Abstract: Predicting soil water characteristic curve (SWCC) from particle size distribution was an easy, rapid and economical method. However, the traditional estimate of water content in this method didn't consider the adsorbed water films, which led to an underestimate of water content in the dry range of the SWCC. In this study, a modified method was developed to take the film water into account when estimating water content, and subsequently the SWCC was predicted by combining water content with the matric suction which estimated by the method in HP model and MV model. The modified model was tested on seventeen soil samples, selected from the UNSODA database. The results illustrated that the modified model significantly improved the model performance in the dry range of the SWCC and accurately estimated the SWCC of sand. Furthermore, the modified model was also applicable for loam and clay by introducing a parameter β.