Seasonal responses of land surface temperature to landscape patterns and multi-objective optimization in urban-rural fringe: a case study of Changqing District of Jinan City, China.

Landscape patterns significantly influence the surface thermal environment in urban-rural fringe, yet research on the Pareto-optimal landscape patterns for optimizing thermal environment across all seasons remains relatively limited. Using Changqing District of Jinan, China, a typical urban-rural fringe area, this study employed a random forest regression model to analyze the responses of land surface temperature (LST) to landscape patterns and their seasonal variations. A non-dominated sorting genetic algorithm was applied to calculate the Pareto-optimal landscape patterns solutions for achieving a suitable thermal environment across all seasons. Results showed clear seasonal differentiation in LST across landscape types. Percentage of landscape (PLAND) and largest patch index (LPI) showed higher predictive importance than patch density (PD) and mean perimeter-area ratio (PARA_MN). The response of LST to cohesion index (COHESION) was mainly more evident in the medium and high value ranges of COHESION. After landscape pattern optimization, only 5.285% of the study area changed landscape type, and the dominant transition types were from cropland and grassland to forestland. LST generally moved closer to the optimization objective intervals in all four seasons, but the deviation of optimized LSTs in summer from the optimization objective interval remained generally higher than those in the other seasons. Among Pareto solutions, increases in PLAND, PD, PARA_MN, core area percentage of landscape (CPLAND), and LPI of forestland, decreases in PLAND, CPLAND, LPI, and COHESION of cropland, and increases in PD and PARA_MN of grassland were generally beneficial for surface thermal environment optimization.