Construction Of Atmospheric Water Vapor Field With High Spatiotemporal Resolution Based On GNSS/MODIS Multi-fusion Data

Water vapor,also referred to as precipitable water vapor,represents an essential component of the atmosphere and one of its more active and difficult to describe parameters.Knowing the distribution of water vapor in the atmosphere is important for researching the local climate and conducting productive activities.At present,numerous techniques are available for the inversion of precipitable water vapor,and the inversion principles and accuracy vary.The ground-based GNSS inversion of precipitable water vapor has the characteristics of all-weather,high resolution,and not affected by clouds and rain,however the water vapor values obtained from the inversion are not continuous.MODIS inversion of precipitable water vapor can obtain a continuous distribution of water vapor,but the accuracy of the data is limited by the constraints of clouds and aerosols.By summarizing the water vapor inversion methods,the advantages are complemented by the fusion study of multi-source data.In this thesis,the PWV calibration model is constructed for the characteristics of high accuracy temporal resolution of GNSS PWV and spatial continuity of MODIS PWV.The GNSS PWV is used to calibrate the MODIS PWV,so as to construct a spatially continuous,high precision and high temporal resolution atmospheric water vapor field in the Jinan region,and beneficial conclusions are obtained to support the applications related to the inversion of atmospheric precipitable water in the Jinan region.The major researches and findings of this thesis are as follows:(1)Introducing domestic and international GNSS and MODIS atmospheric precipitable water inversion studies.Verifying the possibility of using GNSS PWV to calibrate MODIS PWV.To present the related basic principles and methods of RS,GNSS and MODIS inversion of precipitable water vapor.(2)Atmospheric weighted mean temperature,as an essential parameter affecting the GNSS precipitable water vapor inversion,directly influences the GNSS PWV inversion accuracy.In consideration of the absence of an atmospheric weighted mean temperature model applicable to the Jinan region,an atmospheric weighted mean temperature model suitable for the Jinan region was established using data from the Zhangqiu RS station from 2015-2019 and ERA5 reanalysis data.By comparing with Bevis model and the model in eastern China,RMSE accuracy of the Jinan weighted mean temperature model is improved by 20 % and 21 %,respectively.MAE accuracy is improved by 25 % and 23 %,respectively.The accuracy of Bias,RMSE and MAE of PWV of Jinan weighted mean temperature model is improved by 0.34 mm,1.10 mm and 0.54 mm,respectively.Compared with that of Bevis model by using GAMIT10.75 for PWV inversion simulation.The Jinan weighted mean temperature model constructed in this thesis is more suitable for PWV inversion locally.(3)Inversions of GNSS PWV and MODIS PWV were obtained by using the high-precision GNSS observations of 15 CORS stations in Jinan area provided by Shandong Continuously Operating Reference Stations(SDCORS)and MOD05 products released by MODIS.It finds that MODIS PWV had underestimated values.By analyzing the correlation between GNSS PWV and MODIS PWV,the correlation coefficient of the two is 89.91%,and there is a good linear correlation,and the GNSS PWV can be used to calibrate the MODIS PWV.(4)Linear regression calibration model,nonlinear calibration model and BP neural network calibration model were constructed to calibrate the MODIS PWV from GNSS PWV.The RMSE of the linear regression calibration model is 4.15 mm,which improves the accuracy by 26.93%.The RMSE of the nonlinear calibration model is 3.49 mm,which improves the accuracy by 38.55%.The RMSE of the BP neural network calibration model is 3.30 mm,which improves the accuracy by 41.9%.By analyzing the accuracy of the three models using CQRS and ZQRS stations,the RMSE values of CQRS and ZQRS stations were reduced after calibration by the three methods,in which the RMSE values were reduced by 1.5mm and 2.68 mm using the linear regression calibration model,respectively.Using the nonlinear calibration model,the RMSE values were reduced by 2.33 mm and 2.9 mm.Using the BP neural network calibration model,the RMSE values were reduced by 2.86 mm,respectively.All three calibration models can improve the accuracy of MODIS PWV.The calibration effects of the nonlinear calibration model and the BP neural network calibration model are basically the same,and the BP neural network calibration model is a little more accurate and better than the fitting results of the linear regression calibration model.(5)Constructing the precipitable water vapor field in the Jinan region.Through the analysis of the mean monthly distribution of water vapor,the overall distribution of water vapor in Jinan region meets the trend of increasing from southeast to northwest.The distribution of precipitable water vapor is obviously influenced by factors such as altitude and precipitation.Analysis of water vapor distribution in different seasons shows that the water vapor variation in each season also meets the variation of high in the west and low in the east,the precipitable amount of water varies greatly between different seasons.Topography,water source,season and climate all affect the distribution of atmospheric precipitable water.By analyzing the distribution of water vapor and rainfall before and after a precipitation process,fused and calibrated PWV data can be used to understand the water vapor variation in the region and support meteorological applications.