← Back to the map

Brisbane public transport coverage by suburb

Generated 2026-08-19T03:01:03+00:00 from the TransLink SEQ GTFS feed valid 20260818-20261017, BCC Suburb Boundaries, and ABS Census 2021 (mesh block counts and boundaries, with GCP G01 at SAL level for reconciliation).

CANONICAL NUMBERS

Every publishable figure from this analysis, one line each. These supersede all earlier versions of this document; if a number appears elsewhere in the piece and disagrees with this block, this block wins.

Scope

190 suburbs analysed, 1,242,389 residents (mesh-block basis). BCC's Suburb Boundaries dataset holds 195 localities; the five its own metadata names as lying within Moreton Bay (Bulwer, Cowan Cowan, Kooringal, Moreton Bay, Moreton Island) are excluded, so this is a mainland analysis, not the full Brisbane LGA - those five are islands inside the LGA, not another council's, and dropping them removes roughly 186 km2 of the council's 1,343 km2.

Headline: per-resident coverage

**476,945 Brisbane residents - 38.4% of the 1,242,389 living in the 190 mainland suburbs - are beyond a 400 m walk of any public transport stop with a scheduled service.** At 800 m that falls to 98,880 (8.0%). Measured mesh block by mesh block: every ABS 2021 mesh block with a resident population is routed over the OSM pedestrian network to the nearest serviced stop in the main GTFS feed, all modes, in EPSG:7856.

The straight-line floor for the same test is 109,732 (8.8%) at 400 m and 17,540 (1.4%) at 800 m. Both are published everywhere: the floor cannot be argued below, the walking figure is the best estimate of what people face.

Coverage by service class

Proximity to a stop is a weak test - a stop with eight morning departures counts the same as a busway platform. Three tighter classes, all measured on the representative term weekday (20260915) from boardable departures in stop_times (set-down-only calls are excluded):

• ANY - at least one scheduled trip. The headline class.
• HOURLY - departures in at least 10 of the 12 clock-hours between 07:00 and 19:00.
• TURN-UP-AND-GO - at least 4 departures in each of at least 10 of those 12 hours.
• SUNDAY - at least one boardable departure on the representative Sunday (20260927).

The distinct-hours test is the point: a stop with a dozen one-way commuter departures crammed into three morning hours fails HOURLY, as it should.

75% of Brisbane residents live beyond a 400 m walk of a turn-up-and-go service, and 619,584 (49.9%) live beyond 400 m of any Sunday service at all - the weekend desert.

How TURN-UP-AND-GO relates to TransLink's own branding

TransLink markets its high-frequency network as 15 minutes or better. 4 departures an hour is a 15-minute average headway, which is the closest test the timetable supports hour by hour - and checking it against the feed shows the two are not identical. Of the 2,519 stops that qualify, 94.7% average a gap of 15 minutes or better across 07:00-19:00, but only 29.6% never exceed 15 minutes; the median worst gap is 18 minutes. So this class is slightly more generous than the branding, and the beyond-turn-up figure is again a floor.

Classifier verification

Both checks run every time and the pipeline aborts if either fails.

• Cultural Centre busway platforms (3 of them) classify as TURN-UP-AND-GO: PASS.
• Sumner's route 452 stops (3 of them, 8 departures spread over only 3 distinct hours) fail HOURLY: PASS.

Why the two bases differ, and why both are still conservative

The floor is straight-line (Euclidean) distance. Real walking routes are longer than the crow flies - rivers, motorways, culs-de-sac, missing footpaths and noise walls all force detours, and Brisbane's river bends make this worse than in a gridded city - so the floor understates the problem by construction: 109,732 against 476,945 once every block is routed.

Three further choices push both figures the same way, toward reporting more coverage than exists: mesh blocks over 0.25 km2 are measured from their polygon edge rather than their centroid, which is the most generous reading available; any stop with a single scheduled trip counts as service; and the 9 suburbs whose OSM network snaps too badly to trust keep their straight-line figures, which reports more coverage than routing them would. Every beyond-the-threshold count here is therefore a floor on its own basis.

How much of this is measurement convention

The headline has moved four times during this analysis, so the question a reader is entitled to ask is not whether the current number is right but how far it could move again. Every measurement convention in the pipeline was varied at once - 81 permutations across four axes - and the whole distribution is reported here rather than a preferred point in it.

A population-weighted origin was not tested, because it is not derivable: the mesh block is the finest population geography ABS publishes, so there is nothing inside a block to weight by.

The three bands

The published figure, 476,945 (38.4%), sits inside all three - at the 68th percentile of the full grid and below the median of the other two. It is not the most alarming reading available; it is close to the middle of the defensible ones.

Two of the settings are bounds, not conventions

The full band only reaches down to 13.1% because two settings do not change how routing works - they switch it off:

• Polygon edge for every block (180,954). It measures each resident from the corner of their own block nearest a stop. The median populated block is 0.029 km2, about 170 m across, so for most of Brisbane this is a systematic best case rather than a choice of centre. It is retained for large blocks only, where a centroid can land in bushland - and that concession is itself measured: it is worth 13,905 residents against a centroid-always rule, which is the only other defensible origin.
• A 50 m snap threshold (257,882). The median block centroid sits 37 m from the nearest network node, which is ordinary suburban geometry, not bad data. A 50 m bar declares roughly 40% of good blocks unplaceable and cuts the routed set from 176 suburbs to 67. What that permutation measures is mostly straight-line distance, which is the floor, reported under another name.

Both are published above because a hostile reader is entitled to see them. Neither is a convention anyone would adopt for a headline.

The one choice that moves the number most

It is the walk threshold - the 400 m bar itself. Holding every other convention at its published setting, moving the threshold by 5% moves the headline by 80,512 residents:

That is 6x the next largest conventional axis (origin, 13,905). Every other convention in this pipeline is worth less than 3% of the headline; the threshold is worth 17%.

The 400 m threshold is not ours. Queensland's Priority Development Area guidelines require 90% of residences in Brisbane to be within 400 m of a public transport stop; RMIT measured against that target and so does this analysis. Holding it fixed is not holding a convenient assumption fixed - it is using the state's own standard. That is why the third band is the one to publish against. A two-in-five framing holds right across it (37.7% to 39.9%). Let the threshold move as well and the band runs 34.4% to 43.0%, and the framing fails at the bottom: at 420 m the honest phrasing is closer to one in three.

Under every measurement convention we tested, holding the 400 m threshold fixed, the figure lands between 468,290 and 495,185 residents (37.7% to 39.9%). We publish 476,945 (38.4%).

Does the unserved list survive the same treatment?

Each of the 10 suburbs at 0% coverage was re-tested in every permutation. A suburb that only reaches 0% under some of them is not a finding, it is a convention.

Ransome is the one entry that moves. It is at 0% under the published convention and in 30 of 36 conventional permutations, but not all of them, and it carries two further independent weaknesses (see the table). It stays on the list - it is 0% on the published basis and on the large majority of alternatives, and dropping it would shrink our own finding in our own favour - but its 268 residents are 3.7% of the list, and it must never be the case cited to illustrate it. The suburbs to name are the ones at 0% in every permutation tested.

8 of the 10 are at 0% in all 81 permutations: Chuwar, Enoggera Reservoir, Karana Downs, Kholo, Larapinta, Mount Crosby, Port Of Brisbane, Upper Brookfield.

Reconciliation against published Brisbane figures

Three independent measurements of Brisbane transport access are in print. Our headline - 476,945 residents beyond a 400 m walk, 38.4% - is far above the number most readers carry around, so each is rebuilt here on its own definition from this pipeline's data. The residual difference is the finding; the arithmetic is in benchmarks.py and runs in the same build.

Podaris Australian Transport Scorecard (14 Nov 2025)

Podaris report 59% of households within 400 m of a stop by isochrone, 78% within 800 m and 83% within 1.2 km, over a "Brisbane" of 2,140 km2 and 2.29 m people. That extent is neither the Brisbane LGA (1,343 km2) nor Greater Brisbane; the post does not name the boundary. Their sample unit is the SA1; ours is the mesh block, roughly a tenth the area. Their unit is households, so the like-for-like row here is dwellings, not residents.

At 400 m the two agree to 5.9 points (64.9% against 59%), and in the expected direction: their extent is nearly twice ours and reaches into sparser fringe, so they should read slightly worse. This is the closest independent check that exists on the headline, and it lands within six points of it - on a walking-network basis, not a straight-line one.

Beyond 400 m the curves separate, and that is a real difference, not a rounding one. Ours is steep (64.9 -> 92.9 -> 96.8); theirs is flat (59 -> 78 -> 83). 17% of their households are beyond 1.2 km against 3.2% of ours. Two things drive it and we cannot separate them from the published post: their extra ~950 km2 of non-BCC fringe, and the SA1 as a sample unit - a large outer SA1's centroid sits far from anything, where our mesh blocks resolve the houses. We do not claim our tail is right and theirs is wrong. We claim the 400 m figures agree, and that the agreement at the threshold the headline uses is what the headline needs.

Podaris also report 47.5% of Brisbane residents at PTAL 1a or no access at all, on the Transport for London methodology in the morning peak. That is a frequency-weighted measure with no direct analogue here, but it is the same shape of finding as our turn-up-and-go tier.

RMIT / Australian Urban Observatory liveability scorecard (Nov 2018)

RMIT measured 61% of residences within 400 m of a stop using pedestrian road-network distance - the same basis as our hero. Ours is 64.9% of dwellings. These are the two closest-matched numbers in this section and they agree to 3.9 points, across an eight-year vintage gap and a larger study region.

Their second indicator - 12% within 400 m of a stop with a scheduled service at least every 30 minutes between 7am and 7pm on a normal weekday - has no tier in this analysis, so one was built to match it. Two readings, because the wording admits both:

This one does not reconcile. On the stricter reading we are 2.5x more generous than RMIT. Study region and vintage explain part of it - their frame is the wider metropolitan area, and Brisbane's network has changed since 2018 - but not all of it, and we cannot close the gap from the published material. Recorded as an unresolved divergence. It cuts against us, not for us: if RMIT's number is the better one, far more Brisbane residents lack a useful service than this analysis reports. It is also the reason the HOURLY tier must never be described as equivalent to the RMIT indicator - it is a one-departure-an-hour test, not a half-hourly one.

GoThere Brisbane (rebuilt 17 Jul 2026)

GoThere is the dataset a reader is most likely to have seen, and it publishes no population share at any radius, so there is nothing to set against our headline. It measures one WorldPop-weighted point per suburb and counts routes within a straight-line 800 m radius from that point, over 205 gazetted suburbs whose centre is within 20 km of Brisbane Central.

Rebuilt on our data - population-weighted centre of each suburb's mesh blocks, 800 m straight line - 9 of our 165 suburbs inside that 20 km ring have no stop within 800 m of that point: Upper Brookfield (5,188 m), Enoggera Reservoir (2,872 m), Pullenvale (1,988 m), Larapinta (1,480 m), Anstead (1,424 m), Brisbane Airport (1,191 m), Ransome (906 m), Hemmant (868 m), Karawatha (861 m).

GoThere name 6 "transport deserts": Camp Mountain, Draper, Jollys Lookout, Mount Coot-Tha, Port Of Brisbane, Upper Brookfield. 3 of them - Camp Mountain, Draper, Jollys Lookout - are not in Brisbane City Council at all, and Mount Coot-Tha has no resident population in the 2021 mesh blocks.

Their 20 km cut also excludes Chuwar, Karana Downs, Kholo and Mount Crosby - 4 of the 6 populated suburbs where not one resident is within a 400 m walk of a stop. The two datasets are not measuring the same city, and any comparison between the counts is a category error.

Worst-covered suburbs (population >= 500)

Ranked on the walking basis, which is what the headline uses; the straight-line column is carried beside it so the reordering is visible.

Full ranking in worst_covered_suburbs.csv.

The unserved list, defined by residents

10 suburbs holding 7,272 residents have not one resident within a 400 m walk of a serviced stop, and 11 holding 7,605 are at or under 5%. The list is defined per resident on the walking basis; the whole-suburb binary below is what it replaced. 6 of the 10 hold 50 residents or more:

Threshold artefact test, run on this list rather than the retired one. Across every populated block in the 10 suburbs, the closest any resident gets to a serviced stop on foot is 424 m - in Eagle Farm, a block of 5 residents, 1.06x the 400 m test.

2 of the 10 sit within 1.5x the threshold and would move if it did. Stated rather than buried, because this is the artefact objection and it is partly right at the tail of the list: Eagle Farm 424 m (5 residents); Ransome 477 m (268 residents). The list's weight is not there. 6,999 of its 7,272 residents are in suburbs whose nearest block is at least 853 m from a stop on foot, and the threshold-sensitive entries are a few hundred people in industrial and reserve blocks.

Secondary: suburbs with no stop at all

5 suburbs, home to 6,960 residents, have no serviced stop anywhere inside them or within 400 m. This was the primary measure in the previous version of this analysis and is now secondary: a whole-suburb binary hides the 469,985 residents who live beyond a walk of the threshold inside suburbs that do have stops somewhere.

None of these five is a threshold artefact either: the closest is 1,483 m from any serviced stop, about 4x the 400 m test.

A further 4 unserved suburbs are effectively unpopulated (under 50 residents): Banks Creek, England Creek, Lake Manchester, Port Of Brisbane. All 9 unserved suburbs together hold 6,963 residents.

Population basis

Every per-resident figure uses ABS 2021 mesh-block counts summed inside each BCC suburb polygon, not the SAL name-join used earlier. The two agree to 0.18% citywide, but the name-join is wrong for any locality that straddles the council border, because it attributes the whole locality's population to Brisbane.

Chuwar is the case in point: its SAL holds 2,178 people but only 93 live inside the Brisbane polygon - the rest are in Ipswich. Reporting the SAL figure would have overstated the unserved population by about 2,085.

Under regular services only

**9 suburbs have no regular service within 400 m - identical to the 9 on the any-service test, because discounting limited routes adds no suburb. No suburb depends on a limited route for its only access. A route is discounted as school_or_limited when it runs fewer than 10 trips in a representative week; the school half of that category is empty by construction, because no main-feed route runs on a school-term-only calendar** - TransLink ships every dedicated school run in a separate feed.

Route patterns across the main feed (by route_short_name): full_week 348, weekday_only 140, weekend_only 17, school_or_limited 8.

Kids can get out, adults can't

No suburb that lacks regular service is reached by a dedicated school run within 400 m either. Same measurement basis.

At 800 m the count rises to 1 (Port Of Brisbane), which is unpopulated. So on this feed no populated suburb without regular service is rescued by a dedicated school run at either radius.

TransLink publishes the school feed as a pilot; treat its coverage as indicative and not as a service guarantee. It is also a superset of the main feed rather than a school-only feed - all 108,236 main-feed trips reappear in it - so the dedicated school runs used here are the 3,096-trip set difference.

Weekday service but no weekend service

3 suburbs have weekday trips but zero Saturday and zero Sunday trips at stops inside their boundary, on the three representative days below: Nudgee Beach, Pinkenba, Sumner.

Representative days

Chosen from inside the feed's own validity window, never hardcoded: within each day type the median-traffic date is taken, so public holidays and event days cannot become the representative. The weekday is restricted to Tue-Thu of a school term, where school term is defined as a date on which dedicated school-feed services actually run.

Boundary caveat

Suburb boundaries follow road centrelines, so a stop serving a suburb often sits just outside it: the stops_outside_walk (400 m) and stops_outside_walk_800 (800 m) columns carry those, and every access test in this document uses inside-or-within-400 m rather than a strict in-polygon count.

Population join

ABS SAL names are matched exactly, never fuzzily. ABS suffixes a locality name only when it is duplicated: (Qld) when the clash is interstate and (Brisbane - Qld) when it is within Queensland, so a suburb is matched against those three exact forms and nothing else. Other LGA suffixes ((Ipswich - Qld), (Toowoomba - Qld)) are deliberately not tried, since they name different places; stripping the two accepted suffixes produces no name collisions in the Queensland SAL list. All 190 suburbs matched, and the run aborts rather than guessing if any fails to.

Stop-name sanity check

Every unserved suburb's name was searched against stop names in both feeds, to catch a stop that is named for one of these places but carries bad coordinates or sits just past the line.

145 stops in either feed carry an unserved suburb's name; 0 of them lie within 800 m of that suburb. Full table in unserved_name_check.csv.

Institutional population

14,498 residents (1.2%) live in non-private dwellings - correctional centres, barracks, halls of residence, hospitals and aged care. ABS classifies prison mesh blocks as Residential, so the category field cannot find them; what does is persons per private dwelling. An ordinary block sits near 2.5 and the 99th percentile is 4.2, so a block with no dwellings or at least 8 people per dwelling is housing people in an institution.

This matters because some people counted as living far from transport are not free to travel. The suburbs whose coverage moves most once institutional blocks are excluded:

Wacol is the case that would have drawn fire. Of its 4,273 residents, 3,526 (83%) are in institutional blocks - the correctional precinct and immigration detention. Counting them, Wacol looks like one of the worst-covered suburbs in Brisbane at 51.3%. Excluding them it is 98.5%, among the better outer suburbs. Do not cite Wacol as a transport-poverty case.

The correction does not always flatter. Nathan goes the other way, 60.6% to 43.7%: its institutional residents (university accommodation) are the well-served ones, and its households are not.

Every worst-covered table carries an institutional column and an excluding-institutional percentage so this can never be read the wrong way round.

Known limits

• The per-resident measure places every mesh block's whole population at one distance from transport. The populated blocks are small - median 0.029 km2 and 97 residents - so this is minor, but within a block no distribution of people is modelled.

• Distances are straight-line, so every coverage figure here is optimistic and every beyond-the-threshold figure is a floor.

• A stop counts as served if it has any scheduled trip. A stop with three trips a week counts the same as a busway station, so proximity here is not frequency; pair it with the trip_visits_* columns.

• trip_visits_* count scheduled stop calls, not vehicles or capacity, and a rail platform counts separately from its neighbour.

• Population is 2021 Census against a 2026 timetable; the fast-growing outer suburbs are undercounted.