Evaluation of remotely-sensed evapotranspiration datasets at different spatial and temporal scales at forest and grassland sites in Italy

Remote sensing measurements offer great potential to quantify evapotranspiration (ET) more accurately. This work assesses several remotely-sensed ET datasets at high, medium and coarse scale through a comparison with eddy covariance (EC) measurements. In particular, daily retrievals obtained from the Sen-ET model and the ECOSTRESS ET product, both having resolutions of tens of meters, were considered. In addition, 3-km daily ET estimates derived from SEVIRI were analyzed, together with 8-day MODIS ET composites at 500-m spatial resolution. The remotely-sensed actual ET retrievals at medium and coarse scale were compared with EC observations at five forest and grassland sites over different climate zones in Italy. Additionally, for one of these sites (Monte Bondone), the high-resolution ET estimates within the EC footprint were also evaluated. Sen-ET retrievals demonstrated a remarkable ability to describe EC latent heat fluxes (e.g., Pearson correlation coefficient, R, equal to 0.83, and unbiased root-mean-square-difference, ubRMSD, equal to 0.85 mm/day), and therefore great potential for producing high-resolution ET estimates, while ECOSTRESS data were less performant (e.g., R = 0.46, and ubRMSD = 1.65 mm/day). Despite the spatial differences between EC footprint size and the medium along with coarse scale of remotely-sensed datasets, the latter both showed a good agreement with the measured EC fluxes. The SEVIRI ET, in particular, proved to be significantly suitable for the ET description also at a local scale (at daily time step, average R = 0.84, and average ubRMSD = 0.83 mm/day over the study sites), despite a tendency to underestimate in-situ measurements (negative biases ranging between 0.2 and 0.85 mm/day were observed). This study contributes to the topic of evaluating specific remotely-sensed ET datasets under different conditions. The obtained results generally confirm the value of satellite-derived retrievals for ET estimate and monitoring, as well as their potential to improve ET modeling.