How far from a railway station?
For every census since 1831: the distance from where people lived that year to the nearest railway station then open, from the median person to the remotest one in a hundred, computed over 9,175 public passenger stations in Great Britain, living and closed, open on the census night, against a population surface built for each census year from its own returns. The shaded ribbons are the modelled error band of each line, a ceiling; the numbers behind them, and how they were measured, are in the validation notes below.
How sure are we?
Two reconstructions meet here and each carries its own doubt. The station list: positions from OpenStreetMap and Wikidata, dates mostly from Quick's Chronology and station articles, some closure years inherited from the line the station stood on (flagged per row in the download). It is known to lack stations: 220 closed stations named in Wikipedia's county lists have no position anywhere, and those lists cover only some counties, so every historical line here is too long by an amount that is not fully known. Dropping the same share of known closed stations at random moves the p90 line by up to 0.03 mi in the worst census year; that is a lower bound on the effect, not its size. The population surface: each census year's people placed on parishes, registration sub-districts or districts as the records allow. The ribbons are the modelled placement band, calibrated on 1971 enumeration districts and applied to every modelled year; where a tier can be checked directly with 1841 parish points (1851, 1861, 1921, 1931) the check comes out well inside the ribbon, so the ribbon is a ceiling and the check a floor; the years with no check have only the ceiling, and the direction of its error there is not known. The ribbons do not include the missing-station effect. The modern end has an external check of its own: today's network against the DfT NaPTAN station list (2,639 stops, no station of ours involved) and today's population, computed at every publish, gives a median of 1.05 mi and p90 of 4.05 mi. This table's own present-day series lands at 1.00 and 4.03, and the 2021 census row at 1.01 and 4.07. Distances are straight-line, not travel distance.
Download
- gb-stations.csv: every station with coordinates, opened, closed and sources
- station-distance-percentiles.csv: the present-day check series, every year against today's population, in miles and km
- README.md: schema, method, limitations
- METHODS.md: how the datasets were made and what to doubt: who did the work, every hand decision counted, thresholds measured and chosen, watched and targeted checks, input pins; with the pipeline source and the input pins (stations, population, routes)
The population behind the series, as points:
- Every census year's population-weighted points for Great Britain, gzipped CSV in four era files: early (1801 to 1841), victorian (1851 to 1911), interwar (1921 to 1961), modern (1971 to 2021); the same columns in each
- distance-percentiles.csv: the chart's series by census year, with the error band per percentile, the direct 1841 check and the missing-station sensitivity
- VALIDATION.md: totals against the published censuses, coverage and measured error per year
- README.md: columns, method, licences; SHA256SUMS
- The transcriptions of the printed census volumes, row by row with their proof flags, so the figures can be checked against the page images: 1831 Comparative Account, 1841 Enumeration Abstract, 1921 and 1931 County Reports Table 3, Scotland 1921 to 1971 (NRS Table 8)
The same points as four era GeoPackages (about 740 MB) are kept in the project archive rather than served here; ask if you want them.
Licence: ODbL (positions from OpenStreetMap); dates CC0/CC BY-SA per source. Attribute "Mind the Gates railway atlas / © OpenStreetMap contributors". Corrections welcome: get in touch, ideally with a source.