利用MATLAB实现UNB3m、GPT2w+Hopfield、GPT2w+Saastamoinen、GPT3+Hopfield、GPT3+Saastamoinen等5种模型，分析它们在陕西地区的适用性。结果表明，5种模型结果普遍偏小。GPT2w+Saastamoinen和GPT3+Saastamoinen模型整体精度相当，且优于其他3种模型，bias为1.41 cm，RMS分别为4.68 cm和4.67 cm，且随着高程增加精度越来越高。5种策略精度均随季节变化而变化，其中UNB3m变化最为明显，夏冬2季bias差达到7.92 cm，RMS差达到7.67 cm。更高精度计算时，秋季应使用GPT3，而春夏2季时使用GPT2w效果更好。选用同样的气象参数模型时，Saastamoinen模型比Hopfield模型更适用于陕西地区，并且陕北地区精度最好。对比最新的全球气压温度模型GPT3与GPT2w发现，2种模型算得的地面气压P、地面温度T、地面水汽压e、大气加权平均温度Tm等4种气象参数均相差细微，所以在陕西地区利用GPT2w或GPT3分别算得的对流层总延迟ZTD和对流层干延迟ZHD相差很小，通过对流层湿延迟ZWD算得的PWV也几乎相当。"/>
Accuracy and Applicability of Different Tropospheric Zenith Delay Models in Shaanxi Province" />
In this paper, we use MATLAB to realize five models of UNB3m, GPT2w+Hopfield, GPT2w+Saastamoinen, GPT3+Hopfield, GPT3+Saastamoinen. We analyze their applicability in Shaanxi. The results show that the model calculation results are generally small in Shaanxi province. The overall accuracy of the GPT2w+Saastamoinen and GPT3+Saastamoinen models is comparable and superior to the other three models, with bias of 1.41 cm and RMS of 4.68 cm and 4.67 cm respectively; the precision improves as elevation increases. The accuracy of the five models varies with the seasons; UNB3m has the most significant change. The bias difference in summer and winter is 7.92 cm, and the RMS difference is 7.67 cm. For higher precision calculation, GPT3 should be used in autumn, while GPT2w works better in spring and summer. When using the same meteorological parameter model, the Saastamoinen model is more suitable for Shaanxi than the Hopfield model, and northern Shaanxi has the best longitude. In the comparison between two latest models GPT3 and GPT2w, we find that the four meteorological parameters calculated by the two models are slightly different, including surface air pressure, surface temperature, surface water pressure and weighted average air temperature. The difference between the total tropospheric delay ZTD and the tropospheric dry delay ZHD calculated by GPT3 or GPT3 is small, and the PWV calculated by the tropospheric wet delay ZWD is almost the same."/>
不同对流层天顶延迟模型在陕西地区的精度及适用性分析
Abstract：In this paper, we use MATLAB to realize five models of UNB3m, GPT2w+Hopfield, GPT2w+Saastamoinen, GPT3+Hopfield, GPT3+Saastamoinen. We analyze their applicability in Shaanxi. The results show that the model calculation results are generally small in Shaanxi province. The overall accuracy of the GPT2w+Saastamoinen and GPT3+Saastamoinen models is comparable and superior to the other three models, with bias of 1.41 cm and RMS of 4.68 cm and 4.67 cm respectively; the precision improves as elevation increases. The accuracy of the five models varies with the seasons; UNB3m has the most significant change. The bias difference in summer and winter is 7.92 cm, and the RMS difference is 7.67 cm. For higher precision calculation, GPT3 should be used in autumn, while GPT2w works better in spring and summer. When using the same meteorological parameter model, the Saastamoinen model is more suitable for Shaanxi than the Hopfield model, and northern Shaanxi has the best longitude. In the comparison between two latest models GPT3 and GPT2w, we find that the four meteorological parameters calculated by the two models are slightly different, including surface air pressure, surface temperature, surface water pressure and weighted average air temperature. The difference between the total tropospheric delay ZTD and the tropospheric dry delay ZHD calculated by GPT3 or GPT3 is small, and the PWV calculated by the tropospheric wet delay ZWD is almost the same.
