Search

This study evaluates the accuracy of daily reference evapotranspiration (ET o), computed according to the FAO Penman-Monteith equation by using a set of input weather variables obtained by blending ERA5-Land (ERA5L) reanalysis data with surface incoming solar radiation (R s) provided by the instruments on board the Meteosat geostationary satellites, operationally delivered by the Satellite Applications Facility on Climate Monitoring (CM-SAF). Performance assessment was carried out in Sicily (southern Italy) by using data from 38 automatic ground weather stations (AWSs) for years 2003–2020. ERA5L and CM-SAF data were first downscaled and bias-corrected with a calibration dataset; ERA5L air temperature data were also downscaled by lapse-rate correction. ET o estimates obtained with the blended ERA5L and CM-SAF validation dataset (ERA5L+CM-SAF) were compared with two other ET o estimates respectively obtained by using ERA5L and interpolated ground weather data (IGD). Performance indicators of the IGD dataset were evaluated by recursively applying universal kriging or ordinary kriging to the observed weather data, according to a cross-validation procedure. R s provided by CM-SAF outperformed R s obtained by ground interpolation, thus confirming the convenience of using bias-corrected CM-SAF data even when ground observations are available in the study area. ET o estimates with ERA5L+CM-SAF showed a normalized RMSE of 12%, outperforming ERA5L ET o estimates while performing comparably to ET o estimates obtained with the IGD dataset. The results suggested that the proposed blended dataset is a good proxy for interpolated ground weather observations in the assessment of ET o at regional scale when weather measurements cannot be easily gathered or in data-sparse regions. • Shortwave radiation by CM-SAF outperforms ERA5-Land shortwave radiation. • ET o is estimated by blending CM-SAF satellite-based radiation and ERA5-Land data. • ET o computed by the blended dataset is as accurate as ET o by interpolated ground data. • Blended datasets are good proxy for interpolated observations in data-sparse regions. • Shortwave radiation by CM-SAF outperforms spatially interpolated pyranometer data. [ABSTRACT FROM AUTHOR]