The impact of sub-pixel scale urban function on urban heat island: Insights derived from its decomposition

Abstract

Urban functional land use significantly impacts the Urban Heat Island (UHI) effect. While many studies have explored this relationship, research on mixed functional land use and its effects on UHI is limited. This study uses remote sensing imagery and multisource data to identify urban mixed-function structures at the sub-pixel scale. Focusing on the Wuhan Urban Agglomeration (WUA), the study utilizes eXtreme Gradient Boosting (XGBoost) to establish the mapping relationship between mixed land use and UHI. Shapley Additive exPlanations (SHAP) method is applied to explain the mechanism by which mixed functional land use affects UHI. The findings show accuracies exceeding 80% in identifying mixed-function structures, validating the effectiveness of the mixed pixel decomposition method. SHAP analysis reveals that the abundance of industrial land, residential land, commercial land, and public land all positively correlate with UHI. Industrial land has the highest abundance impact, followed by residential land, with public and commercial land having the least. Moreover, SHAP plots show that combining high-fraction industrial and residential land intensifies UHI, as does mixing residential with high-fraction commercial land. This study provides crucial insights for urban planning and sustainable development.