Capturing urban recreational hotspots from GPS data: A new framework in the lens of spatial heterogeneity

Recreational activities are heterogeneously distributed throughout urban space, with far more low-density areas than high-density ones. Identification of recreational hotspots, or high-density areas, plays a critical role in urban planning. Nevertheless, from the perspective of urban heterogeneity, recreational hotspots remain inadequately understood for further theoretical and empirical investigations. Hence, based on the volunteered GPS trajectory data, we established a novel framework for effectively capturing recreational hotspots. The entire process can be divided into three steps: extracting stay points from individuals' tracks; clustering points by using heavy-tailed distribution statistics of the point-point proximities based on triangular irregular network (TIN); and generating the hotspots and integrating them with street segments. To assess the proposed framework, we started by introducing it in three typical Chinese cities and analyzing the reliability of the capturing process. Furthermore, taking one of the three cases as an example, we compared the proposed framework with current widely-used clustering methods, namely, K-means, DBSCAN and CFSFDP. The results show that the proposed framework performs well in both the empirical investigations and methodological comparisons, as it not only highlights the existing hotspots that are in line with general public perceptions, but also outperforms the three clustering algorithms in terms of fitness of the purpose, rapidity, and accuracy. Overall, this study extends urban heterogeneity to the application of the urban recreational system and provides potentials for its redesign and improvement. •Proposes a novel framework for capturing recreational hotspots in lens of the spatial heterogeneity.•Use stay/stop behavior as a probe to detect recreational hotspots.•Introduces ‘mean’ effect to capture the clusters of geo-data with heavy-tailed distributed pattern.•Conducts a practicalcomparative analysis between the proposed framework and three traditional clustering methods.