pith. sign in

arxiv: 2606.23616 · v1 · pith:SUFPQJH6new · submitted 2026-06-22 · ⚛️ physics.soc-ph · cs.CY

One country, multiple portraits: representativeness in GPS-based mobility data is source-specific and spatially dependent

Carmen Cabrera , Francisco Rowe , Miguel Gonz\'alez-Leonardo , Juan Ignacio Vilchis-Garc\'ia , Elisa Omodei , Maribel Hern\'andez-Rosales This is my paper

Pith reviewed 2026-06-26 06:06 UTC · model grok-4.3

classification ⚛️ physics.soc-ph cs.CY
keywords coverage biasmobile phone dataGPS mobilityspatial dependenceMexicorepresentativenessdigital access
0
0 comments X

The pith

GPS mobility data coverage bias differs by source and follows spatial patterns across Mexican municipalities.

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

The paper compares population estimates derived from two anonymised GPS mobile phone sources against the 2020 Mexican Population Census in 2478 municipalities. It establishes that Facebook data yield higher and more evenly distributed coverage, whereas multi-app aggregator data over-represent larger, wealthier, and more digitally connected places. Bias levels cluster spatially so that neighbouring municipalities tend to share similar over- or under-coverage. Different covariates dominate in each source: digital access and material resources for the multi-app data, demographic structure for Facebook. Explicitly including spatial dependence improves model performance yet leaves an appreciable share of variation unexplained.

Core claim

Coverage bias in GPS-based mobile phone data is source-specific and spatially dependent, with Facebook providing higher and more evenly distributed coverage while multi-app data concentrate users in larger, wealthier and more digitally connected places; neighbouring municipalities exhibit similar bias levels, and digital access drives bias for multi-app data whereas demographics drive it for Facebook.

What carries the argument

Direct comparison of source-derived population counts to the 2020 Mexican Census, followed by explainable machine learning for driver identification and spatial dependence modeling.

Load-bearing premise

The 2020 Mexican Population Census constitutes an unbiased and complete ground truth for measuring coverage bias in the mobile phone datasets.

What would settle it

Observing identical magnitude and spatial distribution of coverage bias across the two sources, or finding that spatial dependence vanishes once observed covariates are controlled, would falsify the source-specific and spatially structured claims.

Figures

Figures reproduced from arXiv: 2606.23616 by Carmen Cabrera, Elisa Omodei, Francisco Rowe, Juan Ignacio Vilchis-Garc\'ia, Maribel Hern\'andez-Rosales, Miguel Gonz\'alez-Leonardo.

Figure 1
Figure 1. Figure 1: Overview of the analytical workflow used to assess bias in GPS-based mobile phone data (MPD). (A) Population estimates from single-app and multi-app MPD are analysed together with census-based variables to generate municipal-level population counts and covariates. (B) Bias bi is then estimated for each municipality i and dataset D as the deviation between MPD-based population P D i and census population co… view at source ↗
Figure 2
Figure 2. Figure 2: Comparison of population coverage across single-app (Facebook) and multi-app GPS-based mobile phone datasets in Mexico. (A) Scatterplot comparing multi-app coverage (x-axis) and single-app coverage (y-axis) for each municipality, both on logarithmic scales; the dashed line denotes the β = 1 proportionality benchmark. (B) Relationship between municipal population size (x-axis) and coverage in each dataset (… view at source ↗
Figure 3
Figure 3. Figure 3: Spatial patterns of coverage bias across municipalities for single-app and multi-app datasets. (A) Municipal-level maps of coverage bias (bi = 1−ci ), with darker shades indicating higher bias (i.e. lower coverage relative to the census population). (B) Local Indicators of Spatial Association (LISA) cluster maps identifying high–high (HH), low–low (LL), high–low (HL), low–high (LH) and non-significant (NS)… view at source ↗
Figure 4
Figure 4. Figure 4: Improved model performance after including spatially lagged bias as a model input suggests that neighbouring municipalities provide information about coverage bias beyond that captured by local demographic, socioeconomic and digital-infrastructure covariates. Radial plots summarise relative changes in model performance metrics for (A) the multi-app dataset and (B) the Facebook dataset when spatially lagged… view at source ↗
Figure 5
Figure 5. Figure 5: Spatially explicit interpretation of XGBoost models using SHAP values. Maps show municipality-level SHAP values for the six most influential covariates identified by the XGBoost models for the multi-app (A–F) and single-app (G–L) datasets. Models are estimated using block cross-validation and include a spatially lagged bias term based on a 10-nearest neighbours specification. SHAP values represent the cont… view at source ↗
Figure 1
Figure 1. Figure 1: Pairwise correlations between municipal-level coverage bias and demographic, socioeconomic and technology access model covariates, for the multi-app and Facebook datasets. Correlations are measured with Spearman’s correlation coefficient. Colour intensity indicates the magnitude and direction of the correlation, with negative values corresponding to lower bias (higher coverage) in municipalities with highe… view at source ↗
Figure 2
Figure 2. Figure 2: Global Moran’s I statistics for municipal-level coverage bias, reported for the multi-app and Facebook datasets using queen contiguity and k-nearest neighbours (kNN) spatial weight specifications. Reported values correspond to the mean Moran’s I across permutations, with standard deviations in parentheses. Statistical significance is assessed using permutation-based tests. 3/6 [PITH_FULL_IMAGE:figures/ful… view at source ↗
Figure 3
Figure 3. Figure 3: Relative importance of input variables in XGBoost models estimating municipal-level coverage bias for the multi-app (blue) and Facebook (red) datasets. Importance is measured using mean absolute SHAP values. Panels (A–D) show results from models estimated using spatial block cross-validation, while panels (E–H) show results from random hold-out validation with a 70–30 training–test split. For each validati… view at source ↗
Figure 4
Figure 4. Figure 4: Robustness test of Facebook user counts to the order of temporal and spatial aggregation. The municipal-level Facebook user counts obtained by first computing temporal averages at the tile level and then aggregating spatially to municipalities are shown on the x-axis; counts obtained by first aggregating tile-level observations to municipalities for each day and then computing temporal averages are shown o… view at source ↗
read the original abstract

Anonymised GPS-based mobile phone data are increasingly used to estimate population distribution and human mobility, supporting applications across disaster response, public health, urban planning and migration research. Yet whether these data fairly represent the populations they describe, particularly outside high-income countries, remains poorly understood. We quantify coverage bias for 2,478 municipalities in Mexico by comparing population estimates from a single-platform source (Facebook) and a multi-app aggregator (Veraset) against the 2020 Mexican Population Census. We find that the magnitude and spatial distribution of coverage bias differ substantially across sources. Facebook provides higher and more evenly distributed coverage, whereas the multi-app data concentrate users in larger, wealthier and more digitally connected places. Coverage bias is also spatially structured, with neighbouring municipalities showing similar levels of over- or under-coverage. Using explainable machine learning, we show that digital access and material resources are the dominant drivers of bias for the multi-app data, while demographic and population structure dominate for Facebook. Explicitly modelling spatial dependence improves the performance of statistical models for explaining bias and reveals that an appreciable share of spatial variation remains unexplained by observed covariates. These findings show that coverage bias is source-specific and spatially dependent, and provide a foundation for adjustments that improve the representativeness of mobile phone data in unequal, data-scarce settings.

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

2 major / 1 minor

Summary. The manuscript quantifies coverage bias in two GPS-based mobile phone datasets (Facebook single-platform and Veraset multi-app aggregator) for 2,478 Mexican municipalities by direct comparison to 2020 census population counts. It reports that bias magnitude and spatial distribution differ substantially by source, with Facebook showing higher and more uniform coverage while Veraset concentrates in larger, wealthier, digitally connected places. Bias exhibits spatial autocorrelation; explainable ML attributes drivers (digital access/resources for Veraset, demographics for Facebook); and explicit spatial modeling improves explanatory performance while leaving appreciable variation unexplained. The central claim is that coverage bias is source-specific and spatially dependent.

Significance. If the results hold after addressing ground-truth concerns, the work provides a useful empirical demonstration that representativeness of mobility data is not interchangeable across commercial sources and carries spatially structured patterns. Credit is due for the dual-source comparison, municipal-scale analysis, incorporation of spatial dependence in the explanatory models, and use of interpretable ML to separate drivers. These elements supply a concrete basis for source-aware adjustments in applications such as public health or disaster response in middle-income settings.

major comments (2)
  1. [Data and Methods] Data and Methods sections: All coverage-bias quantities (magnitude, spatial distribution, and ML driver attributions) are defined as deviations from the 2020 Mexican Population Census municipality totals. No sensitivity analysis, post-enumeration adjustment, or comparison to independent population estimates is described to address documented census undercounts in rural, low-income, or indigenous areas. This assumption is load-bearing for the reported source-specific differences and driver rankings.
  2. [Results] Results section on explainable ML and spatial models: The claim that 'explicitly modelling spatial dependence improves the performance of statistical models' and that 'an appreciable share of spatial variation remains unexplained' is central, yet the manuscript provides no table or text reporting the specific spatial specification (e.g., spatial lag, error, or CAR model), the quantitative improvement (ΔR², AIC, or Moran’s I reduction), or cross-validation details for the ML models. Without these, the improvement and residual-variation statements cannot be evaluated.
minor comments (1)
  1. [Abstract] Abstract and Methods: Sample sizes, exact coverage fractions per dataset, and the precise definition of the bias metric (e.g., ratio, difference, or log-ratio) are not stated, making it difficult to reproduce the reported percentages or spatial patterns.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed review. The comments highlight important issues regarding ground-truth assumptions and reporting of model details. We respond to each major comment below and will make targeted revisions to improve clarity and transparency while preserving the core findings on source-specific and spatially dependent coverage bias.

read point-by-point responses

  1. Referee: [Data and Methods] Data and Methods sections: All coverage-bias quantities (magnitude, spatial distribution, and ML driver attributions) are defined as deviations from the 2020 Mexican Population Census municipality totals. No sensitivity analysis, post-enumeration adjustment, or comparison to independent population estimates is described to address documented census undercounts in rural, low-income, or indigenous areas. This assumption is load-bearing for the reported source-specific differences and driver rankings.

    Authors: We agree this is a substantive concern. The 2020 Mexican Census is the official ground truth, but undercounts in rural, low-income, and indigenous municipalities are documented in the literature. We lack access to post-enumeration survey data or alternative population estimates for a formal sensitivity analysis. In revision we will add an explicit Limitations subsection in the Discussion that (i) cites studies on census accuracy in Mexico, (ii) discusses how differential undercount could affect the two data sources, and (iii) notes that the source-specific patterns we report should be interpreted with this caveat. This addresses the load-bearing nature of the assumption without overstating what the data allow. revision: yes

  2. Referee: [Results] Results section on explainable ML and spatial models: The claim that 'explicitly modelling spatial dependence improves the performance of statistical models' and that 'an appreciable share of spatial variation remains unexplained' is central, yet the manuscript provides no table or text reporting the specific spatial specification (e.g., spatial lag, error, or CAR model), the quantitative improvement (ΔR², AIC, or Moran’s I reduction), or cross-validation details for the ML models. Without these, the improvement and residual-variation statements cannot be evaluated.

    Authors: We accept that the quantitative details were insufficiently reported. The manuscript used a spatial error model; the revised version will add (i) the exact specification and estimation method, (ii) a table showing R², AIC, and Moran’s I before/after spatial terms, (iii) 5-fold cross-validation results for the explainable ML models, and (iv) the proportion of residual spatial variation left unexplained. These additions will make the performance claims directly verifiable. revision: yes

Circularity Check

0 steps flagged

Empirical comparison against external census benchmark with no reduction to fitted parameters or self-citations

full rationale

The paper quantifies coverage bias by direct comparison of Facebook and Veraset user counts to 2020 Mexican Population Census municipality totals, then applies explainable ML to attribute observed deviations to covariates such as digital access and demographics. No equations, model fits, or claims reduce by construction to the inputs (e.g., no parameter fitted on a subset then relabeled as a prediction of a related quantity). No load-bearing self-citations or uniqueness theorems are invoked. The derivation chain consists of standard statistical comparisons and feature-importance analysis against an external benchmark, remaining self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Analysis rests on the census as ground truth and on the assumption that mobile data can be aggregated to municipality level without unmeasured selection effects; no free parameters or invented entities are described in the abstract.

axioms (1)
  • domain assumption The 2020 Mexican Population Census provides an accurate and complete population benchmark.
    Used directly as the reference for calculating coverage bias.

pith-pipeline@v0.9.1-grok · 5797 in / 1036 out tokens · 26736 ms · 2026-06-26T06:06:20.747335+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

57 extracted references · 23 canonical work pages · 2 internal anchors

  1. [1]

    Green, M., Pollock, F . D. & Rowe, F .New Forms of Data and New Forms of Opportunities to Monitor and T ackle a Pandemic, 423–429 (Springer International Publishing, 2021)

    2021

  2. [2]

    C., Hidalgo, C

    Gonzalez, M. C., Hidalgo, C. A. & Barabasi, A.-L. Understanding individual human mobility patterns.nature453, 779–782 (2008)

    2008

  3. [3]

    S., Greco, K., Sobolevsky , S

    Kung, K. S., Greco, K., Sobolevsky , S. & Ratti, C. Exploring universal patterns in human home-work commuting from mobile phone data.PloS one9, e96180 (2014)

    2014

  4. [4]

    & Jing, Y

    Liu, Y., Fang, F . & Jing, Y. How urban land use influences commuting flows in wuhan, central china: A mobile phone signaling data perspective.Sustain. Cities Soc.53, 101914 (2020)

    2020

  5. [5]

    & Neville, R

    Rowe, F ., Cabrera-Arnau, C., González-Leonardo, M., Nasuto, A. & Neville, R. Medium-term changes in the patterns of internal population movements in Latin American countries: effects of the COVID-19 pandemic. InPopulation and Development series, No. 139 (LC/TS.2024/71) (Economic Commission for Latin America and the Caribbean (ECLAC), Santiago, 2024)

    2024

  6. [6]

    & Blanchard, P

    Gubert, F . & Blanchard, P . Temporary migration response to rainy season conditions in senegal: New evidence using mobile phone data (2024)

    2024

  7. [7]

    & Rubrichi, S

    Blanchard, P . & Rubrichi, S. A highly granular temporary migration dataset derived from mobile phone data in senegal. Sci. Data12, 1051 (2025)

    2025

  8. [8]

    Using digital footprint data to monitor human mobility and support rapid humanitarian responses.Reg

    Rowe, F . Using digital footprint data to monitor human mobility and support rapid humanitarian responses.Reg. Studies, Reg. Sci.9, 665–668 (2022)

    2022

  9. [9]

    & Pietrostefani, E

    Iradukunda, R., Rowe, F . & Pietrostefani, E. Estimating internal displacement in ukraine from high-frequency gps mobile phone data.Humanit. Soc. Sci. Commun.12, 1863 (2025). 12/15

    2025

  10. [10]

    M., Spinelli, S

    Rotondi, V ., Kashyap, R., Pesando, L. M., Spinelli, S. & Billari, F . C. Leveraging mobile phones to attain sustainable development.Proc. Natl. Acad. Sci.117, 13413–13420 (2020). 11.Rowe, F .et al.Methodological guide on the use of mobile phone data: Migration statistics introduction (2022)

    2020

  11. [11]

    & Watts, D

    Barreras, F . & Watts, D. The exciting potential and daunting challenge of using gps human-mobility data for epidemic modeling.Nat. Comput. Sci.4, 398–411 (2024). 13.Rowe, F .9.: Big data, 42 – 47 (Edward Elgar Publishing, Cheltenham, UK, 2023)

    2024

  12. [12]

    M., Snow, R

    Wesolowski, A., Eagle, N., Noor, A. M., Snow, R. W . & Buckee, C. O. The impact of biases in mobile phone ownership on estimates of human mobility .J. The Royal Soc. Interface10, 20120986, 10.1098/rsif.2012.0986 (2013). https://royalsocietypublishing.org/doi/pdf/10.1098/rsif.2012.0986

    work page doi:10.1098/rsif.2012.0986 2012

  13. [13]

    & Rowe, F

    Cabrera, C. & Rowe, F . A systematic machine learning approach to measure and assess biases in mobile phone population data.arXiv preprint arXiv:2509.02603(2025)

    arXiv 2025

  14. [14]

    & Pham, T

    Suich, H., Yap, M. & Pham, T . Coverage bias: the impact of eligibility constraints on mobile phone-based sampling and data collection.Int. J. Soc. Res. Methodol.25, 849–860 (2022)

    2022

  15. [15]

    O., Engø-Monsen, K

    Wesolowski, A., Buckee, C. O., Engø-Monsen, K. & Metcalf, C. J. E. Connecting mobility to infectious diseases: the promise and limits of mobile phone data.The J. infectious diseases214, S414–S420 (2016)

    2016

  16. [16]

    Lai, S.et al.Exploring the use of mobile phone data for national migration statistics.Palgrave communications5, 1–10 (2019)

    2019

  17. [17]

    & De Domenico, M

    Gallotti, R., Maniscalco, D., Barthelemy , M. & De Domenico, M. Distorted insights from human mobility data.Commun. Phys.7, 421 (2024)

    2024

  18. [18]

    N., Benevenuto, F

    Ribeiro, F . N., Benevenuto, F . & Zagheni, E. How biased is the population of facebook users? comparing the demographics of facebook users with census data to generate correction factors. InProceedings of the 12th ACM conference on web science, 325–334 (2020)

    2020

  19. [19]

    & Zagheni, E

    Gil-Clavel, S. & Zagheni, E. Demographic differentials in facebook usage around the world. InProceedings of the International AAAI Conference on Web and Social Media, vol. 13, 647–650 (2019)

    2019

  20. [20]

    Zhao, Z.et al.Understanding the bias of call detail records in human mobility research.Int. J. Geogr. Inf. Sci.30, 1738–1762 (2016)

    2016

  21. [21]

    & Viotti, L

    Milusheva, S., Bjorkegren, D. & Viotti, L. Assessing bias in smartphone mobility estimates in low income countries. In Proceedings of the 4th ACM SIGCAS Conference on Computing and Sustainable Societies, 364–378 (2021)

    2021

  22. [22]

    & Lessani, M

    Li, Z., Ning, H., Jing, F . & Lessani, M. N. Understanding the bias of mobile location data across spatial scales and over time: a comprehensive analysis of safegraph data in the united states.Plos one19, e0294430 (2024)

    2024

  23. [23]

    A., Gibbs, H., Yabe, T ., O’Brien, D

    Sanchez, S. A., Gibbs, H., Yabe, T ., O’Brien, D. T . & Moro, E. Correcting temporal bias in mobility data using time-use surveys.arXiv preprint arXiv:2601.22330(2026)

    arXiv 2026

  24. [25]

    Ecography , volume =

    Roberts, D. R.et al.Cross-validation strategies for data with temporal, spatial, hierarchical, or phylogenetic structure. Ecography40, 913–929, https://doi.org/10.1111/ecog.02881 (2017). https://nsojournals.onlinelibrary .wiley .com/ doi/pdf/10.1111/ecog.02881

    work page doi:10.1111/ecog.02881 2017

  25. [26]

    & Fortin, M

    Legendre, P . & Fortin, M. J. Spatial pattern and ecological analysis.Vegetatio80, 107–138, 10.1007/BF00048036 (1989)

    work page doi:10.1007/bf00048036 1989

  26. [27]

    Spinons and triplons in spatially anisotropic frustrated antiferromagnets

    Bahn, V . & McGill, B. J. Testing the predictive performance of distribution models.Oikos122, 321–331, https: //doi.org/10.1111/j.1600-0706.2012.00299.x (2013). https://nsojournals.onlinelibrary.wiley.com/doi/pdf/10. 1111/j.1600-0706.2012.00299.x

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1111/j.1600-0706.2012.00299.x 2012

  27. [28]

    J.et al.Assessing the performance of spatial cross-validation approaches for models of spatially structured data (2023)

    Mahoney, M. J.et al.Assessing the performance of spatial cross-validation approaches for models of spatially structured data (2023). 2303.07334

    arXiv 2023

  28. [29]

    & Sekara, V

    den Nijs, K., Omodei, E. & Sekara, V . Data bias in human mobility is a universal phenomenon but is highly location- specific (2025). 2508.00149. 13/15

    arXiv 2025

  29. [30]

    Sinclair, M.et al.Assessing the socio-demographic representativeness of mobile phone application data.Appl. Geogr. 158, 102997, https://doi.org/10.1016/j.apgeog.2023.102997 (2023)

    work page doi:10.1016/j.apgeog.2023.102997 2023

  30. [31]

    & Zangerling, B.Mexico urbanization review: Managing spatial growth for productive and livable cities in Mexico (World Bank Publications, 2016)

    Kim, Y. & Zangerling, B.Mexico urbanization review: Managing spatial growth for productive and livable cities in Mexico (World Bank Publications, 2016)

    2016

  31. [32]

    & Mora-Rivera, J

    Garcia-Mora, F . & Mora-Rivera, J. Exploring the impacts of internet access on poverty: A regional analysis of rural mexico.New media & society25, 26–49 (2023). 36.NapoleonCat. Mexico social media statistics (2026). Accessed January 2026

    2023

  32. [33]

    & Holasut, K

    Sitthiyot, T . & Holasut, K. A simple method for estimating the lorenz curve.Humanit. Soc. Sci. Commun.8, 268, 10.1057/s41599-021-00948-x (2021). 38.Rey , S. J., Arribas-Bel, D. & Wolf, L. J.Geographic Data Science with Python(Chapman and Hall/CRC, 2023), 1 edn

    work page doi:10.1057/s41599-021-00948-x 2021

  33. [34]

    & Messina, J

    Busso, M. & Messina, J. The inequality crisis: Latin america and the caribbean at the crossroads, 10.18235/0002629 (2020)

    work page doi:10.18235/0002629 2020

  34. [35]

    & Mora-Rivera, J

    Martínez-Domínguez, M. & Mora-Rivera, J. Internet adoption and usage patterns in rural mexico.T echnol. society60, 101226 (2020)

    2020

  35. [36]

    & Martinez Dominguez, M

    Díaz de León Castañeda, C. & Martinez Dominguez, M. Factors related to internet adoption and its use to seek health information in mexico.Heal. Commun.36, 1768–1775 (2021)

    2021

  36. [37]

    Nonequilibrium work performed in quantum annealing

    García-Mora, F ., Cavazos-Castro, M. & Mora-Rivera, J. Does mobile internet reduce social inequality? evidence from mexican villages.Inf. T echnol. for Dev.31, 1399–1424, 10.1080/02681102.2025.2488325 (2025). https: //doi.org/10.1080/02681102.2025.2488325

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1080/02681102.2025.2488325 2025

  37. [38]

    & Cabrera, C

    Renninger, A. & Cabrera, C. Extreme heat reduces and reshapes urban mobility .PNAS Nexus5, pgag078, 10.1093/ pnasnexus/pgag078 (2026). https://academic.oup.com/pnasnexus/article-pdf/5/4/pgag078/68031014/pgag078. pdf

    2026

  38. [39]

    & Cabrera-Arnau, C

    Wang, Y., Zhong, C., Gao, Q. & Cabrera-Arnau, C. Understanding internal migration in the UK before and during the COVID-19 pandemic using twitter data.Urban Informatics1, 10.1007/s44212-022-00018-w (2022)

    work page doi:10.1007/s44212-022-00018-w 2022

  39. [40]

    Mobility Patterns Are Associated with Experienced Income Segregation in Large

    Moro, E., Calacci, D., Dong, X. & Pentland, A. Mobility patterns are associated with experienced income segregation in large us cities.Nat. Commun.12, 10.1038/s41467-021-24899-8 (2021)

    work page doi:10.1038/s41467-021-24899-8 2021

  40. [41]

    & Rowe, F

    Cabrera, C., González-Leonardo, M., Nasuto, A., Neville, R. & Rowe, F . Sustained changes to urban mobility after covid-19 amplified socio-economic inequalities in latin america (2025). 2504.15871

    arXiv 2025

  41. [42]

    N., Benevenuto, F

    Ribeiro, F . N., Benevenuto, F . & Zagheni, E. How biased is the population of facebook users? comparing the demographics of facebook users with census data to generate correction factors, 10.1145/3394231.3397923 (2020). Proceedings of the 12th ACM Conference on Web Science, Southampton, United Kingdom, pp. 325–334. Association for Computing Machinery , N...

    work page doi:10.1145/3394231.3397923 2020

  42. [43]

    & Zagheni, E

    Gil-Clavel, S. & Zagheni, E. Demographic differentials in facebook usage around the world.Proc. Int. AAAI Conf. on Web Soc. Media13, 647–650, 10.1609/icwsm.v13i01.3263 (2019)

    work page doi:10.1609/icwsm.v13i01.3263 2019

  43. [44]

    F .et al.Mapping subnational gender gaps in internet and mobile adoption using social media data.Proc

    Breen, C. F .et al.Mapping subnational gender gaps in internet and mobile adoption using social media data.Proc. Natl. Acad. Sci.122, e2416624122, 10.1073/pnas.2416624122 (2025). https://www.pnas.org/doi/pdf/10.1073/ pnas.2416624122

    work page doi:10.1073/pnas.2416624122 2025

  44. [45]

    & Rowe, F

    González-Leonardo, M., Cabrera, C., Neville, R., Nasuto, A. & Rowe, F . Beyond the immediate impacts of covid-19 on internal population movements in mexico: Facebook data reveal urban decay and slow recovery—a research note. Demography62, 1171–1184, 10.1215/00703370-12183205 (2025). https://read.dukeupress.edu/demography/ article-pdf/62/4/1171/2307728/117...

    work page doi:10.1215/00703370-12183205 2025

  45. [46]

    PLoS ONE9(1), e86041 (2014-01-27).https://doi.org/10.1371/journal.pone

    Li, Z., Ning, H., Jing, F . & Lessani, M. N. Understanding the bias of mobile location data across spatial scales and over time: A comprehensive analysis of safegraph data in the united states.PLOS ONE19, 1–23, 10.1371/journal.pone. 0294430 (2024)

    work page doi:10.1371/journal.pone 2024

  46. [47]

    Duan, Q.et al.Identifying counter-urbanisation using facebook’s user count data.Habitat Int.150, 103113, https://doi.org/10.1016/j.habitatint.2024.103113 (2024)

    work page doi:10.1016/j.habitatint.2024.103113 2024

  47. [48]

    2601.22104

    Andrich, P .et al.Social media data for population mapping: A bayesian approach to address representativeness and privacy challenges (2026). 2601.22104. 14/15

    arXiv 2026

  48. [49]

    Pelletier, F .et al.Census counts, undercounts and population estimates: The importance of data quality evaluation. T ech. Pap.2(2020)

    2020

  49. [50]

    Census in crisis—further erasure of indigenous peoples?Science389, eaea0932, 10.1126/science.aea0932 (2025)

    Kukutai, T . Census in crisis—further erasure of indigenous peoples?Science389, eaea0932, 10.1126/science.aea0932 (2025). https://www.science.org/doi/pdf/10.1126/science.aea0932

    work page doi:10.1126/science.aea0932 2025

  50. [51]

    Global Demographic Data Project, funded by the Bill & Melinda Gates Foundation (INV-045237)

    Bondarenko, M.et al.The spatial distribution of population in 2015–2030 at a resolution of 30 arc (approximately 1 km at the equator): R2025a version v1, 10.5258/SOTON/WP00845 (2025). Global Demographic Data Project, funded by the Bill & Melinda Gates Foundation (INV-045237)

    work page doi:10.5258/soton/wp00845 2015

  51. [52]

    & MacManus, K

    Schiavina, M., Freire, S., Carioli, A. & MacManus, K. Ghs-pop r2023a: Ghs population grid multitemporal (1975–2030), 10.2905/2FF68A52-5B5B-4A22-8F40-C41DA8332CFE (2023). 58.Maas, P .et al.Facebook disaster maps: Aggregate insights for crisis response and recovery (2019)

    work page doi:10.2905/2ff68a52-5b5b-4a22-8f40-c41da8332cfe 1975

  52. [53]

    Bing Maps Tile System - Bing Maps

    Microsoft. Bing Maps Tile System - Bing Maps. https://learn.microsoft.com/en-us/bingmaps/articles/ bing-maps-tile-system.[Accessed 07-01-2026]

    2026

  53. [54]

    & Ukkusuri, S

    Verma, R., Mittal, S., Lei, Z., Chen, X. & Ukkusuri, S. V . Comparison of home detection algorithms using smartphone gps data.EPJ Data Sci.13, 6, 10.1140/epjds/s13688-023-00447-w (2024)

    work page doi:10.1140/epjds/s13688-023-00447-w 2024

  54. [55]

    S., Wang, Y

    Zhong, C., Zhou, Z., Aslam, N. S., Wang, Y. & Enaya, A. Anonymised human location data in england for urban mobility research.Sci. Data12, 2040, 10.1038/s41597-025-06323-8 (2025)

    work page doi:10.1038/s41597-025-06323-8 2040

  55. [56]

    & Eagle, N

    Blumenstock, J. & Eagle, N. Mobile divides.Proc. 4th ACM/IEEE Int. Conf. on Inf. Commun. T echnol. Dev.1–10, 10.1145/2369220.2369225 (2010)

    work page doi:10.1145/2369220.2369225 2010

  56. [57]

    & Garcia-Herranz, M

    Schlosser, F ., Sekara, V ., Brockmann, D. & Garcia-Herranz, M. Biases in human mobility data impact epidemic modeling (2021). 2112.12521

    arXiv 2021

  57. [58]

    Representativeness in GPS-based mobility data is source-specific and geographically uneven

    Valavi, R., Elith, J., Lahoz-Monfort, J. J. & Guillera-Arroita, G. blockcv: An r package for generating spatially or environmentally separated folds for k-fold cross-validation of species distribution models.Methods Ecol. Evol.10, 225–232, https://doi.org/10.1111/2041-210X.13107 (2019). https://besjournals.onlinelibrary .wiley .com/doi/pdf/ 10.1111/2041-2...

    work page doi:10.1111/2041-210x.13107 2041