The Longest Line

Walk the longest perfectly straight line you can draw across a state’s public land without ever crossing a road. No pavement. No gravel. No dirt two-track. No forest road. Just you, a compass bearing, and the wildest unbroken corridor the map allows.

Inspired by Alasdair Rae’s longest straight line in Great Britain, this is its human-powered cousin: not just where the line is, but a standing challenge to follow it, end to end, on foot.

The Challenge

For a given state, the line is fixed: it is the single longest straight segment that fits entirely on Federal or State public land and crosses no road of any kind. Your job is to travel it.

• Follow the line. Stay on the straight-line bearing, corridor-style, from one published endpoint to the other. By construction you will never cross a road. If you hit one, you have drifted off the line.
• Single push. One continuous effort. Sleep and rest along the way are fine; the clock keeps running.
• Human power only. Hiking, scrambling, climbing, skiing, packrafting, muscle and gravity, nothing motorized.
• Either direction, solo or as a team (no relays; everyone covers every mile).
• Leave no trace. These corridors are roadless for a reason. Travel like it.

The line is a geographic abstraction; the terrain is not. Expect off-trail, unmarked, committing ground: canyons, talus, blowdown, river crossings, and long waterless stretches. The straight-line distance is the floor; the real distance over the ground is always longer (see table below).

The Lines

Eight states computed so far. Straight-line distance is the published length; “over the ground” is the terrain-corrected distance walked along the actual ground profile, from a 10m elevation model.

Rank	State	Straight (mi)	Over ground (mi)	Corridor	Endpoints (lat, lon)
1	Alaska	531.2	542.8	Interior — Brooks Range / North Slope to the Yukon–Kuskokwim Delta	68.4820, −152.9480 61.9316, −162.5376
2	California	158.6	167.0	High Sierra — John Muir Wilderness / Sequoia–Kings Canyon	37.8742, −119.3451 35.8915, −117.9198
3	Washington	107.3	115.0	North Cascades — Pasayten / Okanogan	48.7823, −122.0354 48.9984, −119.7297
4	Idaho	98.8	105.4	Frank Church–River of No Return / Selway–Bitterroot	45.6967, −114.5310 44.3689, −115.2783
5	Montana	98.3	103.2	Bob Marshall — Great Bear — Scapegoat	48.4045, −113.2395 47.0124, −112.7390
6	Arizona	55.9	65.5	Grand Canyon / Kaibab Plateau	36.5158, −112.7097 35.9842, −111.9525
7	Nevada	50.1	52.6	Desert NWR / Nellis Range	37.1511, −115.0185 36.4533, −115.2450
8	Oregon	48.8	49.9	Three Sisters — Waldo Lake — Diamond Peak	44.3042, −121.7270 43.6615, −122.1345

Notes:

• Arizona’s +17%: the line drops into and climbs out of the Grand Canyon twice!
• The desert states (Nevada, Arizona) lose despite enormous public acreage, because dirt and 4WD roads shred them into small cells. The winners are long, linear roadless corridors. Nothing in the Lower 48 beats the High Sierra crest, where no road crosses the range for ~150 miles between Tioga Pass and Walker Pass.
• Alaska is the national champion by ~3.4×! The line crosses major rivers (the Yukon and Kuskokwim), following it would demand serious packrafting.

Maps

Each line shown on a topographic basemap.

State by state

Download the lines

Each line as a straight segment and as 1-mile waypoints, in GPX (GPS units & watches), KML (Google Earth) and GeoJSON (GIS / QGIS). All WGS84.

State	Mi	Line	Waypoints (1 mi)
Alaska	531	GPX · KML · GeoJSON	GPX · KML · GeoJSON
California	159	GPX · KML · GeoJSON	GPX · KML · GeoJSON
Washington	107	GPX · KML · GeoJSON	GPX · KML · GeoJSON
Idaho	99	GPX · KML · GeoJSON	GPX · KML · GeoJSON
Montana	98	GPX · KML · GeoJSON	GPX · KML · GeoJSON
Arizona	56	GPX · KML · GeoJSON	GPX · KML · GeoJSON
Nevada	50	GPX · KML · GeoJSON	GPX · KML · GeoJSON
Oregon	49	GPX · KML · GeoJSON	GPX · KML · GeoJSON

How the Lines Were Found

Computation was done based on a geospatial stack bundled with QGIS (GDAL/OGR, PROJ, GEOS) plus SciPy/NumPy. The entire stack was built with Claude Code. The method used was a raster rotate-and-scan:

1. Define the free space. A point is “free” if it is on Federal or State public land, inside the state, and not on (nor immediately adjacent to) any road.
2. Burn three masks — public land, state boundary, and roads — onto a grid in an equal-distance projection (EPSG:5070 for the Lower 48, EPSG:3338 for Alaska, EPSG:2992 for Oregon). Roads block a cell wherever they touch it, so a straight line can never jump one.
3. Scan every bearing. For each compass heading (1° steps, then a 0.1° refine), rotate the free-space grid and find the longest unbroken run of free cells in any row. That run is the longest straight line at that heading; the maximum over all headings is the answer.
4. Refine & locate. Re-run the winner on a fine grid (down to 30 m) and convert the endpoints to latitude/longitude.
5. Terrain-correct. Sample a 10m elevation model along the line and integrate the 3-D profile, i.e., the true over-the-ground distance.
6. Verify. Prove it is not just a long line but the longest by ruling out every rival corridor that could possibly beat it (see below).

What “verified” means

A scan returns a long line; verification proves it is the longest in the state. It rests on one fact: a coarse grid can only over-measure a corridor, i.e., it may skip a small inholding or a thin road and report a line a little longer than reality, but it can never make a corridor look shorter than it really is. Equivalently, a corridor’s measured length only shrinks or holds as the grid gets finer. That one-way relationship is what makes verification possible, in two flavors:

• Convergence (the Lower-48). Re-run the entire state at several resolutions. If the same corridor wins every time and its length settles on a stable value, it is the global maximum, i.e., any truly longer rival would have to overtake it as the coarse-grid bias shrinks, and none ever does.
• Elimination (Alaska). Where many corridors compete, list every distinct corridor that scores above the current best on the coarse grid, then re-measure each on a fine grid. Since a fine length can only be ≤ its coarse length, any corridor whose coarse score is already below the best fine result cannot win and is dismissed without further work. Whatever survives that gauntlet is verified.

Alaska is the cautionary tale. Two corridors were a coin-flip apart on the 500 m grid (533.4 vs 531.9 mi). Refining only the front-runner gave 444 mi, but that corridor was a decoy, riddled with inholdings that a coarse grid had papered over. Verification forced its runner-up to be measured too, and that one held firm at 531 mi. Without the elimination step, the published answer would have been wrong by ~90 miles.

The one thing verification cannot do is make the answer better than the data it runs on. It proves a line is the longest in the datasets, not necessarily the longest on the actual ground. Public land comes from USGS PAD-US: a missing parcel, a mis-tagged owner, or a stale inholding boundary moves the result. Roads come from OpenStreetMap: an unmapped backcountry track would let a line run across ground that is not really road-free. So every figure here is the verified maximum given PAD-US and OSM as they stand today, no more, no less. Better sources (for Alaska, the State of Alaska DNR land-status records; for roads, an authoritative state centreline layer like Oregon’s) would sharpen, and occasionally shift, these numbers.

How this differs from Rae’s original

Alasdair Rae’s Great Britain analysis was a vector method: he converted the road network into polygons, then tested straight lines drawn between the vertices of those polygons and kept the ones that crossed no road, homing in by eye on the obvious big gap (the north-west Scottish Highlands). This project takes three deliberate turns away from that:

• Raster, not vector. Rather than test vertex-to-vertex lines, the free space is rasterized and every bearing is swept. That sidesteps the geometry-engine breakdowns and the “one giant connected blob” problem that a state- or nation-scale vector search runs into, and it makes the result exhaustive rather than heuristic.
• No region picked by hand. Every heading is scanned across the entire state automatically, so the winning corridor is found by the algorithm, not chosen by eye, and a multi-resolution re-run verifies it as the true global maximum.
• A public-land filter. Rae’s line could run across any ground between roads; here it must also stay on Federal/State public land. That turns “longest road-free line” into “longest line you could actually, legally walk,” the whole point of this follow-the-line challenge.

Definitions. “Public land” = Federal + State managed parcels. “Any road counts” = paved, gravel, dirt, forest, and 4WD vehicle ways (foot trails and paths are not roads and are excluded). Private inholdings, water, tribal, and local-government parcels are not public land for this challenge.

Datasets & Tools

Public land (ownership)

• USGS PAD-US (Protected Areas Database of the U.S.), Manager Type — Federal+State, used for CA, NV, AZ, ID, MT, AK.
• BLM Oregon/Washington Surface Management / Ownership — used for OR and WA.

Roads

• OpenStreetMap via Geofabrik state extracts (vehicular highway classes) — CA, NV, AZ, ID, MT, WA, AK. Map data © OpenStreetMap contributors.
• Oregon “All Public Roads” (OR-Trans / ODOT TransGIS) — the authoritative compiled all-authorities layer used for Oregon.

Boundaries & elevation

• U.S. Census TIGER/Line — state boundaries.
• USGS 3DEP 10 m elevation, sampled via OpenTopoData (ned10m) — terrain correction.

Software

• QGIS · GDAL/OGR · PROJ · GEOS · SciPy / NumPy.

Method inspired by Stats, Maps & Pix — “The longest straight line in Great Britain”. Lengths carry roughly ±0.2 mi of grid uncertainty; within the data, the winning corridor and its identity are settled for the verified states (subject to the data caveat above).

Disclaimer

The information provided on this website does not identify possible dangers. When you are attempting these routes, you assume responsibility for your own actions and safety.