Assessing uncertainties in satellite-based estimation of evapotranspiration

Abstract

Evapotranspiration (ET) is a major water flux in the terrestrial water balance and a key link between water and surface energy balances. In water sciences and management, the quantification of ET is required but challenging to gauge in situ, leading to the popularity of models based on satellite-derived data. However, uncertainties in satellite-based estimation arise from both methodological and technical factors. This study examines and assesses uncertainties in satellite-based estimation of ET. Part I provides a systematic quantitative literature review, showing the diversity of approaches and constraints arising from the availability and quality of reference data. A meta-analysis of in-situ validations against eddy covariance measurements quantifies the status of uncertainty in terms of reported performance metrics. Part II focuses on the assessment of a satellite-based ET data product for monitoring water productivity from field to global scales. Technical uncertainties through ex-ante and ex-post methods, including error propagation, in-situ validation, and triple collocation analysis are provided. The results highlight spatial variability in uncertainty, limitations of validation data, and challenges in dry and tropical regions providing guidance to users of such products. Finally, this thesis reflects on methodological uncertainties arising from problem framings, model choices, and configurations.