热带地理2025，Vol.45Issue(8)：1344-1357,14.DOI:10.13284/j.cnki.rddl.20240777
多尺度道路网络特征对交通事故的非线性效应研究
Nonlinear Effects of Multiscale Road Network Features on Traffic Accidents
摘要
Abstract
Amidst rapid motorization,urban road traffic safety has become increasingly critical.As a fundamental framework of traffic systems,road networks strongly influence the occurrence of traffic accidents.The impacts of urban road networks on traffic accidents exhibit considerable spatial heterogeneity.However,the nonlinear relationship between road networks and traffic accidents,and the moderating effect of land use,have been overlooked.We focused on the main urban area of Wuhan and used a Gradient Boosting Decision Tree model(GBDT)to explore the nonlinear relationships between road networks and traffic accidents,while also identifying the synergistic effects of land use and road networks on traffic accidents.The geometric and topological features of urban road networks generally have nonlinear and threshold effects on traffic accidents.The total contribution rate of road network characteristics to traffic accidents reached 51.2%,highlighting their key role in traffic accidents.Average geodesic distance,mesh density,and global clustering coefficient showed a nonlinear positive correlation with traffic accidents.Meanwhile,the density of the road network and the density of intersections showed a nonlinear negative correlation with traffic accidents.The assortativity index showed a nonlinear U-shaped relationship with traffic accidents.Closeness centrality,betweenness centrality,and detour rate have nonlinear effects on traffic accidents,with spatial scale differences observed at 400,1 700,3 000,and 7 500 m.The influence of the 7 500 m scale is greater.In terms of near centrality,there is a nonlinear positive correlation and threshold effect between near centrality at the 400 and 1 700 m scales and the number of traffic accidents.In contrast,the near-centrality at the 3 000 m scale has an approximate linear positive correlation with the number of traffic accidents.Meanwhile,the near-centrality at the 7 500 m scale has a right-biased U-shaped relationship.For the betweenness centrality,the betweenness centrality of 400,1 700,and 3 000 m scales also showed a nonlinear positive correlation and threshold effect.However,the betweenness centrality of 7 500 m scale and the number of traffic accidents showed a U-shaped relationship.In terms of the detour rate,there is a nonlinear negative correlation and threshold effect between the detour rate at the 400 m and 7 500 m scales and the number of traffic accidents.There is a non-linear positive correlation and threshold effect between the detour rate of 1 700 and 3 000 m and the number of traffic accidents.Land use and road networks exhibit complex synergistic effects on traffic accidents.First,moderate land use mixing produces positive traffic safety benefits under the combined effects of road networks.However,when the land-use mixing degree exceeds a certain threshold,excessive mixing reduces the positive traffic safety benefits caused by road networks.Second,there was a positive correlation between floor area ratio and traffic flow.Increasing the floor area ratio may inhibit the positive effect of road network characteristics on traffic safety.However,on a global scale,traffic flow in a high plot ratio area may be more positively affected by the optimization of road network characteristics,thus reducing the risk of traffic accidents.This study used machine-learning methods to examine the nonlinear relationships between urban road network characteristics and traffic accidents.By identifying nonlinear,scale,and threshold effects,this study overcomes the limitations of traditional studies.It not only enriches the dimensions and depth of research on the relationships between road networks and traffic accidents,but also accurately identifies the thresholds of key road network indicators,providing strong support for the in-depth analysis of the driving mechanisms behind traffic accidents.This study further investigated the synergistic effects of urban land use and road networks on traffic accidents and proposed collaborative optimization strategies.These strategies offer theoretical guidance and technical support for urban road traffic planning and contribute to the enhancement of urban traffic safety.关键词
道路网络/交通事故/土地利用/非线性关系/尺度效应/梯度提升决策树模型/武汉市主城区Key words
road network/traffic accidents/land use/nonlinear relationship/scale effect/gradient boosting decision tree model/Wuhan main urban area分类
管理科学引用本文复制引用
周芳羽,谢波,肖扬谋..多尺度道路网络特征对交通事故的非线性效应研究[J].热带地理,2025,45(8):1344-1357,14.基金项目
国家自然科学基金项目(42371252、41971179) （42371252、41971179）
