Exploring the impact of seasonal forest landscapes on tourist emotions using Machine learning

Abstract

This study investigates the emotional changes of tourists experiencing natural forest landscapes in different seasons, which plays a crucial role in the sustainable development of forest landscape systems and forest tourism. However, existing research has limited focus on the temporal impact of detailed elements of forest landscapes on tourists' emotions. We integrated a research framework utilizing methods such as Object Semantic Attention Network (OSANet), High-Resolution Net (Hrnet), EXtreme Gradient Boosting (XGBoost), and SHapley Additive exPlanations (SHAP), thereby uncovering the nonlinear impact of detailed forest landscape indicators on tourist emotions. The results revealed(1) In the past three years, the emotional score assessments of tourists' experiences in natural forest landscapes showed a prevalence of positive emotions at 85.7%, with negative emotions constituting 14.3%. The emotional distribution characteristics varied significantly across seasons, with kernel density peak values of 4.16 in spring, 2.43 in summer, 5.33 in autumn, and 1.74 in winter. (2) The landscape indicators with the most significant impact on emotions across different seasons were Openness in spring, summer, and autumn (with contribution values exceeding 0.16, 0.13, and 0.175, respectively) and Ground exposure in winter (with a contribution value greater than 0.2). The SHAP waterfall charts indicated that the indicators providing a stable positive effect varied by seasonTrees in spring (0.361), Openness in summer and autumn (0.276 and 0.24, respectively), and Mountains in winter (0.145). This study provides a data foundation for the sustainable development of forest landscapes and tourism, aiding planners and managers in understanding the nonlinear impact of detailed elements of forest natural landscapes on tourists' emotions across different seasons. It also broadens the methodologies and perspectives for researching seasonal forest landscape preferences.