NYC Commute Network / 15–90 Minutes
NYC
TIME FIELD
Where can 30 minutes take you? Compare subway, driving, and walking through one shared New York.
MQBKBXSI
01
Interactive model
Commute Map Lab
Click to place a destination, then switch to Inspect Homes to read the best route from any residential hex.
Calculating the time field…
SubwayWalk + wait + ride + transfer + walkHeight: time remaining · pause on a hex for route steps
shorternear limit
Residential reach
Neighborhood field
| Rank | Neighborhood | Time | Reach |
|---|
| Rank | Neighborhood | Time | Reach |
|---|
02
Transparent by design
How the field is built
This is a comparative accessibility study, not a live journey planner or rental-listing service. Every assumption stays visible.
Subway schedules, entrances, pedestrian paths, and roads.
Subway, walking, and directed free-flow driving graphs.
Residential lots allocate official ACS rent to each H3 cell.
A worker changes mode while height changes independently.
Model snapshot
— H3 cells
Resolution 9 cells are joined to 2020 Neighborhood Tabulation Areas and weighted by MapPLUTO residential units.
Transit graph
— stations
Direction-aware GTFS graphs are rebuilt for weekday peak, weekday midday, and combined Saturday–Sunday service. Published GTFS transfer times are used before the documented fallback.
Road + walking graphs
OpenStreetMap networks
Walking follows the pedestrian network at 4.8 km/h. Driving follows directed roads with OSM speed data and excludes live traffic and parking.
Research goal
What does one time budget mean across three travel modes?
Subway, driving, and walking are shown as separate maps. The interface does not hide their different assumptions inside a composite score.
Pedestrian model
OSMnx shortest paths
H3 centers and valid station entrances or exits are snapped to the OpenStreetMap walking graph. Direct walking and first/last mile access use network distance at 4.8 km/h.
Rent height
H3-level median gross rent estimate
MapPLUTO residential units allocate official Census block-group values to each cell. Missing observations use the official tract or borough value—never a neighboring-area interpolation. Each distinct rent value receives a distinct ranked height.
Model date
Loading model…
GTFS feed —–— · generated —.
Validation
Geometry, indexes, paths, and matrices
Automated checks cover CRS, unique H3 and station indexes, residential NTA coverage, entry/exit rules, OSM connectivity, GTFS path sums, weekend routes, and Staten Island separation.
Analysis notebook ↘System diagram ↗Data dictionary ↗Rent data table ↘Rent data audit ↗Full limitations ↗Final reflection ↗
What this map does not know
Schedules are not lived experience.
The model does not include live delays, crowding, stairs, elevator outages, construction, safety perception, fare cost, or the emotional texture of a commute.
Buses and ferries are excluded. Staten Island Railway remains disconnected from the subway because the model will not invent a ferry edge. NTA names are statistical approximations, not final definitions of neighborhood identity.
OpenStreetMap paths and MTA schedules are still models: walk connectors are snapped to nearby graph nodes, expected waiting is half the scheduled headway, and temporary service conditions are absent.
Driving uses free-flow road speeds and omits traffic, parking, incidents, toll delay, and curb access. The H3 rent surface allocates five-year ACS estimates by residential units and carries sampling error; it is not a live market price.
Reflection and next questions
The field is precise about its assumptions, not about lived certainty.
Building the project showed that “30 minutes” is not a neutral measure. Express corridors stretch it, walking compresses it, and roads create a third field whose apparent speed excludes the cost of congestion and parking. Rent height adds housing context while the commute threshold still controls which homes appear. A future version should add observed reliability, live traffic, accessibility, and asking-rent data with an explicit license.