A global implementation of single- and dual-source surface energy balance models for estimating actual evapotranspiration at 30-m resolution using Google Earth Engine

Authors: JAFAAR, H. - American University Of Beirut; MOURAD, R. - American University Of Beirut; Kustas, William - Bill; Anderson, Martha

Submitted to: Water Resources Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/31/2022
Publication Date: 11/2/2022
Citation: Jafaar, H., Mourad, R., Kustas, W.P., Anderson, M.C. 2022. A global implementation of single- and dual-source surface energy balance models for estimating actual evapotranspiration at 30-m resolution using Google Earth Engine. Water Resources Research. 58. Article e2022WR032800. https://doi.org/10.1029/2022WR032800.
DOI: https://doi.org/10.1029/2022WR032800

Interpretive Summary: Accurate and frequent estimation of actual plant water use or evapotranspiration (ET) at sub-field scales provides essential information for agricultural water management and is critical in quantifying the earth’s water, energy, and carbon cycles. Therefore, remote sensing of ET has become one of the most effective methods for ET mapping, complementing point-based ET observational methods. This study presents a first implementation of the Two-Source Energy Balance (TSEB) model within Google Earth Engine (GEE) environment using satellite data applied globally and compared to 33 tower sites measuring ET across the US, Europe, and Australia. In addition, a model inter-comparison is conducted between TSEB and a single-source energy balance model HSEB (Hybrid Surface Energy Balance). Both models produced satisfactory estimates of daily ET for all biomes combined, with comparable results at weekly and monthly timescales. The TSEB model tended to perform better for croplands while HSEB had slightly better performance over natural ecosystems. Based on these results, HSEB and TSEB ET models show tremendous potential for water resource management applications on a global scale using GEE.

Technical Abstract: Evapotranspiration (ET) provides a robust connection between hydrological cycles and surface energy. Remote sensing of ET has substantially developed over the last few years, allowing spatial and temporal assessments at unprecedented resolutions. Accurate and near-daily continuous ET estimation is critical for multiple water resources and agricultural management applications, water budget studies, and crop yield and drought monitoring. This study presents a first implementation of the Two-Source Energy Balance Model (TSEB-PT) within Google Earth Engine (GEE) environment. The algorithm uses the harmonized Landsat-Sentinel earth observations, 20-30 m sharpened MODIS land surface temperatures, along with remotely sensed vegetation cover information and ERA5- reanalysis data as meteorological inputs for near-daily ET retrievals. First, an overview of the TSEB-PT framework is introduced. Then, TSEB-PT model is evaluated across multiple biomes and climatic zones across the US, Europe, and Australia by comparing daily ET estimates against eddy covariance (EC) data forced with energy balance closure from 33 flux tower sites (3 are without energy closure). Daily ET estimates of TSEB yielded an RMSE of 1.32 mm/d, a correlation (r) of 0.77, a Nash-Sutcliff efficiency of 0.56, and Pbias of -1% for all sites combined compared to EC data. The uncertainty of the TSEB-ET estimate varies across climate zones and biome types. TSEB-PT ET produced the best results in humid environments and sites surrounded by woody savannas and savannas. TSEB-PT daily ET multi-model comparison with HSEB-ET (hybrid surface energy balance model) against EC data from the same 33 flux tower sites was also performed to assess relative advantages and disadvantages. Both models produced approximately similar results for all biomes combined at a daily time step, with comparable results at weekly and monthly timescales. Our findings show models’ suitability for specific cover types and climatic conditions. No model outperformed the other for all conditions, and both models performed poorly for sites in certain conditions. Based on these results, HSEB and TSEB ET algorithms emanate as potential tools for water resources management on a global scale. They can be adapted, considering the intrinsic characteristics of climate and ecosystems.