Development of a Satellite-based evapotranspiration algorithm: A case study for Two Deciduous Forest Sites

We introduce a new methodology to estimate 8-day average daily evapotranspiration (ET) using both routinely available data and the Penman-Monteith (P-M) equation. Our algorithm considers the environmental constraints on surface resistance and ET by (1) including vapor pressure deficit (VPD), incoming solar radiation, soil moisture, and temperature constraints on stomatal conductance; (2) using leaf area index (LAI) to scale from the leaf to the canopy conductance; and (3) calculating canopy resistance as a function of environmental variables such as net radiation, precipitation index, and VPD. Remote sensing data from the Moderate Resolution Spectroradiometer (MODIS) and the Advance Microwave Scanning Radiometer-EOS (AMSR-E) were used to estimate ET by using MODIS land surface temperature (LST) to estimated VPD, AMSR-E soil moisture to estimate canopy conductance, and MODIS surface emissivity and albedo to estimate shortwave and net radiation. The algorithm was evaluated using ET observations from two AmeriFlux Eddy covariance flux towers located at the Morgan Monroe State Forest (MMSF) in Indiana and the Harvard Forest (HarvF) in Massachusetts for the periods of 2003-2008. ET estimates from our algorithm was compared to the flux observations. Results indicated a root mean square error (RMSE) of the 8-day average ET of 0.57 mm for the HarvF and 0.47 mm for the MMSF. A significant correlation was found between the estimated 8-day average ET and the observed 8-day average ET with r2 of 0.84 for the HarvF and 0.88 for the MMSF. Using tower meteorological data, the r2 slightly increased to 0.90 for the MMSF. The algorithms for VPD and radiation were tested against flux observations and found a strong correlation with r2 ranging from 0.68 to 0.82. Sensitivity analysis revealed that the modeled ET predictions are highly sensitive to changes in the canopy resistance values, so accurate estimates of canopy resistance is essential for improve ET predictions. Our algorithm matched the magnitude and detects the seasonal variation on the flux tower observed ET. In this paper, we have demonstrated that our algorithm was successful in estimating ET without the need for site based meteorological observations. The significant relationship between our algorithm estimates and flux tower observations implies the potential for applying our model to different ecosystems