TriMet Stops — agent guide

Active transit stops. 6,316 Point features, WGS84, one GeoParquet file of 287.9 KB in 1 row group(s).

Access

Reads stream over HTTP range requests — query in place, do not download. The file is Hilbert-ordered and carries a GeoParquet 1.1 bbox covering column, so a spatial filter on geometry_bbox prunes row groups from metadata alone. Note the column name: GDAL writes the covering as <geometry column>_bbox, so it is geometry_bbox here, not bbox.

The recipes below use bare relative paths ('stops/stops.parquet') for readability. Prefix them with https://data.source.coop/cholmes/trimet/ to run remotely.

Other formats: stops.pmtiles for map display (layer name stops), and TriMet's original Shapefile and KML, linked from collection.json as source_shapefile and source_kml.

Coordinate system, and what follows from it

The GeoParquet is in EPSG:2913 — NAD83(HARN) / Oregon North, international feet — which is what TriMet surveys and publishes in. It is deliberately not reprojected. Only the PMTiles are in WGS84, because vector tiles are Web Mercator by definition.

The practical consequence is a good one: measurements just work, in feet.

Useful conversions: × 0.3048 for metres, ÷ 5280 for miles, ÷ 43560 for acres, ÷ 27878400 for square miles. 400 m is 1312.34 ft, a quarter mile is 1320 ft.

Two things to watch:

• Coordinates are not longitude and latitude. An x of 7633099 is feet east of the projection's false origin, not a degree. Anything expecting lon/lat — a web map, a geocoder, most GeoJSON consumers — needs a transform first.
• always_xy := true when you do transform. Without it DuckDB honours EPSG:4326's authority-declared latitude-first axis order and every result comes back Infinity:

The bbox covering column is in the same feet, so a spatial filter written against it uses projected coordinates, not degrees. The collection's STAC extent stays in WGS84, because STAC requires that regardless of the data's own CRS.

Nothing here carries a pre-computed length or area except the district boundary, which has TriMet's own area_sq_mi and acres.

Quirks and caveats

stop_id joins cleanly to route-stops — exactly

All 6,316 stop_id values here appear in route-stops, and route-stops has exactly 6,316 distinct stop_id values. The two agree perfectly, and stop_name never disagrees between them. This is the catalog's most reliable join.

type has a value here that the routes layer does not

BSC, "Shared Bus and Streetcar", appears on 14 stops. It is documented on the stops metadata page but not on the routes or route-stops pages, and it does not occur in those layers. Code that maps type through a lookup built from routes will miss it.

zipcode is a string; stop_id is an integer

zipcode is text (leading zeros are not an issue in Oregon, but the column is still typed as a string). stop_id is int32 — do not quote it when joining.

The same stop_id is used by TriMet's arrivals API

stop_id is the public stop number. It is the locIDs parameter of TriMet's arrivals endpoint, which is how you get from a point in this catalog to live data. That API needs a registered app ID and is governed by separate terms.

Query recipes

Stops by jurisdiction and mode

Stops within 400 m of a transit center — note the reprojection

How many routes serve each stop, by joining to route-stops

Related collections

• route-stops — the same stops exploded by route and direction; join on stop_id
• rail-stops — rail platforms with real station names, but a different id space
• transit-centers — the 15 major transfer hubs

Provenance

TriMet publishes this as tm_stops at developer.trimet.org/gis, last updated July 30, 2026. Every column description in collection.json comes from TriMet's metadata page — that page, not this catalog, is the authority on what a field means.

Conversion: ogr2ogr -f Parquet -t_srs EPSG:4326 -lco COMPRESSION=ZSTD -lco SORT_BY_BBOX=YES -lco WRITE_COVERING_BBOX=YES, then tippecanoe with -r1 --no-feature-limit so no feature is dropped at any zoom. See tools/convert.py.

License, and a warning

TriMet distributes these datasets free of charge, but the terms of use do not grant redistribution rights. Section 4 of TriMet's Terms of Use reserves reproduction, modification, distribution and republication of site Content without written consent; the redistribution license in section 5 covers the Web Services API, not these GIS downloads. The collections therefore declare "license": "other" with a link to those terms, rather than claiming an open license the source does not offer.

Practically: use the data, and contact gis@trimet.org before redistributing it or building a product on it. If you need transit data under clear open terms, TriMet's GTFS feed is the better starting point.