Abstract: To investigate the distribution characteristics of tunnel leakages and their correlation with construction methods, statistical analysis and comparative research based on measured data from operational subway tunnels in a city is systematically conducted in this paper. The objective is to reveal the differences in leakage distribution patterns under various construction conditions, providing a basis for tunnel structural health assessment and maintenance decision-making. A total of over 10 000 leakage samples are collected from tunnels constructed using the cut-and-cover method, the mining method, and the shield tunneling method. These samples cover multidimensional data, including defect type classification, seepage gap width, and construction methods. Frequency statistics are used to identify the characteristics of leakage types, gap width distribution patterns, and their correlation with construction methods. The results show that: ①in terms of defect types, the distribution of leakage defects in cut-and-cover and mining method tunnels are similar, primarily consisting of leakage throw construction joints, expansion joints, and cracks, accounting for approximately 94.60% and 96.28% of the total, respectively. In contrast, shield tunneling tunnels exhibit distinct leakage types, with segment joint leakage and hole leakage being the main types, accounting for about 79.84% of the total. ②In terms of gap width distribution, the width of construction joint leakage and crack leakage are both distributed in the range of 0.2 mm to 1.5 mm, with a peak frequency occurring in the 0.4 mm to 0.7 mm range, which accounted for 71.77% and 76.44% of the total samples respectly for construction joint leakage and crack leakage. It is shown that construction method had no significant impact on the range of leakage gap widths. ③Based on the statistical results, a simplified Bayesian network is established, with construction methods as parent nodes and leakage types, leakage crack width grades, and construction joint width grades as child nodes. This network can be used for inference calculations related to leakage defects. The study confirm that differences in lining structure characteristics caused by different construction processes are important reasons for the variation in leakage defect types and forms. The findings provide directions for optimizing construction processes to prevent leakage in subway tunnels, guide the prioritization of defect prevention under different construction methods, and offer practical value for defect risk classification and maintenance resource allocation through its Bayesian association network.