Abstract: A Micro-Lysimeter based experiment was conducted to determine daily soil evaporation (Es) in the artificial grassland of Hunshandake sandy terrain in crop growing season of 2005, followed by the identification and analysis of the main factors affecting Es. Es BP network forecast models are compared between using six input vector factors (daily mean temperature, net radiation, relative humidity, wind speed at 2m height, soil moisture in 0～30cm soil depth and LAI) and four input vector factors (daily mean temperature, net radiation, soil moisture in 0～30cm soil depth, and LAI). The research results indicate that BP network model is suitable for Es forecasting, and the six and four input vector factors of BP network models are both convenient and feasible in Es forecasting at production levels. The six input vector factor network model is more accurate than the four input vector factor network model. The research provides supplement to traditional Es calculation methods.