Spherical Cap Harmonic Modeling Of Water Vapor By Fusing Multi-Source Data From GNSS/MODIS/ERA5

Atmospheric precipitable water vapor(PWV)is an important parameter of the earth’s atmosphere,and its spatial and temporal changes affect the earth’s energy transfer and weather changes.PWV can be monitored and inverted by a variety of different modeling techniques,with different spatial and temporal resolutions and/or accuracies.In this thesis,the spherical cap harmonic model is used to fuse the PWV data from Global Navigation Satellite System(GNSS),Moderate resolution imaging spectroradiometer(MODIS)and European Centre for Medium-Range Weather Forecasts ERA5 reanalysis data to generate PWV data sets with high spatial resolution and high precision.The main contents and results of this thesis are as follows:(1)In this thesis,the PWV retrieved from sounding data is used to evaluate the PWV from GNSS data.It is shown that GNSS PWV and sounding PWV have comparable accuracy in the study area;and GNSS PWV is used to analyze the accuracies of ERA5 PWV and MODIS PWV.According to the analysis,ERA5 PWV has higher accuracy and more uniform spatial distribution than MODIS PWV.(2)Based on the annual and semi-annual periodic functions combined with ERA5 PWV,a regional MODIS PWV correction model is constructed,and the accuracy was compared with the MODIS PWV corrected by the linear model.The experimental results show that the periodic model is better than the linear model.The average error and root mean squared error are 3.22 mm and 5.63 mm,respectively,which proves that the periodic model has better applicability in the study area.Through seasonal analysis,the original MODIS data in winter is replaced with the periodic correction data set.The MODIS PWV dataset with higher accuracy is obtained.(3)The spherical cap harmonic function is used to construct the PWV fusion model,and the spherical cap harmonic coefficients are estimated by two methods,which are Helmert variance component estimation and Tikhonov regularization,respectively.The final fusion models from the two methods are compared;note that for both models best model orders have been determined by model selection methods.The results show that the fusion performance of Tikhonov regularization is better than that of the Helmert variance component estimation.The accuracy is evaluated by using the reserved GNSS PWV,and the average absolute errors and root mean squared errors are1.73 mm and 2.09 mm,for the two methods respectively.To sum up,the fused PWV with spherical cap harmonic model has higher accuracy and better spatial resolution than either MODIS PWV or ERA5 PWV.