Abstract:The combined model based on the quadratic surface model and BP neural network model is applied to the GPS height anomaly fitting, while the combination is determined from the variance reciprocal method and general regression neural network (GRNN). The GPS elevation data in a certain area is used, the results show that both the accuracy and reliability with the combined model are more superior to the single models, and the fitting accuracy with the combined model based on general regression neural network (GRNN) is better than that with the combined model based on the variance reciprocal method.