Nonlinear meteorological controls on coastal air quality along the Indian East Coast: Insights from a multi-city...

This study investigates the role of meteorological variability in shaping ambient air quality across five coastal cities along the Indian East Coast (Haldia, Balasore, Visakhapatnam, Chennai, and Puducherry). Despite their shared coastal setting, the cities exhibit pronounced meteorological heterogeneity, leading to a clear latitudinal gradient in air pollution, with northern cities (Haldia and Balasore) experiencing substantially higher particulate matter concentrations and more frequent exceedances than southern cities (Chennai and Puducherry). Across all cities, PM₁₀ is the dominant contributor to variability in Air Quality Index (AQI) variability. A Random Forest model reveals that meteorological parameters explain a substantial fraction of particulate matter variability (R² up to ∼0.6), whereas gaseous pollutants show lower predictability (R² < 0.35). Using Shapley Additive exPlanations (SHAP) analysis, we identify critical nonlinear thresholds and binary regime switches governing pollutant accumulation. A consistent regional specific humidity (SH) threshold of ∼15 ± 2 g/kg is identified as a key regime-switching threshold, with values below this level consistently associated with PM₁₀ accumulation across all cities. In southern cities, a Sea Breeze Component (SBC) > 3 m/s is the threshold for coastal ventilation; below this level, pollutants remain trapped. Furthermore, we uncover a monsoon break thermal trap, in which temperatures exceeding 32 °C flip the expected seasonal dispersion into an accumulation regime. These findings demonstrate that coastal air quality is governed by discrete meteorological regime transitions rather than linear responses. The identified thresholds provide a physically interpretable framework for improving localized air quality forecasting and mitigation strategies in complex coastal environments.