Atlas
22 rail networks, 2,670 stations and 38,258 scheduled trips. Every map here was generated by the same code from the agency’s own published schedule, with no per-city drawing and no manual correction.
Each network links to a running map: a real service day, at whatever speed you want to watch it. Los Angeles and Mexico City come first, because they are the two the essay argues from. The rest follow from the cleanest result to the roughest — the networks the pipeline placed every stop and traced every trip on, then those it had to fudge, with the caveats under each one saying exactly how.
-
LA Metro Rail
- 7 station names had nowhere to sit without overlapping another and are not drawn
-
Mexico City Metro
- This is a 2025 snapshot: the published feed's service period ran to December 2025, so the date above is from its own calendar rather than this week.
- The operator publishes headways rather than timetabled times, so the trains here run at the scheduled interval for each period of the day, evenly spaced. It shows: on the time chart this network is a solid band rather than the peaks and troughs the timetabled cities draw, because the published interval barely varies.
- The station icons the Metro is known for are not in the data, and are the thing this pipeline cannot generate.
- 2,090 trips (25%) run part of the way on a neighbouring line's track, because the schematiser did not attribute that segment to their line. They follow the right corridor, but not always the right parallel track
- 3 station names had nowhere to sit without overlapping another and are not drawn
-
Chicago 'L'
-
UTA TRAX & FrontRunner
-
DART Light Rail
-
Sound Transit Link & Sounder
-
Pittsburgh Light Rail
-
BART
-
MARTA Rail
-
Metro Transit Light Rail
-
Metra
- 4 station names had nowhere to sit without overlapping another and are not drawn
-
Cleveland RTA Rapid
- 1 station name had nowhere to sit without overlapping another and is not drawn
-
MBTA Subway
- 4 station names had nowhere to sit without overlapping another and are not drawn
-
SEPTA Regional Rail
- 7 station names had nowhere to sit without overlapping another and are not drawn
-
RTD Denver Rail
- 6 station names had nowhere to sit without overlapping another and are not drawn
-
Long Island Rail Road
- 17 station names had nowhere to sit without overlapping another and are not drawn
-
Valley Metro Rail
- 1 of 124 stops could not be placed on the map, so trains pass straight through them
- 138 trips (13%) skip a stop the map does not carry
- 1 station name had nowhere to sit without overlapping another and is not drawn
-
NJ Transit Rail
- 2 of 231 stops could not be placed on the map, so trains pass straight through them
- 516 trips (28%) skip a stop the map does not carry
- 22 station names had nowhere to sit without overlapping another and are not drawn
-
New York City Subway
- 2 of 989 stops could not be placed on the map, so trains pass straight through them
- 577 trips (7%) skip a stop the map does not carry
- 6,627 trips (78%) run part of the way on a neighbouring line's track, because the schematiser did not attribute that segment to their line. They follow the right corridor, but not always the right parallel track
- 88 station names had nowhere to sit without overlapping another and are not drawn
-
TriMet MAX
- 4 of 234 stops could not be placed on the map, so trains pass straight through them
- 380 trips (37%) skip a stop the map does not carry
- 60 station names had nowhere to sit without overlapping another and are not drawn
-
Miami Metrorail & Metromover
- 25 of 89 stops could not be placed on the map, so trains pass straight through them
- 31 trips could not be traced across the network at all and are not shown
- 743 trips (38%) skip a stop the map does not carry
-
Muni Metro
- 34 of 292 stops could not be placed on the map, so trains pass straight through them
- 1,210 trips (89%) skip a stop the map does not carry
- 7 station names had nowhere to sit without overlapping another and are not drawn
On the imperfections
Where a network above lists caveats, they are real, and I have left them visible rather than quietly dropping the affected trains. Most come from the same two places: an agency numbering its platforms in a way the schematiser merges differently than the timetable expects, or a network dense enough that some station names have nowhere to sit without overlapping another.
Miami is the weakest result here, and the reason is interesting — Metrorail and Metromover are almost separate systems sharing one feed. Washington DC is missing entirely, because WMATA publishes its GTFS only behind an API key.