Research On Regional Ionospheric Modeling Of Ground-Based GNSS

Ionospheric delay limits the measurement and positioning accuracy of single-frequency receivers,and slows down the single-point positioning convergence speed of dual-frequency receivers.When the satellite's elevation angle is low,the measurement error caused by the ionosphere can be up to meters or even hundreds of meters.Therefore,it is of great theoretical significance and practical value to carry out research on ionospheric modeling in order to reduce measurement and positioning errors.This paper intensive studies the modeling of regional ionosphere based on ground-based GNSS observation data.The main research works and results are summarized as follows:(1)Establish a regional ionospheric grid model.Based on the polynomial ionospheric model and GNSS observation data,the ionospheric grid models of China's land area(15°N?55°N,700°E?140°E)on April 25,2015 and Oceania region(10°S?450°S,113°E?154°E)on December 5,2016 are established,respectively.(2)Study the influence of internal factors such as model order and satellite elevation angle on the modeling accuracy,model correction rate and modeling time of regional ionosphere to determine the best modeling method.The results show that increasing the model order in China's land area can improve the modeling accuracy and correction rate,but the operation time increases,and the 7*8 order model is the best in this region.Increasing the satellite cut-off height angle can reduce the modeling time and improve the modeling accuracy,but the model correction rate is decreasing,when the cutoff altitude angle of satellite is set to 15° the best model is built.Increasing the model order in the Oceania region can also improve the accuracy of modeling,but the model correction increases first and then decreases,and the operation time also increases,and the 3*3 order model is the best in this region.The results of establishing these two ionospheric grid models based on the best modeling strategy show that:the distribution of VTEC increases gradually from high latitude to low latitude areas;the accuracy of modeling in the middle region is higher than that in the edge region,the marginal effect in polynomial model.(3)Sudy the measured data,CODE and other external data to evaluate the model.According to the interpolation function,the VTEC under the model built by the GNSS checkpoint is compared with the data from the actual measurement,CODE,IGS,JPL,and the institute of seismology of China Earthquake Administration to evaluate the external accuracy and reliability of the model.The results show that the modeling consistent with the measured data and data published by each center,and the model is closer to the measured data.The model built in China's land region has higher accuracy and model correction rate than the ISCEA and JPL,but slightly lower than CODE.The model built in the Oceania region has higher accuracy and correction rate than CODE,IGS,and JPL.(4)Establish ionospheric grid model based on GPS/GLONASS combined data,and analyze its modeling accuracy and advantages and disadvantages to explore the application of multi-source data fusion in ionospheric modeling.The results show that GPS/GLONASS combined modeling can express more detailed ionospheric distribution characteristics and reduce the marginal effect of polynomials to a certain extent.However,its modeling accuracy and model correction rate are lower than GPS modeling.(5)Propose an improved polynomial ionospheric modeling method based on Wavelet Neural Network.We calculate the VTEC difference between the measured data and the model parameters,and predict the VTEC difference of the grid points in the modeling area by Wavelet Neural Network to improve the polynomial model.The results show that,the improved model can improve the overall correction rate of the model,so that the average correction rate of inspection stations in Oceania can be increased from 91.44%to 92.9%.The improved model can describe more details of ionospheric spatial distribution and reduce the marginal effect of polynomial modeling to a certain extent.However,the modified model destroys the overall regularity of the polynomial model and makes VTEC excessively uneven in the changing region.In addition,the wavelet neural network has a large prediction error for larger values,so this method can only improve the polynomial model within a limited range.