Advances in validation of remotely sensed land surface evapotranspiration

Abstract

Evapotranspiration (ET) is an important component in the soil-vegetation-atmosphere continuum. Remotely Sensed ET (RS-ET) provides multi-scale and spatiotemporally continuous information over the land surface, and has become an effective approach to obtain ET. Due to the heterogeneity of land surface and complexity of meteorological conditions at the near-surface layer, there exist various uncertainties derived from the model mechanism, parameterization scheme, input data and time scale conversion, which hindered correct estimations of ET, and further effect on its application. Therefore, it is essential to validate RS-ET to optimize the models and improve the associated products. This paper evaluated a group of validation methods for RS-ET (including evaporation and transpiration), which usually consists of direct validation and indirect validation. An overview of the principles, applicability, advantages, and disadvantages for all the validation methods were summarized. Direct validation is based on in situ measurements (including (micro-) lysimeter, stem sap flow, bowen ratio energy balance system, eddy covariance, and scintillator) to get the ground truth value, which can be used as the primary and reliable method to validate RS-ET and usually employed at the pixel and regional (or basin) scales. In the absence of ground truth ET, indirect validation becomes feasible, which can be classified into (1) cross-validation, (2) Multi-scale validation based on high spatial resolution remote sensing data, and (3) spatiotemporal variation analysis that combines multiple ET impact factors. Nevertheless, there are still a series of theoretical and methodological challenges in the validation of RS-ET, such as the scale mismatch between in situ measurement and remote sensing pixels due to the land surface heterogeneity. It is well-acknowledged that how to get the ground truth value at pixel and regional scales is the core issue of validation. This study demonstrated that validations of RS-ET products can be not only applied over homogeneous land surface but also heterogeneous surface with further development, which may at least but not limited to quantification of the spatial heterogeneity of land surface hydrothermal conditions, optimization of the experimental sites for validation over heterogeneous land surface, multi-scale measurements of ET on heterogeneous surface, acquisition of ground truth ET at pixel and regional scales, validation demonstration and uncertainty analyses of the validation process. Moreover, this study also proposed a generalized validation framework to validate RS_ET products at different scales (pixel scale and regional scale), which included direct validation (as the priority method) and indirect validation methods (as the auxiliary method), multiple validation data (i.e., ground truth ET at the pixel and regional scale, ET reanalysis data, various ET products, estimated ET from models and ET impact factors). The current framework focused on evaluating the accuracy and the spatiotemporal variations, identifying the error sources of the RS-ET products and analyzing the uncertainties during the validation process. This work is expected to improve the land surface remote sensing products and promote the development of quantitative remote sensing science.