Quantifying anisotropic impacts of LCZ landscape patterns on NO2 variabilities in the Yangtze River Delta, China

Urban landscape patterns strongly affect air quality, yet existing studies often neglect anisotropic processes, rely on arbitrary fixed-radius buffers, and lack standardized classification frameworks. To overcome these limitations, we developed a novel analytical framework that integrates high-accuracy Local Climate Zone (LCZ) classification, multi-scale spatial modeling, and a shape-adaptive influence zone. Central to this framework is the Adaptive Directionally Weighted Sector Influence Zone (ADWSIZ), a new method that explicitly captures the directional heterogeneity of landscape effects on pollutants. Applied to station-based NO 2 measurements in the Yangtze River Delta (YRD) urban agglomeration, we used multi-scale Geographically Weighted Ridge Regression (GWRR) to identify a 4000 m circular buffer as the optimal modeling scale, revealing a clear decoupling from the intrinsic static LCZ pattern scale (approximately 1000 m). Building upon this optimal distance, the ADWSIZ model was developed to refine the influence zone’s shape. The application of ADWSIZ improved model performance (Adjusted R 2 : 0.464 → 0.506 ; RMSE: 2.950 → 2.834 µg/m 3 ) and reduced residual spatial autocorrelation. Our findings highlight that the anisotropic influence of features like industrial land and compact high-rise fabrics is a critical, previously underexplored factor in explaining NO 2 variation. This study establishes a replicable and transferable methodology that advances the conceptualization of influence zones from simple distance-based buffers to scale- and shape-adaptive forms, offering a robust tool for refining spatial coupling analyses between urban morphology and air pollutants.