Objective Rapid urbanization coupled with global climate warming has continuously intensified the urban heat island (UHI) effect across the Yangtze River Delta. Plant phenology acts as a sensitive bioindicator that intuitively reflects regional climatic shifts and anthropogenic environmental modifications. Exploring the spatiotemporal response mechanisms of woody ornamental plant phenology to varying UHI intensities has become a core scientific prerequisite for landscape architecture practitioners to mitigate urban thermal stress and adapt urban green space design to climate change. Existing local studies in Suzhou mainly focus on single tree species or fragmented seasonal phenological records, while few comprehensive full-year comparative analyses simultaneously cover flowering and leaf phenology of multiple landscape plants across graded heat island zones. Therefore, this research aims to quantify how flowering and leaf phenological metrics of common park woody plants respond to gradient land surface temperature (LST) in central Suzhou, reveal interspecific divergent response characteristics, and further provide quantitative support for heat-adaptive tree species selection in urban park green spaces. Understanding and identifying the spatiotemporal feedback of plant phenology to UHI gradients is critical to solving climate-adaptive landscape configuration problems for subtropical urban areas.

Methods This study took the central urban area of Suzhou as the research scope. Landsat remote sensing images were used to invert spatial LST distribution, and the study area was classified into cold island and heat island zones based on standardized LST gradients to generate a heat island intensity zoning map. Combined with the spatial layout of urban park green spaces and the actual application status of landscape vegetation, seven typical parks covering both cold and heat island gradient zones were selected as field sampling sites. A total of 22 healthy dominant woody ornamental tree species with coverage proportion not less than 50% in each park were screened as research objects, including 14 flowering ornamental species and 8 foliage ornamental species. Continuous fixed-point in-situ phenological monitoring was conducted for a full annual cycle from September 2023 to August 2024, with unified observation standards defined for four core flowering phenological indicators (first flowering date, full flowering date, end flowering date, flowering duration) and three key leaf phenological indicators (leaf-out onset date, peak growth date, leaf fall onset date). Two statistical approaches were adopted for quantitative analysis: independent sample t-test was applied to identify significant phenological date differences between cold island control plots and heat island plots, and Pearson correlation analysis was used to calculate the linear correlation coefficient and temperature response sensitivity k between each phenological metric and plot-scale LST, to distinguish interspecific differences in thermal response intensity. Hierarchical cluster analysis was additionally performed to classify tree species based on their integrated phenological thermal response features.

Results 1) Phenological disparities and temperature sensitivity of flowering species between thermal zones. For all 14 flowering tree species, the first flowering date, full flowering date and end flowering date in heat island plots were significantly earlier than those in cold island control plots (p<0.05). The average inter-zone difference of flowering duration between heat and cold island zones reached (-1.86 \pm 1.01) d, which indicated that rising surface temperature shortened the overall flowering period of all ornamental flowering trees. The average LST response sensitivity of first flowering, full flowering, end flowering and flowering duration were ( - 0.84 \pm 0.38) d/°C, \( - 1.25 \pm 0.69) d/°C, ( - 1.27 \pm 0.44) d/°C and ( - 0.48 \pm 0.19) d/°C, respectively. Among all flowering species, Prunus subhirtella (Japanese Late Cherry) exhibited the weakest thermal response across all four floral phenological indicators. All flowering phenological indicators presented significant linear correlations with plot LST, yet correlation coefficients and sensitivity k showed obvious interspecific gradients; hierarchical clustering further categorized the 14 flowering trees into multiple independent thermal response groups. 2) Leaf phenological variations and differential LST correlation across thermal gradients. Obvious group differences of leaf-out onset date, peak growth date and leaf fall onset date were detected between cold and heat island plots among the 8 foliage ornamental species. Leaf-out onset and peak growth phases were generally advanced under higher LST, while leaf fall onset was delayed for most foliage trees in heat island areas. The average LST response sensitivity of peak growth period was ( - 1.25 \pm 4.79) d/°C, with Koelreuteria paniculata (Chinese Flame Tree) as the most sensitive species and Ginkgo biloba (Ginkgo) the least sensitive. The mean sensitivity of leaf-out onset was ( - 2.55 \pm 1.12) d/°C, where Liquidambar formosana (Sweet Gum) had the highest thermal response magnitude and Zelkova schneideriana (Hornbeam) the lowest. The average response sensitivity of leaf-fall onset reached (7.71 \pm 1.05) d/°C. The linear correlation strength between leaf phenological metrics and LST varied remarkably among different foliage tree species. Partial foliage species with extremely small inter-group phenological differences failed to pass the significance test, mainly due to tiny numerical gaps between cold and heat island samples and high discrete degree of field observation data.

Conclusion This study systematically summarized the divergent response patterns of ornamental flowering and foliage woody plants to urban heat island gradients in central Suzhou, and clarified the spatial differentiation characteristics of plant phenology under different thermal environments. Distinct thermal response types of landscape trees were classified by quantifying phenological advance/delay days and linear temperature response sensitivity, which fully explained interspecific discrepancies in phenological feedback to urban thermal environment. Based on these quantitative phenology-LST coupling results, targeted tree species matching and collocation optimization strategies were proposed for green spaces under different heat island intensity gradients, forming climate-adaptive landscape construction schemes. The full-year observation quantitative dataset can provide reliable data support for seasonal landscape construction, heat-adaptive vegetation allocation and UHI mitigation-oriented park green space planning in Suzhou, as well as other subtropical river delta cities suffering severe urban heat island effects