Research And Application Of Zenith Tropospheric Delay Correction Model

With the rapid development of the Global Navigation Satellite System(GNSS),it has achieved fruitful results in the fields of precision single point positioning(PPP,Precise Point Position)and GNSS meteorology.When the satellite signal passes through the atmosphere,the delay caused by the influence of the non-dispersive medium is the tropospheric delay.For a long time,tropospheric delay has always been an important factor affecting GNSS navigation positioning and precise single point positioning.Therefore,studying tropospheric delay is of great significance for improving GNSS navigation accuracy and PPP positioning.There are many methods for tropospheric delay correction,including parameter estimation method,external correction method,difference method and model correction method.The model correction method has been widely used due to its cheapness and convenience,but it also has the defect of low model accuracy.With the rapid development of neural network technology,genetic algorithms and data assimilation technology,these technologies have presented many advantages in many fields.The tropospheric delays in other directions can be obtained by the tropospheric delay solution in the zenith direction.Therefore,the above techniques are mainly used for the study of the zenith Tropospheric Delay(ZTD).The specific research contents and results are as follows:(1)The basic structure of the atmosphere and meteorological parameters affecting atmospheric refraction,such as air pressure,temperature,relative humidity and water pressure,are introduced.On the basis of analyzing the tropospheric delay with time,the time series variation of ZTD is summarized.(2)Based on the Hopfield model,a high precision tropospheric delay fusion model(BP-Hopfield model)is established by using error compensation technology.The verification data used meteorological sounding data of more than 120 stations in the world provided by the University of Wyoming in the United States in 2010.The results show that the average accuracy of the Hopfield model is ±35.3 mm,and the average accuracy of the fusion model(BP-Hopfield model)is ±23.5 mm.The accuracy of the fusion model(BP-Hopfield model)is about 33% higher than that of the Hopfield model.(3)Based on the EGNOS model,a high-precision tropospheric delay fusion model(GA-BP-EGNOS model)was established by using genetic algorithm and BP neural network combination technique.Using the ZTD data from 20 stations in North America from 2011 to 2014 provided by the IGS Center,the results show that the average accuracy of the fusion model(GA-BP-EGNOS model)is ±34.4 mm,and the average accuracy of the EGNOS model is ±80.4 mm.Compared to the EGNOS model,the accuracy of the fusion model(GA-BP-EGNOS model)was increased by approximately 58%.(4)Aiming at the defects of low quality of meteorological data and periodic consideration of modeling data in UNB3 m model,a high-precision fusion model(ERA-UNB3 m model)was established by assimilation of ERA-Interim atmospheric reanalysis data.Multi-dimensional verification was carried out using the ZTD data of 40 stations in the 2016 European region provided by the IGS Center.The results show that the accuracy of the fusion model is ±31.4mm and the accuracy of the UNB3 m model is ±46.3mm.The results show that the accuracy of the fusion model(ERA-UNB3 m model)is about 31% higher than that of the UNB3 m model.(5)Through the precision single point positioning verification,it is found that the higher the accuracy of the ZTD model,the better the positioning effect.The parameter estimation method is slightly better than the ZTD model.For the error in the three directions of XYZ,the Z direction has a larger error than the X and Y directions.For the error of the NEU in three directions,the U direction error is the largest.