Work

Impact of vegetation landscape on PM2.5 and spatial heterogeneity in China

Year: 2026

Type: article

Abstract: PM 2.5 poses a significant threat to public health and ecological stability, particularly in rapidly urbanizing China. Although vegetation is recognized as an effective nature-based solution for air pollution mitigation, the combined effects of vegetation type, amount, and spatial configuration at the national scale remain poorly understood. This study systematically evaluates the impacts of vegetation types, spatial configurations, and greenness levels on PM 2.5 concentrations across China. The relationships and spatial heterogeneity between PM 2.5 and vegetation landscape metrics (including Class Area (CA) and Fragmentation Index (Fra) of forests, shrublands, and grasslands, as well as the Normalized Difference Vegetation Index (NDVI)) were quantified using Pearson correlation analysis, multivariate linear regression controlling for meteorological and anthropogenic factors, and machine learning models. Results indicate that vegetation landscape metrics explain more than 25% of the spatial variation in PM 2.5 . Pronounced spatial heterogeneity was observed: PM 2.5 exhibited significant negative correlations with NDVI in 91.23% of the study area ( p < 0.05), while positive correlations occurred in limited arid inland basins and high-mountain border regions. Forest CA and Fra were generally negatively correlated with PM 2.5 in the eastern-western transitional zones, whereas Forest Fra tended to aggravate pollution in densely forested eastern China. In contrast, shrubland and grassland CA and Fra showed positive correlations with PM 2.5 in over 50% of the regions, particularly in ecologically fragile areas dominated by vegetation degradation and wind-blown dust. These findings demonstrate that vegetation effects on PM 2.5 are highly area-dependent and context-specific, providing important implications for regionally differentiated landscape planning and air quality management in China.

Source: Ecological Indicators

DOI: 10.1016/j.ecolind.2026.114665