Development Of Real-time And High-precision Global Models For The Key Parameters Of Ground-based GNSS Water Vapor Monitoring

As an important part of the Earth’s atmosphere,atmospheric water vapor plays a vital role in many atmospheric processes,such as global climate change,hydrological cycle and atmospheric circulation.Tropospheric precipitable water vapor(PWV)is an important indicator for measuring the amount of atmospheric water vapor and an important parameter in numerical weather forecast and climate research.Recently,using PWV derived from ground-based GNSS(Global Navigation Satellite System)to monitor the climate change has become a relatively common and effective method.The atmospheric weighted mean temperature(T_m),as a key parameter in GNSS PWV retrieval,the accuracy of T_maffects the calculation and acquisition of GNSS PWV.The purpose of this work is to addresses the shortages of existing global T_mmodels and PWV vertical correction models,such as low spatial resolution as well as only single gridded data is used for modeling,finally,the real-time and high-precision global models of key parameters for ground-based GNSS water vapor monitoring are developed using the latest ERA5(the fifth generation of European Centre for Medium-Range Weather Forecasts Reanalysis)reanalysis,MERRA-2(the Second Modern-Era Retrospective analysis for Research and Applications)reanalysis,the GGOS(Global Geodetic Observing System)Atmosphere gridded products and other data sources.The main research and contributions of this work are as follows:1.A global T_mvertical stratification grid model considering T_mnonlinear elevation reduction with three different grid spatial resolutions,named as IGGNTm-H model,is developed by using the ERA5 reanalysis dataset from 2012 to 2017 based on the sliding window algorithm.Multi-source T_mdata are used to evaluate the performance of the new model in vertical interpolation.The results show that:(1)treating the ERA5 T_mlayered profiles as reference values,the mean bias and root mean square(RMS)error values of the three IGGNTm-H models are about-0.04 K and 2.73 K,respectively.Treating T_mlayered profiles derived from radiosonde as reference values,the mean bias and RMS error values of the three IGGNTm-H models are about-0.1 K and 3.7 K,respectively.(2)Both three IGGNTm-H models show good performance in T_mspatial interpolation for MERRA-2 and GGOS gridded datasets with different grid spatial resolutions against the surface T_mdata derived from radiosonde,and the interpolation accuracy has improved by approximately 5.7%to 23.1%compared with the condition without considering T_mvertical correction.(3)The three IGGNTm-H models all have good T_minterpolation accuracy in the global scale,and show better performance in low latitude and polar regions.In addition,the larger the height difference,the lower the interpolation accuracy.2.A global T_mrefined grid model considering fine spatiotemporal variation with two different grid spatial resolutions,named as IGGNTm model,is developed by using the hourly ERA5 reanalysis dataset from 2012 to 2017 based on the sliding window algorithm.Multi-source T_mdata are used to evaluate the performance of the new model.Besides,the performance of IGGNTm model is compared to the widely used GPT3(Global Pressure and Temperature 3)model.The results show that:(1)treating the ERA5 surface T_mdata as reference values,the mean bias and RMS error values of the two IGGNTm models are about-0.23 K and 3.12 K,respectively,which has improved the accuracy(in RMS)by approximately 1.6%compared with the GPT3 model.Treating the GGOS T_mdata as reference values,the mean bias and RMS error values of the two IGGNTm models are about-0.01 K and 3.08 K,respectively,which has improved the accuracy(in RMS)by approximately 5%compared with the GPT3 model.Treating surface T_mdata derived from radiosonde as reference values,the mean bias and RMS error values of the two IGGNTm models are about0.14 K and 3.87 K,respectively,which has improved the accuracy(in RMS)by approximately5%compared with the GPT3 model.(2)Compared with the GPT3 model,both two IGGNTm models have better performance in T_mestimation in the global scale,especially in high altitude regions.Moreover,IGGNTm model can also well simulate the high frequency diurnal variation of T_m.(3)The impact of the IGGNTm model on GNSS PWV retrieval is analyzed theoretically.The mean theoretical RMS error and relative error of PWV caused by IGGNTm model are 0.26 mm and 1.41%,respectively.3.A global PWV vertical correction model considering time-varying lapse factor is developed according to the 15°global latitude zone division by using the ERA5 reanalysis dataset from 2012 to 2017,named as GPWV-H model.Multi-source PWV data are used to evaluate the performance of the new model in vertical interpolation.Besides,the performance of GPWV-H model is compared to a widely used PWV vertical correction model(named as EPWV-H).The results show that:(1)treating the ERA5 T_mlayered profiles as reference values,the mean bias and RMS error values of the GPWV-H model are-0.1 mm and 1.07 mm,respectively,which has improved the accuracy(in RMS)by approximately 9.3%compared with the EPWV-H model.Treating PWV layered profiles derived from radiosonde as reference values,the mean bias and RMS error values of the GPWV-H model are-0.35 mm and 1.43mm,respectively,which has improved the accuracy(in RMS)by approximately 5.9%compared with the EPWV-H model.(2)GPWV-H model shows excellent performance in PWV spatial interpolation for MERRA-2 gridded dataset with different grid spatial resolutions against the surface PWV data derived from radiosonde,and the interpolation accuracy has improved by approximately 15.1%～17.1%and 0.8%～1.6%compared with the condition without considering PWV vertical correction and using EPWV-H model,respectively.(3)Compared with the EPWV-H model,GPWV-H model has better PWV interpolation accuracy in the global scale,especially in low latitudes.Additionally,the interpolation error of GPWV-H model in middle and high latitudes is relatively small,and it reduces with the increase of height difference.In summary,the three new models developed in this work have excellent performance on a global scale without extra meteorological parameters as input,and provide model parameters with multiple grid spatial resolutions to meet the needs for different users.These conclusions indicate that the new models can have important applications in global real-time and high-precision ground-based GNSS water vapor monitoring.