Abstract: A Flood Loss Prediction Model for insurance based on terrain condition and domain knowledge is proposed and implemented.By this model the terrain and drainage ability factor is generated based on Digital Elevation Model（DEM）,and Neural Network is used with its powerful non-linear mapping ability. A method to incorporate domain knowledge with this model is given. Experiments demonstrate that the model can generate dynamic predictions and has good generalization ability．It could provide an operable flood loss prediction platform for insurance and related industries.