How we calculate this

The duration formula

Adapted, with attribution, from Den Norske Turistforening and UT.no, who publish theirs: 3.5 km/h, plus 15 minutes per 100 m of ascent, plus an hour of breaks every five hours, children under 10 double. We converted the constants rather than re-guessing them.

Then, for every mode

• Breaks: 12 minutes added per hour of moving time — an hour of breaks in every five hours of walking.
• The line is shorter than the walk: distance is increased by 4% before anything else. UT.no make the same point about their own geometry; ours is traced from planning figures at ±15–25 m, so the effect here is larger, not smaller.
• Steep descent counts: descent on stretches steeper than 12% adds time rather than saving it.
• Who is walking: with a child under 10 — ×2; with a child 10 or older — ×1.5.
• Rounding: to the nearest five minutes. "1 h 47 min" claims a precision the inputs do not have.

Worked example

Cold Spring to Little Stony Point and back is 4.2 miles with 150 ft of ascent, on foot:

The difficulty bands

The grade is derived, never hand-assigned, and every place it appears links here. The numbers are the truth; the grade is a convenience — that ordering is the whole argument against a community-editable star rating.

• Demanding — anything steeper than 12% at any point.
• Moderate — over 5 miles, or over 700 ft of ascent, or steeper than 8.33% (the US Access Board's outdoor limit).
• Gentle — none of the above.

The thresholds are adapted from Norway's Merkehåndboka, scaled to a riverside linear park: this whole corridor is 6.8 miles long and its highest point on the mainline is about 60 ft.

Where the slope figures come from, and what is wrong with them

No slope on this site was measured on the ground. The trail is not built, so there is nothing to measure. Every grade is modelled from USGS 3DEP elevation, sampled along an alignment we traced from the published FGEIS figures of January 2026 — a ~2.2 m georeferencing fit, but ±15–25 m of real-world uncertainty, because the alignment itself is not final.

There is a second problem, and it is the more interesting one. A bare-earth elevation model does not know about structures. Much of the Fjord Trail mainline is planned as elevated boardwalk and bridge, and 3DEP measures the ground underneath a structure, not the deck on top of it. That is why the traced mainline shows an 18.6% "maximum" where the design intent is a boardwalk under 5%.

So the planner does two things, and shows you both:

• On mainline sections it plans on the length-weighted typical grade, and prints the model maximum beside it with this explanation attached to the number.
• On the connectors — natural ground on existing state-park paths, never built and never surveyed — it plans on the model maximum, because there is no structure to be wrong about.

The day HHFT's survey lands, both are replaced by measured values and the badge on every number changes from "from design documents" to "measured on site" without anyone editing a page.

The limits you can set yourself

The defaults are standards, not judgements. Change them on the results page and the answers change.

• Running slope, default 8.33% — the ABA §1017.7.1 figure, which permits 8.33% for up to 200 ft.
• Tread width, default 36 in — the ABA minimum clear width. Recorded and displayed, but it cannot filter anything yet, because no width has been measured anywhere on this trail.
• Gap between resting places, default 1320 ft — a quarter of a mile.

What each option means

A promise without a definition is how trust gets broken at mile 3. These are the published definitions the planner uses.

Step-free the whole way

No steps, no stiles and no kerbs without a ramp, from the parking space or the station platform to the far end of the route and back. Where a route is step-free only as far as a point, we say where that point is rather than calling the whole route step-free.

Somewhere to sit at least every quarter mile

A bench or a resting place with a level surface beside the path, at intervals of no more than 1,320 ft (400 m), with space beside it for a wheelchair or a pushchair so that resting does not mean blocking the path.

An accessible restroom on the route

A restroom on or within 100 yards of the route with a step-free approach, a door at least 32 in clear, and a transfer space beside the toilet. "Accessible" here does not mean "there is a portable toilet in the parking lot".

Shade for most of the route

Tree canopy or built shade over more than half the length. This matters more than it sounds: the National Weather Service recorded a wet-bulb globe temperature of 28.3 °C in this corridor on 7 August 2026, which is the level at which organised outdoor activity is normally curtailed.

Quiet — away from traffic noise

Not within earshot of Route 9D or the Metro-North line for most of the route. The Hudson Line runs the length of this corridor and freight trains use it at night, so "quiet" here means quieter, not silent.

I am travelling with a service animal

Service animals are welcome everywhere the public may go, so this never removes a route. Selecting it adds the relief areas, the water points and the surface temperature note to each answer, which is what a handler actually needs to plan around. This one never removes a route.

I will have a stroller or a wagon

Wide enough for a stroller to pass another person, firm enough for small wheels, and no steps. A route can be fine for a wheelchair and awkward for a wagon, and the other way round, so this is a separate question.

The return train

Departures come from Metro-North's published GTFS timetable — https://rrgtfsfeeds.s3.amazonaws.com/gtfsmnr.zip, downloaded 2026-08-07, covering 2026-08-01 to 2026-11-22. Two service days were sampled: Saturday 2026-08-08 for the weekend pattern and Wednesday 2026-08-12 for the weekday one.

Scheduled times from the published timetable, not a live departure board. Two service days were sampled from the feed — a Saturday and a Wednesday — so weekday and weekend patterns are real but public-holiday and engineering-work variations are not. Check before you travel.

Public holidays run a weekend timetable and are not modelled here, because the sampled feed covers 1 August to 22 November 2026 and no holiday fell inside the two days we captured.

Breakneck Ridge station has zero rows in that feed. Not a reduced service — none. It closed on 21 April 2025 for a rebuild delivered as part of this project, and 511NY carries the closure as scheduled through 21 April 2027. A trip planner that queried it and rendered the empty result would show you nothing, forever, and look like it was working. So the planner reports the suspension and names the nearest station that has trains.

The "leave by" time is the departure, minus the walk to the platform, minus the slack recorded on the itinerary. Missing a train at Cold Spring can mean an hour on a platform, so the slack is deliberately generous.

Walking to the station

Straight-line distance from the published coordinates, multiplied by 1.3 for the street network, walked at 2.5 mph. It is an estimate of a walk nobody has timed.

In full: straight-line distance between two published coordinates, multiplied by 1.3 for the street network, walked at 2.5 mph — slower than the trail figure, because of crossings. Where a coordinate is itself uncertain, the route says so: Dutchess Manor's position is ±150 m, taken from Wikipedia, because there is no record of it in Nominatim.

What we will not do

• Publish a slope or a width we cannot source.
• Call anything "field checked" until somebody has walked it with a level and a tape.
• Show a scheduled departure as though it were a live one.
• Round an unknown up to a yes. Where a record does not answer a question, the planner says "not yet known" and keeps saying it until a survey exists.