Contrasting changes in fine-scale land use structure and summertime thermal environment in downtown Shanghai

Abstract

• Changes in patterns of fine-scale land use structures and intensity of intra-SUHI were analyzed. • The associations between changes in land use structures and intensity of intra-SUHI were analyzed. • Comparison of two regression models quantifies the importance of interactions of the predictors. • PLSR models perform slightly better than RFR models, but their combined usage will perform better . Understanding the associations between fine-scale land use structures and the surface urban heat island effect within the cities (synonymously defined as intra-SUHI) is vital for decision-making on sustainable land development, UHI mitigation, and climate adaptation. Taking downtown Shanghai as a case, this study quantified the associations between the changes in land use structures (indicated by eleven metrics of urban morphology, land use composition, and urban green infrastructure) and the intensity of intra-SUHI (intra-SUHII) at the land parcel level. Two model types, including Type-1 models that merely consider the main effect of the factors, and Type-2 models that consider both the main effect and potential interactions, were employed for analysis. The performance of each model type was parallelly assessed using the partial least square regression (PLSR) and random forest regression (RFR) methods. Our results revealed that Type-2 models exhibited higher interpretation power (PLSR:55.604–55.887% and RFR:53.841–55.739%) than Type-1 models (PLSR:47.531–47.591% and RFR: 47.111–50.300%) for explaining the variance in parcel-level intra-SUHIIs. This finding indicated the non-negligible importance of potential interactions of the factors in determining the models' interpretation power. Our study can be generalizable to many large cities with similar climatological conditions.