Methodology

Where the facts come from

Timetables, station positions, and route geometry come from each operator's GTFS feed, the machine-readable timetable operators publish. Amtrak's feed is checked daily and stored when it changes. Named stretches and their one-sentence descriptions are paraphrased from the operator's own route guides and route pages, and each is cited on the page where it appears.

Sources are ranked. Numbers come only from primary structured data such as a GTFS feed. Descriptions of what is visible come from the operator's own material or from the National Park Service. Reference sites are used for context only, and forums are never cited.

What counts as a named stretch

A stretch is admitted when it appears in an Amtrak route guide, an NPS Trails and Rails description, or two independent operator descriptions. A route gets its own page once it has three or more such stretches. Routes below that line are listed on the routes page without a page of their own.

How the daylight table is computed

For each stretch and each direction, the site takes one representative trip from the current timetable: the trip that runs on the most days over the feed's calendar (its days per week times the weeks it runs, since the feed splits a daily train into many calendar pieces), then the one with the most stops, then the lowest train number. The stretch's scheduled time is the midpoint between the departure from the first of its two stations and the arrival at the second, converted from the feed's zone to the station's local zone.

Sunrise and sunset are computed for the 15th of each month at the midpoint of the two stations' coordinates, using the standard sunrise equation, which is accurate to about two minutes. The scheduled time is then classified for each month:

• daylight: more than 30 minutes after sunrise and more than 30 minutes before sunset
• dawn: within 30 minutes of sunrise
• dusk: within 30 minutes of sunset
• dark: otherwise

The twelve monthly results are summarized into a phrase such as "daylight May to Aug; dusk Sep; dark Oct to Apr". Only these four words and month ranges are used. Scheduled times are used as published; a late train moves evening stretches into darkness and morning stretches into daylight. A buffer for each route's average lateness, from the Federal Railroad Administration's quarterly reports, is planned and not yet applied.

How the "what to see" pages rank routes

A "what to see" page lists the routes with at least one named stretch tagged with that interest. The order is by the number of such stretches on each route, and the number is shown. For two interests together, a route counts only when it has a stretch for each, and its score is the sum. A page is listed for search engines once three or more routes qualify; below that it exists as a finder state only. The other measures named in the plan, miles of a feature in daylight and stations near a park, are added when the geometry and station work is done.

A comparison page puts two routes side by side on duration, frequency, named stretches, what each shows, and the daylight phrase for each route's highlighted stretch. A comparison exists for every pair of routes with pages that share at least one station.

Seasons and times

Winter is December to February, spring March to May, summer June to August, fall September to November. Times are shown in the local time of the place, with the zone named on the page. Station times marked "+1d" are on the day after the train left its first station.

What this site does not do

It does not sell tickets, show fares or availability, or report where a train is now. It does not recommend. It describes what the timetable and the sources say, and leaves the choice to you.

Review and corrections

Written descriptions are drafted by an AI model from the cited sources and reviewed by a person before publication; the reviewer's name is shown once the site's byline is set. A copy of each cited document is stored when it is fetched, monthly, and the route page shows the date of the stored copy. The corrections page lists every factual correction made after publication, with the date. Anyone named on a page can request a review through the contact page. Every quarter, 50 pages are checked at random against their sources and the result is posted here.

For language models and agents

An overview of the site in the llms.txt format, listing every route page and interest page with what each shows, is at /llms.txt; every page links to it with a describedby relation. Every crawler is allowed by robots.txt, which states the site's Content Signals. The metrics page shows which agents and bots fetch the site and in what order.