Firn

Six steps

How a Firn forecast is made

1. 01

   Several weather models, not one

   In the US, Firn starts from NOAA’s models (the high-resolution HRRR for the next day or two, then the GFS). Everywhere else it uses the best model for that place. It also reads the European (ECMWF), German (ICON) and Canadian (GEM) models for the same days, so it knows when they agree and when they don’t.

2. 02

   Snow at the height you ski

   A forecast for the town at the bottom of the hill can be rain while it snows up top. Firn forecasts at a point on the mountain: mid-mountain for the headline number, plus the base and the summit where we know the heights.

3. 03

   Snow, not just water

   Models forecast the water that falls. How deep that gets depends on the cold: an inch of water can be 5 inches of wet spring snow or 18 inches of cold, light powder. Firn turns water into snow hour by hour, from the temperature in the hour it falls, instead of one fixed ratio.

4. 04

   One forecast run for all 15 days

   Every day of the 15-day outlook comes from the same forecast run, so a storm that shifts a day between runs is never counted twice or dropped.

5. 05

   A check against the other models

   Sometimes one model puts out a storm far bigger than any other model expects. From the day after tomorrow on, when the main model is well above every other model, Firn shows the blend of the models for that day instead. A big day the models agree on is left alone: Firn has no cap on how much snow it will forecast.

6. 06

   How sure it is

   Where the models disagree, Firn shows a range and a confidence label for the next few days, so you can tell a sure thing from a maybe.

This winter Firn starts learning from that record. Three things are being tested now, and each can go live region by region only after it beats the current forecast on real storms it hadn’t seen, and we switch it on:

1. 1

   Trusting the better models more

   If one model keeps overdoing storms in the Sierra and another nails Utah, Firn leans on whichever has been closest in each region.

2. 2

   Learning each mountain

   Some resorts sit where storms stall; others get blown dry. When the record shows Firn running high or low at a resort, it adjusts, a little at a time, and only as the evidence builds.

3. 3

   Learning the snow itself

   Sierra snow is often wet and heavy; Utah and Colorado snow is often light. Firn checks the snow it predicts against the snow that falls in each region.

1. Runs every day behind the scenes
2. Compared with the live forecast on the same days
3. Beats it on real storms it hadn’t seen
4. We switch it on, region by region

In testing means these run every day behind the scenes and are compared with the live forecast on the same days. You see the live forecast until a step has proven itself. We’ll update this page as each one goes live.

What Firn can’t tell you

• Forecasts are estimates. Mountains make their own weather, and the far days change a lot.

• Firn doesn’t rate avalanche danger or say whether terrain is safe. Read your local avalanche center before heading out of bounds.

• Wind can move a lot of snow after it falls; what you find on a given run can differ from the number.

Questions

The data

Weather data

Open-Meteo.com (CC BY 4.0)

Weather models

NOAA, ECMWF, DWD and Environment Canada

Snow stations

NRCS SNOTEL

Snow reports

Each resort’s own

The Stoke score that sits on top of Firn is explained on How scores work.