Evaluation of gridded and reanalysis vapour pressure deficit datasets over East Africa

Gridded TerraClimate and ERA5-Land reanalysis datasets provide high-resolution climate data of either direct vapour pressure deficit (VPD) or its input variables at 0.04 and 0.1° spatial resolution respectively, and over extended time periods. However, there are gaps in their applicability over data-sparse regions such as East Africa. This study aims to (i) validate VPD from both gridded TerraClimate VPD and ERA5-Land reanalysis using station-derived VPD for the period 2014–2023 and (ii) compare and contrast the two datasets at a monthly timescale over the period 1984–2023 in a spatiotemporal context. We employed the root mean square error (RMSE), scatter index (SI) and correlation coefficients for accuracy evaluation against 20 meteorological stations across East Africa. For spatiotemporal comparison between the two datasets, we used the Mann–Kendall trend test, Student’s t-test and correlation analysis. Results showed that ERA5-Land reanalysis generally performed better across most parts of East Africa, as evidenced by strong correlations and low SI scores ranging within 0–30% for 95% of the stations, compared to a good-fair performance of gridded TerraClimate with SI scores of 0–50% for 95% of the stations. Finally, comparative analysis revealed strong agreement between the two datasets on trend magnitude, direction and significance as well as June-to-August seasonal means. Additionally, very strong positive correlations (0.75–1.0) were observed for most parts of East Africa, except over areas dominated by very high humidity such as mountainous areas and those surrounding lakes. Finer temporal- and spatial-scale studies are recommended, with emphasis on exploring the added value of combining the two datasets.