See it coming.
Watch it change.
The Storm Desk picks storms out of every model run days ahead, then follows each one run by run: the likely range at your places, the ways it could go, the one thing that decides between them, and a briefing where every number says where it came from. When it's over, it checks what fell.
Snow arrives Tuesday morning and peaks in the afternoon at 1–2″ an hourHRRR 18z. Most likely 6–10″ at your placeEPS 12z · 51 members, with a 1 in 6 chance of a foot. The models moved the low 30 miles southeast since the overnight run, which added about an inchECMWF 12z vs 00z.
How much, how sure, and which kind of storm.
Each storm opens on an answer card for your place: the most likely amount, a low end and a high end, and the chances of the amounts that matter. Tabs switch the whole page between the storm's hazards, so a nor'easter, an ice storm and a summer soaker each lead with the right thing.
Not one forecast. The likely ones.
The ensembles' members are sorted into two or three named scenarios, each with its chance, its map and its totals. They're matched from one run to the next, so you can watch a scenario grow or fade.
Scrub through the runs, not just the hours.
Playback moves two ways: through the storm's hours, and back through the runs that forecast it. Earlier runs' tracks fade behind the newest, the ensemble's lows spread into a cone, and arrows show how far the low moved since the last run.
Plain words, every number sourced.
The briefing is built from fixed rules, not free text: when it starts, how hard it comes down, how sure the models are, and what changed. Each number is tied to the run it came from. For big storms, an AI-written draft is available on request.
The desk covers all your places at once, with a region map you can click anywhere: the most likely total, the chances, the gusts and what falls where, and each model's own map beside it.
Snow arrives Tuesday morning and peaks in the afternoon at 1–2″ an hourHRRR 18z. Most likely 6–10″ at your placeEPS 12z · 51 members, with about a 1 in 6 chance of a footEPS 12z. The models moved the low 30 miles southeast since the overnight run, adding about an inchECMWF 12z vs 00z. Sleet may mix in near the coast until middayHRRR 18z.
EPS and ECMWF © ECMWF, CC BY 4.0.
Then it grades the forecast.
After the storm, the desk gathers what was measured: the NOHRSC snowfall analysis, storm reports, CoCoRaHS and co-op gauges, radar rainfall and peak gusts. It writes what happened against what it said, whether the chances verified, and which model got closest at each lead time.
Over a season that becomes a record: which models to trust for which storms, and how early.
During the storm: NowThe snow band came in 25–35 miles south of where the models had it five days out. ECMWF moved it south first; GFS caught up only a day ahead. From three days out, the likely range included what fell, near its top end.
© marks ECMWF data (CC BY 4.0). Measured: NOHRSC snowfall analysis, storm reports, CoCoRaHS and co-op gauges.
Data on this page: © ECMWF (IFS, AIFS and the EPS ensemble, CC BY 4.0); NOAA GFS, GEFS, HRRR and the National Blend of Models; Environment and Climate Change Canada (GEM, GEPS); Deutscher Wetterdienst (ICON). Measured totals: NOHRSC snowfall analysis, NWS storm reports via the Iowa Environmental Mesonet, CoCoRaHS, co-op gauges via RCC-ACIS, MRMS. Pictures are illustrations: the storm, places and numbers are made up.