Abstract. Estimating evapotranspiration (ET) beyond the local or point scale is critical for water resources and ecosystem studies. Remote sensing offers a unique advantage by enabling ET monitoring at larger spatial scales than in-situ instruments. By leveraging relationships between surface biophysical parameters and terrestrial thermal emission, continuous ET can be retrieved across diverse landscapes. Herein, we apply the EVapotranspiration Assessment from SPAce (EVASPA) contextual tool over southern France, using MODIS-derived land surface temperature / emissivity (LST/E), NDVI and albedo products. The dataset spans 2004–2024, yielding 972 instantaneous ET estimates. The EVASPA ensemble integrates multiple member outputs generated from: 1) alternative formulations of the evaporative fraction (EF) and ground heat flux (G), and 2) different LST and radiation inputs. Evaluation against flux tower data shows that even a simple ensemble average provides reasonable agreement, though individual member performance varies substantially. Uncertainty analyses were also performed where we looked at how each of the distinct variables (i.e., LST, radiation, evaporative fraction (EF), and ground heat flux (G) methods) influenced the modelled ET. The analyses reveal that LST inputs and EF formulations are the dominant sources of variability, with seasonal dependence–uncertainties peak during summer (tending to follow the annual cycle of radiation) and are partly influenced by satellite characteristics. Generally the satellite's overpass time introduces more incertitude to the gap filled daily ET estimates compared to the LST/LSE separation methods; as such, the uncertainties in LST could, by extension, be partially attributed to uncertainties in the radiations during the acquisition time. Radiation inputs also contribute to the variations in the ensemble, while G flux methods exert comparatively minor influence, especially for estimates derived from TERRA morning overpasses. Overall, our results demonstrate that ensemble-based contextual modelling can provide both reliable flux estimates and a meaningful uncertainty spread. By allowing optimal member selection according to surface and climatic conditions, ensemble modelling using EVASPA enhances ET retrieval robustness thus providing more resilient and informative estimates. Such ensemble frameworks are especially valuable for forthcoming missions like TRISHNA, where consistent and accurate, high-resolution ET monitoring will be crucial for operational water and ecosystem management.
Mwangi et al. (Mon,) studied this question.