Error analysis of ionospheric VTEC inversion over China via a single-parameter directional perturbation method

Abstract

Disentangling individual error sources in regional ionospheric modeling remains difficult because multiple uncertainties are strongly coupled during the Vertical Total Electron Content (VTEC) inversion. To address this issue, we develop a single-parameter directional perturbation framework and apply it to GNSS observations from more than 250 stations of the Crustal Movement Observation Network of China (CMONOC), with the aim of identifying and quantifying the dominant drivers of inversion error. The results show that phase-smoothed pseudorange (P₄) noise and elevation mask angle (EMA) dictate inversion stability, revealing a critical trade-off between low-elevation noise suppression and spatial coverage. Specifically, increasing P₄ noise (0.1 to 0.5 m) escalates the Mean Absolute Error (MAE) from 0.88 to 5.66 TECu (1 TECu = 10¹⁶ el⋅m⁻²), while raising the EMA (20° to 30°) increases the MAE from 0.44 to 1.05 TECu. Furthermore, the Spherical Harmonic Order (SHO) exhibits a non-linear impact: the K=M=8 configuration is optimal, whereas over-parameterization (K=M=10) triggers severe boundary oscillations, driving the MAE up to 19.45 TECu under weak observation geometry. Our results also indicate a strong environmental dependence of systematic biases. In the low-TEC morning sector (06:00–08:00 LT), Differential Code Bias (DCB) perturbations inflate the MAPE to 24.85%, and this sensitivity is further amplified during geomagnetic disturbances, causing a significant drop in inversion accuracy (MAE=0.60 TECu). Under these same conditions, the optimized high-order SH scheme proves its robustness, maintaining an MAE of only 0.08 TECu. These results provide a quantitative assessment of the error budget in regional VTEC inversion and offer practical benchmarks for improving regional ionospheric modeling under different space weather conditions.