Forecast Flakes is the winter room of the Forecast network — the place snow, ice and freezing rain get their own treatment instead of being one card on a state portal. It covers the United States and Canada, and it is built to answer the four questions people actually have in a winter storm: when do the first flakes fall, how does the event unfold, how much has to be plowed, and how does this compare to a normal winter.

The design

Same Interstate sign system as the rest of the network, turned down to a winter night. The paper goes from reflective buff to packed snow and the ink to midnight, and blue leads instead of green, because winter weather is a services problem — what closes, what needs plowing, when it starts. The accent is the six-inch band of the National Weather Service snowfall scale, so the site's accent colour and its map legend are literally the same blue.

What is wired in today

• NWS NDFD — snowfall amount, ice accretion, snow level, temperature and wind on the gridded forecast, in six-hour blocks out about seven days. This is what the engine reads.
• NWS NBM — the National Blend, carrying temperature, wind and precipitation further out. Its probabilistic snow fields are not in this store yet.
• Environment and Climate Change Canada — current conditions and public forecast text for 844 city points, captured when the site is built.
• ECMWF ENS — a true 50-member ensemble read straight off the weatherzarr Zarr archive, driving the two-week outlook: the temperature plume, the share of members below freezing, and the share below freezing while precipitation accumulates.
• NOAA nClimGrid — observed monthly temperature and precipitation against 1996–2025 normals, degree days, and threshold-day counts back to 1951. This drives the season replay.

What is not, and why the site says so

A winter site is only as honest as its gaps. Rather than fill a range with a model that is not actually running, the status board on the front page marks each one:

• HRRR and RRFS — both publish to the weatherzarr archive and are current, but carry only a handful of fields between them: HRRR has 2-metre temperature, composite reflectivity and 10-metre wind; RRFS has temperature, dewpoint and precipitable water. Neither publishes a snow variable, so the convective-scale snow-band tier is connected but not yet useful.
• GEFS and ECMWF AIFS — both publishing and current. GEFS is an ensemble mean rather than members; AIFS carries temperature and total precipitation. Again, no snow variable.
• MJO and ENSO — the subseasonal drivers behind a seasonal outlook. Not on this host.
• Observed snowfall — there is no snowfall observation, normal, or season-to-date dataset here at all, and no model in the archive publishes a snow variable either. That is why the replay is scored on temperature, precipitation and degree days rather than on snow totals, and why forecast verification — how much fell against how much was called for — cannot be shown yet.

The replay

Because there is no winter weather in August, the front page leads with the 2025–26 season replay: November through March observed against normal, month by month and state by state, with each state's December–February mean ranked against every winter back to 1951. Nationally that winter finished 74th of 75 counting from the coldest — the second-warmest in the record.

The network

Forecast Flakes is a keystone of the Forecast network and will branch into the state and metro sites — Forecast Indiana, Ohio, Texas and the rest — alongside weather.city and squall.cloud.

forecastflakes.com · © 2026