Daily rhythm of urban space usage: insights from the nexus of urban functions and human mobility

The study addresses the limitation of traditional urban zoning that emphasizes spatial heterogeneity of functions but overlooks temporal dynamics of how spaces are used throughout the day. Increasing urban density leads to multiple functions coexisting within the same zones, making human activity patterns more complex (e.g., daytime work, noon catering, nighttime entertainment). The research questions are: (1) What are the daily rhythms of urban space usage? (2) How do the distribution and combinations of urban functions shape these temporal usage patterns? Using Beijing as a case, the study proposes zoning based on daily usage rhythms derived from human mobility and examining determinants via multinomial logistic models. Understanding these rhythms is important for fine-grained urban management, planning, and transport operations.

The review covers three strands: (1) Methods of urban zoning: Historically, zoning used field observation and remote sensing to map land cover/land use but lacked functional specificity and temporal detail. With big data, POIs, mobility, and social media have been integrated to infer functional zones via clustering, yet most work still targets spatial heterogeneity and static functions. (2) Spatio-temporal dynamics of human mobility: Human-tracking data (mobile phones, GPS, transit, social media) reveal intra-urban mobility patterns at fine temporal scales. Prior studies have extracted activity types and examined temporal variation (e.g., commuting, recreation, health-seeking), but comprehensive typologies of daily space-usage rhythms remain limited. (3) Factors associated with human mobility: Classical theories (Central Place Theory, location theory, New Economic Geography, Urban Ecology) explain distribution of functions; ICT-era datasets allow micro-scale analysis of space-time trajectories. Existing work shows mobility concentrates around certain functions but often neglects hour-by-hour interplay and coexisting functions. The conceptual framework argues for zoning by daily usage rhythm and modeling how both functions and their interactions shape these rhythms.

Study area and units: The area within the Fifth Ring Road of Beijing was divided into 10,889 grid cells at 250 × 250 m resolution. Data: (1) POIs (Baidu Map, Aug 2017): ~180,000 POIs with name, coordinates, and category mapped to urban functions: working (enterprises, government), recreation/entertainment, residing, education (schools), seeking healthcare (hospitals), shopping (shops), catering (restaurants), touring (parks/tourism), etc. Table 1 reports per-cell statistics (e.g., max and averages for each type). (2) Mobile signaling data (China Unicom): Five weekdays, Jan 9–13, 2017. Records include anonymized user ID, timestamp, and location. Trips reconstructed and aggregated to grid cells; privacy preserved by aggregation. Filtered trips less than 30 minutes to retain meaningful activity stops, yielding about 6.8 million trips between grid cells. For each grid cell, computed hourly average number of destinations, forming a 24 × 10,889 time–space matrix. Clustering daily usage rhythms: To minimize magnitude effects, normalized data using a standard deviation multiplier. Used hierarchical cluster analysis to explore structure and select k=7 clusters, then applied k-means in IBM SPSS Statistics 26.0 to assign each grid cell to one of seven daily usage rhythm types. Modeling determinants: Applied multinomial logistic (MNL) regression in Stata 16.0 with the seven rhythm types as the dependent variable (low equilibrium as reference). Explanatory variables: counts of facilities by function and interaction terms capturing co-occurrence effects among functions; distance to city center as control. Diagnosed multicollinearity (high VIFs for full interaction set); addressed by organizing variables into nine models/groups where each includes one function’s interactions with the set of other functions (see Table 2), enabling inclusion of interactions without multicollinearity. Model fit assessed via chi-square (p≤0.01), log likelihood, AIC, BIC, and McFadden’s R2. Validation: Examined three typical mixed-use zones (Wangjing, Sanlitun, Shichahai) to compare observed rhythms with local function mixes.

- Seven distinct daily usage rhythm types emerged: high equilibrium and low equilibrium (relatively uniform usage at high vs. low levels); diurnal (5:00–14:00 activity concentration); nocturnal (16:00–5:00 next day); and three peak-centric types—morning (7:00–9:00), noon (10:00–14:00), and evening (16:00–21:00). - Spatial patterns: High equilibrium, diurnal, morning, and noon types concentrate in core urban areas; evening and low equilibrium types are more peripheral. High equilibrium zones cluster around major business areas and stations (Sanlitun, Chaoyangmen, Guomao; train stations). Diurnal zones are common in the city center and job clusters (Zhongguancun, Wangjing). Evening-type zones align with entertainment areas (Shichahai, Sanlitun). Low equilibrium zones are widespread on the periphery and interspersed elsewhere. - MNL model significance and fit: All models significant at p≤0.01 with good fit indicators (improved log likelihood, AIC, BIC; higher McFadden’s R2 across models), confirming explanatory power of function distributions and their interactions. - Associations between functions and rhythm types (relative to low equilibrium): • Working: increases odds of high equilibrium, diurnal, and morning types. • Residing: increases high equilibrium, diurnal, nocturnal, and evening types. • Catering: increases high equilibrium, diurnal, morning, evening, and noon types. • Seeking healthcare: increases high equilibrium, diurnal, nocturnal, morning, and noon types. • Entertainment: increases high equilibrium, diurnal, nocturnal, and evening types. • Education: increases high equilibrium, diurnal, and morning types. • Shopping: increases high equilibrium, diurnal, and evening types. • Touring: increases high equilibrium, diurnal, morning, evening, and noon types. - Interaction effects and dominance: Coexisting functions modify impacts. Working, residing, healthcare, education, and touring often act as dominant functions with cumulative effects when paired (e.g., coexisting working and residing amplify likelihoods of high equilibrium, diurnal, and nocturnal types). Catering frequently acts as an accompanying function alongside work, residence, healthcare, education, and touring, but can be co-dominant with shopping and recreation. Shopping tends to be dominant when coexisting with most functions except when paired with touring. - Validation in typical zones: Wangjing (enterprises, malls, residences) shows diurnal, morning, evening, and nocturnal patterns consistent with work- and commerce-led mixes; Sanlitun (work, catering, shopping, entertainment) exhibits high equilibrium, diurnal, nocturnal, and evening patterns; Shichahai (tourism, bars) primarily diurnal with nocturnal/high equilibrium near bar clusters and low equilibrium in residential pockets.

The findings address the research questions by demonstrating that urban spaces exhibit consistent daily usage rhythms that can be systematically identified via mobility data, and that these rhythms are strongly shaped by both the presence and co-occurrence of urban functions. The typology clarifies how specific functions activate zones at particular times (e.g., morning peaks around work and education; evening/nocturnal activity around entertainment and residence; midday activity around catering and healthcare). Interaction analyses reveal dominance and accompaniment relationships among functions, explaining cumulative or moderated effects in mixed-use settings. Validation in Wangjing, Sanlitun, and Shichahai corroborates model interpretations with observed function mixes and temporal patterns. Policy implications include designing and coordinating mixed-use developments to exploit synergistic functions, and implementing time-sensitive infrastructure and service provision (e.g., peak-hour transport support in working zones; 24-h services in residentially active areas). These insights refine fine-grained planning, transport management, and urban governance by integrating temporal dynamics with spatial function distributions.

The study proposes an urban zoning approach based on daily usage rhythms derived from mobile signaling data and k-means clustering, and explains these rhythms using multinomial logistic models of function distributions and co-occurrences. It identifies seven rhythm types—high/low equilibrium, diurnal, nocturnal, morning, evening, and noon—and links them to specific functions and their interactions. Dominant functions (work, residence, healthcare, education, touring) and accompanying roles (notably catering) shape temporal patterns in mixed-use contexts. Validation in typical zones supports the framework. The contributions include: (1) a fine-grained, time-aware zoning typology; (2) empirical evidence on how function mixes and interactions drive temporal usage; and (3) practical implications for time-sensitive planning and integrated mixed-use design. Future research could extend to weekends and seasonal variations, incorporate additional data sources, and explore causal mechanisms and policy simulations.

- Temporal and scope constraints: Mobility data cover five weekdays (Jan 9–13, 2017) within Beijing’s Fifth Ring Road; weekend/seasonal variations and outer areas are not analyzed. - Data source coverage: Mobile signaling data are from a single operator (China Unicom), which, despite wide coverage, may not capture all population segments uniformly. - Modeling considerations: Initial multicollinearity between interaction terms and base variables required grouped modeling; potential endogeneity (reverse causality and omitted variables) is discussed and argued to be limited but cannot be completely ruled out. - Functional proxies: POIs (Aug 2017) serve as proxies for functions and may not fully represent intensity or temporal operation schedules of facilities.