COVID-19 confines recreational gatherings in Seoul to familiar, less crowded, and neighboring urban areas

Human activities within urban environments significantly influence social phenomena, including city growth, economic patterns, and the spread of infectious diseases. Urban areas often exhibit a hierarchical structure, where higher-level regions attract more people for recreational purposes than lower-level ones. This hierarchy poses challenges during pandemics, as outbreaks originating in high-level hotspots can spread rapidly, and the economic impact varies across different levels. This study addresses the gap in understanding how urban hierarchies influence behavioral changes during pandemics by examining recreational gathering patterns in Seoul before and after the COVID-19 outbreak. The researchers used motel booking data, as motels in Korea are often situated near recreational venues and frequently used for leisure activities by young adults. The data provide a detailed picture of individual-level movement patterns in Seoul, allowing for a nuanced analysis of the impact of the pandemic on recreational behavior.

Existing research highlights the importance of urban hierarchies in understanding various social phenomena. Studies have shown how urban structures influence city growth, economic activity, and epidemic spread. The concentration of activity in high-level areas creates vulnerabilities during pandemics, as outbreaks in these regions can spread rapidly. Economic impacts also differ across hierarchy levels, with lower-level areas often disproportionately affected. Despite the importance of urban hierarchy, its role in analyzing behavioral changes during pandemics has received limited attention. This study builds upon existing work by explicitly incorporating urban hierarchy into the analysis of behavioral responses to COVID-19.

The study utilized anonymized customer-level accommodation reservation data from a leading Korean online platform (Goodchoice Company LTD.) covering the period from January 2019 to November 2020. This data included geographic locations of motels in Seoul. To validate the use of motel reservations as a proxy for recreational gatherings, the researchers compared reservation counts with mobility inflow data from the Seoul Open Data Plaza, finding a significant correlation, particularly for nighttime inflows. The data was divided into pre-COVID-19 (January 21, 2019 – November 3, 2019) and post-COVID-19 (January 20, 2020 – November 1, 2020) periods. A hierarchical structure of recreational areas was established by assigning hotspot levels to Google S2 cells based on aggregated reservation counts. Two measures were proposed to characterize location trajectories: hotspot entropy (Shannon entropy of hotspot levels) and radius of recreational activities (variance of geographic distances from the most frequent cell). A mechanistic model was developed to reproduce observed movement patterns, incorporating factors such as recreational hierarchy, geographic distance, and attachment to a location. The model parameters were estimated by minimizing the Jensen-Shannon divergence between simulated and empirical distributions of hotspot entropy, reservation count, and radius of recreational activities. The model allowed for the comparison of behavioral changes between the two periods.

The analysis revealed a significant drop in motel reservations around the time of the first COVID-19 case and the implementation of social distancing measures in South Korea. While reservation counts gradually recovered, the overall pattern showed a clear change in recreational behavior. The study confirmed the importance of three factors in shaping recreational gatherings: recreational hierarchy, geographical distance, and attachment to a location. Post-COVID-19, individuals demonstrated a stronger preference for familiar recreational areas, a reduced tendency to visit high-level areas, and a greater inclination to stay closer to their homes. The impact of COVID-19 was uneven across the urban hierarchy. While the proportion of visits to the highest-level areas decreased, the decrease was even more pronounced in low-level areas, suggesting a worsening of inequality. The model developed successfully reconstructed the flow matrix, distributions of hotspot entropy and radius of recreational activities, and the retention of attachment to location, both before and after the COVID-19 outbreak. Parameter estimation from the model indicated a decrease in the likelihood of exploring new places, a reduced preference for high-level recreational areas, and an increased tendency to stay closer to home after the outbreak. The changes in model parameters highlight the significant influence of COVID-19 on people's recreational behavior.

The findings indicate that the COVID-19 pandemic significantly altered recreational gathering patterns in Seoul. Individuals responded to the pandemic by limiting their exploration of new areas and favoring familiar, less crowded locations. The model successfully captured these changes, highlighting the critical role of recreational hierarchy, geographic distance, and attachment to location. The observed shift toward familiar and nearby destinations is likely due to a combination of individual choices to minimize contact risk and the effect of public health policies such as social distancing and capacity restrictions. This shift also contributed to the worsening of inequality across the urban hierarchy. The model's ability to reproduce these patterns supports the hypothesis that these factors are key determinants of recreational activity, and changes in these factors can be used to predict shifts in patterns.

This study provides valuable insights into the impact of a pandemic on recreational gatherings within a hierarchical urban structure. The findings emphasize the importance of considering individual movements and location choices when designing public health policies. The developed model serves as a useful simulation tool for anticipating behavioral changes in response to future events. Future research could explore the combined effects of voluntary behavior changes and public health guidelines on recreational activity patterns using higher-resolution data and advanced models.

Several limitations should be acknowledged. The study relies on motel booking data, which may not fully capture all recreational activities. The model assumes agents start with an empty reservation history, simplifying the complexities of individual behavior. Geographic distance was simplified to distances between cells, neglecting the complexities of urban transportation systems. The findings primarily reflect the behaviors of the primary users of the studied platform, likely low- and middle-income populations. Further research should address these limitations to gain a more comprehensive understanding.