ERA5 is a climate reanalysis produced by European Centre for Medium-range Weather Forecasts (ECMWF) and suggested to overcome this constraint. This study assessed and compared over the Benin basins at spatial and monthly time scale, the quality of ERA5 and its variant ERA5-Land (namely LAND). ERA5 relies on the single-levels version with 0.25° x 0.25° resolution while LAND is the land surface version with 0.1° x 0.1° resolution. Four variables were collected including runoff, evapotranspiration (ETR), water table depth (WTD), and soil water content (SWC). Point station data were analyzed using the correlation performance evaluators, Mean Absolute Error (m) and Relative Mean Absolute Error (r). The results showed that LAND simulates well the peaks of mean runoff. It showed the best runoff performance in terms of correlation (~0.61) compared with ERA5 (correlation ~0.49). Both reanalysis showed high correlations (generally > 0.80) for SWC, but the correlations obtained from ETR are slightly lower (ERA5~0.58 vs. ERA5-Land~0.54). Correlations were below 0.5 on both reanalyses for WTD with slight overestimation (m=4.73 m for ERA5 vs. m=3.13 m for LAND). This study does not identify any reanalysis that is better than another, both spatially and monthly scale. Nevertheless, this study indicated that the choice of reanalyses must rely on their performance and the given water cycle element. Correcting the variables of these reanalysis could also improve their performance. [ABSTRACT FROM AUTHOR]