Navigation risk assessment for the Three Gorges section of the Yangtze River based on multi-model integration

Since the completion of the Three Gorges Dam, navigation conditions in the Three Gorges section of the Yangtze River have become increasingly complex, and the assessment of navigation risk is critical to the safe operation of the reservoir area. Existing studies have primarily focused on environmental, managerial, vessel, and human factors, relying on theoretical assumptions and expert experience, which may introduce subjective bias. Additionally, previous assessment studies have not given sufficient attention to vessel navigation and accident data in the Three Gorges section of the Yangtze River. To address these problems, this study collects data on the waterway conditions of the Three Gorges section and expert judgments regarding its navigational conditions to establish a risk assessment indicator system. Subsequently, on the one hand, data from the Automatic Identification System (AIS) are collected to extract vessel count and design deadweight tonnage, which are incorporated into the risk assessment indicator system. The Analytic Hierarchy Process (AHP) is employed to determine the weights of individual evaluation indicators. Combined with the Fuzzy Comprehensive Evaluation (FCE) model, and through calculation of a three-tier evaluation matrix, the navigational risk value for the Three Gorges section using the original model is derived. On the other hand, accident data are systematically collected, cleaned, sorted, and subjected to preliminary statistical analysis. An Information Value-Logistic Regression (IV-LR) integrated model is then introduced to analyze the accident data, enabling quantification of the correlation between influencing factors and accident occurrence. These quantified correlations are integrated into the indicator system to form more objective factor weights, thereby improving the fuzzy comprehensive evaluation model. The three-tier evaluation matrix is then recalculated to obtain the improved navigational risk value for the Three Gorges section using the improved model. Finally, a comparative analysis is conducted on the navigational risk assessment results for the Three Gorges section obtained respectively from the original and improved models used in this study. The evaluation results of the original model showed that the overall navigation safety level of the Three Gorges section in 2024 was rated as “Basic Safety,” while the segmental risk assessment found that the safety levels of the Chongqing–Fuling section and the Fuling–Fengdu section were rated as “Needs Attention,” indicating that the model failed to fully reflect the actual conditions of local navigation segments in the overall risk assessment. The evaluation results of the improved model revealed that the integration of the IV-LR model indicated the overall navigational safety level of the Three Gorges section in 2024 was at the “Needs Attention” level, which more accurately reflected the actual conditions of local segments. The model also identified meteorological conditions, waterway complexity, and vessel design tonnage as major factors significantly affecting navigation safety. Additionally, safety awareness and the completeness of the supervision system were identified as potential high-risk factors requiring continuous attention. Factors such as vessel composition, anchorages, berthing areas, and port operation zones were classified as extremely high-risk sources in Section IV. However, their odds ratios showed negative correlations, indicating that existing control measures had already mitigated the risks associated with these factors. The results demonstrate that the improved model can provide evaluation outcomes that account for both overall and local risks, offering stronger practical guidance for navigation safety management in the Three Gorges section. This approach integrates objective accident data with traditional expert knowledge, thereby enhancing the objectivity of risk assessment. The proposed methodology provides valuable support for decision-making related to waterway safety management, early warning system development, and accident prevention strategies. Furthermore, this method is applicable to navigational risk assessment in other complex inland waterway systems, contributing to the broader field of waterway transportation safety research.