Abstract: Based on the time series 10-day composite NOAA Pathfinder AVHRR Land (PAL) dataset (8 km×8 km), and by using land surface energy balance equation and “VI-T_s” (vegetation indexland surface temperature) method, a new algorithm of land surface evapotranspiration (ET) was constructed. This new algorithm did not need the support from meteorological observation data, and all of its parameters and variables were directly inversed or derived from remote sensing data. A widely accepted ET model of remote sensing, i.e., SEBS model, was chosen to validate the new algorithm. The validation test showed that both the ET and its seasonal variation trend estimated by SEBS model and our new algorithm accorded well, suggesting that the ET estimated from the new algorithm was reliable, being able to reflect the actual land surface ET. The new ET algorithm of remote sensing was practical and operational, which offered a new approach to study the spatiotemporal variation of ET in continental scale and global scale based on the long-term time series satellite remote sensing images.

Key words: land surface evapotranspiration, PAL data set, VI-T_s method, remote sensing model, bamboo-shaped rainwater harvesting ditch, water holding and impounding characteristics, soil water content, hilly loess region.