Unraveling associations between grey-green spatial trade-offs and PM_2.5 concentrations using explainable machine...

Urban air quality is increasingly associated with spatial balance between grey and green spaces in rapidly urbanizing cities. However, previous studies have primarily examined grey or green spaces independently, with limited attention to their joint spatial dynamics and nonlinear associations with air pollution at intra-urban scales. This study addresses this gap by investigating the relationship between grey-green spatial coordination and fine particulate matter (PM_2.5) concentrations in Vilnius during 2016-2024 using the Grey-Green Space Coordination Index (GGSCI), Local Indicators of Spatial Association (LISA) spatial clustering, and an interpretable eXtreme Gradient Boosting-SHapley Additive exPlanations (XGBoost-SHAP) framework. Results showed a consistent negative association between GGSCI and PM_2.5, where elderships classified as "General imbalance" and "Highly imbalance" (GGSCI < -0.125) correspond to elevated PM_2.5, while "General balance" and "Highly balance" (GGSCI > 0.125) classes are associated with lower pollution levels. Persistent high-high PM_2.5 clusters correspond to imbalanced grey-green structures, whereas low-high clusters are primarily associated with balanced GGSCI classes. Moreover, SHAP analysis revealed heterogeneous predictor contributions across different value ranges and grey-green coordination conditions. Among the PM_2.5 predictors, Normalized Difference Vegetation Index (NDVI) values greater than ≈ 0.3 were more frequently associated with negative SHAP contributions in balanced GGSCI classes, whereas population density (POP) greater than ≈ 2000 people km^-2 and Nighttime Light intensity (NL) exceeding ≈ 18 were more frequently associated with positive SHAP contributions in imbalanced GGSCI classes. These findings show how PM_2.5 predictive patterns vary across grey-green coordination conditions and may inform coordinated urban development and air-quality planning in Vilnius.