Research On Accuracy Evaluation And Model Establishment Of ZTD/PWV Based On GNSS And Reanalysis

Although water vapor accounts for less than 4%of the total neutral atmosphere,it is an important part of the Earth's neutral atmosphere.On the one hand,it is the basis of energy transfer in the atmosphere.Water is often evaporated at low latitudes,and water vapor is transported to high latitudes to condense and release a lot of heat.On the other hand,water vapor is also one of the most important greenhouse gases on the Earth,which has a significant impact on global climate change.In addition,Water vapor also plays a key role in the hydrological cycle of the atmosphere,as it moves around the Earth under the influence of wind,feeding the formation of clouds and precipitation.Therefore,it is very important to understand the change of water vapor for the study of climate change characteristics and meteorological monitoring.This paper focuses on establishment of tropospheric delay model and water vapor model,as well as fusion and global variation trend of water vapor.The main research contents and results are as follows:(1)The accuracy and applicability of temperature,pressure and humidity meteorological products from different reanalysis data as well as ZTD derived from reanalysis on a global scale are analyzed based on surface meteorological observation data and GNSS ZTD.Firstly,atmospheric pressure,temperature and humidity observed by 377 meteorological stations around the world in 2016 provided by SOPAC are taken as true values.The accuracies of surface temperature,pressure and relative humidity obtained from ERA5,MERRA-2,ERA-Interim,JRA55,NCEP/NCAR and NCEP/DOE reanalysis data are compared and analyzed.The RMS errors of surface pressure,temperature and relative humidity of ERA5 are 1.24 hPa,3,54? and 18.57%,respectively,indicating ERA5 reanalysis is the best among the six reanalysis data.Secondly,taking ERA5 reanalysis data as an example,the accuracies of atmospheric pressure,temperature and relative humidity obtained from atmospheric pressure level and surface level are compared and analyzed.The RMS errors between atmospheric pressure,temperature and relative humidity obtained from ERA5 and those from meteorological stations are 3.03hPa,3.69? and 13.95%,respectively.In terms of relative humidity and pressure,the accuracy of pressure level from ERA5 reanalysis is obviously better than that of surface level,but the temperature accuracy of pressure level from ERA5 reanalysis is slightly worse than that of surface level.Finally,taking ERA5 reanalysis data with high accuracy as an example,the accuracy of ZTD derived from pressure level and surface level is compared and analyzed.The RMS errors of ZTD derived from pressure level and surface level inversion ZTD and IGS ZTD are 14.73 mm and 36.58mm,respectively,indicating that the accuracy of ZTD derived from pressure level is significantly better than that derived from surface level.(2)The global tropospheric delay model is established based on ERA5 reanalysis data.Periodic item of ZTD from ERA5 reanalysis data from 2015 to 2018 can be derived from with the Fourier transform analysis.The global troposphere delay ERA5 grid model is established by the spectrum analysis and least squares method,and the accuracies of the GZTDERA5 and GPT2 model troposphere model are verified using IGS ZTD in 2019.The root mean square errors of the GZTDERA5 and GPT2 troposphere model are 38.03 mm and 42.43 mm respectively,thereby indicating that GZTDERA5 has a better accuracy.(3)The global grid model of PWV and ZTD varies with height is established based on the ERA5 reanalysis.The variation rule of ZTD with height is analyzed,and the global exponential grid ZTD model is established based on the ZTD results of ERA5 reanalysis data from 2015 to 2018.The model accuracy is verified with the 2019 ERA5 ZTD results.The results show that the root mean square errors of the model 31-37 level and 36-37 level are 23.71 mm and 4.81mm,respectively.The absolute accuracy of the model 36-37 level is better than that of the model 31-37 level.According to the variation rule of PWV with height,the cubic polynomial monthly model,the cubic polynomial year model and the exponential function year model are respectively used to fit the variation of PWV with height.The root mean square of the exponential model,cubic polynomial yearly model and monthly model are 0.50mm,0.38mm and 0.14mm,respectively,indicating that the accuracy of the cubic polynomial monthly model with height variation is obviously better than that of the cubic polynomial year model and exponential model.(4)The conversion coefficient grid model between global PWV and ZTD is established based on ERA5 reanalysis data.The correlation of PWV and ZTD from reanalysis data from 2016 to 2018 is made on a global scale,and most of their correlation coefficients are greater than 0.9.Scatterplot showes that there is a good linear regression relationship between PWV and ZTD.The conversion factors of the grid model between PWV and ZTD are established and the model accuracy is verified with the PWV and ZTD from ERA5 inversion in 2019.The root mean square error and mean bias of the model are 0.97 mm and 0.01 mm,thereby indicating that the model has a good accuracy and can be used to convert directly GNSS ZTD into PWV when there are no meteorological observation data available.(5)The global accuracy and applicability of PWV from different reanalysis data and COSM-2 PWV are analyzed based on PWV from GNSS,radiosonde and microwave radiometer.Taking the PWV of GNSS inversion from 2016 to 2018 as the true value,the root mean square errors of PWV products from ERA5,ERA-interim,JRA55,NCEP/NCAR and NCEP/DOE reanalysis data and GNSS PWV are 1.84mm,2.32mm,2.53mm,3.34mm and 3.51mm,respectively,thereby indicating the accuracy of the latest reanalysis data,ERA5,is the best.In addition,the accuracies of PWV products from the five reanalysis products in different climate zones are verified.The root mean square errors of PWV products from the five reanalysis data in the polar region are small,but the relative root mean square errors are large.The results indicate that the absolute accuracy is good,and the relative accuracy is poor in the polar region.However,the absolute and relative accuracies of the PWV products from the five reanalysis data in the tropics are opposite to those in the polar region,and the absolute accuracy is poor while the relative accuracy is better.Compared with the PWV from radiosonde and GNSS,the root mean square errors of COSMS-2 inversion PWV are 3.92 mm and 3.28 mm,respectively,which indicates that the PWV from COSMIC-2 from 46°N to 46°S is in good agreement with PWV from GNSS and radiosonde.The COSMIC-2 PWV value is compared with those from SSMIS,AMSR-2 and GMI.The root mean square errors of COSMIC-2 PWV are 2.64mm,2.62m and 2.70mm,respectively,indicating that there is a good consistency between the COSMIC-2 PWV and PWV from the three kinds of microwave remote sensing over the ocean.In addition,the accuracies of COSMS-2 PWV in tropics,subtropics,Southern and Northern hemispheres are analyzed,and the results show that the accuracies of COSMS-2 PWV in the Northern and Southern hemispheres is similar,but the accuracy of COSMS-2 PWV in the subtropical region is better than that in the tropical region.In addition,we also analyzed the relationship of PWV,surface temperatures,wind speed,cloud cover and rain rate from the microwave remote sensing.The results show that PWV accuracy from microwave remote sensing has a negative correlation between surface temperature,cloud cover and rain rate,and the PWV accuracy is positively correlated with wind speed.(6)We analyze the global mean,variation and trend in PWV are analyzed based on PWV from GNSS,microwave remote sensing,ERA5 and MERRA-2 reanalysis data.GNSS PWV data from 2000 to 2014 are used to compare and analyze global mean PWV value,variability and trend of ERA5 PWV and MERRA-2 PWV over the land.The results show that ERA5 and GNSS have better consistency in the long-term variation trend in water vapor over the land.The PWV variation trends of microwave remote sensing,ERA5 and MERRA-2 reanalysis data over the ocean from 1988 to 2019 are compared and analyzed.The results show that the PWV from ERA5 reanalysis data has a better consistency with that from microwave remote sensing data in the trend in water vapor over the ocean.Finally,the PWV from ERA5 reanalysis data from 1980 to 2019 are used to analyze the trend in global water vapor in the past 40 years,as well as the water vapor variation from 1980 to 1989,from 1990 to 1999,from 2000 to 2009 and from 2010 to 2019.The results show that the trend in water vapor in the short term is mainly influenced by El Nino and La Nina events.And the relationship between temperature and PWV trend from ERA5 reanalysis data is analyzed.The results show that there is a good consistency between PWV and temperature trend.Overall,the global trend in PWV and temperature is upward.(7)A water vapor fusion model of GNSS and ERA5 reanalysis is established based on support vector mechanism.The ERA5 reanalysis PWV and GNSS PWV are fused to construct the PWV temporal and spatial fusion model with the support vector machine method.The results of the PWV temporal fusion model show that the average root mean square error of the ERA5 reanalysis PWV is reduced by 0.37mm,the maximum accuracy is improved by 18.09%,and the average accuracy is improved by 12.32%.The results of PWV spatial fusion model show that the average root mean square error of ERA5 reanalysis PWV product is reduced by 0.49mm,and the average accuracy is improved by 15.22%.Both spatial and temporal fusion models show that the accuracy of GNSS PWV and ERA5 reanalysis PWV fusion using support vector machine is significantly improved.