Understory vegetation affects the richness and stability of urban forest ecosystems. To investigate the influence of soil physicochemical properties on the diversity of understory plants in urban forests, this study used 30 urban forest communities in the Beijing Plain area as the research object and analyzed the correlation between understory plant diversity and soil factors by correlation analysis. Furthermore, pH, soil bulk density (SBD), total soil porosity (TSP), soil water content (SWC), soil organic carbon (SOC), soil organic matter (SOM), total nitrogen (TN), total phosphorus (TP), effective phosphorous (AP), and effective potassium (AK) were determined in this study. The Shannon diversity index (H’), Pielou evenness index (E), Simpson dominance index (C), and Margalef richness index (DMG) of understory plants were calculated. The soil nutrient contents and the understory plant diversity indices of the different community types showed significant differences. There was a strong correlation between soil properties and the diversity index of understory vegetation. SOM and SOC were the main factors affecting the Shannon-Wiener index, Pielou index, Simpson index, and Margalef richness index of the understory plants. We conclude that soil properties were one of the primary drivers of the formation of understory vegetation diversity. The results of the study can provide scientific guidance for the management of urban forests.

Climate change has presented considerable challenges in the management of urban forests and trees. Varieties of studies have predicted the potential changes in species distribution by employing single-algorithm species distribution models (SDMs) to investigate the impacts of climate change on plant species. However, there is still limited quantitative research on the impacts of climate change on the suitable distribution ranges of commonly used urban tree species. Therefore, our study aims to optimize traditional SDMs by integrating multiple machine learning algorithms and to propose a framework for identifying suitable distribution ranges of urban trees under climate change. We took Michelia chapensis, a tree species of particular significance in southern China, as a pilot tree species to investigate the evolution of its suitable distribution range in the context of two future climate scenarios (SSP126 and SSP585) across four periods (2030s, 2050s, 2070s, and 2090s). The findings indicated that the ensemble SDM showed strong predictive capacity, with an area under the curve (AUC) value of 0.95. The suitable area for Michelia chapensis is estimated at 15.9 × 10⁵ km² currently and it will expand in most areas under future climate scenarios according to the projection. However, it will contract in southeastern Yunnan, central Guangdong, the Sichuan Basin, northern Hubei, and Jiangxi, etc. The central location of the current suitable distribution area is located in Hengyang, Hunan (27.36° N, 112.34° E), and is projected to shift westward with climate change in the future. The migration magnitude is positively correlated with the intensity of climate change. These findings provide a scientific basis for the future landscape planning and management of Michelia chapensis. Furthermore, the proposed framework can be seen as a valuable tool for predicting the suitable distribution ranges of urban trees in response to climate change, providing insights for proactive adaptation to climate change in urban planning and landscape management.

The rapid advancement of AI-driven generative design brings new opportunities, but its application in landscape garden design remains limited by two gaps: (1) semantic misalignment between generated images and the designer’s intent, and (2) low-resolution outputs with insufficient details. To address these gaps, we developed GardenDiff, a domain-adapted diffusion model trained via parameter optimization and a specialized landscape garden dataset. Central to this approach is Structured Design Captioning (SDC), a hierarchical annotation system specifically designed for garden design that encodes design elements, style features, and auxiliary scene information. To develop this model, we designed a three-stage experimental framework. In Stage 1, we examined the effects of training caption systems and training resolution on generated landscape garden imagery by controlled experiments. In Stage 2, we conducted joint training across five garden styles (Chinese, Japanese, Mediterranean, Nordic, and English) based on the optimized parameter settings from Stage 1 to construct the GardenDiff model. In Stage 3, we validated the model performance through expert evaluation (N = 36) and public evaluation (N = 136) and analyzed style-specific variations in the generated outcomes. Research results showed that Structured Design Captioning (SDC) improved Spatial Rationale by 19.67–39.46% compared with generic captions, and training at 1536 × 1536 pixels improved image quality by 23.2% compared with 768 × 768-pixel training. GardenDiff trained with these optimized parameters showed notable advantages. Its overall scores (5.06) exceeded those of Stable Diffusion XL base 1.0 (SDXL 1.0) by 16.4% and DreamShaper XL by 22.4%. The model improved across four dimensions, including Design Rationale, Design Professionalism, Design Accuracy, and Design Satisfaction. Our study offers a new model to improve the perspective visualization of generative garden design and provides insights into AI-informed landscape and urban design.