Abstract:Campus roads serve as vital spaces for college students to connect with nature and gain restorative benefits. To explore the nonlinear influence of campus road space characteristics on restorativeness from both perceptual and material levels. A total of 105 road images from 35 universities in China were sampled. A questionnaire (n=525) was used to evaluate their restorativeness and perceptual features, and semantic segmentation was employed to calculate their material features. An XGBoost model was fitted and the SHAP tool was used to interpret the results. Compared with physical features, perceptual features contribute more to restorativeness—the changes in SHAP values of each spatial feature present different linear patterns. The SHAP values of naturalness, beauty, and harmony increase approximately monotonically. The SHAP values of colorfulness, layering, and rise in environmental brightness are approximately in an inverted U-shaped pattern, with an optimal feature value range. The SHAP values of novelty, landscape complexity, and building presence increase approximately in an inverted L-shaped pattern, with a saturation point of contribution. The SHAP value of quietness increases approximately in a U-shaped pattern, with an unsuitable feature value range. The SHAP values of the green view ratio and enclosure increase approximately in an S-shaped pattern, with both suitable and unsuitable feature value ranges. The influence of spatial features on restorativeness is nonlinear, and features should be regulated within the most effective range.
