Support Vector Machine for Regional Ionospheric Delay Modeling

Area of Science:

• Geophysics
• Space Physics
• Atmospheric Science

Background:

• Ionospheric total electron content (TEC) distribution is complex and difficult to model accurately.
• The traditional polynomial (POLY) model struggles to represent TEC variations, especially during active ionospheric periods.

Purpose of the Study:

• To develop an improved ionosphere modeling method that overcomes the limitations of the regional POLY model.
• To enhance the accuracy of TEC modeling and its application in precise point positioning (PPP).

Main Methods:

• Established a regional POLY model using continuously operating reference station (CORS) observations.
• Developed a support vector machine (SVM) regression model to compensate for POLY model errors.
• Combined POLY and SVM models to create the novel TEC SVM-P model.

Main Results:

• The SVM-P model achieved a root mean square error (RMSE) of 0.980 TEC units, a 17.3% improvement over the POLY model (1.185 TEC units).
• Single-frequency precise point positioning (PPP) accuracy improved by over 40% using SVM-P models compared to POLY models.
• The SVM-P model demonstrated superior performance compared to the back-propagation neural network combined with POLY (BPNN-P) model (1.070 TEC units).

Conclusions:

• The proposed SVM-P model offers a significant advancement in regional ionosphere modeling.
• This method enhances the accuracy and reliability of TEC predictions for geodetic applications like PPP.
• The SVM-P model presents a more effective approach for ionosphere modeling than traditional POLY or BPNN-P methods.