Assessment of Multi-Source Evapotranspiration Products over China Using Eddy Covariance Observations

Shijie Li,Peng Bai,Peng Deng,Guojie Wang,Jie Wang,Haishan Chen,Shujia Zhou,Yong Huang,Shanlei Sun

doi:10.3390/rs10111692

Abstract

As an essential variable in linking water, carbon, and energy cycles, evapotranspiration (ET) is difficult to measure. Remote sensing, reanalysis, and land surface model-based ET products offer comprehensive alternatives at different spatio-temporal intervals, but their performance varies. In this study, we selected four popular ET global products: The Global Land Evaporation Amsterdam Model version 3.0a (GLEAM3.0a), the Modern Era Retrospective-Analysis for Research and Applications-Land (MERRA-Land) project, the Global Land Data Assimilation System version 2.0 with the Noah model (GLDAS2.0-Noah) and the EartH2Observe ensemble (EartH2Observe-En). Then, we comprehensively evaluated the performance of these products over China using a stratification method, six validation criteria, and high-quality eddy covariance (EC) measurements at 12 sites. The aim of this research was to provide important quantitative information to improve and apply the ET models and to inform choices about the appropriate ET product for specific applications. Results showed that, within one stratification, the performance of each ET product based on a certain criterion differed among classifications of this stratification. Furthermore, the optimal ET (OET) among these products was identified by comparing the magnitudes of each criterion. Results suggested that, given a criterion (a stratification classification), the OETs varied among stratification classifications (the selected six criteria). In short, no product consistently performed best, according to the selected validation criterion. Thus, multi-source ET datasets should be employed in future studies to enhance confidence in ET-related conclusions.

Highlights

As an essential component of water balance, evapotranspiration (ET) can directly impact both regional and global hydrological processes
The negative Mean Error (ME) of EartH2Observe-En is mainly because of its underestimates in higher ET (Figure 3d), while the highest and the moderate ME (RME) for Modern Era Retrospective-Analysis for Research and Applications (MERRA)-Land and GLEAM3.0a are closely associated with systemic biases and overestimates in lower ET, respectively (Figure 3a,b)
Based on TS, the worst, the moderate, and the best overall performances in estimating monthly ET were found to correspond to the MERRA-Land, GLDAS2.0-Noah and EartH2Observe-En, and GLEAM3.0a products, respectively

Summary

Introduction

As an essential component of water balance, evapotranspiration (ET) can directly impact both regional and global hydrological processes. Numerous remote sensing [19,20,21,22,23], reanalysis [24,25,26,27], and land surface model (LSM)-based ET products [28,29,30], as well as estimates from empirical up-scaling of in situ observations [31] with different spatio-temporal resolutions and spans have recently been developed While these datasets provide an opportunity for use in long-term and large spatial ET-related studies, validations and inter-comparisons of the data are necessary. To reduce the impacts of ET product uncertainties on the degree of confidence for ET-related results (e.g., hydrological cycle, land-atmosphere interaction, agriculture, and ecosystem), we should assess the suitability of the ET products

Objectives

Results

Conclusion