How We Calculate National Park Crowd Scores (Our Method)

The Short Answer

Every crowd score you see on this site comes from three years of real NPS visitation data, averaged by month, then divided by that same park's own busiest month to get a percent-of-peak figure. That percent-of-peak is converted into a 1–10 score and labeled Quiet, Moderate, Busy, or Peak. Nothing is guessed or estimated — every number traces back to a figure the National Park Service actually published.

• Scores are built from NPS Visitor Use Statistics, averaged over the last three full calendar years to smooth out one-off spikes like a wildfire closure or an anniversary surge.
• Each month is compared only against that same park's own busiest month — never against other parks.
• The percent-of-peak figure is divided by ten, rounded up, and clamped to a 1–10 range, then bucketed into 1–3 Quiet, 4–6 Moderate, 7–8 Busy, and 9–10 Peak.
• The method only understands months, not days — it cannot tell you that a specific Tuesday is quiet or that a three-day weekend will spike.
• The weather shown next to each score is a 30-year NOAA climate normal, not a forecast — see how that differs on the climate normals vs. forecast guide.

Step 1: Three Years of NPS Visitation Data, Averaged

The starting point for every score is "recreation visits" — the National Park Service's own count of visits made for recreational purposes, which specifically excludes administrative and other non-recreational traffic. The NPS publishes this figure monthly for every park unit through its Visitor Use Statistics program, and it's the same dataset researchers and park planners use, not a proprietary count. You can see the raw sourcing behind every figure on this site at our data sources page.

Rather than using a single year, Park Crowd Calendar averages the last three full calendar years for each park and month. A single year can be misleading — a government shutdown, a wildfire closure, a road washout, or a pandemic-era visitation collapse can make one October look nothing like a typical October. Averaging three years smooths those anomalies out without erasing genuine seasonal patterns, so the resulting number reflects what a typical month at that park actually looks like rather than what happened to happen in one unusual year.

Step 2: Why Each Park Is Normalized Against Its Own Peak

Raw visitor counts are almost useless for trip planning once you start comparing parks. A park like Gates of the Arctic, which has no roads and receives only a few thousand visits a year, would look permanently empty next to Grand Canyon, which draws millions. But that comparison answers the wrong question. Someone planning a trip to Gates of the Arctic doesn't care how it stacks up against Grand Canyon — they want to know whether June or September is the relatively busier month for that specific park.

That's why every score is normalized against the park's own busiest month rather than against any other park. Each month's three-year average is divided by that same park's highest three-year average month, producing a percent-of-peak figure — a number that answers "is this a relatively quiet time for this park," which is the question a visitor is actually asking.

A worked example makes this concrete. Pull up the month-by-month crowd calendar for Zion and its two quietest months, January and February, both score 5 out of 10 — which lands in the Moderate bucket, not Quiet. May through September, by contrast, all rate a full 10, Peak. That 5 is a real, honest number: even Zion's calendar floor is a solidly average month by national park standards, because Zion is one of the most consistently visited parks in the system. Normalizing per-park doesn't manufacture an artificially quiet month where none exists — it just tells you, accurately, that this is as quiet as Zion gets.

Step 3: Converting Percent-of-Peak Into a 1–10 Score

Once a month's percent-of-peak figure exists, turning it into a score people can scan at a glance takes four steps:

1. Take the percent-of-peak figure (for example, a month running at 43% of the park's busiest month) and divide it by ten.

2. Round the result up to the nearest whole number.

3. Clamp the number so it never falls below 1 or rises above 10, even at the extremes.

4. Assign a label based on the final score: 1–3 Quiet, 4–6 Moderate, 7–8 Busy, and 9–10 Peak.

Rounding up rather than down is a deliberate choice. A month sitting at 31% of peak becomes a 4 (Moderate) rather than a 3 (Quiet), which errs toward not understating how busy a place will feel. The goal of the whole system is to be a reliable planning signal, not to flatter any particular month.

What Crowd Scores Don't Capture (and Where to Get the Rest)

Being upfront about the limits of this method matters as much as explaining how it works. Three things are worth stating plainly.

First, the source data is monthly, not daily. NPS Visitor Use Statistics don't break visitation down by day of the week or by specific date, so no crowd score here can tell you that a Tuesday in July is quieter than a Saturday, or that a holiday weekend inside an otherwise quiet month will spike well above that month's average. Think of the score as describing the season, not the specific day.

Second, the weather figures shown alongside crowd scores are NOAA 30-year climate normals — long-run averages of what a given month is typically like — not a forecast for any upcoming trip. Where it's available, a separate live 7-day forecast is shown for more immediate planning; the two numbers answer different questions, and mixing them up is a common mistake. The full breakdown of that distinction lives in a separate guide to climate normals versus forecasts.

Third, "visitation" specifically means NPS recreation visits — a defined, published statistic that excludes administrative and non-recreational traffic. It's a consistent yardstick across every park, but it isn't a measure of trail congestion, parking lot fullness, or shuttle wait times on any single day.

This process is maintained and reviewed by Sharon Ben Moshe, Park Crowd Calendar's founder — it isn't an auto-generated black box. For the complete technical breakdown of the pipeline, including exactly how the underlying database is structured, see the full methodology page.

Frequently Asked Questions

What does a crowd score of 7 actually mean?

A 7 falls in the Busy bucket (7–8), meaning that month's average visitation sits at roughly 61–70% of that park's own busiest month over the last three years. It's a relatively high-traffic month for that specific park, though not its absolute peak.

Does Park Crowd Calendar use real-time or live crowd data?

No. Crowd scores are built from historical, monthly NPS Visitor Use Statistics averaged over three years — they describe a typical month, not what's happening in the park right now. For day-of conditions, check the park's own alerts page and the live 7-day weather forecast shown on each park's month page.

Why do two different parks with the same score of 5 not feel equally crowded in person?

Because every score is relative to that park's own peak, not to other parks. A 5 at Yellowstone, which draws millions of annual visits, represents far more actual people on the ground than a 5 at a small, lightly visited park — the score tells you where a month sits on that park's own calendar, not its absolute traffic compared to anywhere else.

How often is the underlying visitation data updated?

The National Park Service publishes new Visitor Use Statistics on an ongoing basis, and Park Crowd Calendar's three-year rolling window is recalculated as new NPS data becomes available, so older, less representative years roll out of the average over time.

Is this the same method used on the methodology page?

Yes — this article is a plain-language walkthrough of the same four-step process described in full technical detail on the methodology page, which also documents how the site sources and structures its underlying data, covered further on the data sources page.

Why average three years instead of just using the most recent year?

A single year can be distorted by one-off events — closures, construction, unusual weather, or a post-pandemic rebound — that don't represent a typical year. Averaging three full calendar years keeps the seasonal pattern intact while smoothing out those anomalies.