How PeakCam Measures Snow
Most snow reports online repeat what a resort's marketing department publishes. PeakCam reads from government sensors where they exist, uses a weather model where they don't, and tells you which is which. This page documents each step, including the parts that are imperfect, so you can judge the numbers yourself.
The sensors: NRCS SNOTEL and SCAN
SNOTEL (“snow telemetry”) is a network of over 900 automated stations run by the USDA's Natural Resources Conservation Service across the western United States, with sister SCAN stations elsewhere. Each station weighs the snowpack sitting on a fluid-filled pillow to derive snow water equivalent (SWE) — how much water the snow would melt into — and measures depth with an ultrasonic sensor, plus temperature and precipitation. Where a resort has an NRCS station assigned, PeakCam pulls that station through the AWDB API every six hours. Resorts without one are covered in the “limits” section below.
A sensor is honest but literal: it reports one fixed patch of ground at one elevation, all winter. When PeakCam and a resort disagree, the usual reasons are location (one pillow versus the most flattering slope on the mountain), snowmaking the sensor cannot see, and resorts quoting storm totals rather than a change in settled depth.
Quality control
Raw telemetry misbehaves: sensors drift, ice fools the ultrasonic reading, and a snow-laden branch falling on the pillow can report as a ten-inch storm. PeakCam runs range checks (values a mountain cannot physically produce) and spike checks (jumps too large for the elapsed time). A reading that fails is replaced with the previous valid value and flagged in the internal record; a silent station carries its last good reading forward. One honest caveat: the public page does not yet label carried-forward values — a depth that has not moved in days deserves your suspicion.
Thirty-year normals
A 60-inch base means one thing in Vermont and another in Utah. For context, PeakCam compares each station's current SWE against that same station's 1991–2020 median for the exact day of the water year (which begins October 1). That is the “% of normal” figure — 100% means the snowpack sits at its own historical median, not a regional average. Early and late in the season the median approaches zero and the percentage becomes meaningless, so PeakCam suppresses it rather than print “400% of nothing.”
Ratings, trends, and outlook
The condition rating is a fixed function of the measured data, not editorial judgment: Great is 6 inches of new snow in 24 hours or 12 in 48; Good is 2 inches new, or a 24-inch base at 100% of normal or better; Fair is a 20-inch base at 70% of normal; otherwise Poor. Temperature does not move the rating. The 7-day trend arrow tracks whether the station's SWE is building or shrinking. The outlook label combines that trend with the next 48 hours of forecast snowfall and high temperature — National Weather Service forecasts in the U.S., Open-Meteo elsewhere.
On-mountain reports
Skiers standing on the snow see what a pillow cannot: wind effect, crust, how the groomers held up. Signed-in users can file condition reports. Once a resort has at least two unflagged reports within 24 hours, they are blended into the rating at 30% weight against the sensor's 70%, capped at moving it one tier. A lone report never moves the number, and no report ever changes a measured depth.
Freshness
Sensor syncs run every six hours. Pages rebuild on a rolling hourly window, so what you see can be up to an hour older than the database; forecasts are fetched when a page regenerates, not on every visit. Each resort page's structured data carries the timestamp of its latest reading. A dead feed currently pages the operators — it does not yet put a warning banner on the resort page.
What PeakCam does not measure
Resorts outside the NRCS network — Canada, most of the eastern U.S., and the South American mountains — have no pillow telemetry. Their depth and new-snow figures are weather-model estimates (Open-Meteo) updated every six hours, and those pages show no percent-of-normal figure because there is no station history to compare against. Lift and trail counts, where shown, are resort-reported, not sensed. Webcams are the resorts' own cameras: some live video, some stills on a refresh timer, some links out to the resort's page.