Dataset brief

The campsite cancellation corpus

A continuously collected record of when sold-out North American campsites become bookable again — and how long they stay that way.

Observation window
2026-07-10 → 2026-09-01 (54 days, ongoing)
Events recorded
164,384 availability transitions
Collection
Four scheduled passes daily, 06:00 / 11:00 / 15:00 / 19:00 PT
Geography
38 US states, Canadian provinces and territories
Held by
Daniel Maynard, sole proprietor, Los Angeles, California

At a glance

164,384availability events recorded
2,788campgrounds observed opening or filling
149,610paired open→fill windows
24hmedian time a freed site stayed bookable

What one record contains

Every transition is timestamped to the pass that observed it and carries the campground identifier, operator, state, coordinates, the affected date windows and the number of nights. Because openings and fillings are both recorded, spells can be paired — which is what makes duration measurable rather than inferred.

A record is an event; a spell is one date window at one campground. The 85,372 opening and 79,012 filling events carry 240,027 window-openings, of which 149,610 (62%) have been observed filling again. Every duration figure here comes from those pairs. The remaining 89,230 were still bookable when the window closed, so this rate is a floor on how many openings eventually fill, not a measure of how many are ever taken.

Coverage

4,966 campgrounds are catalogued; 2,788 produced at least one transition in the window. The rest were either continuously full or continuously available.

Reservation systemScopeCampgrounds
Recreation.govUS federal — NPS, USFS, BLM, USACE3,848
ReserveAmericaOR, MT, IN, TX and others175
GoingToCampMulti-province Canada166
BC ParksBritish Columbia136
Alberta ParksAlberta98
Ontario ParksOntario96
Parks CanadaNational parks83
ReserveOhioOhio67
UseDirect — MinnesotaMinnesota65
ReserveCaliforniaCalifornia state parks57
Florida State ParksFlorida49
UseDirect — MissouriMissouri48

How long a freed site lasts

Gone withinShare of 149,610 spells
5 hours7%
24 hours42%
48 hours66%

The median is 24 hours. Read the five-hour row with the sampling caveat below: it is a floor, and the true short-spell rate is higher than 7%.

How often it is sampled

The corpus is not uniformly sampled in time, and that changes what a duration means. Three cadences run: campgrounds a paying subscriber is watching are polled every five minutes (51 of them at present), a larger watched set every ten, and everything else four times a day at 06:00, 11:00, 15:00 and 19:00 Pacific.

The effect is measurable, not theoretical. Splitting the same spells by which lane observed them:

SampledSpellsGone within 5h
Every 5 minutes5,46815%
Four times a day144,1426%

Watch a campground closely and roughly 15% of its openings turn out to last under five hours. Watch it four times a day and you record 6%, because the rest opened and refilled between two passes and were never seen. Two things drive that gap and this data cannot separate them: coarse polling is blind to short spells, and the closely-watched parks are the sought-after ones, which plausibly refill faster on their own. Either way the headline short-tail is an undercount.

What the corpus shows

Cancellations are not distributed the way people search

Turnover concentrates in campgrounds nobody is refreshing. Over this window William Kent Campground recorded 84 openings covering 191 site-nights and Staircase Campground 69, while celebrated destinations produced almost none. Availability advice keyed to famous parks is advice about the rarest events in the dataset. Ranked cuts of this live at the cancellation report and hardest to book.

A freed site has a measurable half-life

Across 149,610 paired spells the median site stayed bookable 24 hours, 42% were gone within a day and 66% within two. The tail is what matters operationally: 7% lasted under five hours.

Regional cuts hold up

Restricted to 206 campgrounds within 150 miles of Portland, a 27.8-day cut taken on 2026-08-07 found 2,438 openings — roughly 88 a day — with Thursday the peak at 18.6% against Saturday's 11.7%. That weekday pattern survived its own falsification test: excluding the overnight detection bucket that could have manufactured it, Thursday rose rather than fell. (Figures from that dated cut, not recomputed here.)

What this dataset is not

Using it

The portable layer is the derived one: per-campground turnover rates, duration distributions, day-of-week and seasonal patterns, regional cuts. If you are researching camping supply, writing about why campgrounds are hard to book, or want a cut for a particular state or province, write to support@campsage.app.

Availability for US federal sites originates with Recreation.gov. CampSage is independent and not affiliated with any booking provider or land manager.

Every figure recomputed from the production logs on each request, covering 54 days of collection to 2026-09-01. They move as collection continues. · Live map · Feed API
CampSage is an independent service and is not affiliated with, endorsed by, or sponsored by recreation.gov, ReserveCalifornia, ReserveAmerica, GoingToCamp, UseDirect, Parks Canada, or any park agency. All trademarks are the property of their respective owners. · Terms · API