Associations of street morphology and local wind environment with annual cumulative ambient PM2.5 exceedance: Evidence from Wuhan, China.

Streets are important urban microenvironments where residents encounter ambient PM2.5, while their pollution conditions are shaped by coupled spatial and aerodynamic processes. However, how street morphology, grey-green elements, and local ventilation are jointly associated with cumulative PM2.5 exceedance remains insufficiently understood. Using 6785 streets in Wuhan, we classified streets into five typologies and developed a multidimensional indicator system integrating street geometry, grey-green elements, and the local wind environment. Annual cumulative excess exposure (ACEE) was defined as an ambient PM2.5 exceedance indicator. Interpretable machine learning (CatBoost-SHAP) and partial least squares path modeling were combined to quantify type-specific predictive contributions, nonlinear response patterns, and direct and indirect pathways. Open street canyons accounted for more than 75% of the sampled streets, and the models explained over 70% of ACEE variation on average. Street geometry and wind environment jointly contributed more than 50%, although their relative importance varied across street types. In enclosed canyons, ACEE was mainly associated with the Building View Index and Sky View Factor, whereas wind-related indicators were more influential in open canyons, with street-wind angle (WA) and street wind speed (WS) contributing 25.28% on average. Model-derived WS breakpoints occurred at approximately 0.33 m/s (95% CI: 0.31-0.44 m/s) in enclosed canyons and 1.19 m/s (95% CI: 0.78-1.31 m/s) in open canyons. Green-space associations were context-dependent: higher visible greenery was positively associated with ACEE in enclosed canyons, whereas canopy ratio showed more consistent negative associations. These findings provide context-specific evidence for typology-sensitive street planning and air-quality management.