Towards climate-responsive communities: Exploring the nonlinear relationship between residential morphology and land surface temperature

Residential morphology strongly influences urban thermal environments , especially under extreme summer heat that amplifies urban heat island effects . However, the ways morphological factors produce nonlinear impacts on land surface temperature (LST) remain less understood. Nonlinear relationships refer to disproportionate or threshold-based changes in LST in response to variations in morphological indicators. This study analyzes 857 residential areas within Beijing's Fifth Ring Road using summer LST from Landsat 9 and 14 morphological indicators derived from OpenStreetMap and AOI data. A LightGBM model combined with SHAP values quantifies indicator importance and reveals nonlinear response patterns. The findings indicate a “warmer inside, cooler outside” spatial distribution pattern of residential LST across the study area. Among residential types, LST levels rank as Enclosed > Hybrid > Row > Point group. Building construction shows the highest influence on LST, followed by scale and location, natural environment, and development intensity. Key morphological indicators include building height, green space ratio, surrounding land surface temperature, and building spacing. The analysis reveals a nonlinear relationship between residential morphology and thermal environments, with threshold effects or inflection points observed in the influence curves. By identifying these nonlinear relationships and key indicators, the study offers a scientific basis for targeted planning strategies to enhance thermal comfort and promote sustainable residential environments. • Temperature levels vary by type: Enclosed > Hybrid > Row > Point group. • Constructing accurate nonlinear explanatory model by machine learning algorithms. • The morphology of residential areas crucially shapes their LST. • Nonlinear effects and thresholds provide insights for residential planning.