Nonlinear Effects of Human Settlements on Seasonal Land Surface Temperature Variations at the Block Scale: A Case Study of the Central Urban Area of Chengdu

The land surface temperature (LST) in the central urban area has shown a consistent upward trend over the years, exacerbating the surface urban heat island (SUHI) effect. Therefore, this study focuses on the central urban area of Chengdu, using blocks as the research scale. The Gradient Boosting Decision Tree (GBDT) model and SHAP values are employed to explore the nonlinear effects of human settlements (HS) on LST across different seasons. The results show that (1) At the block scale, the overall impact of HS on LST across all four seasons tracks the following order: built environment (BE) > landscape pattern (LP) > socio-economic development (SED). (2) LP is the most important factor affecting LST in summer, while the BE has the greatest influence on LST during spring, autumn, and winter. (3) Most HS indicators exhibit seasonal variations in their impact on LST. The impervious surface area (ISA) exhibits a significant positive impact on LST during spring, summer, and autumn. In contrast, the nighttime light index (NTL) and functional mix degree (FMD) exert a significant negative influence on LST in spring, autumn, and winter. Additionally, the normalized difference vegetation index (NDVI) negatively affects LST in both spring and summer. Moreover, connectivity (CNT) and functional density (FPD) demonstrate notable threshold effects in their influence on LST. (4) Certain HS indicators exhibit interaction effects, and some combinations of these indicators can effectively reduce LST. This study reveals HS–LST interactions through multidimensional analysis, offering block-scale seasonal planning strategies for sustainable urban thermal optimization.