Study On Multi-source Precipitable Water Vaper Fusion Method And Its Climatic Applications In China

Water vapor is one of the most important greenhouse gasses,the variation of PWV content is an important factor that deeply influences the global climatic sensitivity.Precipitable water vapor(PWV)is the most used indicator for quantifying the content of water vapor.Therefore,it’s of great importance to the comprehensive analyses of the spatiotemporal characteristics of PWV and the further cognition of the atmospheric variation under the context of global warming.With the rapid development of space geodesy techniques,a large amount of PWV products derived from different techniques have been effectively accumulated.However,there are obvious discrepancies among multi-source PWV products.The fusion of multi-source PWV data can make full use of the advantages and alleviate the disadvantages of different PWV products,which is also beneficial to the comprehensive monitoring of PWV as well as the in-depth research of climate change.Accordingly,this dissertation focuses on the fusion of multi-source PWV data based on the PWV derived from ground-based GNSS,radiosonde,satellite remote sensing images and reanalysis dataset.The main research tasks and conclusions are summarized as follows:(1)Homogeneity test for the GNSS-PWV time series in China.To detect and correct the potential inhomogeneity in GNSS-PWV time series,a new method named“adaptive absolute homogeneity test(AAHT)” is proposed by combining STL and PMFred methods.The success rate of AAHT is better than 74.9%,the false alarm rate is less than 12.2%,and the accuracy of the detected changepoint is 8.3 days.Moreover,the homogeneity of the GNSS-PWV time series over 207 GNSS stations from 2008 to2018 in China is tested using AAHT,and the origins of the detected changepoints are verified with the help of GNSS coordinate and ERA-PWV time series.The test results indicate 110 inhomogeneous GNSS-PWV time series,containing 193 non-climatic changepoints.Among these non-climatic changepoints,5 and 44 are related to the changes in hardware and geological activities.Finally,the values and directions of the offsets at these non-climatic changepoints are estimated and further used to correct these inhomogeneous GNSS-PWV time series.A comparison with the ERA-PWV over the same GNSS stations illustrates that the homogenization can reduce 22% of the discrepancy between ERA-PWV and GNSS-PWV and improve the correlation coefficient of the linear trends of these two PWV datasets from 0.75 to 0.98.(2)Performance evaluation of the multi-source PWV time series.The GNSS-PWV,ERA-PWV and RS-PWV from 2008 to 2018 in China are firstly compared to calculate the correlation coefficient,bias and RMS of the difference between each two of the three PWV time series.The comparison shows a good agreement among these three PWV time series and a slight dry bias in the RS-PWV.Then,the magnitudes of the observation noise in these three PWV are calculated based on the statistical results,which are 0.49,1.26 and 1.09 mm,respectively.Next,the performance of PWV derived from MODIS Near-Infrared PWV(MOD-NIR-PWV)is evaluated based on the hourly GNSS-PWV and ERA-PWV time series from 2003 to 2018.The RMSs are 3.8 and 3.0mm,and the biases are 2.9 and 2.1mm for GNSS-PWV and ERA-PWV,respectively,which illustrates a significant wet bias of MOD-NIR-PWV in China.To reduce the systematic bias,a gridded calibration model is constructed based on the differences between hourly ERA-PWV and MOD-NIR-PWV over the grid points.After the calibration,the RMSs are decreased to 1.0 and 1.6 mm with respect to ERA-PWV and GNSS-PWV,which are reduced by 71% and 53%,respectively.(3)Investigation of the multi-source PWV fusion method and construction of the fused PWV dataset.The abovementioned homogenization and calibration provide highprecision original data for the PWV fusion.The spherical cap harmonic(SCH)model is used as the function model to fuse the daily time series of the homogenized GNSSPWV,calibrated MOD-NIR-PWV and ERA-PWV.The degree of the SCH model is set to 60 for a better representation of the detailed spatial characteristics of PWV.To solve the multi-collinearity of the model parameters and the potential ill-posed problem of the normal equation when using the least-squares method,the elastic net regularization method is introduced to estimate the SCH model parameters.Then,an initial fused PWV time series with a spatial resolution of 0.05° is generated based on the fusion model.Next,principle component analysis is used to extract the potential unmodeled signals in the SCH modeling residuals.Finally,the extracted unmodeled signals are used to correct the initial fused PWV time series for the construction of the 0.05°×0.05°fused PWV time series with high precision and spatiotemporal continuity in China.To evaluate the performance of the fused PWV time series,they are compared with the original modeled data(16897 ERA-PWV and 216 GNSS-PWV time series)and test data(20 ERA-PWV and 8 GNSS-PWV time series).Results show that the precision is0.94 and 1.79 mm with respect to the modeled ERA-PWV and GNSS-PWV,and the accuracy is 0.82 and 1.87 mm with respect to the test data,respectively.(4)A new method is developed to effectively extract the time-varying signals in the PWV time series.The seasonal asymmetry and inter-annual difference in PWV are pointed out as the main factor of high RMS of the modeling residuals based on the conventional time-constant harmonic model that is solved by the ordinary least-square estimator(OLS-based model).Therefore,a time-varying harmonic model is proposed considering the temporal variations of multiscale signals in PWV time series,which is solved by a total Kalman filter(TKF-based model).The modeling results based on the new TKF-based model are compared with the OLS-based model using the GNSS-PWV time series over 91 global IGS stations from 2000 to 2018.The results show that the TKF-based model can reduce the RMS of modeling residuals by 14%.(5)The spatiotemporal characteristics are analyzed based on the fused PWV time series and TKF-based time-varying harmonic model.Results show that the mean values of PWV in the study period are decreased from the southeast to the northwest region.The maximum and minimum value of mean PWV is about 50 mm in south China and less than 10 mm in the Tibetan plateau.The trend of PWV time series over China in the study period varies from-0.52 to 0.57 with a mean of 0.17 mm/year.The relative trend varies from 1.8～5.0 with a mean of 1.3 %/year.The results indicate a positive trend in the PWV time series in China.In addition,the annual periods are the main component in the PWV time series in China,the mean ratio of the magnitudes of the annual periodic variations to the total variations in PWV is about 76.7%.Based on the modeling results using the TKF-based time-varying model,the annual and semi-annual periodic amplitudes anomaly time series are calculated by subtracting the OLS-based timeconstant amplitudes from the TKF-based time-varying amplitudes,and the correlations with Oceanic Ni?o Index(ONI),a predictor of the ENSO phenomenon,are calculated.Results show that the mean correlation coefficients are 0.48 and 0.41 for the annual and semi-annual periodic amplitude anomaly,respectively.This suggests that ENSO is one of the reasons for the time-varying signals in PWV over China.In addition,the correlation coefficients in southeast coastal regions are significantly positive(0.6 to0.8),and the response of the annual periodic amplitude anomaly is faster than ONI with a lag time from 7 to 12 months.These results suggest that the periodic amplitude anomaly in PWV time series can be a new predictor of the ENSO phenomenon.This dissertation aims at dealing with several key issues in the climatic applications based on multi-source PWV data,including the accuracy improvement of original PWV data by the homogenization of GNSS-PWV and the error calibration of MOD-NIR-PWV,the fusion of the multi-source PWV based on the regularization and bias correction,the extraction of the time-varying PWV signals considering the temporal variations of PWV time series over China,the correlation analyses of the PWV variations and the ENSO phenomenon.It is believed that this research can provide some valuable theoretical and application references for the extension of the GNSS applications and the analysis of the mechanism of the PWV and climatic change.This dissertation contains 64 figures,17 tables,and 185 references.