Abstract
<title>Abstract</title> <p>Ionospheric disturbances significantly affect the performance of network RTK (NRTK), particularly at low latitudes where ionospheric gradients may vary rapidly across regional reference networks. This study evaluates the use of the I95 index, derived from double-differenced ionospheric residuals, as an empirical indicator of NRTK performance degradation. Full-year GPS data from a 13-station CORS network in Hong Kong for 2019 and 2024 were processed using a leave-one-out virtual reference station (VRS) processing strategy. The relation between I95 and NRTK performance was assessed by the quantile regression. The results reveal a systematic degradation of NRTK performance with increasing I95, characterized by lower ambiguity-fixing rates, broader positioning error distributions, and increased sensitivity of the upper error quantiles. These effects were markedly stronger in 2024 than in 2019. For the 2024 dataset, the probability of 3D-RMS errors exceeding 10 cm increased from 11.3% for I95 values of 0–1 ppm to 88.5% for values above 20 ppm. Over the same range, the mean cumulative positioning error increased from 0.07 to 1.56cm/km, corresponding to a factor of about 22. The fitted exceedance model yielded probability levels of 40%, 60% and 80% at I95 values of approximately 4.16, 8.28 and 14.83 ppm, respectively. Based on the combined exceedance and cumulative-error characteristics, four empirical performance regimes were defined for the Hong Kong network, with I95 intervals of <4, 4–10, 10–16 and >16 ppm. The 2019 results preserved the general ordering of these regimes but exhibited a weaker response and limited sampling at high I95 values. The derived thresholds provide a practical regional link between residual ionospheric monitoring and user-domain NRTK performance under low-latitude conditions.</p>