London · Census 2021 · study published 27 August 2026
The same access points. The same 300 m threshold. Two results.
For the same official mapped access points in London, the result is 87.61% by straight line and 50.66% on the walking network.
straight line
walking network
difference36.95 pp
3,251,463 of the 8,799,776 represented residents enter the aggregate “outside the threshold” set. The calculation uses 4,994 population-weighted centroids.
Check the primary study. The plan below explains the method; it does not plot a real route.
What stays fixed
The parks do not change. The ruler does.
The main comparison holds the represented population, official pedestrian-capable access points and 300 m threshold constant. Only the distance definition changes.
Straight-line proximity
Measures the geometric distance from a centroid to the nearest access point. It does not check whether the street network offers an equally short walk.
Walking-network distance
Connects the centroid and access point to network nodes and totals the walking route. Street connections and detours enter the measurement.
The study publishes a third method too: straight-line distance to the green-space polygon. That answers a different question because a polygon edge is not necessarily an entrance.
The nine published values
Choose a method and threshold, not an invented point between them.
The control switches only among cells in the supplementary package. It does not interpolate, calculate routes or classify a neighbourhood.
| Method | 300 m | 500 m | 1,000 m |
|---|---|---|---|
| Site polygon, straight line | |||
| Access point, straight line | |||
| Access point, walking network |
Matrix source: software package v2.1, London method-threshold and entrance-penalty files.
Deterministic recomputation
The people count comes from two totals, not the rounded percentage.
The reclassification table separately publishes the population outside the threshold under each method. Subtracting those counts gives the study's exact displayed value.
Walking network, outside threshold4,341,752
Straight line, outside threshold− 1,090,289
Enters the aggregate “outside threshold” set= 3,251,463
Share check: 3,251,463 ÷ 8,799,776 × 100 = 36.949383…%, displayed as 36.95%.
Not the reverse: 8,799,776 × 36.95% = 3,251,517, which is 54 higher. A percentage shown to two decimal places cannot reconstruct the exact count.
The same network model50.66% → 99.48%from 300 to 1,000 m: +48.82 percentage points
What the population represents
4,994 statistical points stand in for 8,799,776 people.
The authors use population-weighted centroids for London's 4,994 small statistical areas. Each area's Census population is assigned to its representative point.
Consequence: if the point crosses the threshold, the area's entire population receives the same aggregate status. The page does not know each resident's route and does not show variation within the area.
A small aggregation difference: the analysis uses 8,799,776 residents, 48 fewer than the regional Census 2021 total published by ONS. This difference is compatible with ONS's “cell key” statistical protection, which can perturb published cells slightly. The comparison on this page remains unchanged.
Before interpretation
The indicator measures a cartographic definition of access.
- Mapped access
- Ordnance Survey warns that including an access point in the product does not guarantee public access. Gates, opening hours and restrictions may change real conditions.
- Representative origin
- A population-weighted centroid represents an LSOA. It is not a person's address and does not describe an individual route.
- Mutable sources
- OS Open Greenspace and OpenStreetMap are updated. These values freeze the published analysis, not the current state of every street or entrance.
- Experience not measured
- Safety, usability for different mobilities, ecological quality, crowding and realised use do not enter the percentage.
- No causal conclusion
- The comparison is descriptive. It does not say that the method causes access or that changing a threshold would produce a social outcome.
- No uncertainty interval
- Contrasts between methods and thresholds are not confidence intervals. The study treats them as sensitivity to definition.
Deliberately absent: no neighbourhood map, no LSOA label, no invented route and no ranking of policies.
How the comparison was built
From the Census to the nine cells.
- 01Demand
Census 2021 TS001 population is represented at population-weighted centroids for LSOAs in the ONS London region.
- 02Destinations
The authors clip OS Open Greenspace polygons and official pedestrian-capable mapped access points to the region boundary.
- 03Distances
Geometric proximity is compared with distance on OpenStreetMap walking graphs calculated in five buffered borough or local-authority tiles.
- 04Classification
Each centroid is checked at 300, 500 and 1,000 m. The area's population follows its status, then the totals are weighted.
Evidence trail
Article, tables, definitions.
- Primary studyZhou and Geng, “Auditing urban nature access claims for sustainable city governance”npj Urban Sustainability, Article in Press, 27 August 2026, CC BY 4.0
- Supplementary tablesSupplementary informationTables 8, 9, 14 and 16
- Data and codeSoftware package v2.1Derived CSVs and a manuscript-output script; it does not rebuild the raw GIS workflow
- Entrance definitionOS Open Greenspace: data creationThe official public-access warning
- Population and pointsONS LSOA Population Weighted Centroids · Census 2021 TS001 v3ONS also explains the cell-key perturbation protection here
- Normalised evidenceDownload the JSON used by local checksNine cells, counts, formulas and archive hashes
The local archive preserves the article, supplement, software package and OS/ONS snapshots with SHA-256 hashes. Third-party files are not served as public assets. This edition attributes the facts and uses original graphics.