Seasonal Variations in the Restorative Quality Of Campus Landscapes: A Machine Learning Approach Using Crowdsourced Data

Abstract

This study proposes a seasonal machine learning model using crowdsourced data to explore differences in restorative quality between Spring and Summer at Jiangnan University. Using 800 street view images and the Perceived Restorativeness Scale-11 (PRS-11), we trained a Random Forest model to investigate the relationship between restoration and seasonal changes in campus landscapes. The results revealed four key findings: (a) spring campus landscapes are more conducive to students' mental restoration than summer landscapes; (b) waterfront areas exhibit significant seasonal differences in restorative quality, while teaching and sports zones show minimal variation; (c) positive landscape perceptions significantly enhance attention restoration quality, particularly in spring, whereas negative emotions correlate negatively with restoration in both seasons; and (d) artificial elements (e.g., buildings) and low-level landscape visual features serve as critical predictors of restorative quality. This study provides a seasonal perspective to examine the mental health benefits of campus landscapes.