Spatiotemporal dynamics and nonlinear landscape-driven mechanisms of urban heat islands in a winter city: A case study of Harbin, China

With increasing global climate change, the urban heat island (UHI) effect poses substantial challenges to urban sustainability, particularly in winter cities. However, the interactions among climate, urbanization, and landscape changes influencing UHI dynamics in winter cities remain inadequately understood. To address this gap, Harbin—a representative winter city—was selected as a case study. This study used multi-temporal remote sensing data from 2002 to 2023 and developed an integrated analytical framework. The framework was designed to examine spatiotemporal patterns, quantify changes, and classify dynamic expansion types of UHIs. In addition, a random forest model combined with SHapley Additive exPlanations (SHAP) analysis was applied to reveal the nonlinear impacts of landscape changes on UHI dynamics. Results showed that daytime UHI extent fluctuated (364 to 547 km²), while nighttime UHI expanded from 369 km² in 2002 to 698 km² in 2020. Overlapping UHIs dominated both periods, but nighttime UHI exhibited a stronger and more consistent expansion than daytime. At night, UHI patches frequently emerged along major rivers, and strong UHI zones expanded towards suburban forest areas. Landscape transformations showed clear thresholds. For example, impervious surface ratios exceeding 20 % and population density increases above 2500 people/km² significantly intensified daytime UHI expansion. In contrast, increases in vegetation cover(EVI_change>0) and small water body changes (±5 %) were associated with lower nighttime land surface temperature (LST). These findings provide key insights into the nonlinear drivers of UHI dynamics in winter cities, supporting evidence-based urban planning and climate adaptation strategies.