Abstract:In view of the problems that the existing regional zenith tropospheric delay(ZTD) model belongs to function or grid type, the parameters are fixed, and it is difficult to express the characteristics of rapid spatio-temporal change of ZTD, we propose a new combined prediction model based on wavelet transform, Fourier series fitting, AR and SVR. In the time domain, the ZTD sequence is decomposed into low-frequency and high-frequency sequences by wavelet transform. The low-frequency sequence is fitted with Fourier series as a function of time, and the high-frequency sequence is predicted by AR. The mapping of position parameters to Fourier series parameters is established by SVR in spatial domain. Inputting time and location information in the model, the corresponding ZTD prediction value can be obtained. The two-year ZTD data of 94 GNSS stations are used for modeling, and one-year ZTD data of 24 GNSS stations are used for model prediction. The results show that the bias and RMSE between the measured value and prediction value is -2.02 mm and 3.07 cm, which is better than most of regional ZTD models. The model can significantly improve the positioning accuracy in pseudorange single point positioning. The experiments show that the combined model has high prediction accuracy and reliability, and that it has certain application value.