Analysis of the global and local effects of urban green space vegetation types on early-summer near-surface O3 concentrations based on street view imagery – A case study of hangzhou

With the expansion of urban development, air pollution has become increasingly severe. In recent years, ozone (O3) pollution has emerged as a major threat to air quality in China. Existing studies have rarely explored the impact of vegetation types on O3 from a micro-scale perspective. Taking Hangzhou, a city with severe O3 pollution, as an example, street-view images from eight districts were first collected. A DeepLabV3+ semantic segmentation model trained on the ADE20K dataset was employed to accurately classify trees, shrubs, and grasses in the images. Meteorological and urban structural factors were incorporated, and Bayesian-optimized Optimization-based Extreme Gradient Boosting (XGBoost), SHAP (SHapley Additive exPlanations), and Geographically Weighted Random Forest (GWRF) models were applied to analyze the independent effects of the three vegetation-type factors on O3 concentration and their synergistic interactions from both global and local scales. The results indicate: (1) Xiaoshan District and the core urban areas such as Xihu District and are severely polluted, with an overall decreasing trend in O3 concentration (the area with O3 concentration in the range of 120.01-160.00 μg/m3 decreased from 2,100 km2 to 0 km2). (2) Meteorological factors dominate O3 variation, while the contributions of vegetation-type factors are relatively low. Nevertheless, vegetation-type factors still exhibit stable and differentiated contributions to O3 concentration. At the global scale, tree area proportion (TAP) shows either positive or negative effects on O3 concentration at high levels (average contribution = 0.002), shrub area proportion (SAP) has a positive influence, and grass area proportion (GAP) demonstrates a negative influence (average contribution = 0.006 for both). (3) At the local scale, the extremely high-importance areas of TAP shifted from northern Xihu District and Gongshu District in 2014 to southern Xihu District and Binjiang District in 2017. SAP exhibited relatively high and extremely high importance in southern Xihu District and western Xiaoshan District. The relatively high and extremely high importance areas of GAP were distributed in core urban zones such as eastern Xihu District, southern Gongshu District, Binjiang District, and Shangcheng District. The extent of extremely high-importance areas for all vegetation-type factors showed a decreasing trend. (4) High-importance areas of building density (BD) and industrial impervious surface percentage (ISP) spatially overlapped with those of vegetation-type factors, concentrating in core urban areas and the industrial zone of Xiaoshan, whereas meteorological factors played an independent role mainly in the peripheral areas of the city. This study aims to provide a scientific basis for precise O3 pollution control and optimal configuration of urban green spaces.