Visual Perception Within the Built Environment of Traditional Villages in Karst Areas: Exploring Multidimensional Associations of Landscape Features

Abstract

Traditional villages in karst natural heritage areas are distinctive rural built environments where natural, cultural, and landscape elements intertwine. Identifying relationships between built-environment characteristics and visual preference is important for improving spatial quality and supporting sustainable tourism. This study examined Jiuxian Village in Guilin, China, using stimulus–organism–response (S-O-R) theory to develop an indicator framework. We collected 194 panoramic images and recruited 72 university students to evaluate visual quality using Scenic Beauty Estimation (SBE) and Semantic Differential (SD) methods. Image features were quantified through computer vision, and machine-learning models combined with SHapley Additive exPlanations (SHAP) were applied for interpretation. Visual order and immediate affective responses were positively associated with SBE, whereas sense of history showed a weaker direct association with SBE than these two dimensions. Visible natural elements and color richness contributed positively, while building coverage, hardscape, and public facilities contributed negatively. Natural and anthropogenic elements displayed nonlinear predictive patterns, with SHAP contributions varying across proportional ranges, as well as positive, negative, and conditional joint predictive patterns. These findings provide guidance for landscape planning and spatial-quality improvement in traditional villages within karst environments.