The Modeling And Research Of China GNSS Tropospheric Wet Delay And Weighted Mean Temperature

Tropospheric zenith wet delay(ZWD)and weighted mean temperature(T_m)are important parameters for conducting atmospheric scientific research,as well as key factors affecting GNSS high-precision navigation and positioning.Aiming at the current model equations of tropospheric zenith wet delay and weighted mean temperature that do not take into account issues such as latitude,elevation and seasonal changes at the same time,this paper combines the MERRA-2 reanalysis data and radiosonde data in China to study the change characteristics of ZWD and T_m.Using machine learning algorithms as a modeling tool,researches on model construction of ZWD and T_m in China were carried out.The major research work and contributions of this study as follows:1.The current research status of ZWD and T_m models are elaborated in detail,and analyze the current problems,introduce the basic principles of key tropospheric parameters,modeling methods and the data sources involved in this article.2.Research the relationship between ZWD,water vapor pressure and surface temperature.Because water vapor has the characteristics of irregular and nonlinear changes,a water vapor attenuation factor that can describe the vertical decline of water vapor is added to the modeling.A tropospheric wet delay model that takes into account time,space,and meteorological factors simultaneously is established.Both the MERRA-2 atmospheric reanalysis data and radiosonde data in China are treated as reference values,to evaluate the performance of the ZWD models.3.The accuracy test results of the ZWD model show that:(1)ZWD has a strong correlation with water vapor pressure,followed by the correlation between ZWD and surface temperature.Both meteorological parameters can be used as input parameters to participate in ZWD model construction;(2)The Saastamoinen model and Hopfield model have large negative deviations in China.The GPT2w-1 model,BP_ZWD model,GB_ZWD model and RF_ZWD model all have small deviations,indicating that the ZWD model constructed based on machine learning algorithms has small system deviations and Reliability is good;(3)When MERRA-2 reanalysis data and radiosonde data are used as reference values,the RF_ZWD model has achieved good accuracy compared with Saastamoinen model,Hopfield model,GPT2w-1 model,BP_ZWD model and GB_ZWD model,respectively.And it shows the high-precision and stable performance in China.4.Study the relationship between T_m,surface temperature and elevation and analyze the characteristics of the vertical decline rate of T_m.Based on the characteristics of T_m affected by seasonal factors and latitude,a weighted mean temperature model in China that takes meteorological parameters,vertical decline rates,elevation,latitude,and seasonal factors simultaneously|is established.Both the MERRA-2 atmospheric reanalysis data and radiosonde data in China are treated as reference values,to evaluate the performance of T_m models.5.The accuracy test results of the T_m model show that:(1)There is a strong correlation between T_m and surface temperature and elevation;(2)The vertical decline rate of T_m has annual and semi-annual changes,and the vertical decline rate function formula can be established;(3)When MERRA-2 reanalysis data and radiosonde data are treated as reference values,the RFT_m model achieves good accuracy compared with the Bevis model,GPT2w-1 model and BPT_m model,and the model has no obvious seasonal changes.The impact is small,especially in high altitude areas,showing a small RMS error.Thus,the RFT_m model has high-precision and stable performance in China.