Comparing ERA5 and model data to observations for wind resource assessment – a case study from Tanzania

The ERA5 reanalysis is established as a key source of gridded wind-speed information for much of the world, having better performance and higher resolution (30km) than other reanalysis products. It’s used as a source of truth for driving wind-power models, and verifying ML predictions, especially in the absence of observational measurements. However, verification of ERA5 has mostly focused on Europe and the northern hemisphere, not examining performance in low-observation regions such as East Africa, where wind power investment and green energy provision is crucial to climate goals. As part of the FOCUS-AFRICA project, this study investigates how well the ERA5 reanalysis represents the climatology of 19 different observation sites in Tanzania and compares them to 3 CORDEX-Africa models driven by ERA-Interim at a similar resolution to ERA5, and one convection-permitting 4.4km resolution model (CP4A). ERA5 is shown to perform poorly at representing inland wind climatologies in Tanzania, with Perkins skill scores of 0.21-0.62 (1 is perfect), in comparison to European inland stations showing an average a score of ~0.8 in previous studies. This is caused by underestimations of mean wind speed compared to observations for inland sites.  The CP4A model performs best with scores of 0.54-0.79, despite the fact this model is not forced by real-world conditions.

These results show the need for caution when using ERA5 as a basis for any wind resource assessment or model validation. It also indicates that wind resource in East Africa may be underestimated, which would have negative impacts on investment decisions in the region.