Hydrological performance evaluation of temperature reanalysis products for the Ouémé River Basin in West Africa

West Africa has a complex climate regime. It affects hydrological predictivity in the region where the majority population depends on agriculture. With a warming planet, the challenge is further exacerbated by frequent hydroclimatic extremes. To achieve secured livelihoods and resilience, hydrological understanding is a key. However, there is an elemental challenge of missing measured weather data. Weather variables that are drivers of the water and energy balance are necessary for the setup of robust hydrological models. We focus on the Ouémé River Basin in Benin, which lacks spatially representative in-situ temperature observations. To fill this gap, the study evaluates global earth datasets in the form of reanalysis products that are emerging useful for hydrological modeling. We perform an intercomparison of five temperature reanalysis datasets for the basin using the hydrological model Soil and Water Assessment Tool (SWAT). These datasets are CFSR, CPC, ERA5, EWEMBI, and PGFv3, available at a daily temporal resolution. We test their performance on the simulation of hydrological processes in the Ouémé basin.

To evaluate each temperature data, a multi-site calibration is performed in SWAT using daily discharge time series. Validation is carried out as a two-fold process. The first is point validation performed using discharge data at five gauge sites and the second is spatial validation on the sub-catchment level conducted using satellite-derived actual evapotranspiration (AET) data from GLEAM v3.5b. This multi-gauge and multi-variable approach is used to minimize uncertainties associated with the application of SWAT.

This study is one of a kind for the basin, testing the datasets for their hydrological performance and overcoming a major gap toward achieving robust models. Temperature reanalysis products provide high temporal resolution, long time series, and spatially representative datasets. However, the response to input data errors can vary significantly given the non-linear interaction of parameters in a hydrological model. Therefore, hydrological evaluation is an important step before reanalysis data can be used for modeling and decision-making. We also demonstrate the significance of testing multiple water fluxes to assess the performance of climate datasets. A higher variation in performance for temperature datasets is observed for AET than for the streamflow component. It is an important outcome to determine the most suitable temperature product for the Ouémé basin.