A data-driven framework for navigation risk identification and assessment in complex bridge waters using AIS trajectories

Introduction Navigation risk assessment in complex bridge waters is challenged by intersecting traffic flows, localized risk hotspots, and inconsistencies between annual vessel movements and collision risk. Methods AIS trajectories collected from 2022 to 2024 around Zhanjiang Bay Bridge and Tiaoshun Bridge were processed and matched to route legs. An AIS-informed route network was reconstructed using traffic density and observed crossing, merging, and diverging patterns. IWRAP MKII was used to estimate ship -ship collision and ship -bridge allision frequencies. A unified fixed-route network and sequential control scenarios were used to examine interannual differences. Results AIS-observed vessel movements decreased from 7,199 in 2023 to 6,613 in 2024, whereas ship -ship collision frequency under the fixed-route network increased from 0.037870 to 0.062410 events/year, an increase of approximately 64.80%. Cumulative route-leg passage frequency increased by 31.36%. High-risk route legs and waypoints were concentrated near bridge approaches and intersections between the main channel and port-access routes. AREA-6188 and AREA-6190 exhibited relatively high ship -bridge allision frequencies. In the sequential control scenarios, adjusting annual vessel movements alone reduced the modeled collision frequency, while subsequent adjustments to ship-type composition and other AIS-derived characteristics produced smaller incremental changes. The C4 estimate was 0.064760 events/year, approximately 3.8% higher than the C5 reference value of 0.062410 events/year. Discussion The results indicate that annual vessel counts alone do not fully represent traffic exposure within the route network. The proposed framework supports retrospective risk-hotspot identification and the prioritization of traffic management and bridge-pier protection, rather than real-time, vessel-specific collision prediction.