Ground-based GNSS Water Vapor Inversion And Construction Of Three-dimensional Tomography Model

As an unstable component in the atmosphere,water vapor affects the changes of weather all the time.Therefore,the distribution and change of water vapor is a very important information that meteorology pays attention to.With the continuously development of satellite navigation and positioning technology,the application of Global Navigation Satellite System(GNSS)is no longer limited to PNT(Positioning,Navigation,Timing)services.GNSS meteorology is a new application field which be expand by the combination of the navigation and positioning technology and meteorology.Although GNSS meteorology has gradually matured after considerable development,there are still many problems to be solved urgently.Based on the continuous operating reference stations(CORS)in Hong Kong,We will gradually conduct in-depth research from two-dimensional ground-based GNSS water vapor detection to three-dimensional ground-based GNSS water vapor detection and be directed against the problems found in the research process to discuss and solve it.The specific problems and the research contents and results obtained are aimed at these problems are as follows:(1)The weighted mean temperature(T_m)is an important variable in the ground-based GNSS water vapor inversion process.For the sake of simplicity,T_m is generally calculated by the model be established between T_m and other meteorological factors which have strong correlation with T_m,but the premise of the model establishment is to have an accurate sample value of T_m.The calculation of the sample value is calculated by numerical integration using high-altitude weather data.Saturated Water Vapor(es)as an important variable in the calculation formula of T_m numerical integration affects the accuracy of the T_m sample value.The error of the T_m sample value will cause the reliability of the T_m model to decrease,so this paper discusses and compare the influence of different saturated water vapor pressure models involved in the calculation on T_m and the finally retrieved Precipitable Water Vapor(PWV).(2)The existing T_m models are mainly divided into single-factor T_m model and multi-factor T_m model,but the existing single-factor models do not fully consider all the atmospheric parameters that have strong correlation with T_m and the multi-factor models do not take into account the multicollinearity among the independent variables in the model.The multicollinearity among independent variables will affect the stability of the model and result in large differences between the coefficients of independent variables of the T_m model established using the sample values of different periods,and even find the positive and negative of coefficients is inconsistent with the positive and negative correlations among independent variables and dependent variables.In response to the above problems,the optimal variable selection criteria are used to select independent variables that can make the model optimally fit,and the ridge regression method is used to weaken the multicollinearity among independent variables to establish a stable multi-factor T_m model.(3)Ground-based GNSS three-dimensional water vapor detection is based on two-dimensional water vapor detection,using slant wet delay(SWD)which be calculated through GNSS observation equation and tomography technology to detect the distribution and changes of water vapor at different altitudes.In this paper,the traditional method of calculating the intercept of the GNSS signal line by the voxel block is abandoned.The new solution is to project the GNSS signal line to a two-dimensional projection surface to obtain the intersection point information,and then according to the two Dimensional intersection information to reconstruct the line information in three-dimensions.According to the new idea,proposing a straight line voxelization method based on three-dimensional reconstruction of a space straight line.(4)Aiming at the problem that there is no GNSS signal line passing through some voxel blocks in the tomographic area near the edge area,this paper proposes an adaptive selection method of the tomographic area boundary,which determines the tomography Boundary of each layer according to the real-time distribution of GNSS signals.This method reasonably eliminates the non-informative voxel blocks in the tomographic area,reduces the amount of calculation,and avoids the problem of calculating the water vapor parameters compulsorily in the voxel blocks with no signal to pass through the voxel based on the spatial correlation.