Day-night visitor usage of urban parks: Exploring temporal dynamics and driving factors through mobile phone big data

Urban parks serve as prominent venues for nature-based recreational activities. However, current research does not adequately address how temporal factors influence enjoyment of urban park services and fails to delineate variations in recreational demand at different times. To address this research limitation, we developed a novel methodology using mobile phone data to accurately estimate the duration of stay of park visitors in Sapporo, Japan, providing insights into people's recreational preferences. We employed Random Forest regression and Negative Binomial regression models to identify variations in daytime and nighttime recreational activity levels, as well as their main influencing factors across 28 urban parks in Sapporo from 2019 to 2020. Research findings indicate a significant reduction in visitor numbers by 59.83 % from pre-pandemic levels in 2019, accompanied by an increase in the average duration of stays. Comparative data from 2019 and 2020 underscore a decrease in nighttime activities, from 72.41 % in 2019 to 62.07 % in 2020. Additionally, supported by an R-squared value of 0.695, the Random Forest model identified points of interest (23.6 %), NDVI (24.1 %), and distance to the CBD (17.8 %) as the key variables predicting visitor duration of stay, with points of interest being particularly prominent in 2020. Transportation services (β=0.9236) and public services (β=0.8958) significantly encourage longer stays during the daytime. Conversely, educational services (β=1.9333) and entertainment services (β=1.7023) exhibit a substantial positive impact at nighttime, highlighting pronounced temporal differences in visitor preferences. These results suggest that targeted enhancements in park amenities could significantly improve urban recreational spaces, effectively catering to diverse lifestyle and leisure needs.