Inversion Method for GNSS Interference Spatial Extent Based on Multi-source CORS Observations

Objectives: Large-scale GNSS navigation anomalies degrade positioning, navigation, and timing (PNT) over dense urban environments and stress applications that depend on open civil signals. After the major navigation anomaly event, operational stakeholders still need reproducible, observation-based evidence on how different constellations, frequency bands, and observation types respond, together with a geographic depiction of where signal conditions deteriorate most severely. The aim is to build a CORS-network workflow that automatically flags anomalous epochs, quantifies interference severity in one consistent scoring framework, and outputs estimated interference centers, effective spatial extent (radius), and a confidence measure suitable for monitoring and contingency planning. Methods: Full-day (1 Hz) multi-constellation, multi-frequency GNSS data from twelve CORS sites were synchronized to a common time base. Interference detection combines an elevation-dependent dynamic carrier-to-noise-density (C/N₀) threshold with a composite interference score that sums C/N₀ shortfall, multipath excess above a 60 cm threshold, band-count shortfall with N_min = 1, and observation-gap penalties. Gap logic suppresses false alarms from normal satellite masking by requiring mid-sky geometry (15°-85° elevation at both gap endpoints), gap durations between 0.5 and 30 minutes, and at least five consecutive missing epochs for outage-style events. Events are aggregated in fixed-length windows (e.g., 30 minutes), and spatial solutions are computed only when at least two stations contribute events in the same window. Interference centers are obtained from C/N₀- multipath-band-gap score-weighted station coordinates and, whenever at least three stations participate, complemented by azimuth-ray intersections using up to the four highest-score sites. The influence radius is the maximum great-circle distance from the estimated center to affected stations; a bounded confidence index blends the fraction of participating stations and the mean interference score in each window. Results: During the main disturbance interval, GPS and BeiDou lose multifrequency tracking and visible satellites, GLONASS band availability weakens, whereas Galileo largely preserves stable multifrequency service. At site S05, severe C/N₀ loss, band-count collapse, and episodic multipath spikes up to hundreds of centimetres coincide on GPS L1/L2/L5-class signals plus GLONASS G1 and BDS B1, indicating power-dominated suppression with intermittent tracking rather than benign environmental multipath alone. Automated processing detected 16 387 interference events network-wide and produced five high-confidence spatial-extent solutions. Spatially, 16 279 events cluster at S05 (mean score 0.93), while other sites register only scattered hits. Constellation-wise, GPS contributes 11 389 events (mean score 1.19) versus 1 210 for GLONASS and 3 788 for BeiDou. Band-count-below-threshold cases dominate (11 274 events), followed by C/N₀ gaps (5 760) and multipath gaps (3 431), underscoring band suppression and signal interruption as the leading signatures. Sky plots and heatmaps align the strongest disturbance with the southwestern urban sector, radiating along multiple satellite arcs and decaying outward from the S05 neighborhood. Conclusions: Coupling elevation-adaptive C/N₀ gating with unified multipath, band-count, and gap scoring yields a transferable pipeline for cataloguing realworld interference manifestations from dense CORS archives. Windowed multi-station aggregation with weighted centroids, azimuth intersections, explicit radii, and confidence metrics furnishes quantitative spatial-extent products that support interference localization studies, spectrum enforcement dialogue, and resilient PNT operations in future urban anomalies.