Driving factors of summer diurnal surface thermal environment in subtropical coastal cities: a Shenzhen case study

Subtropical coastal cities face increasing heat load risks due to rapid urbanization. However, the complex nonlinear relationships between urban thermal environments and their driving factors are not fully understood. This study utilizes land surface temperature data from the Ecosystem Spaceborne Thermal Radiometer Experiment on Space Station to analyze the urban surface thermal patterns in Shenzhen. Extreme Gradient Boosting, SHapley Additive exPlanations, and Partial Dependence Plot analyses are applied to evaluate the contributions and marginal effects of various driving factors on land surface temperature. The findings highlight that community-scale models provide more stable predictions compared to grid-based models, particularly in terms of error metrics. Ecological coverage and terrain elevation emerge as the dominant factor categories influencing land surface temperature. Factors such as fractional vegetation coverage, Forest and Buildings removed Copernicus Digital Elevation Model, and average building height significantly cooling both daytime and nighttime land surface temperature. In contrast, factors like average housing price, distance from the coastline, population heat density, and point of interest density exhibit variable correlations with land surface temperature across different value ranges, revealing complex marginal effects. This study provides actionable insights for urban planners and policymakers aiming to reduce heat load risks in subtropical coastal cities, contributing to more sustainable and resilient urban development.