Investigating the quantitative impact of the vegetation indices on the urban thermal comfort based on machine learning: A case study of the Qinhuai River Basin, China
Jianqing Zhao, Chunguang Hu, Zhuoqi Li, Maomao Zhang, Houbao Fan, Kaili Li, Ruidi Yuan
Abstract
• Apply 14 multidimensional vegetation indices for TC analysis. • Advanced machine learning algorithms (RFR, SHAP) improve prediction accuracy. • Dynamic analysis of vegetation indices and TC over time reveals long-term trends. • Refines the impact of vegetation indices within urban functional areas for planning purposes. Rapid urbanization and economic development in China have reshaped land cover and impacted urban Thermal Comfort (TC). This study examines the relationship between Vegetation Indices (VIs) and TC in the Qianhai River Basin using multi-temporal remote sensing data and Machine Learning (ML) models. TC classifications were derived through inversion and analyzed using the standard deviation ellipse and Moran's I. This study uses Random Forest Regression (RFR) to evaluate the predictive performance of 14 vegetation indices. The results revealed that in 2022, high TC areas (917.52 km²) were mostly found in green areas, farmlands, and water bodies, while low TC areas were concentrated in urban and industrial zones. VIs such as the Normalized Difference Moisture Index (NDMI), Shortwave Infrared Vegetation Index (SWIRVI), and Global Environment Monitoring Index, which reflect vegetation moisture and growth, were key in characterizing TC. The study refines the mechanism of VIs on the scale of urban functional areas, which fills the deficiency of neglecting the differences of functional areas in previous studies. SWRWI and Vegetation Growth Period Index were more relevant in agricultural zones, whereas indices emphasizing impervious surfaces better represented TC in urban areas. Identifying the most impactful VIs provides valuable insights for developing targeted urban planning strategies.