Campus squares are significantly associated with students’ emotions: A multimodal field experiment approach

University students are vulnerable to psychological distress under intense academic and social pressure. Although campus spaces are recognized as relevant to student well-being, evidence for campus-square environments remains limited. This study investigated emotional responses in 12 squares at Tsinghua University through 516 multimodal field experiment sessions. Physiological monitoring, PANAS emotion scales, questionnaires, and first-person visual recordings captured both session-level emotional change and momentary visual exposure. A two-scale framework was applied: session-level OLS models assessed associations between spatial characteristics and emotional changes, while within-session centered XGBoost models tested whether 5-s visual-scene deviations explained short-term SCL fluctuations. Long-term models identified significant associations involving road density, floor area ratio, car-related visual composition, and participant characteristics. Appraisal-augmented models showed that subjective evaluations offered complementary explanations for self-reported valence. Short-term analyses revealed strong session-level dependence in relative SCL; current visual elements added limited predictive value, but sky view factor and green view index were significantly and negatively associated with relative SCL after accounting for participant and square clustering. Overall, under real campus field conditions, emotional responses were shaped mainly by broader spatial context, while momentary visual exposure provided modest, interpretable physiological signals for emotionally supportive campus design.