Residents’ seasonal behavior patterns and spatial preferences in public open spaces of severely cold regions: Evidence from Harbin, China

In severely cold regions with distinct seasons, understanding the dynamic behavior patterns can provide a year-round reference for urban issues such as spatial vitality assessment, quality optimization, and promotion of public health. However, traditional methods for identifying typical behavior patterns from irregular or mixed behaviors are laborious and difficult to accurately determine the proportion of specific behaviors and their spatial preferences. Therefore, a computer vision technology-based system was developed to reveal the typical behavior patterns and their dynamic change in cold regions seasonally. Firstly, we collected behavioral data by conducting longitudinal video observations of a residential square in Harbin, and extracted trajectories of each season. Then, hierarchical clustering of trajectories was performed by calculating the similarity between trajectory pairs in each season. Afterwards, geographically weighted regression analysis was used to explore the spatial preference characteristics of different behavioral patterns. The results showed that there were five specific behavior patterns, and the overall accuracy of the behavior pattern extraction system could reach 87.5%. The functional characteristics of the square changed slightly in different seasons. In spring and autumn, optional activities or social activities account for 96%, while in winter and summer they account for 80% and 67% respectively. Additionally, specific behaviors exhibit seasonal distribution characteristics, and the impact of sky view factors (SVF), facilities, greenery, and shading on behavioral patterns varies seasonally. These findings, we hope could facilitate urban designers and planners to explore behavior-specific fine-grained information at the micro-scale for building all-season-friendly cold cities.