A global 30-m ET model (HSEB) using harmonized Landsat and Sentinel-2, MODIS and VIIRS: Comparison to ECOSTRESS ET and LST

Abstract

Advances in earth observation science in recent years have contributed to improving the quantification of evapotranspiration (ET) at field, regional and global scales. Many studies have stressed the need for a high temporal and spatial resolution ET product that minimizes the bias between modeled and actual water use for proper water accounting. We present a hybrid single-source energy balance (HSEB) model that calculates evapotranspiration at the field-scale based on the synergistic use of Sentinel-2, Landsat, Visible Infrared Imaging Radiometer Suite (VIIRS) and (Moderate Resolution Imaging Spectroradiometer) MODIS land surface temperature products. The model operates in Google Earth Engine as a time series using global atmospheric variables and 100-m Copernicus Land cover data. Evaluation of HSEB for calculating evaporation over 29 flux tower sites within an extensive range of climatic conditions and biomes over the US, Europe and Australia for 2018–2020 shows that the model significantly improves the temporal and spatial components of ET mapping. Overall, HSEB performed well in all considered biome types and climatic conditions (r = 0.81, 0.74, and 0.8, a Nash-Sutcliff efficiency of 0.6, 0.74, 0.8, and a bias of 4%, 1%, and − 0.9% at the daily, weekly, and monthly scales, respectively). Root Mean Square Error averaged at 1.31 mm/day. Comparison of instantaneous latent heat fluxes of the ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) against tower data shows that HSEB produces acceptable results, although ECOSTRESS had a slightly lower bias (3.8% for HSEB vs. -0.8% for ECOSTRESS). HSEB performed better over croplands. We also discuss comparisons of sharpened LST from VIIRS VNP02, VNP21, MODIS, Landsat, Sentinel-3, and LST from ECOSTRESS versus ground LST measurements, and briefly discuss the sensitivity of HSEB to the thermal data used. Both ECOSTRESS and Landsat showed better performance at different LST ranges and time of day when compared to LST observation collected over a small potato field in Lebanon. The analysis of impact of LST product used in HSEB on ET results at US-ARM site showed that HSEB with MODIS LST outperforms HSEB ET from VNP02 (an overestimate) and VNP21 (an underestimate). We conclude that HSEB can be used as an operational global model for monitoring evaporative stress and evaporation at the small-agriculture field level with higher temporal and spatial resolution utilizing the wide suite of available satellite data. © 2022 The Authors