Reassessing the role of urban green space in air pollution control

Abstract

The assumption that vegetation improves air quality is prevalent in scientific, popular, and political discourse. However, experimental and modeling studies show the effect of green space on air pollutant concentrations in urban settings is highly variable and context specific. We revisited the link between vegetation and air quality using satellite-derived changes of urban green space and air pollutant concentrations from 2,615 established monitoring stations over Europe and the United States. Between 2010 and 2019, stations recorded declines in ambient NO₂, (particulate matter) PM₁₀, and PM_2.5 (average of -3.14% y^-1), but not O₃ (+0.5% y^-1), pointing to the general success of recent policy interventions to restrict anthropogenic emissions. The effect size of total green space on air pollution was weak and highly variable, particularly at the street scale (15 to 60 m radius) where vegetation can restrict ventilation. However, when isolating changes in tree cover, we found a negative association with air pollution at borough to city scales (120 to 16,000 m) particularly for O₃ and PM. The effect of green space was smaller than the pollutant deposition and dispersion effects of meteorological drivers including precipitation, humidity, and wind speed. When averaged across spatial scales, a one SD increase in green space resulted in a 0.8% (95% CI: -3.5 to 2%) decline in air pollution. Our findings suggest that while urban greening may improve air quality at the borough-to-city scale, the impact is moderate and may have detrimental street-level effects depending on aerodynamic factors like vegetation type and urban form.

Keywords: ecosystem service; green infrastructure; public health; urban planning; vegetation.

Fig. 1.

Distribution of the air quality–monitoring stations across biomes in Europe (n = 2,127) and the United States (n = 488) (A). Inset histograms show the proximity of stations to roads and the building footprint within 30 m. Air pollutant time series along with linear trends are shown in (B).

Fig. 2.

Empirical estimates of the association between annual mean (A) and maximum (B) air pollutant changes, and changes in green space, emissions, and climate predictor variables. Estimates and 95% CI are indicated with points and error bars and are expressed as percentage changes in air pollutant concentrations per SD (δ) increase in the predictor variable. The size of the point indicates the magnitude of the effect.

Fig. 3.

Empirical estimates of the association between annual mean air pollutant changes, and changes in green space predictor variables including NDVI (total green space) and fractional tree cover. Separate models and effect sizes were estimated for each spatial scale over which green space was aggregated (A) and each biome within which stations were located (B). Estimates and 95% CI are indicated with points and error bars and are expressed as percentage changes in air pollutant concentrations per SD (δ) increase in the predictor variable. The size of the point indicates the magnitude of the effect.

Fig. 4.

Example of an extreme increase (A−C) and decrease (D−F) in green space within a 60-m buffer (street-level) of two air quality–monitoring stations. Aerial photographs from Google Earth Pro shown for reference.

Fig. 5.

Changes in air pollutant concentrations at street-level (within 60 m) between 2010 and 2019 for a sub-set of air quality stations with the greatest gains (n = 37) and losses (n = 65) in green space (A) and tree cover (B) over the same period. P-values derived from linear mixed-effects models are shown to indicate no significant difference (P = 0.05) in air pollutant changes between stations with extreme gains and losses of green space.