A data-driven approach for determining navigation waters influenced by bridges – A framework and a case study

This study proposes a quantitative method for determining the bridge area waters using Automatic Identification System (AIS) data. A multi-dimensional behavioral feature analysis framework is developed, systematically integrating trajectory preprocessing, centerline extraction, and behavioral feature modeling. The methodology employs Gaussian Process Regression (GPR) to extract a unified navigation centerline that represents the dominant vessel passage path. Subsequently, Kernel Density Estimation (KDE), acceleration analysis, and course difference analysis are applied to describe vessel spatial distribution, speed variation, and course adjustment near the bridge. A comprehensive case study of the Nanjing Yangtze River Bridge demonstrates that vessels exhibit noticeable changes in spatial distribution, speed variation, and course adjustment within specific areas approaching the bridge. Based on the analysis of these behavioral characteristics, the bridge area waters arequantitatively identified and determined. The bridge area waters for downstream navigation extend approximately 2.5 km upstream and 2 km downstream from the bridge axis, while for upstream navigation, it ranges from 2 km downstream to 1 km upstream of the bridge. This study provides a scientific basis for intelligent navigation supervision and risk prevention in bridge areas, offering data-driven decision support for maritime safety management. • A data-driven framework is developed to quantitatively determine bridge-influenced waters. • Spatial, speed, and course behaviors are modeled to identify significant vessel maneuvering changes. • Case study reveals upstream and downstream bridge area waters with clear quantitative boundaries. • The method provides a replicable basis for intelligent supervision and risk prevention in bridge areas.