Development and Validation of a Multi-Source Data-Fused Empirical TEC Model for North America with a Winter Anomaly Correction

Existing global ionospheric models have limited capability in representing the winter anomaly over North America. This study develops a North American Ionospheric TEC Empirical Model (NAI-TECEM) based on multi-source TEC data fusion. GPS-TEC, CODE-GIM, IRI2020, Jason-TEC, and COSMIC-TEC data from 2010 to 2021 were integrated using 15th-degree and 15th-order spherical harmonic expansion with Helmert variance component estimation. Under geomagnetically quiet conditions, a 19-parameter model was constructed using F10.7p (the average of the daily F10.7 value and its 81-day centered moving average), local time, day of year, geographic latitude, and modified dip latitude to describe diurnal, seasonal, geomagnetic latitude, solar activity, and winter anomaly variations. For internal consistency, parameters were estimated using nonlinear least squares with the Levenberg–Marquardt algorithm, yielding a mean residual of 0.42 TECU and an RMSE of 3.29 TECU. Additional statistical diagnostics and sensitivity analyses confirmed the robustness and stability of the estimated model coefficients. Pearson correlation coefficients between the fitted model and the fused TEC dataset exceed 0.95 at mid-latitude test points and 0.92 at high-latitude test points, indicating strong internal consistency. For external performance, three independent GPS stations (TUKT, IQAL, and NWOT) were used to compare observed TEC, modeled TEC, and reference products, together with a Winter Anomaly Index (WAI) to quantify seasonal anomaly intensity. The model achieved superior agreement with observations compared to IRI2020. Additional sensitivity tests under representative extrapolation and extreme-condition scenarios showed that the modeled TEC remained physically plausible, further supporting the robustness of NAI-TECEM beyond its calibration conditions. Validation for 2022–2024 further confirms that the model accurately reproduces the spatial structure, solar-activity dependence, and WAI-based winter anomaly characteristics over North America.