Mapping evapotranspiration at a high resolution using the surface aerodynamic temperature model and airborne multispectral remote sensing data

Abstract

Irrigation is the largest single consumer of water in the world, and with the increasing population, limitation of natural resources, climate change, and global warming, the pressure on water resources has become more significant and attention to agriculture is increasing daily. The limitation of agricultural areas requires efficient use of these areas to obtain a maximum yield. Evapotranspiration (ET) is a major component of the water budget and energy balance. Therefore, exact measurement of plant water use (and thus ET) is vital for efficient use of water resources, planning, and management purposes, especially for arid and semiarid regions. Many methods have been developed for estimating crop ET on a small field scale, such as the Bowen Ratio (BR), the Eddy Covariance (EC), and Lysimeter systems; however, remote sensing-based ET methods have been developed for estimating crop water needs on a regional scale. The energy balance (EB)-based ET algorithms require the computation of net radiation (Rn), soil heat flux (G), and sensible heat flux (H) to solve for latent heat flux or ET as a residual. Values of Rn and G can be estimated with an acceptable accuracy. However, estimation of H is not straightforward. This is because surface aerodynamic temperature (To) is difficult to measure or estimate. Instead, radiometric surface temperature (Ts) is generally used in the estimation of H. However, using Ts may cause overestimation of H, and thus underestimation of ET. To account for those differences between To and Ts, several remote sensing-based algorithms have been developed for mapping ET. The Surface Aerodynamic Temperature (SAT) model is one of them, and was used in this study to estimate sensible heat flux (H) for cotton fields and calculate ET as a residual of the EB for research fields located at the USDA-ARS Conservation and Production Research Laboratory (CPRL) near Bushland, Texas in 2008. By using the SAT model, ET results obtained from the multispectral airborne remote sensing data were compared with ET calculated with model input data collected at the large weighing lysimeters site . Resulting SAT ET values were obtained with a Mean Biased error (MBE) and a Root Mean Squared error (RMSE) of 2.67% and 8.61%, respectively. Then, actual crop ET from the SAT model were compared to measured values from the large lysimeter mass balance. This evaluation resulted in 25.9% MBE and a 44.07% RMSE for the east irrigated fields while for the west dryland fields the error obtained was 42.13% MBE and 42.91% RMSE. In addition, the crop water stress index (CWSI) was used to calculate actual ET using remote sensing inputs and results were also compared to lysimeter measured ET values. Results indicated that the errors were MBE value of 3.77% and an RMSE value of 10.76% for the east fields and 0.89% MBE and 6.0% RMSE for the west fields of the research area, respectively. The results show that the SAT model that was used in this study may not be appropriate for sparse vegetation and heterogeneous surface conditions and that further improvement of the model is required with the application of remote sensing data. On the other hand, the CWSI method performed better than the SAT model for estimating ET and crop water stress levels.