A New Method to Estimate Evapotranspiration from the Combination of Modis Aqua and Terra Measurements

Evapotranspiration(ET) is one of the key fluxes in all water budget studies, satellite remote sensing is a promising tool to establishment operational frame works for estimating ET at large extent. One of the most common ways is to solve LE, as a residual in the land surface energy balance equation: RN G = H + LE. And the largest uncertainty in estimating H comes from the acquisition and spatialization of near-surface air temperature, Ta. In order to eliminate the input of Ta, we test a combination model of PBL and SVAT, and use two time measurement of surface radiometric temperature introduced from MODIS products. In this study, such model was applied for calculating sensible heat fluxes in a semiarid region on July to August, 2003. The study was conducted in Tongyu site, Jilin Province, China, which is located at 4.416N, 122.867E. Field data from CEOP plan is used for validation. Three MODIS land products are used in this study: the 8-day leaf area indices product (MOD15) and the daily TS product (MOD11A1) at 1-km resolution and the daily atmosphere profile product (MOD07). The sensible heat results show relatively small RMSE compared to eddy-covariance measurements, and Bowen ratio show a promising correlation coefficient(r=0.76) in clear-sky cases with field data. Because of the limited verification data and dates, the model results are preliminary and need further testing. As operational production of ET is our goal, the model is encouraging in the following aspect: 1). Uncertainty in the retrieval of TR and the interpolation of air temperature from the meteorological point measurements is alleviated; 2). Only temporal changes in radiometric temperatures is used in this model rather than absolute real temperature, biases of the derived surface temperature is not as destructive as previous; 3) LAI is used in a simplified function to decompose soil evaporative from plant transpiration. Future work will carry out in temporal-scaling to obtain gap-free dataset caused by cloud contamination and parameterization optimization, and it will provide more insight into the optimal approach for diagnosis of land surface fluxes using remote sensing observations.