Study And Methodology For Real-time Ionosphere Modeling

The ionosphere plays an important role in the space environment between the sun and the earth,which can influence the human life and production.The electromagnetic wave broadcasted by Global Navigation Satellite System(GNSS)can be affected by free electrons in the ionosphere,which will induce a few meters to tens of meters delay.It has been one of the main error sources that restrict the high-precision positioning of GNSS users,especially for single-frequency GNSS users.The Total Electron Content(TEC)is a key parameter to describe the variation of ionosphere.The construction of real-time ionosphere model can provide ionospheric delay corrections for real-time navigation and positioning users,improve the convergence time of Precise Point Positioning(PPP)and realize the ionosphere monitoring for space weather.Recently,the International GNSS Service(IGS)Ionosphere Analysis Centers have begun to pay attention to construct the real-time ionosphere model,and some preliminary results have been made.Therefore,modeling the ionosphere in real-time has become a hot topic for ionospheric research.To construct real-time ionosphere model,this paper studied the construction of ionosphere model by normalizing multi-source ionospheric data and the construction of a high-precision ionosphere prediction model.The main research contents and contributions are listed as follows:(1)In order to understand the accuracy of IRI-2016 model and NeQuick2 model for a long period,which is used to determine whether they can meet the requirements of normalization,different methods were employed to evaluate the accuracy of IRI-2016 model and NeQuick2 model from day of year(DOY)001,2002 to DOY365,2018.The ionosphere final product provided by IGS(IGSG)is used for assessment as well.The results showed that the accuracy of IRI-2016 model and NeQuick2 model is nearly the same,but both worse than IGSG.The mean RMS values of dSTEC assessment method for IGSG,IRI-2016 and NeQuick2 model were1.8,2.7,and 2.6 TECU,respectively.Meanwhile,the values of IRI-2016 and NeQuick2 model were smaller than that of IGSG.The RMS values of them with respect to Jason-2/3 VTEC were 4.8/6.2,6.0/6.0,and 5.9/6.0 TECU,respectively.(2)To achieve the optimal fusion for different kinds of ionospheric observation data and overcome the limitation caused by the traditional methods,which treat the values of un-observational ionospheric region as a constant.In this study,the single-layer normalization method is proposed.For this method,the NeQuick2 model is used to calculate the vertical ionospheric observation data at corresponding ionospheric pierce points for different observation ranges.Then,the scale factors of different observation ranges are used to make different kinds of ionospheric data have the same observation ranges.Meanwhile,the normalized data from January 1,2015 to September 27,2015 were used to construct the Global Topside Ionospheric Model(GTIM)by different strategies,which is used to verify the feasibility of the presented method.The results showed that the accuracy of GTIM constructed by using LEO satellite observation at different altitudes(MA-LEO)is better than that by using LEO satellite observation data at similar orbital altitudes(SA-LEO)and can better reflect the variation of topside ionosphere.The RMS values of LEO-based GPS receiver DCB(LEO DCB)at 500 and 800 km estimated by MA-LEO has an improvement of23.1% and 40.7% compared with those estimated by SA-LEO,respectively.For GTIM-500,the RMS values of GPS DCB estimated by MA-LEO decreased about12%.For GTIM-800,the RMS values of estimated GPS DCB reduced about 20%.Meanwhile,accuracy of GTIM estimated by MA-LEO was improved 24.2%.(3)The estimation method and processing steps which were used to model the ionosphere by using different kinds of ionospheric data were described in detail.Meanwhile,global ionosphere model from DOY 050,2017 to DOY 080,2017 were constructed by the ground-based multi-GNSS observation data(Ground-Based)and the combination of normalized multi-source data by single-layer normalization method and the ground-based multi-GNSS observation(Multisource),respectively.The results showed that the accuracy of the ionospheric model constructed by the Multisource solution was better than that by Ground-Based solution.The differences for ionospheric models constructed by the two solutions mainly ranged from-1.0 to0.0 TECU and appeared over the oceanic region.The mean RMS values of the ionospheric model constructed by Multisource solution over the ocean is 5.03 TECU,which is about 15% smaller than that of Ground-Based solution.At the same time,there was little differences between the estimated satellite DCB and receiver DCB for different solutions.The RMS value of dSTEC assessment method was 2.41 TECU for Multisource solution,which was 3.60% smaller than that of the Ground-based solution.The maximum improvement was about 15% compared with that of the Ground-Based solution.(4)The principle of the semi-parametric model and the way to predict TEC values were introduced in detail.Meanwhile,the Long Short-Term Memory(LSTM)network and its processing steps to predict the ionospheric spherical harmonic coefficients were introduced as well.Subsequently,different methods were used to predict the TEC values from DOY 005,2020 to DOY 100,2020.The results showed that the accuracy of ionospheric prediction model constructed by semi-parameter model was slightly better than that by LSTM network.For different ionospheric prediction models,the deviations between the ionospheric prediction products provided by CODE and the GIM file provided by CODE(CODG)as well as the semi-parametric TEC prediction model and CODG are nearly the same,which were slightly better than that between the prediction model estimated by LSTM network and CODG.The deviations between different ionospheric prediction models and CODG were in accordance with normal distribution and more than 97% of predicted residuals are between-5 and 5 TECU.The correlations between one day prediction results and CODG are larger than 0.95,and those between two days prediction results and CODG are larger than 0.94.(5)To understand the accuracy of the real-time ionospheric products provided by different IGS ionosphere analysis centers comprehensively,the accuracy of them from DOY 001,2017 to DOY,365 2018 was assessed.The results showed that the accuracy of the real-time ionospheric product is worse than that of the one day ionospheric forecast product provided by CODE(C1PG),and the accuracy of the product is not stable.The real-time ionospheric product provided by the Chinese Academy of Sciences(CAS RT-GIM),Wuhan University(WHU RT-GIM),the Polytechnic University of Catalonia(UPC RT-GIM)and C1 PG has a good agreement with IGSG.Their differences with respect to IGSG were-0.97,-0.90,-0.77,and-0.80 TECU,respectively.When compared with the Jason-2/3 VTEC,the RMS values of UPC RT-GIM,CAS RT-GIM,WHU RT-GIM,C1 PG,and IGSG were 4.44/5.71,3.59/5.21,3.71/5.28,3.64/5.28,and 3.70/5.70 TECU,respectively.The RMS values evaluated by dSTEC method were 4.02,1.72,2.20,1.86,and 1.49 TECU for corresponding products.Finally,different real-time ionospheric products were used to correct the ionospheric delay,and the deviation of kinematic SF-PPP was decimeter level in the horizontal direction as well as meter level in the vertical direction.(6)The estimation method and modeling processing steps as well as the strategy for real-time ionosphere modeling were introduced.Meanwhile,the real-time ionospheric data derived by the zero-differenced ambiguity fixed solution and the real-time ionospheric prediction model from DOY,300,2020 to DOY 366,2020 were used to construct real-time ionosphere model(SGG RT-GIM)together.At the same time,the differences between the ionospheric model constructed by the ionospheric data derived from the zero-differenced ambiguity fixed solution method(SGG Post-GIM)andSGG RT-GIM were calculated.The results showed that the mean deviations between SGG Post-GIM and IGSG as well as SGG RT-GIM and IGSG were-0.45 and-0.57 TECU,respectively.Meanwhile,the RMS values evaluated by dSTEC method were 1.19,1.15,and 1.57 TECU for IGSG,SGG Post-GIM,andSGG RT-GIM,respectively.Finally,IGSG,SGG Post-GIM,andSGG RT-GIM were used to correct the ionospheric delay and the static SF-PPP was employed for test.The deviations in the three dimensional direction were 0.3?0.6,0.3?0.6,and 0.3?0.6 m,respectively.Their deviations were 0.14,0.19,and 0.21 m in the horizontal direction,respectively,while the corresponding values were 0.36,0.33,and 0.38 m in the vertical direction.While the ionospheric delay was corrected by Klobuchar model,deviation in the three dimensional direction were 0.6?0.9 m,and the deviation is 0.25 m in the horizontal direction,and 0.64 in the vertical direction.