Research On Atmospheric Water Vapor Of Ground GNSS Based On ERA-5 Reanalysis And Sounding Data

With the development of GNSS(Global Navigation Satellite System)meteorology,water vapor inversion technology based on GNSS is becoming more and more mature.At present,there are many calculation methods of atmospheric water vapor inversion,which have their own advantages but limitations.Therefore,it is necessary to make innovation and comprehensive application on the basis of full research on various methods.In this paper,the accuracy and spatial-temporal resolution of GNSS water vapor inversion are further improved through the correction and optimization of the original weighted average temperature model by using reanalysis data and sounding data in combination with the calculation methods of previous scholars.The main work and achievements of this paper are as follows:(1)In this paper,we use the latest generation reanalysis grid data(ERA5)from 2008 to2017 provided by ECMWF in Jiangsu and its surrounding areas,and use linear regression method to establish single factor and multi factor models to realize the weighted mean temperature modeling of Jiangsu and its surrounding areas,and use the sounding data to modify them respectively.Compared with the sounding data,the experimental results show that the RMS error range of₈)calculated by single factor and double factor models is 2.60L?2.97K and 2.52K?2.92K respectively.The results show that the accuracy of the single factor model established in this paper is equal to or even slightly better than that of the local model in Jiangsu Province,and the accuracy of the double factor model can be increased up to 10.52%,which proves the applicability of the era5 reanalysis data and radiosonde to establish the Jiangsu regional₈)model.(2)Combining the double factor correction model and BP neural network,a new weighted mean temperature model is established.The results show that the root mean square error of the stations in Jiangsu region is predicted to be within 2.52K?2.72K in 2018 by using the two factor correction model,while the root mean square error of the stations in 2018 predicted by using BP neural network model is within 2.29K?2.60K,and the average optimization degree of the five stations is about 10%.Experimental results show that the neural network can effectively establish the relationship between the weighted average temperature and other nonlinear influence factors,and improve the accuracy of the calculation of₈).(3)Through comparative analysis of PWV-sounding and PWV-GNSS in Sheyang station,Jiangsu Province,the results show that the correlation between PWV-sounding and PWV-GNSS is 0.95,the average deviation between PWV-sounding and PWV-GNSS is 1.55mm,and the root mean square error is 2.86mm.The results show that the atmospheric precipitable water calculated by JSCORS data and ERA5 reanalysis data meets the accuracy requirements and greatly improves the spatial-temporal resolution.(4)Based on the comparison between the actual precipitation in the summer of 2018(June?August)in Sheyang,Jiangsu Province and GNSS-PWV,it can be inferred that when the water vapor content keeps a certain value(about 70mm),a relatively large change may cause precipitation after 2-3 hours.