Decoding Urban Dynamics: Contextual Insights from Human Meta-Mobility Patterns

Abstract

Research on capturing human mobility patterns for efficient and sustainable urban planning has been widely conducted. However, studies that unveil spatial context beyond macro-level mobility patterns are relatively scarce. This study aims to analyze the spatiotemporal human meta-mobility patterns with rich context using POI data in Seoul from comprehensive perspectives. As a result, the floating population of Seoul exhibits regular and irregular cyclical mobility patterns on weekdays and weekends, respectively, stemming from the periodicity of the dominant POIs. Additionally, graph construction based on mobility similarity and their regional clustering show clusters vary by POIs but are generally divided into peripheral and central regions of Seoul. This indicates that socioeconomic factors cannot be ignored when understanding human mobility patterns. This helps to provide scientific evidence to support policy recommendations towards greenways and sustainable urban mobility systems, such as quantitative disparity of greenways, qualitative issues of greenways in the central areas, and inequality in cultural consumption. Addressing key considerations through targeted policies could significantly improve the overall quality of life for urban residents. We expect this study to lay the groundwork for future research that aims to understand realistic human mobility patterns with a rich context.

1. Introduction

Modern cities have complex structures due to the dense concentration of population and social capital, which results in various socio-economic phenomena. Particularly, Seoul is one of the largest urban agglomerations in the world [1] and is characterized by a high population density per unit area and a significant number of floating populations relative to its area. Here, the floating population refers to individuals who move to and from a specific area within a particular time frame or period, including not only residents but also visitors, commuters, shoppers, tourists, and others. That is, the floating population is not limited to individuals who temporarily stay in the area but includes all people who regularly move within or through the area. Analyzing the mobility data of urban populations reveals diverse movement patterns shaped by various activities, offering crucial insights into the dynamics and characteristics of the city. This analysis can be utilized in urban planning and policy-making to provide a basis for more efficient and sustainable urban mobility systems. Specifically in metropolitan areas like Seoul, analyzing the floating population plays a crucial role in managing transportation systems, designing public services, and preventing epidemics [2,3,4,5,6,7].

Accordingly, various attempts have been made to understand the macroscopic mobility patterns of floating populations. The Seoul Life Movement Data [8] provided by the Seoul Metropolitan Government defines “life movement” as “the movement of all populations in Seoul between specific points in time and regions”. However, actual floating populations move for various purposes even within the same time and region. Thus, if we limit the life movement to uniform and macroscopic spatiotemporal dimensions, it becomes challenging to understand the structural, functional, and contextual characteristics of the city inherent in the urban mobility dynamics.

In this regard, this study aims to analyze and understand the spatiotemporal human mobility patterns combined with rich situational context from both macroscopic and microscopic perspectives. To achieve this, we analyze the daily life movement patterns of the floating population in Seoul by investigating the Foursquare mobility data in which check-ins, geographic coordinates, and POI (Point of Interest) information of users are recorded. Additionally, POI meta-information and regional visitation statistics are combined in order to unfold meta-mobility patterns of the city floaters. In our study, ‘meta-mobility’ refers to human mobility combined with POI meta-information, allowing us to understand movements within their contextual framework. Most POI-based mobility studies provided POI-specific movement patterns within a region or in a short term, which is limited to obtaining insights into hyper mobility from macroscopic perspectives. We then apply graph clustering across the 25 districts of Seoul based on mobility networks generated by the similarities of meta-mobility probability distributions. Furthermore, we explore greenways and sustainable urban mobility systems by focusing on the radius of life activities, greenways, and transportation systems based on POI check-ins and meta-information.

The analysis results indicate that the periodic mobility patterns of Seoul’s floating population are largely bifurcated into weekdays and weekends, reflecting regular and irregular lifestyle patterns, respectively. By incorporating spatial characteristics using POI meta-information, we observed that the periodic patterns during weekdays are further reinforced around the top POI metas with high visit probabilities. The similarity of lifestyle mobility patterns among Seoul’s districts varies by POI meta, and overall there is a tendency for the districts to be clustered into central and peripheral regions. Such polarization reflects the disparities in housing and commuting location choices based on economic circumstances and implies that socioeconomic factors have a strong influence on urban mobility patterns. In terms of the radius of life activities, the center and outskirts of Seoul exhibit different levels of inequality: compared to residents in the center, residents in the outskirts have lower accessibility to greenways, workplaces, and high culture (fine arts) in that order. While central districts benefit from higher accessibility, they also experience higher levels of air pollution, partly due to the density of the transportation system and the floating population. This helps to provide scientific evidence to support policy recommendations towards greenways and sustainable urban mobility systems, such as quantitative disparity of greenways, qualitative issues of greenways in central areas, and inequality in cultural consumption. Addressing key considerations through targeted policies could significantly improve the overall quality of life for urban residents.

This study contributes to a multifaceted understanding of spatiotemporal meta-mobility patterns of Seoul’s floating population by contextualizing POI check-in time series records, reflecting spatial characteristics, from both macroscopic and microscopic perspectives. POI-based meta-mobility exhibits more dynamic clustering of regions that transcends geographical proximity, highlighting structural, functional, and contextual aspects in the dynamics of urban mobility. This study shows an analytical methodology that can derive contextual insights into human mobility from general POI movement data and presents ways of supporting policy recommendations based on scientific evidence. We anticipate that the research outcomes will serve as a foundational study with applications across a broad spectrum of disciplines.

2. Related Work

With the development of transportation systems, the boundaries between regions have become increasingly blurred, and urban living zones have expanded, leading to a dynamic evolution in the distinction between physical and functional boundaries of cities. Understanding and applying such phenomena to urban planning has become more significantly challenging. However, the rise of lifelogging through various mobile devices and the consequent large-scale collection of daily mobility data has facilitated numerous studies aimed at analyzing human mobility patterns to understand urban characteristics and inform urban planning [9,10,11,12].

Studies comparing and analyzing the hourly mobility patterns of the floating population across urban areas worldwide have shown differences in mobility patterns specific to each city [13,14,15]. These analyses, conducted at a macroscopic level, focus on the volume of floating population movements during specific time periods. Furthermore, examining the average mobility radius of urban areas in connection with local characteristics has revealed that analyzing human mobility patterns can provide a deeper understanding of urban socio-economic features [16,17,18].

Point of Interest (POI) data are crucial for analyzing human mobility patterns from a microscopic perspective. It has been demonstrated that considering POI information can enhance next-location prediction models and develop personalized location recommendation systems based on mobility [19,20,21,22,23]. Additionally, by adding meta-information to the edges connecting movement paths, studies have identified socio-economic factors such as wealth disparity [24,25,26]. This underscores the importance of POI data analysis for a multifaceted understanding of human mobility patterns [27,28].

Nevertheless, there is a lack of studies that combine POI metadata and spatial information to analyze human mobility patterns from a comprehensive perspective that includes situational context. Therefore, this paper aims to conduct a step-by-step and multifaceted analysis of the spatiotemporal meta-mobility patterns of the floating population inherent in POI data and integrate these with socio-economic factors in order to elucidate the structural and functional characteristics of urban districts in Seoul, one of the largest urban agglomerations in the world.

3. Data

3.1. Data Collection

To analyze the spatiotemporal mobility patterns in urban areas, we focused on Seoul, which accounts for over 90% of South Korea’s floating population, utilizing POI visit data. For this purpose, we collected check-in data from Foursquare users worldwide, as disclosed in previous studies [29,30]. The dataset spans approximately 22 months from April 2012 to January 2014, including service usage logs from 114,324 users. It comprises 22,809,624 check-ins across 3,820,891 locations (an average of 6 check-ins per POI). The data include the timestamp of each check-in (based on UTC), the latitude and longitude coordinates of the check-in locations, and the POI category information visited by users. Each user is assigned a unique anonymized ID, allowing us to track individual mobility paths. That is, Foursquare data have the unique advantage of providing not only the precise coordinates of the locations visited by users but also detailed POI information about these locations.

Although there is a temporal gap between the data collection period and the time of this study, the collected data hold significant research value considering the practical difficulties of obtaining large-scale POI data over a long-term period and the representativeness of Foursquare as a location-based service platform with a large global user base during the data collection period. Furthermore, this study focuses on an analytical methodology that can derive contextual insights into mobility patterns from general POI movement data, independent of the Foursquare platform.

3.2. Data Preprocessing

Based on the latitude and longitude coordinates of Seoul’s district boundaries, we extracted the relevant datasets from the Foursquare data, which contain check-in information from users worldwide, using the Python Shapely module, as shown in Figure 1. This figure represents POIs as nodes (black dots) and user movements as edges (gray lines). We applied the GeoPandas module to add administrative district information of Seoul corresponding to the check-in coordinates. Consequently, we constructed a dataset of floating population POI check-ins across Seoul’s 25 administrative districts from 3 April 2012, to 29 January 2014. For accurate time classification, we converted the original check-in times to Korea Standard Time (KST). Through preprocessing, the dataset was refined to 124,186 check-ins in Seoul, recorded by 1992 Foursquare users. Specifically, the locations visited by users totaled 29,970, categorized into 386 POIs. For contextual purposes, these POIs were reclassified into 16 POI metas according to an industry classification standard [31]. The basic statistics of the Seoul Foursquare dataset used in this study are shown in

Table 1. Basic statistics of preprocessed and integrated Foursquare datasets used in this study.

Information Items	Number of Records
#Users	1992
#Check-ins	124,186
#Places	29,970
#POI (Point of Interest)	386
#POI (Point of Interest) metas	16
Latitude range	[37.413294, 37.75133]
Longitude range	[126.734086, 127.269311]

.

We utilize Open Street Map (OSM) to verify that the latitude and longitude coordinates of POIs correspond to the relevant regions on OSM. Foursquare POIs provide POI names based on user-generated keywords along with their respective categories, whereas OSM lacks user visitation intentions. To address this, we employed Foursquare POIs to obtain contextual mobility information and used OSM to verify the accuracy of POI coordinates by marking them on OSM via Folium in accordance with their names.

4. Methods

In this section, we outline our analytical approaches based on POI data and interpret the main results. Figure 2 illustrates the key analytical processes employed in this study, along with corresponding figure numbers.

4.1. Clustering of Seoul Districts Based on Macroscopic Movement Patterns

Based on the preprocessed dataset, we first represent an anonymized individual user i’s movement paths as a graph $G_i=(V_i,E_i)$ to collectively understand the macroscopic movement patterns of the floating population in Seoul. In this graph, the vertices $V_i=\{u,v|(u,v)\in E_i\}$ represent the districts that contain the points of interest (POIs) visited by a user i, and the edges $E_i$ denote the travel paths between these POIs in sequential order. Specifically, an edge $(u,v)\in E_i$ is formed when a user i moves from a POI located in one district u to a POI in another district v. Subsequently, we combine the POI visit networks $G_i$’s of each individual user i to construct a directed weighted graph $G={\left\{G_i\right\}}_{i=1}^n=\{V,E|V_i\in V\}$ for the n entire users. This graph reflects the movement flow between districts in Seoul as recorded by Foursquare users. To capture the strong connectivity between regions with high floating population movement, we apply the Louvain algorithm to cluster the combined graph G [32,33].

Figure 3a,b displays the results of graph clustering, distinguishing between (a) weekdays and (b) weekends, while Figure 3c illustrates the division of living zones in reference to the ‘2030 Seoul Urban Master Plan’ established by the Seoul Metropolitan Government [34]. From the overall clustering patterns observed in Figure 3a,b, it is evident that there is a strong connectivity between adjacent areas. Notably, there is a clear distinction between the northern and southern regions of the Han River, with minimal cross-river movement, indicating that geographical proximity significantly influences inter-regional mobility. However, the fact that not all adjacent areas form a single cluster suggests that regional characteristics, such as population density and transportation systems, serve as boundaries to daily movement.

The structural characteristics of mobility patterns become more noticeable when cluster analysis is conducted separately for weekdays and weekends. During weekdays (Figure 3a), the cluster sizes are in general uniformly distributed, whereas on weekends (Figure 3b), the cluster sizes expand, particularly around the Southwest and Northwest Living Zones. This suggests that the regular movement patterns such as commuting and school attendance during weekdays result in a mobility range limited to within 2–3 districts. In contrast, on weekends, when time constraints are relatively less stringent, the mobility range of the floating population expands.

The delineation of living zones in Seoul, established based on the physical environment and the functions of districts as shown in Figure 3c, exhibits a pattern similar to the expanded mobility range observed in the weekend clusters in Figure 3b. However, it shows a significant difference in cluster size, approximately twice as large, compared to the reduced mobility range observed in the weekday clusters (Figure 3a). This indicates that while the actual daily mobility of Seoul’s floating population differs from the statically defined living zones, it reflects the boundaries shaped by homogeneous characteristics of the regions, such as population density and transportation systems.

4.2. Temporal Mobility Patterns of the Floating Population

Using the clustering of districts based on mobility patterns, we examined the comprehensive weekday and weekend mobility ranges of Seoul’s floating population. Next, to explore the periodic characteristics of mobility patterns, we first divided the time series data on a weekly basis. To capture and magnify the movements by time of day, we further subdivided the week into 168 h (24 h/day × 7 days) and calculated the probability of POI visits by the floating population for each time slot, as illustrated in Figure 4.

In Figure 4, it can be observed that the periodicity of mobility patterns is bifurcated into weekdays $(0\le h\le 120)$ and weekends $(120<h\le 168)$. This reflects (1) the regular daily routines such as commuting and school attendance during weekdays and (2) the more autonomous and irregular activity patterns during weekends, characterized by visits to various points of interest (POIs) at more flexible times. A closer examination of the weekday plot reveals three sharp peaks in activity occurring at 9 a.m., 1 p.m., and 8 p.m., corresponding to increases in mobility associated with commuting to work, lunchtime, and post-work movements, respectively. In contrast, the weekend plot shows a more gradual increase in POI visit probability, peaking at 7 p.m., without the concentration of activity at specific times seen during weekdays. This suggests that the simultaneous events that drive weekday mobility are less evident on weekends, allowing for a more relaxed and autonomous schedule for various activities.

4.3. Temporal Mobility Patterns of the Floating Population by POI Meta

To consider the spatial characteristics of the hourly movements of Seoul’s floating population, the visit probabilities in Figure 4 were recalculated based on POI meta categories. Figure 5 shows the probability distributions of visits, incorporating both the spatial characteristics of the POI metas and the periodic time information, by categorizing all POIs into 16 different meta categories.

As shown in Figure 5, while the overall movement patterns during weekdays and weekends maintain a periodicity consistent with the earlier analysis, it is evident that each POI (Point of Interest) meta-category exhibits its own unique periodicity.

This trend is particularly pronounced for higher-ranking POI metas with higher visitation probabilities compared to lower-ranking ones. Specifically, this includes the restaurant, transportation system, culture and entertainment, cafe, shop, business, and education meta categories. Notably, the higher usage probabilities of restaurants and transportation systems correlate strongly with the sharp increases in mobility probability observed at 9 a.m., 1 p.m., and 7–8 p.m. in Figure 4. This suggests that these peaks are driven by activities related to lunch and dinner times as well as commuting.

4.4. Similarity of Movement Patterns across POI Meta Categories

To analyze the similarity of periodic mobility patterns among the top 7 POI meta categories with the highest visit probabilities, we calculated the similarity using Kullback–Leibler (KL) divergence [35,36].

The following Equation (1) calculates the similarity between the visit probability distributions $P^r\left(\tau \right)$ and $P^s\left(\tau \right)$ over time $\tau$ of POI meta categories r and s as:

$$D_{\mathrm{K}L}(r,s)=\sum _{\tau }{P}^r\left(\tau \right)log{\displaystyle \frac{P^r\left(\tau \right)}{P^s\left(\tau \right)}}$$

(1)

Since KL divergence is directional, we used Equation (2) in order to ensure that the similarity matrix between POI meta categories is symmetrical as:

$$D_{\mathrm{K}LS}(r,s)={\displaystyle \frac{1}{2}}\left(D_{\mathrm{K}L}(r,s)+D_{\mathrm{K}L}(s,r)\right)$$

(2)

Figure 6 presents the clustering results of the POI meta-similarity matrix, calculated using Equation (2) and clustered using the Louvain algorithm, revealing three major clusters. The similarities among POI meta categories reflect the common patterns of daily life movement. In the figure, the size of the nodes and the thickness of the edges are proportional to the number of visitors to the POI metas and the similarity between the metas, respectively. Edges with low similarity were filtered to extract clusters of POI metas with strong similarities.

Based on the clustering results in Figure 6, the averages and variations in the daily visit probability distributions of POI meta categories during weekdays and weekends were separated by cluster and shown in Figure 7a,b, for each. The figure allows us to obtain hourly visit probabilities throughout the day. For the detailed probability distributions of each POI meta-category, refer to Figure A1 and Figure A2 for weekdays and weekends, respectively.

Summarizing the findings, the first cluster, the restaurant meta, has the highest usage probability, exhibiting regular peak times during both weekdays (Figure 7a) and weekends (Figure 7b). This suggests that meal times remain consistent, maintaining their periodicity. The second cluster, centered on culture and entertainment in Figure 6, shows high similarity in periodic mobility patterns but differences between weekdays (Figure 7a) and weekends (Figure 7b). This can be interpreted as reflecting the limited time available for these activities during weekdays, whereas more flexible timing is possible on weekends. Lastly, the third cluster, comprising business and education metas Figure 6, exhibits periodic mobility patterns related to commuting and school attendance during weekdays in Figure 7a, while showing irregular mobility patterns without specific peaks during weekends in Figure 7b.

4.5. Spatiotemporal Similarity of Mobility Patterns across Districts in Seoul

Previously, we examined the similarity of hourly mobility patterns among the top 7 POI meta categories. Now, for each POI meta category, we aim to analyze the spatiotemporal similarity of mobility patterns among districts in Seoul by extracting weekday POI visit data for each district. To achieve this, we calculated the hourly visit probabilities $P_i^m\left(\tau \right)$ for a POI meta category m located in district i. To calculate the similarity of mobility patterns between districts for each POI meta category, Equation (1) is refined into Equation (3), representing the Kullback–Leibler (KL) divergence between the probability distributions of regions i and j for the POI meta category m as:

$$D_{\mathrm{K}L}^m(i,j)=\sum _{\tau }{P}_i^m\left(\tau \right)log{\displaystyle \frac{P_i^m\left(\tau \right)}{P_j{i}^m\left(\tau \right)}}$$

(3)

Since KL divergence is directional, Equation (4) is used to ensure that the similarity matrix between districts is symmetrical as:

$$D_{\mathrm{K}LS}^m(i,j)={\displaystyle \frac{1}{2}}\left(D_{\mathrm{K}L}^m(i,j)+D_{\mathrm{K}L}^m(j,i)\right)$$

(4)

Using Equations (3) and (4), we generate an undirected weighted graph based on the similarity between the 25 districts of Seoul. We then perform cluster analysis through graph clustering. Figure 8 visualizes the clustering results for POI meta categories with high modularity (greater than 0.3), specifically (a) restaurant, (b) transportation, and (c) culture and entertainment, with different colors representing the clusters. Additionally, Figure 9 provides supporting data for Figure 8, showing the average probability distributions for each cluster within the respective POI meta category.

In Figure 8a, the restaurant POI meta category shows a distinct division between the outskirts (brown) and the central area (gray) of Seoul. Both clusters exhibit sharp peaks at 12 p.m. and 7 p.m. in Figure 9a; however, the probability of restaurant visits is higher in the central area compared to the outskirts, with lower variation between regions in the center. This can be attributed to the influence of working hours on employees in the central area, where industries are concentrated, resulting in more consistent restaurant visit times and relatively higher visit probabilities compared to the outskirts.

The transportation POI meta category in Figure 8b shows that all three clusters experience a sharp increase in usage probabilities during commuting hours as shown in Figure 9b). However, the outskirts of Seoul (brown) have earlier usage times and significantly higher usage rates during the morning commute compared to the other two clusters located in the central area. In the evening commute, the peak usage time is later for all clusters. Additionally, Figure 9b reveals that the usage rate of the transportation system in the outskirts (brown) is relatively lower than in the central area during the afternoon period between commutes. This reflects the structural characteristic of Seoul where residential and work areas are different, especially in the outskirts.

For the culture and entertainment POI meta category in Figure 8c, the clusters are relatively evenly distributed compared to other meta categories. In Figure 9c, all clusters show an increase in usage after 6 p.m., indicating that the floating population disperses to various districts in order to engage in cultural and entertainment activities after work hours. Furthermore, the later usage times in the outskirts (brown) are linked to the earlier findings in the transportation meta category, suggesting a time lag as people travel home from the central area after work.

All in all, different POI meta categories reveal structural, functional, and contextual characteristics of the 25 districts in Seoul.

4.6. Meta-Mobility Patterns and Demographic Variables

In Figure 10a, Seoul’s districts are clustered based on the similarities in visit probabilities for the integrated top 7 POI meta categories. Comparatively, Figure 10b presents the clustering results of the average apartment sale prices per square meter in Seoul, collected during the same period as the dataset [37]. Figure 10c shows the clusters based on low-income household rates among the districts [38].

As the figure shows, the spatiotemporal meta-mobility patterns (Figure 10a) of Seoul’s floating population reveal a polarization between central and peripheral areas, closely mirroring the distribution of apartment sale prices (Figure 10b) and low-income household rates (Figure 10c) in Seoul. This indicates a significant separation between residential and commuting areas in the outskirts of Seoul, reflecting the disparity between living and economic activity locations. In other words, the mobility patterns of the floating population, based on the major POI meta categories, are influenced by socio-economic factors. Such bifurcation of clusters particularly highlights the urban characteristics where the floating population is drawn from the peripheral areas to the central areas, which are densely equipped with various economic infrastructures.

The clustering of spatiotemporal mobility patterns based on POIs differs from that of movements between geographical statistical units (e.g., administrative districts). The latter tends to form clusters among neighboring regions, while POI-based patterns exhibit more dynamic clustering that transcends geographical proximity. With this advantage, in the following section, we will conduct a multifaceted analysis of the structural, functional, and contextual aspects of Seoul’s districts from the perspective of greenways and sustainable urban mobility systems. This analysis will focus on the socio-economic clusters that reflect the dichotomous nature of the most frequent top 7 POI meta-mobility patterns, particularly between the central and peripheral areas of Seoul, as illustrated in Figure 10b.

5. Towards Greenways and Sustainable Urban Mobility Systems

In this section, we aim to explore greenways and sustainable urban mobility systems through POI-based meta-mobility patterns. This involves focusing on the radius of life activities, greenways, and transportation systems based on POI check-ins and meta-information.

5.1. Radius of Life Activities

The source and destination of movements convey important information such as the directionality of human mobility and the the frequency of visits to various destinations. Understanding the origin, particularly the home, is essential for comprehending the radius of life activities, as it reflects the socioeconomic conditions in urban environments. In our dataset, Foursquare users rarely checked in at home, with approximately 200 users doing so. Although the number of users is limited, we can still discern tendencies in life activities away from home. For example, we can analyze the distances users are willing to travel to POIs, which can be interpreted as the influence or attractiveness of target destinations.

In this regard, we focus on POIs related to workplaces, greenways, and cultural centers, reflecting economic activity hubs, recreational spaces, and cultural spaces, respectively. Specifically, cultural spaces are categorized into ‘popular culture’ (entertainment), such as movies, performances, and theater, and ‘high culture’ (fine arts), including art exhibitions, opera, and orchestra performances. Figure 11 presents the cumulative distribution functions (CDFs) of distance in kilometers from home to visit destinations related to (a) workplaces, (b) greenways, (c) high culture, and (d) popular culture for the center (red) and outskirts (blue) of Seoul, as shown in Figure 10b. Note that 3 kilometers indicate the average radius of Seoul districts.

In general, the majority of residents in the outskirts of Seoul tend to visit high culture, workplaces, popular culture, and greenways that are farther away compared to residents of central districts, in that order. For instance, 50% of residents in central districts are likely to commute to work within 7.5 km (within first-order neighboring districts), whereas those in outskirts tend to commute within 9 km (within second-order neighboring districts). Regarding cultural spaces, residents in the outskirts of Seoul exhibit distinct mobility patterns for consuming high culture and popular culture; for high culture, the majority travel about 12 km (within third-order neighboring districts), while for popular culture, they travel less than 6 km (within second-order neighboring districts). Figure 12 accordingly illustrates the radius of the four types of life activities for the center and outskirts of Seoul.

In Figure 12, concentric circles represent the radius of life activities, with circle colors indicating the distributions of residents (from the inside out: 25%, 50%, 75%) for each center and outskirts of Seoul. As shown in the figure, the radii of life activities for residents in central and peripheral districts of Seoul differ significantly. Residents in central districts generally have a uniform radius, whereas residents in peripheral districts exhibit highly uneven radii depending on the purpose of travel, with a larger radius compared to central residents.

This imbalance in peripheral areas is particularly pronounced for workplace and high culture activities. For workplaces, the commuting radius for most peripheral residents exceeds 9 km, indirectly suggesting that the psychological threshold for commuting is around one hour by public transport. For high-culture activities, the travel radius for peripheral residents is the largest, with a minimum distance of 9 km. This indicates a lack of spaces for consuming high culture in peripheral areas.

5.2. Greenways and Sustainability

It has been often assumed that there would be a significant correlation between the geographic distribution of natural greenery and greenways, but we observed that this is not the case in Seoul. Figure 13 illustrates several key aspects of green spaces and urban living in Seoul: (a) the actual distribution of open space, primarily natural greenery, in Seoul [39], (b) the distribution of greenway-related POIs, such as hiking trails, fields, river, parks, and outdoor recreations, (c) the levels of satisfaction with the green environment across Seoul’s 25 districts, and (d) clusters grouped by one-person household rates.

As shown in the figure, there is a discrepancy between (a) the actual green spaces and (b) POI check-in records related to greenways (visit frequency is represented by a heat map). Specifically, there are more visit records in the central part of Seoul, while the outskirts, which are richer in green spaces, have relatively fewer visit records. The discrepancy may be partly due to the higher concentration of the floating population in central Seoul, including those working in the area who can easily access greenways during their break time. Figure 13c clusters the top 10 districts with the highest satisfaction in green space environments, and Figure 13d clusters the top 10 districts with the highest proportion of single-person households, provided by [38]. It can be observed that peripheral districts, which have a higher number of multi-person households, generally show lower satisfaction with green space environments compared to central districts. This supports the need for greater accessibility to green spaces in the outskirts of Seoul to facilitate family-oriented activities and community life, leading to increased overall satisfaction and quality of life for residents.

According to a survey conducted by the Seoul Institute [38], 54.3% of Seoul residents reported that their primary exercise locations, whether for regular or irregular exercise, are parks and trails such as greenways. This suggests that greenways are a significant factor influencing the health and well-being of Seoul residents. Additionally, the survey found that the proportion of individuals who do not avoid exercise (“exercise irregularly” or “exercise regularly”) reaches approximately 80% among those with higher income and educational levels. This indicates that socioeconomic factors are closely linked to human movement patterns.

Figure 14 represents the utilization levels of various modes of transportation ((a) subway, (b) bus, (c) private vehicle) based on POI check-in records, depicted through a heatmap. As shown in the figure, the utilization of the transportation system is significantly higher in the central area of Seoul compared to its outskirts. This higher utilization is associated with dissatisfaction regarding air pollution (Figure 13d), provided by [38]. While the central area benefits from the infrastructural advantages of the transportation system, it also suffers from the disadvantages of accompanying air pollution. In other words, although access to greenways is high, the issue of air pollution, which is contrary to green environments, coexists.

5.3. Scientific Evidence for Policy Suggestions

This study presents scientific evidence to support policy recommendations aimed at enhancing urban well-being in Seoul, based on an analysis and understanding of human meta-mobility patterns. POI-based spatiotemporal data reveals several key considerations for urban policy-making:

• Quantitative Disparity of Greenways: there is a significant imbalance in the distribution of open space and greenways across Seoul, necessitating more equitable allocation to ensure all residents have sufficient access to green spaces.
• Qualitative Issues of Greenways in Central Areas: this higher utilization of the transportation systems is associated with dissatisfaction regarding air pollution, indicating a need for improvements in qualitative greenway policies.
• Inequality in Cultural Consumption: there is a marked disparity in consumption opportunities of high culture between the central districts and the outskirts of Seoul, suggesting the need for policy measures to promote a balanced cultural experience citywide and ease socio-economic disparities.

Addressing these issues through targeted policies could significantly improve the overall quality of life for Seoul’s residents. First, on quantitative regional disparity of greenways, closely interconnected greenway network policies are required. As shown in Figure 13b, it is evident that certain regions, especially the outskirts of Seoul, are marginalized in terms of access to greenway POIs. Since having even access to greenways is crucial for the citizens’ overall well-being [40], a policy that creates a dense urban greenway network is highly necessary. Furthermore, greenway POIs provide supportive evidence to effectively find and connect blind spots of the greenway network beyond geographical structure.

Second, on the qualitative issues of greenways in central areas, policies implementing urban ventilation corridors are necessary. Mobility pattern and satisfaction surveys reveal an ironical relationship between access to green spaces and the well-being of residents. Central districts benefiting from better access to greenways and also experiencing higher levels of air pollution are attributed to the dense distribution of transportation POIs in the region. This contrast underscores the necessity for policies on air circulation that bring qualitative improvement of greenways and the layout of roads and buildings.

Finally, the inequality of cultural consumption suggests that policies to narrow the socio-economic disparities are required. Residents in central Seoul have more opportunities to experience high culture due to highly accessible amenities of high culture, while residents in the outskirts must travel significantly farther to reach them, leading to far fewer opportunities for cultural consumption. An integration of greenways and culture as mixed-used greenway facilities would be an ideal solution as it can effectively increase both the traveling exhibition space of the cultural activity and the sustainability of limited urban areas. Such multifaceted greenways enable the cultural integration of Seoul, leading to the reconfiguration of spaces such as media art parks.

6. Discussion

The novelty of this study lies in incorporating the meta-mobility of city floaters into spatiotemporal movements based on POI visitations. This approach helps to reveal urban dynamics with rich context, as POI check-ins convey time-evolving user intentions of visitations and latent demands on POI types (i.e., POI meta-categories) and thus provide a realistic portrayal of urban life.

Contextual human meta-mobility elucidates the structural, functional, and socioeconomic characteristics of urban cities, in contrast to context-free collective human mobility [10,41,42,43] and mobility snapshots without a time dimension [44,45]. As shown in

Table 2. Description of features and characteristics.

Features	Characteristics	Description
Intuitive Features	Temporal characteristics	Mobility patterns are largely bifurcated into weekdays and weekends, reflecting regular and irregular life movements.
	Spatial characteristics	Collective mobility are clustered among adjacent districts (especially, the Han River acts as a geographical barrier).
	Spatiotemporal characteristics	While regular movement patterns during weekdays maintain a consistent range (2–3 districts), irregular movement patterns during weekends expand the movement range (2–7 districts).
Counterintuitive Features	Gap between static and dynamic distributions	The distribution of natural green spaces differs from the distribution of POI-based visits related to greenways. This suggests that static indicators like a geographical distribution may overlook demand for visits.
	Dynamic clustering beyond geographical boundaries (association with socio-economic factors)	Meta-mobility based on the top 7 POI meta categories shows a polarized trend towards the center and outskirts of Seoul, differing from the existing clustering tendency among adjacent areas. Additionally, these polarized clusters show a strong correlation with socio-economic indicators (e.g., real estate prices, low-income rates).
	Dynamic radius of life activities (association with socio-economic imbalance)	The radius of life activities varies between the center and outskirts depending on context such as workplaces, greenways, high culture, and popular culture. Movement range in central Seoul is generally uniform, whereas it is greatly uneven in the outskirts. The uneven movement range in outskirts implies an imbalance in related infrastructure (e.g., lack of cultural facilities related to high culture (fine arts) in the outskirts). Most residents in the outskirts share a psychological threshold for commuting within about 9 km (a 1-h distance by public transport), indicating a separation between economic activity areas and residential areas.
	No winners	Central areas have relatively well-balanced infrastructure such as greenways, cultural facilities, economic activity centers, and transportation systems compared to outskirts, but also suffer from air pollution caused by high density of population and transportation systems.

, the main findings of the study are two-fold: intuitive (common) features and counterintuitive (special) features of human meta-mobility.

7. Conclusions

This study attempted to understand mobility patterns by incorporating POI meta-information and quantifying similarities in cyclic movement probability distributions across regions over a two-year period, which enables the analysis of spatiotemporal mobility with context. It also provides scientific evidence to support policy recommendations regarding greenways and sustainable urban mobility systems, such as quantitative disparity of greenways, qualitative issues of greenways in central areas, and inequality in cultural consumption. Addressing key considerations through targeted policies could significantly improve the overall quality of life for urban residents.

Collective mobility patterns are largely bifurcated into weekdays and weekends, showing different clusters among the 25 districts in Seoul. Specifically, the uniform cluster sizes during weekdays reflect efficient commuting ranges for regular activities such as school and work commutes, while the uneven cluster sizes during weekends indicate more autonomous mobility radii centered around transportation infrastructure.

Furthermore, by dividing the time series data on a weekly basis and estimating the probability distributions of hourly POI visitations, we identified both regular and irregular patterns of mobility during weekdays and weekends. By calculating and clustering the similarity of cyclic movements across Seoul districts for each POI meta category, we gained insight into a realistic portrayal of urban life.

Overall, the analysis indicates a pronounced bifurcation between the central and peripheral areas of Seoul, reflecting the degree of separation in the radius of life activities. This suggests that the mobility patterns of the floating population are influenced by socio-economic factors. The study highlights the importance of analyzing spatiotemporal mobility patterns to understand contextual situations.

Future research plans include developing methodologies for more flexible and dynamic definitions of POI meta categories, moving beyond static classifications, and applying these methodologies to time series models to predict population movement.