Assessing the landscape visual quality of urban green spaces with multidimensional visual indicators

High-quality urban green spaces (UGSs) are crucial for human well-being. Proactive assessment of the landscape visual quality (LVQ) before construction ensures proper alignment with future expectations and reduces resource wastage. In this study, we aimed to use virtual reality (VR) and the multidimensional visual indicators, eye-tracking, image segmentation, and spatial features, to assess the human perception of the LVQ of UGSs. This study encompassed three diverse UGS types across northern, central, and southern China, rendering 30 panoramic images for assessment. Sixty participants assessed these images across seven dimensions: beauty, comfort, color, complexity, liveliness, greenness, and safety. Integrating these multidimensional visual indicators, a generalized estimating equations (GEE) logistic regression model demonstrated superior performance over existing traditional models focusing only on spatial features, facilitating more accurate evaluations of LVQ perception. Moreover, herb plants (eye-tracking indicator), water ratio (image segmentation indicator), and number of materials (spatial feature indicator) were the most positive factors affecting human perception. Isolated planting style positively impacted the perception of greenness, and sky ratio negatively correlated with beauty perception. Additionally, openness levels of 20–80 % enhanced beauty perception, while openness above 80 % decreased liveliness but improved safety perceptions. Shrub species diversity positively correlated with perceptions of greenness and complexity, whereas single and dense shrub arrangements diminish perceptions of greenness and liveliness. Overall, this study provides valuable insights for urban planning at the design stage to enhance decision-making and visual quality of UGSs, thereby contributing to the establishment of more sustainable urban development strategies.