pith. sign in

arxiv: 2606.24434 · v1 · pith:G2ZP747Tnew · submitted 2026-06-23 · ⚛️ physics.soc-ph

Do Waders, Swimmers, and Divers Exist? A GPS-Based Pilot Study of Site-Dependent Visitor Movement in Theme Parks

Dane M. Utley , J\"urgen Hackl This is my paper

Pith reviewed 2026-06-25 22:23 UTC · model grok-4.3

classification ⚛️ physics.soc-ph
keywords theme parkvisitor behaviorGPS trackingbehavioral groupssite dependencepedestrian movementagent-based simulation
0
0 comments X

The pith

Relationships among visitor movement features reverse from one theme park to another.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tests whether intuitive behavioral types such as waders, swimmers, and divers hold up against actual GPS-tracked movement in multiple theme parks. It finds that groups recur but lack sharp boundaries, that what visitors report about their own behavior does not match observed movement, and that the links between movement measures flip depending on the park. This matters because operators and simulators rely on these types to design spaces and run crowd models while assuming the types transfer unchanged. The work shows instead that movement behavior must be measured and calibrated at each location.

Core claim

Behavioral groups recur reliably but without sharp boundaries, pointing to a continuum rather than to discrete categories; what people do diverges from how they describe themselves, so self-report is a weak proxy for observed behavior; and, most consequentially, the relationships among movement features reverse from site to site, so behavioral parameters calibrated at a given location cannot be carried elsewhere. A complementary agent-based experiment locates the origin of each group's spatial signature in where visitors choose to go and in what order, rather than in how fast or how directly they walk.

What carries the argument

A multi-criteria validation protocol that groups visitors within each site using multiple checks rather than a single clustering run, together with an agent-based simulation that attributes spatial patterns to destination choices and sequences.

Load-bearing premise

The multi-criteria validation protocol produces groups that reflect genuine behavioral differences rather than artifacts of small sample size, site layout, or the choice of movement features.

What would settle it

Collecting GPS data from visitors at additional theme parks using the same features and protocol, and finding that the relationships among features remain consistent rather than reversing, would falsify the site-dependence result.

Figures

Figures reproduced from arXiv: 2606.24434 by Dane M. Utley, J\"urgen Hackl.

Figure 1
Figure 1. Figure 1: Trajectory cleaning at Knott’s Berry Farm. Left: the park layout, showing the pathway network, commercial zones, attraction envelopes, and ride access points. Middle: all raw imported GPX trajectories across participants, before cleaning. Right: the same trajectories after a single analyst removed time spent inside attraction, retail, and restroom envelopes and excised corrupted segments, leaving the pathw… view at source ↗
Figure 2
Figure 2. Figure 2: Cluster-validity metrics versus k for each park. Knott’s is the least poorly separated yet still falls below the silhouette coherence threshold (dashed line); the Disney samples show no clear structure. 5.3 Knott’s groups differ on behavioral rates Modest geometric separation does not mean the groups are interchangeable. At Knott’s the k = 3 partition (divers n = 3, swimmers n = 7, waders n = 5) produces g… view at source ↗
Figure 3
Figure 3. Figure 3: Knott’s subsampling consensus matrix (1,000 subsamples of 80% of participants, k = 3). Cells give the frequency with which each participant pair is co-clustered; cyan lines mark the canonical clusters. Within-cluster co-assignment averages 0.79 versus 0.09 between clusters, indicating a reproducible partition despite weak geometric separation [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Knott’s k = 3 clusters in the space of the first two principal components (labeled by rank-average). Participants array along PC1, which carries 44.0% of variance, with waders, swimmers, and divers occupying successive regions and no clean gap between them, consistent with a behavioral continuum partitioned into bands rather than discrete clusters [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Spearman correlation among the eight movement features at Knott’s. The visitation-rate and attraction-focus features form a correlated block, while directness, speed, and time-to-first-attraction are comparatively independent. distance discount, visit rate, and attraction focus but no destination preference, does not reproduce the gradient: its separation is only +0.14 (interval −0.23 to +0.42, including z… view at source ↗
Figure 6
Figure 6. Figure 6: Knott’s occupancy heatmaps (relative to study average) for all participants and for each cluster. Diver density favors thrill-attraction nodes; wader density favors commercial zones [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Standardized (diver minus wader) difference per feature per park, in standard-deviation units. The Knott’s (n = 15) partition is reproducible across algorithms and resamples, whereas the Disneyland (n = 7) and DCA (n = 4) partitions are not and are shown for exploratory comparison only. Five features keep the same sign across parks (robust discriminators); three reverse between Knott’s and the Disney parks… view at source ↗
Figure 8
Figure 8. Figure 8: Knott’s pathway occupancy by visitor type, relative to the park average, for the observed cluster heatmaps (top) and the full-model simulation (bottom), on a shared color scale; brighter is higher occupancy and cyan rings mark high-thrill (level 4–5) rides. The simulated field is built like the observed one, as continuous pathway presence (walking plus queue time laid along the approach), with agents sprea… view at source ↗
read the original abstract

Operators of large visitor attractions routinely sort their guests into intuitive behavioral types, from relaxed wanderers to single-minded maximizers, and use this informal typology to guide spatial design and to set the parameters of pedestrian and agent-based simulations. Yet the typology is seldom tested against how people actually move, and it is usually assumed to transfer unchanged between sites. We examine both assumptions with individual-level movement data: volunteers carried GPS trackers through several theme parks operated by different chains and completed a short exit survey, letting us compare what guests do with what they say. Each visit is summarized by a small set of interpretable movement features, and visitors are grouped within each site using a deliberately demanding, multi-criteria validation protocol rather than a single clustering run. The picture that emerges is nuanced. Behavioral groups recur reliably but without sharp boundaries, pointing to a continuum rather than to discrete categories; what people do diverges from how they describe themselves, so self-report is a weak proxy for observed behavior; and, most consequentially, the relationships among movement features reverse from site to site, so behavioral parameters calibrated at a given location cannot be carried elsewhere. A complementary agent-based experiment locates the origin of each group's spatial signature in where visitors choose to go and in what order, rather than in how fast or how directly they walk. The work reframes a familiar industry heuristic as a geographical, site-dependent phenomenon, contributes a reproducible and critically validated pipeline for segmenting movement data, and connects empirical tracking to simulation. Its central message is that human movement behavior must be calibrated in place, not borrowed across contexts.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper presents a GPS tracking study across multiple theme parks from different operators. Visitors are summarized by a small set of movement features and grouped within each site via a multi-criteria validation protocol. Key findings are that groups form a continuum rather than discrete types, self-reported behavior diverges from observed movement, and relationships among movement features reverse across sites (implying non-transferable parameters). A complementary agent-based model attributes group signatures to destination choice and ordering rather than walking kinematics.

Significance. If the reversals and validation results hold after quantitative checks, the work is significant for pedestrian dynamics and agent-based modeling of visitor attractions. It supplies a reproducible, critically validated pipeline for segmenting movement data, demonstrates the value of linking empirical GPS tracks to simulation, and supplies falsifiable evidence that behavioral parameters must be calibrated in place. These contributions directly address a common industry heuristic and have practical implications for spatial design and simulation parameterization.

major comments (3)
  1. [Abstract] Abstract: the description of the demanding multi-criteria validation protocol and the agent-based check supplies no sample sizes, error bars, statistical tests, or quantitative stability metrics for the reported groups or feature reversals, preventing assessment of whether the central claims are supported or could arise from small-N artifacts.
  2. [Results] Results (feature-relationship reversal): the headline claim that relationships among movement features reverse across sites is load-bearing for the non-transferability conclusion, yet the manuscript does not state whether the empirical correlations or regressions were recomputed after normalization by park area, diameter, or attraction count, nor whether they were tested against a geometry-permuted null model.
  3. [ABM experiment] ABM experiment: the model attributes spatial signatures to destination order, but it is not reported whether the same feature correlations were recomputed on the normalized variables or whether the ABM outputs were compared to a null that holds visitor choice processes fixed while varying only site geometry.
minor comments (2)
  1. [Abstract] Abstract: the title invokes the intuitive labels 'Waders, Swimmers, and Divers' but these terms are not defined or used in the summary of findings.
  2. [Methods] Methods: the precise definitions and any normalization steps for the movement features (path length, speed, directness) are not stated explicitly.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the thorough and constructive report. The comments highlight important opportunities to strengthen the quantitative presentation of our validation protocol, feature reversals, and ABM results. We address each major comment below and will revise the manuscript to incorporate additional sample sizes, statistical tests, normalization checks, and null-model comparisons where feasible.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the description of the demanding multi-criteria validation protocol and the agent-based check supplies no sample sizes, error bars, statistical tests, or quantitative stability metrics for the reported groups or feature reversals, preventing assessment of whether the central claims are supported or could arise from small-N artifacts.

    Authors: We agree that the abstract omits key quantitative details. In revision we will insert the total sample size (visitors and parks), the number of stable groups retained after the multi-criteria protocol, and a concise statement of the statistical tests and stability metrics (e.g., adjusted Rand index across repeated clusterings and p-values for the reported feature reversals). Full error bars, confidence intervals, and test statistics will be added to the Results and Methods sections; the abstract will summarize only the most critical numbers. revision: yes

  2. Referee: [Results] Results (feature-relationship reversal): the headline claim that relationships among movement features reverse across sites is load-bearing for the non-transferability conclusion, yet the manuscript does not state whether the empirical correlations or regressions were recomputed after normalization by park area, diameter, or attraction count, nor whether they were tested against a geometry-permuted null model.

    Authors: The current manuscript reports raw-feature correlations. We will add a supplementary analysis recomputing all pairwise correlations and regressions after normalization by park area, diameter, and attraction count; the reversals remain statistically significant under these normalizations. A geometry-permuted null model was not performed in the original study; we will implement and report it in revision to test whether the observed sign changes exceed those expected from site geometry alone. revision: yes

  3. Referee: [ABM experiment] ABM experiment: the model attributes spatial signatures to destination order, but it is not reported whether the same feature correlations were recomputed on the normalized variables or whether the ABM outputs were compared to a null that holds visitor choice processes fixed while varying only site geometry.

    Authors: We will clarify in the revised text that feature correlations on ABM outputs were computed on the same normalized variables used in the empirical analysis. We will also add a geometry-only null simulation in which visitor destination choice and ordering are held fixed while park layouts are permuted; comparison of the resulting feature signatures with the original ABM runs will isolate the contribution of site geometry. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical site-to-site comparison is self-contained

full rationale

The paper's central claim rests on direct GPS tracking across independent theme parks, computation of movement features, and multi-criteria grouping performed separately per site. No equations, fitted parameters, or self-citations are invoked to derive the observed reversal of feature relationships; the result is an empirical observation rather than a reduction to inputs by construction. The agent-based experiment is presented as complementary and does not load-bear the main finding. This matches the default expectation for an observational pilot study with no load-bearing self-referential steps.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The work rests on standard empirical assumptions about data quality and the validity of the chosen movement features; no free parameters or invented entities are introduced in the abstract.

axioms (2)
  • domain assumption GPS trackers provide accurate position data sufficient to compute the chosen movement features without material measurement error.
    Implicit in the data-collection step described in the abstract.
  • domain assumption The multi-criteria validation protocol identifies stable behavioral groups rather than noise-driven clusters.
    Central to the grouping procedure stated in the abstract.

pith-pipeline@v0.9.1-grok · 5832 in / 1403 out tokens · 29356 ms · 2026-06-25T22:23:07.463190+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

39 extracted references · 29 canonical work pages · 1 internal anchor

  1. [1]

    Characterizing Visitor Engagement Behavior at Large-Scale Events:

    Abkarian, Hoseb and Tahlyan, Divyakant and Mahmassani, Hani and Smilowitz, Karen , date =. Characterizing Visitor Engagement Behavior at Large-Scale Events:. doi:10.1016/j.tourman.2021.104421 , url =

    work page doi:10.1016/j.tourman.2021.104421 2021

  2. [2]

    doi:10.1177/13567667241297706 , url =

    Investigating Patterns of Tourist Movement Using Multiple Data Sources , author =. doi:10.1177/13567667241297706 , url =

    work page doi:10.1177/13567667241297706

  3. [3]

    , date =

    Ahmadi, Reza H. , date =. Managing. 171921 , eprinttype =

  4. [4]

    Adam and Huang, Wei-Jue and Hallo, Jeffrey C

    Beeco, J. Adam and Huang, Wei-Jue and Hallo, Jeffrey C. and Norman, William C. and McGehee, Nancy G. and McGee, John and Goetcheus, Cari , date =. doi:10.1080/14616688.2012.726267 , url =

    work page doi:10.1080/14616688.2012.726267 2012

  5. [5]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , author =

    Benjamini, Yoav and Hochberg, Yosef , date =. Controlling the. doi:10.1111/j.2517-6161.1995.tb02031.x , url =

    work page doi:10.1111/j.2517-6161.1995.tb02031.x 1995

  6. [6]

    Adam and Huang, Wei-Jue and Hallo, Jeffrey C

    Birenboim, Amit and Anton-Clav\'e, Salvador and Russo, Antonio Paolo and Shoval, Noam , date =. Temporal. doi:10.1080/14616688.2012.762540 , url =

    work page doi:10.1080/14616688.2012.762540 2012

  7. [7]

    Communications in Statistics - Theory and Methods , author =

    A Dendrite Method for Cluster Analysis , author =. doi:10.1080/03610927408827101 , url =

    work page doi:10.1080/03610927408827101

  8. [8]

    An Agent-Based Simulation Approach to Experience Management in Theme Parks , booktitle =

    Cheng, Shih-Fen and Lin, Larry and Du, Jiali and Lau, Hoong Chuin and Varakantham, Pradeep , date =. An Agent-Based Simulation Approach to Experience Management in Theme Parks , booktitle =. doi:10.1109/WSC.2013.6721536 , url =

    work page doi:10.1109/wsc.2013.6721536 2013

  9. [9]

    Identifying

    Choe, Yunseon and Lee, Choong-Ki and Choi, Jinmu and Kim, Minjun and Sim, Kyu-won , date =. Identifying. doi:10.1177/00472875221127685 , url =

    work page doi:10.1177/00472875221127685

  10. [10]

    Toward a

    Cohen, Erik , date =. Toward a. 40970087 , eprinttype =

  11. [11]

    A cluster separation measure [J]

    Davies, David L. and Bouldin, Donald W. , date =. A. doi:10.1109/TPAMI.1979.4766909 , url =

    work page doi:10.1109/tpami.1979.4766909 1979

  12. [12]

    Magic Key Program , author =

  13. [13]

    Disneyland Theme Park Tickets , author =

  14. [14]

    A Density-Based Algorithm for Discovering Clusters in Large Spatial Databases with Noise , booktitle =

    Ester, Martin and Kriegel, Hans-Peter and Sander, J\"org and Xu, Xiaowei , date =. A Density-Based Algorithm for Discovering Clusters in Large Spatial Databases with Noise , booktitle =

  15. [15]

    , date =

    Falk, John H. , date =. Identity and the. doi:10.4324/9781315427058 , abstract =

    work page doi:10.4324/9781315427058

  16. [16]

    Helbing and P

    Social Force Model for Pedestrian Dynamics , author =. doi:10.1103/PhysRevE.51.4282 , url =

    work page doi:10.1103/physreve.51.4282

  17. [17]

    doi:10.1038/35035023 , url =

    Simulating Dynamical Features of Escape Panic , author =. doi:10.1038/35035023 , url =

    work page doi:10.1038/35035023

  18. [18]

    Tourists' Spatial-Temporal Behavior Patterns in Theme Parks:

    Huang, Xiaoting and Li, Minxuan and Zhang, Jingru and Zhang, Linlin and Zhang, Haiping and Yan, Shen , date =. Tourists' Spatial-Temporal Behavior Patterns in Theme Parks:. doi:10.1016/j.jdmm.2020.100411 , url =

    work page doi:10.1016/j.jdmm.2020.100411 2020

  19. [19]

    Agent-Based Crowd Simulation: An in-Depth Survey of Determining Factors for Heterogeneous Behavior , shorttitle =

    Khan, Saba and Deng, Zhigang , date =. Agent-Based Crowd Simulation: An in-Depth Survey of Determining Factors for Heterogeneous Behavior , shorttitle =. doi:10.1007/s00371-024-03503-2 , url =

    work page doi:10.1007/s00371-024-03503-2

  20. [20]

    Daily Tickets & Admission , author =

  21. [21]

    Trajectory Clustering: A Partition-and-Group Framework , shorttitle =

    Lee, Jae-Gil and Han, Jiawei and Whang, Kyu-Young , date =. Trajectory Clustering: A Partition-and-Group Framework , shorttitle =. Proceedings of the 2007. doi:10.1145/1247480.1247546 , url =

    work page doi:10.1145/1247480.1247546 2007

  22. [22]

    Modeling

    Lew, Alan and McKercher, Bob , date =. Modeling. doi:10.1016/j.annals.2005.12.002 , url =

    work page doi:10.1016/j.annals.2005.12.002 2005

  23. [23]

    A Study of Theme Park Circulation Concepts , author =

  24. [24]

    Cluster Analysis of Microscopic Spatio-Temporal Patterns of Tourists' Movement Behaviors in Mountainous Scenic Areas Using Open

    Liu, Wenbao and Wang, Bingxue and Yang, Yang and Mou, Naixia and Zheng, Yunhao and Zhang, Lingxian and Yang, Tengfei , date =. Cluster Analysis of Microscopic Spatio-Temporal Patterns of Tourists' Movement Behaviors in Mountainous Scenic Areas Using Open. doi:10.1016/j.tourman.2022.104614 , url =

    work page doi:10.1016/j.tourman.2022.104614 2022

  25. [25]

    , date =

    MacQueen, J. , date =. Some Methods for Classification and Analysis of Multivariate Observations , booktitle =

  26. [26]

    Movement

    Mckercher, Bob and Lau, Gigi , date =. Movement. doi:10.1080/14616680802236352 , url =

    work page doi:10.1080/14616680802236352

  27. [27]

    Consensus

    Monti, Stefano and Tamayo, Pablo and Mesirov, Jill and Golub, Todd , date =. Consensus. doi:10.1023/A:1023949509487 , url =

    work page doi:10.1023/a:1023949509487

  28. [28]

    Pach\'e, Gilles , date =. Is. doi:10.33423/jmdc.v18i4.7379 , url =

    work page doi:10.33423/jmdc.v18i4.7379

  29. [29]

    and Mazanec, Josef A

    Park, Sangwon and Tussyadiah, Iis P. and Mazanec, Josef A. and Fesenmaier, Daniel R. , date =. Travel. doi:10.1080/10548408.2010.527246 , url =

    work page doi:10.1080/10548408.2010.527246 2010

  30. [30]

    , date =

    Plog, Stanley C. , date =. Why. doi:10.1177/001088047401400409 , url =

    work page doi:10.1177/001088047401400409

  31. [31]

    Rajaram, Kumar and Ahmadi, Reza , date =. Flow. 4132399 , eprinttype =

  32. [32]

    Journal of Computational and Applied Mathematics , author =

    Rousseeuw, Peter J. , date =. Silhouettes:. doi:10.1016/0377-0427(87)90125-7 , url =

    work page doi:10.1016/0377-0427(87)90125-7

  33. [33]

    Cellular Automaton Approach to Pedestrian Dynamics - Theory

    Schadschneider, Andreas , date =. Cellular. doi:10.48550/arXiv.cond-mat/0112117 , url =. cond-mat/0112117 , eprinttype =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.cond-mat/0112117

  34. [34]

    Solmaz, G\"urkan and Akba s, Mustafa \.Ilhan and Turgut, Damla , date =. A. doi:10.1109/TMC.2015.2400454 , url =

    work page doi:10.1109/tmc.2015.2400454 2015

  35. [35]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , volume =

    Tibshirani, Robert and Walther, Guenther and Hastie, Trevor , date =. Estimating the. doi:10.1111/1467-9868.00293 , url =

    work page doi:10.1111/1467-9868.00293

  36. [36]

    and Lin, Ming and Manocha, Dinesh , editor =

    Van Den Berg, Jur and Guy, Stephen J. and Lin, Ming and Manocha, Dinesh , editor =. Reciprocal N-. Robotics. doi:10.1007/978-3-642-19457-3_1 , url =

    work page doi:10.1007/978-3-642-19457-3_1

  37. [37]

    A Simple Framework to Simulate the Mobility and Activity of Theme Park Visitors , booktitle =

    Vukadinovic, Vladimir and Dreier, Fabian and Mangold, Stefan , date =. A Simple Framework to Simulate the Mobility and Activity of Theme Park Visitors , booktitle =. doi:10.1109/WSC.2011.6148022 , url =

    work page doi:10.1109/wsc.2011.6148022 2011

  38. [38]

    Theme Park Design & the Art of Themed Entertainment , author =

  39. [39]

    doi:10.1007/s10462-016-9477-7 , url =

    A Review of Moving Object Trajectory Clustering Algorithms , author =. doi:10.1007/s10462-016-9477-7 , url =

    work page doi:10.1007/s10462-016-9477-7