Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 53 of the 54 references it cites were found in a registry or an index; none of the 42 sources its claims cite is checked for support yet.

The world’s user-generated road map is more than 80% complete

Open access
Barrington-Leigh, Christopher, Millard-Ball, Adam · 2017 · PLOS ONE

Abstract

OpenStreetMap, a crowdsourced geographic database, provides the only global-level, openly licensed source of geospatial road data, and the only national-level source in many countries. However, researchers, policy makers, and citizens who want to make use of OpenStreetMap (OSM) have little information about whether it can be relied upon in a particular geographic setting. In this paper, we use two complementary, independent methods to assess the completeness of OSM road data in each country in the world. First, we undertake a visual assessment of OSM data against satellite imagery, which provides the input for estimates based on a multilevel regression and poststratification model. Second, we fit sigmoid curves to the cumulative length of contributions, and use them to estimate the saturation level for each country. Both techniques may have more general use for assessing the development and saturation of crowd-sourced data. Our results show that in many places, researchers and policymakers can rely on the completeness of OSM, or will soon be able to do so. We find (i) that globally, OSM is ∼83% complete, and more than 40% of countries-including several in the developing world-have a fully mapped street network; (ii) that well-governed countries with good Internet access tend to be more complete, and that completeness has a U-shaped relationship with population density-both sparsely populated areas and dense cities are the best mapped; and (iii) that existing global datasets used by the World Bank undercount roads by more than 30%.

Overview

CoChat Summary

TL;DR

Using visual comparison with satellite imagery and sigmoid saturation fits to OpenStreetMap contribution history, the authors estimate that the global OSM road network was ∼83% complete in January 2016 (95% CI 81–84%). They also estimate a global road stock of ∼39.7×10⁶ km and find that IRF/World Bank road-length data substantially undercount roads.

Key findings

  • Globally, OSM roads were ∼83% complete in January 2016 (95% CI 81–84%) by the multilevel model; country-level parametric fits gave 87% when weighted by estimated road length, while a global sigmoid fit suggested 97% but was not preferred because recent global growth appeared linear. About 42% of countries were more than 95% complete; of 185 countries with both estimates, 77 were more than 95% complete, and 47 had the highest confidence because multilevel, country-level, and sub-national fits all indicated >95%.
  • Completeness was higher in well-governed countries with good Internet access and in small countries; it had a ∪-shaped relationship with population density, with both sparsely populated areas and dense cities best mapped, while smaller towns and villages were most likely to have missing streets. GDP had no clear independent effect except at the lowest densities. Countries such as Kiribati, Afghanistan, Egypt, and China were less than one-third complete, while Cuba, Ecuador, Syria, and most European and North American countries were more than 95% complete.
  • The estimated global road stock was ∼39.7×10⁶ km, or ∼5.6 m per person, with the United States accounting for ∼23.0% of the world’s roads. OSM-based estimates exceeded IRF estimates in the majority of countries; globally they were 134% of the IRF estimate, and in 94 of 190 countries they were more than 150% of IRF. In the poorest countries, road supply was nearly 40% larger than suggested by IRF, and existing global datasets used by the World Bank undercount roads by more than 30%.

Methods

  • Visual assessment: a stratified, probability-weighted sample of 45 points per country—25 points selected proportional to log population density and 20 points restricted to densely populated areas above a country-specific density threshold—yielded 8,370 observations. OSM street edges were overlaid on Google/Bing aerial imagery at 1:5000, missing and present edges were counted, and February 2015 estimates were updated to January 2016. A multilevel regression and poststratification model used local population density and country-level GDP per capita, Internet penetration, population size, and World Bank “voice and accountability,” estimated in PyStan.
  • Saturation method: using OSM full-history data, the authors extracted every version of every node and way tagged “highway,” calculated road lengths, and built cumulative road-length time series by country. They fit flexible sigmoid functional forms with up to four jumps, plus a linear growth model, using nonlinear least squares and selecting the best form by mean-squared error; they also fit sub-national administrative units and country-specific density quintiles.
  • Combining estimates and road length: where the parametric estimate fell within the 95% confidence interval of the multilevel estimate or differed by ≤0.05, the parametric fit was used (61 countries); otherwise the multilevel estimate was used (124 countries). The parametric fit was also used for 68 countries without a multilevel estimate. Total road length was estimated as OSM 2016 road length divided by estimated fraction complete, excluding countries with completeness <0.05 (five small-island states).

Limitations

  • Stated: OSM contributions are not deterministic, so the parametric functional forms are heuristic and have limited predictive power for future contributions in individual countries, which may occur in jumps. The authors also assume urban growth is slow and that additions represent previously missing streets, whereas the street network is physically growing. The relatively small sample size per country means visual-assessment confidence intervals are often wide, though this is partly mitigated by using two independent methods.
  • Stated: the completeness measure considers only the presence of geographic features, not attribute information such as street names, nor other quality measures such as positional accuracy. Country-level factors have only limited predictive power, and wide confidence intervals plus endogeneity concerns mean the results should not necessarily be given a causal interpretation.
  • Evident from the text: possible double-counting of dual carriageways in OSM is noted, though the authors say it is unlikely to account for more than a small part of the difference (one-way streets were only 2.0% of ways as of May 2017). The visual assessment relies partly on Google Maps imagery even though the authors note Google Maps data themselves are not complete in many parts of the world, and manual counting involved some judgment.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1371/journal.pone.0180698

Cites

53 of 53 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig 1 Example visual assessment. Street data from OSM is overlaid on satellite imagery of Kuwait City, Kuwait. Here, the network is 99% complete, with 2 out of 300 edges missing. The red lines indicate street edges in the OSM database. The green lines (highlighted with a white oval) are missing edges. While Google imagery was used in our actual inspection procedure of streets, lower-resolution public-domain imagery is shown in the figure. A version of this figure with the original imagery is available from the authors. Satellite imagery source: Landsat. Road network data source: OpenStreetMap.
Fig 1 Example visual assessment. Street data from OSM is overlaid on satellite imagery of Kuwait City, Kuwait. Here, the network is 99% complete, with 2 out of 300 edges missing. The red lines indicate street edges in the OSM database. The green lines (highlighted with a white oval) are missing edges. While Google imagery was used in our actual inspection procedure of streets, lower-resolution public-domain imagery is shown in the figure. A version of this figure with the original imagery is available from the authors. Satellite imagery source: Landsat. Road network data source: OpenStreetMap.CC BY · PubMed Central
Fig 2 Predicted partial effects in multilevel model. The red line (intercept-only model) shows the baseline predictions, across the density spectrum, when all country-level variables are at their means. Each of the other lines shows the predicted fraction complete after a one standard deviation increase in one country-level predictor. 95% credible intervals are shaded. The thin grey line shows the cumulative distribution of grid-cell level densities in the world.
Fig 2 Predicted partial effects in multilevel model. The red line (intercept-only model) shows the baseline predictions, across the density spectrum, when all country-level variables are at their means. Each of the other lines shows the predicted fraction complete after a one standard deviation increase in one country-level predictor. 95% credible intervals are shaded. The thin grey line shows the cumulative distribution of grid-cell level densities in the world.CC BY · PubMed Central
Fig 3 Visual assessments of completeness: Observations vs multilevel model. The two sets of estimates correlate well at the country level, with no evidence of bias, adding confidence to our model predictions. The multilevel estimates are obtained from poststratification using out-of-sample predictions for each grid cell in a country. The red line indicates equality (i.e., the 45° line).
Fig 3 Visual assessments of completeness: Observations vs multilevel model. The two sets of estimates correlate well at the country level, with no evidence of bias, adding confidence to our model predictions. The multilevel estimates are obtained from poststratification using out-of-sample predictions for each grid cell in a country. The red line indicates equality (i.e., the 45° line).CC BY · PubMed Central
Fig 4 Growth in OSM dataset: Parametric fits and visual assessment. The largest ten countries by road length are shown, along with the global data. The S1 Appendix provides similar plots for all countries. The thick red line shows the actual data for roads, along with the predicted values, asymptote and visual assessment. The thin red line shows that other paths, which are mainly pedestrian routes, continue to grow in some countries even where the road network is complete. The decline in the US is mainly due to the bulk import of TIGER data, which has subsequently been cleaned (e.g. forest tracks are retagged as tracks rather than roads). Years shown indicate January 1.
Fig 4 Growth in OSM dataset: Parametric fits and visual assessment. The largest ten countries by road length are shown, along with the global data. The S1 Appendix provides similar plots for all countries. The thick red line shows the actual data for roads, along with the predicted values, asymptote and visual assessment. The thin red line shows that other paths, which are mainly pedestrian routes, continue to grow in some countries even where the road network is complete. The decline in the US is mainly due to the bulk import of TIGER data, which has subsequently been cleaned (e.g. forest tracks are retagged as tracks rather than roads). Years shown indicate January 1.CC BY · PubMed Central
Fig 5 Completeness of the OSM dataset, by country, January 2016. The fraction complete is estimated by the parametric model, where that estimate falls within five percentage points or the 95% confidence interval of the multilevel model. Otherwise, the multilevel model is used.
Fig 5 Completeness of the OSM dataset, by country, January 2016. The fraction complete is estimated by the parametric model, where that estimate falls within five percentage points or the 95% confidence interval of the multilevel model. Otherwise, the multilevel model is used.CC BY · PubMed Central
Fig 6 Completeness of the OSM dataset, by grid cell, January 2016. The fraction complete is estimated by the multilevel model. The color intensity represents the number of estimated street edges, thus highlighting parts of the world with a denser street network. The full-resolution image is available online.
Fig 6 Completeness of the OSM dataset, by grid cell, January 2016. The fraction complete is estimated by the multilevel model. The color intensity represents the number of estimated street edges, thus highlighting parts of the world with a denser street network. The full-resolution image is available online.CC BY · PubMed Central
Fig 7 Comparison of methods. The largest ten countries by road length are shown, and a similar plot for all countries is provided in the S1 Appendix. The bars indicate the bootstrapped and multilevel model estimates from the visual assessment. The green makers indicate the estimates from the parametric fits at the country level, and by subnational density quintile and subnational administrative geography.
Fig 7 Comparison of methods. The largest ten countries by road length are shown, and a similar plot for all countries is provided in the S1 Appendix. The bars indicate the bootstrapped and multilevel model estimates from the visual assessment. The green makers indicate the estimates from the parametric fits at the country level, and by subnational density quintile and subnational administrative geography.CC BY · PubMed Central
Fig 8 Road length per capita.
Fig 8 Road length per capita.CC BY · PubMed Central
Fig 9 Road length data from the International Road Federation are substantially lower than OSM-based estimates. The red line indicates equality (i.e., the 45° line).
Fig 9 Road length data from the International Road Federation are substantially lower than OSM-based estimates. The red line indicates equality (i.e., the 45° line).CC BY · PubMed Central

Verification

Claims the paper makes

0 Valid · 1 Not valid · 25 Not checked · 89 Own results and interpretations

43 more claims

Create a free account to see all 115 claims and to verify the 41 references still open.

See all claims — free account
  • It may state two separate claims as one. 25
  • Probably not a substantial claim. 52

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Show all 54 references

Research works printed but not linked: 1

  1. 2002, A World Bank Urban Transport Strategy Review

Cited by

Items 501–557 of 557

Metadata

Authors
Barrington-Leigh, C., & Millard-Ball, A.
Affiliations
McGill University (CA)
Corresponding author
Barrington-Leigh, Christopher
Venue
PLOS ONE
Published
2017-08-10
Volume
12
Issue
8
Pages
e0180698
ISSN
1932-6203 (ISSN-L)
OpenAlex
W2744637967
Cited by
553
References
54
Citations by year
2026: 117 · 2025: 74 · 2024: 66 · 2023: 49 · 2022: 71 · 2021: 61 · 2020: 49 · 2019: 46 · 2018: 19 · 2017: 1
FWCI
94.56
Type
journal article
Language
en
Open access
Locations
  • PLoS ONE (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • PubMed Central (publishedVersion, cc-by)
  • eScholarship@McGill (McGill) (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • eScholarship@McGill (McGill) (submittedVersion, cc-by)
Funders
  • Social Sciences and Humanities Research Council of Canada (435-2016-0531)
  • Hellman Fellows Program
  • UC Santa Cruz Committee on Research
Topics
  • Social Sciences › Geography, Planning and Development › Geographic Information Systems Studies
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Automated Road and Building Extraction
  • Physical Sciences › Computer Science › Computer Science Applications › Mobile Crowdsensing and Crowdsourcing
Keywords
OpenStreetMap, crowdsourced geospatial data, roadmapping, data saturation, population density, internet access, satellite imagery validation, developing countries
MeSH terms
Transportation Facilities, Cities, Humans, Maps as Topic, Databases, Factual, Motor Vehicles, Geographic Information Systems, Geographic Mapping, Satellite Imagery

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-11-19, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-19 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.