Impact of green space patterns on PM2.5 levels: A local climate zone perspective

Green spaces have been demonstrated to significantly decrease PM2.5 levels. However, the impact of green spaces on PM2.5 levels in different spatial forms of urban blocks is not yet fully understood. This research utilized ensemble machine learning algorithms to investigate the impact of green space spatial patterns, assessed through landscape pattern analysis, to PM2.5 concentrations during summer and winter based on local climate zones (LCZ). The results revealed significant differences in PM2.5 concentration across the seven types of primary LCZs, primarily influenced by meteorological factors. In winter, eight green space indicators exhibited a more substantial contribution to PM2.5 levels in comparison to the summer season. Significant discrepancies were noted in the contributions of these indicators across different LCZs. Patch density (PD) and landscape shape index (LSI) made a more substantial contribution, while relative patch richness (RPR) and Shannon's evenness index (SHEI) showed a less significant contribution. The spatial pattern of green spaces was significantly related to their contributions. Among the seven predominant LCZs, the five key indicators included PD, LSI, Shannon's diversity index (SHDI), area-weighted mean patch area (AREA_AM), and connectance index (CONNECT). Spatial heterogeneity was further observed in the positive and negative contributions of each green space indicator. This study enhances understanding of the impact of green spaces on PM2.5 across LCZs and provides valuable insights for urban management and planning. © 2024 Elsevier Ltd