Nonlinear Mechanisms Underlying Rural Streetscape Aesthetics: Threshold and Interaction Effects via Interpretable Machine Learning

Aesthetic perception of rural streetscapes reflects individuals’ cognitive responses to their surroundings and is central to understanding how landscape preferences are formed. Existing studies using Scenic Beauty Estimation (SBE) are constrained by incomplete indicator systems and overreliance on linear approaches. This study proposes an interpretable machine learning framework that integrates multi-source data to examine the nonlinear influences of streetscape features on SBE. Using Sanguan Village, a water-networked settlement in Jiangsu, we developed a 24-indicator system spanning color, spatial, natural, artificial, and cultural dimensions. Based on 523 panoramic images and aesthetic ratings from 1175 respondents, we compared OLS, DT, MLP, SVR, RF, and XGBoost models. The best-performing XGBoost, combined with SHAP analysis, revealed threshold effects and interaction patterns among variables. Green visibility, architectural aesthetics, building visibility, sky visibility, environmental coordination, and water are the top six feature variables most strongly associated with rural streetscape aesthetic perception, and each exhibits threshold effects. The saturation threshold for green visibility is 0.153, and architectural aesthetics can only make a positive contribution when its score exceeds 3.815. The appropriate range for building visibility is below 0.452, while the optimal value for sky visibility is approximately 0.194. We also explored the context-dependence of these threshold effects across urban and rural settings. This study proposes streetscape optimization strategies focusing on screening key factors, controlling their thresholds, and coordinating the allocation of streetscape features. The interpretable analytical framework for rural scenic beauty established in this research can facilitate evidence-based landscape optimization and provide scientific support for sustainable rural development and tourism in this case.