Key points are not available for this paper at this time.
Urban greenspace is generally assumed to mitigate air pollution. However, clear large-scale real-world evidence remains limited, particularly regarding how its effects vary with seasons, vegetation types, and street structures. To bridge this gap, this study integrates 4.254 million historical street view images, 1.894 million records of mobile monitoring air pollution data, and three-dimensional urban morphology data in Hong Kong, employing a double machine learning (DML) framework to disentangle the complex effects of urban greenspace. The results show that the DML framework provides more robustness and less biased estimates than conventional models. The estimated greenspace’s effects vary substantially across vegetation types and seasons: grass reduced PM2.5 levels in spring (θ = −8.464) and autumn (θ = −6.506); trees increased PM2.5 levels in summer (θ = 0.762); and mid-level vegetation showed opposite effects in spring (θ = −2.275) and winter (θ = 3.480). Mechanistically, these differences reflect the relative magnitudes of biogenic volatile organic compounds (BVOC) emissions, deposition processes, and aerodynamic effects. Moreover, the street height-to-width ratio significantly moderates greenspace’s effects due to the ventilation alteration within street canyons. These insights advance current knowledge and offer an evidence base for policy-making in urban greenspace construction and renewal and air pollution control.
Yang et al. (Wed,) studied this question.