A simple method to estimate actual evapotranspiration from a combination of net radiation, vegetation index, and temperature

Satellite remote sensing is a promising technique for estimating global or regional evapotranspiration (ET). A simple and accurate method is essential when estimating ET using remote sensing data. Such a method is investigated by taking advantage of satellite measurements and the extensive ground‐based measurements available at eight enhanced surface facility sites located throughout the Southern Great Plains (SGP) area of the United States from January 2002 to May 2005. Data analysis shows that correlation coefficients between ET and surface net radiation are the highest, followed by temperatures (air temperature or land surface temperature, T s ), and vegetation indices (enhanced vegetation index (EVI) or normalized difference vegetation index (NDVI)). A simple regression equation is proposed to estimate ET using surface net radiation, air or land surface temperatures and vegetation indices. ET can be estimated using daytime‐averaged air temperature and EVI with a root mean square error (RMSE) of ∼30 W m −2 and a correlation coefficient of 0.91 across all sites and years. ET can also be estimated with comparable accuracy using NDVI and T s . More importantly, the daytime‐averaged ET can also be estimated using only one measurement per day of temperatures (the daytime maximum air temperature or T s ) with comparable accuracy. A sensitivity analysis shows that the proposed method is only slightly sensitive to errors of temperatures, vegetation indices and net surface radiation. An independent validation was made using the measurements colleted by the eddy covariance method at six AmeriFlux sites throughout the United States from 2001 to 2006. The land cover associated with the AmeriFlux sites varies from grassland, to cropland and forest. The results show that ET can be reasonably predicted with a correlation coefficient that varies from 0.84 to 0.95 and a bias that varies from 3 W m −2 to 15 W m −2 and RMSE varying from ∼30 W m −2 to ∼40 W m −2 . The positive bias partly comes from the energy imbalance problem encountered in the eddy covariance method. The proposed method can predict ET under a wide range of soil moisture contents and land cover types.