Abstract:Based on the zenith tropospheric delay(ZTD) calculated from the data of 23 continuously operating reference stations(CORS) in Anhui province, we analyze the applicability of two tropospheric model schemes, GPT3+Hopfield and GPT3+Sasstamoinen tamoinen. Using sounding data, we analyzed the accuracy of GPT3 model in estimating weighted average temperature (Tm) and the inversion accuracy of the precipitable water vapor (PWV). The results show that: 1) the ZTD accuracy of GPT3 combined with Saastamoinen model is better than that of Hopfield model. At the same time, the ZTD accuracy of GPT3 model has significant spatial-temporal distribution characteristics. The accuracy of southern Anhui is lower than that of northern Anhui, and the accuracy in spring and winter is better than in summer and autumn; 2) In Anhui province, the Tm accuracy of the two grid resolutions of GPT3 model are basically the same; the average deviation is about -2.0 K, and the RMS value is about 4.5 K; 3) In Anhui province, the PWV (GPT3-PWV) inversed based on the meteorological parameters of the GPT3 model has high consistency with the PWV of the sounding station, and also has characteristic spatial-temporal variations. It gradually decreases from the south to the north of Anhui, with the maximum in summer and the minimum in winter.