Mining the driving factors of the urban thermal environment by building semantic information at block level—A case study of Shenyang

Urban blocks are the fundamental units of cities. Understanding the driving factors of urban thermal environments is crucial for environmental protection. Current research focuses more on natural factors like vegetation and land cover rather than social factors such as population activity and building function. Recent studies have started to quantify social factors, including building height, but the relationship between block function types and driving factors remains unclear. This paper proposes an approach to identify thermal environmental drivers in urban blocks by improving functional classification accuracy using building information. Enhanced classification improves feature homogeneity within classes and separability of driving mechanisms between classes. We developed a multidimensional driving factor analysis model and analyzed thermal environmental drivers across different block types using data from Shenyang, China. Results show our method achieves a kappa coefficient of 0.90, 0.18 higher than conventional methods. Incorporating social factors improved the regression model's R² from 0.82 to 0.84. Natural factors influence thermal environments differently based on block functions. Building geometry dominates commercial and residential zones, while land coverage dominates industrial, public service, and scenic areas. Without improved classification accuracy, identifying these dominant factors would be less precise, leading to less effective optimization strategies. Therefore, accurate functional classification is crucial for quantifying thermal environment drivers and formulating precise optimization strategies. The proposed framework, relying on open geospatial data, can be applied to other cities and provides actionable insights for mitigating urban heat islands through targeted planning.