Recurrent visitations reveal selectivity beyond the 15-Minute City vision

Technical Summary: Recurrent Visitations Reveal Selectivity Beyond the 15-Minute City Vision

Problem Statement
The 15-Minute City (15mC) planning paradigm posits that residents should access essential services within a short temporal threshold (e.g., 15 minutes) from home, assuming that local amenity provision translates directly into local use. However, current assessments rely heavily on supply-side accessibility metrics (e.g., isochrones) that quantify the volume of reachable opportunities but fail to capture the behavioral reality of recurrent visitation. Empirical evidence suggests a persistent gap: even in neighborhoods with complete 15-minute accessibility, individuals frequently travel beyond their nearest options. The paper addresses the lack of behavioral evidence regarding whether proximity structures recurrent visitation and to what extent the 15mC's "one-size-fits-all" proximity target aligns with actual mobility patterns, particularly in dense, transit-oriented European contexts.

Methodology
The study introduces K-Visitation, a scalable, behaviorally informed framework designed to compare recurrent visitation patterns against proximity-based expectations. The analysis utilizes an 18-month anonymized mobile phone dataset comprising approximately 720,000 users across Finland.

1. Data Processing: Stay locations were identified using the InfoStop algorithm on raw GPS pings, aggregated to H3 hexagonal grids (Level 9, ~174m). Ten daily amenity categories were derived from Point-of-Interest (POI) data.
2. K-Visitation Framework: For each individual, the framework identifies the minimal subset of visited places ($K$) required to cover all their observed amenity categories. Two distinct sets are constructed:
   • $K_{freq}$: The recurrent set, ordered by visitation frequency, representing the destinations that anchor an individual's routine.
   • $K_{dist}$: The nearest set, ordered by Euclidean distance from home, representing the nearest available options that the individual actually reached.
3. Metrics:
   • Spatial Alignment ($q_K$): A frequency-weighted overlap metric between $K_{freq}$ and $K_{dist}$. A value of 1 indicates perfect alignment; lower values indicate bypassing of nearby options.
   • Travel Time Differences: Absolute ($\Delta T$) and relative ($\Delta T_{rel}$) differences between travel time to $K_{freq}$ and a neighborhood-level proximity baseline ($T_{prox}$), calculated using public transport routing (R5Py).
4. Null Modeling: A density-aware Exploration and Preferential Return (d-EPR) model was employed to generate a null expectation ($q_K^{null}$). This model preserves generic mobility mechanisms (distance decay and density attraction) but removes specific destination preferences, allowing the authors to isolate behavioral selectivity from structural density effects.
5. Predictive Analysis: A gradient boosting classifier (XGBoost) with SHAP (SHapley Additive exPlanations) values was trained to distinguish "overshooting" destinations ($K_{freq} \setminus K_{dist}$) from "aligned" destinations ($K_{freq} \cap K_{dist}$) based on destination attributes, residential context, and travel time.

Key Results

• Systematic Divergence: Across Finnish cities, recurrent visitations systematically diverge from the nearest options. The spatial alignment metric ($q_K$) centers around 0.5, indicating that individuals repeatedly bypass nearby places that would satisfy their amenity needs.
• The "Inversion" in Urban Cores: Contrary to the 15mC hypothesis that denser areas yield higher local alignment, the sharpest divergence occurs in amenity-rich urban cores (e.g., Helsinki, Turku). In these areas, high accessibility correlates with low spatial alignment and high relative travel time differences. Residents in dense cores use local abundance as a platform to reach further, specialized destinations, incurring measurable travel time costs.
• Selectivity vs. Density: The d-EPR null model revealed that empirical alignment ($q_K$) is consistently higher than the density-predicted alignment ($q_K^{null}$). This indicates that while density attracts movement, individuals remain more locally anchored than a pure density-seeking process would predict. The divergence is driven by selective behavior: individuals target destinations that are amenity-rich but functionally specialized, rather than simply the densest or nearest options.
• Amenity Hierarchy: The study identifies a functional hierarchy in visitation patterns:
  • Routine Anchors: Services like bus stops, schools, and daycare show high alignment with proximity, adhering to the 15mC logic.
  • Common Amenities: Groceries and cafes are locally available but often bypassed for alternatives several times farther away, suggesting selective substitution or integration into wider activity chains.
  • Specialized Functions: Cinemas, cultural amenities, and shopping centers exhibit high proportional distance differentials, relying on wider catchments and agglomeration economies, consistent with Central Place Theory.
• Destination Characteristics: Overshooting destinations are characterized by high POI counts but lower functional diversity and specific social profiles (experienced segregation), rather than just high density.

Significance and Claims
The paper claims that proximity is a necessary but insufficient condition for local living. The findings challenge the operationalization of the 15mC as a uniform temporal threshold, arguing that it overlooks the foundational hierarchy of urban centers and the behavioral reality of destination selectivity.

• Theoretical Contribution: The study advances the literature by distinguishing between supply-side accessibility and realized behavioral mobility. It demonstrates that the 15mC premise—that local abundance anchors local mobility—breaks down precisely where physical conditions for proximity are most fully met.
• Policy Implication: The authors argue that 15mC implementation should not aim for a universal localization target for all amenities. Instead, planning should secure routine local access for low-order functions while maintaining sustainable connectivity for higher-order, specialized, and infrastructural functions that inherently depend on wider catchments.
• Methodological Contribution: The K-Visitation framework provides a behavioral baseline for evaluating planning scenarios, moving beyond static accessibility metrics to reveal where proximity is bypassed and where wider connectivity remains necessary.

The authors maintain modesty regarding their claims, acknowledging limitations such as the inability to verify specific amenity usage from stay locations, the exclusion of transient populations, and the specific context of Finland's urban form and transport infrastructure. They suggest that while their findings resonate with US data, social implications may differ in hyper-dense contexts.