Snow Day Cast

How it works

The chance you see is our estimate of whether a typical public school district (US) or school board (Canada) closes, or cancels buses, on the next school day. This page lists exactly how it is built so you can judge it for yourself.

Where the forecast comes from

For a US ZIP code, every number starts with the National Weather Service. Your ZIP code is matched to its map centre using the US Census Bureau's 2020 ZIP Code Tabulation Area gazetteer (33,144 areas, public domain), and the NWS forecast grid covering that point is fetched from api.weather.gov. From it we read the quantitative forecast: snowfall amount, ice accumulation, probability of precipitation, temperature, wind chill and wind speed, each with the hours it applies to, plus the short text forecast you see at the bottom of the result. Results are refreshed at most every 30 minutes per ZIP.

Some ZIP codes belong to a post office box or a single building and have no map area. For those we use the nearest-numbered ZIP in the same area, and the result says so.

Which day, which hours

The target is the next school day, Monday to Friday. Before 6 am on a weekday, that day is the target; after 6 am it is the next weekday. The site does not know your district's holidays.

Snow and ice are totalled from 3 pm on the day before through 9 am on the target day. Snow that falls between 4 am and 9 am is counted separately as commute snow. The lowest temperature and wind chill in that window, and the highest wind speed, feed the model.

The closing point by state

Districts in Georgia close for an inch; districts in Minnesota do not. Each state has a "closing point": the snow total at which a typical public district has about a 50 percent chance of closing. These are our judgements, not measured statistics.

Closing pointStates
0.5 inFL, GA, AL, MS, LA, SC, HI
0.75 inTX
1 inNC, TN, AR, AZ
1.25 inOK
1.5 inNM, CA, OR, VA, KY
2 inNV, WA, MD, DE, DC
3 inWV
3.5 inNJ, MO, KS
4 inPA, OH, IN, IL, ID, UT, CO, NE (and any state not listed)
4.5 inIA
5.5 inCT, RI
6 inNY, MA, MI, WY
6.5 inWI, SD
7 inNH, VT, ME, MN, ND, MT
10 inAK

Canada

For a Canadian postal code, only the first three characters (the forward sortation area) are used. There are 1,643 of them, located with Statistics Canada's 2021 boundary file, and the forecast is Environment Canada's city page nearest the area's centre, fetched from the government's GeoMet API; the result names the city page used. In most of Canada the usual event is a bus cancellation rather than a school closure, so the Canadian number is the chance of buses cancelled or school closed.

Snow comes from Environment Canada's forecast periods: the night before the target day counts in full and 35 percent of the target day's total counts as falling before 9 am. Amounts are read from the forecast's accumulation field or its text: "Amount 5 to 10 cm" is read as 7.5 cm, "flurries" as 1 cm, "periods of snow" with no figure as 3 cm, "blizzard" or "heavy snow" as 15 cm. "Freezing rain" counts as 2.5 mm of ice, or 6 mm if described as significant, heavy or prolonged; ice pellets or freezing drizzle as about 1 mm. Wind chill is taken from the forecast's own field or from the text.

Closing pointProvinces and territories
5 cmBritish Columbia
10 cmNova Scotia, Prince Edward Island
15 cmOntario, New Brunswick, Newfoundland and Labrador
20 cmAlberta, Saskatchewan, Manitoba, Quebec
25 cmYukon, Northwest Territories, Nunavut

The same curve and adjustments below apply, with the cold thresholds of −20 °F and −35 °F being about −29 °C and −37 °C.

The curve

The closing chance follows an S-shaped curve centred on the closing point: roughly 15 percent at half the closing amount, 50 percent at it, 97 percent at double it, and exactly zero with no snow. The chance of at least a delay uses the same curve centred at 45 percent of the closing point, so delays become likely well before closures do. The closing chance can never be higher than the delay chance.

Adjustments

The final figure is capped at 98 percent. Nothing is ever certain until the district says so.

Verdicts

85 percent and above: very likely a snow day. 65 to 84: likely. 40 to 64: toss-up. 15 to 39: unlikely, maybe a delay. 1 to 14: very unlikely. 0: no snow day.

What it cannot know

Whether your district is public or private, its holidays, its bus routes, how its roads were treated, what neighbouring districts decided, and whether a superintendent is cautious or not. Two districts ten miles apart routinely make opposite calls on the same forecast. Treat the number as a well-informed guess and check your district's own announcements.

Checking the code

The model lives in /source/snow.js, with automated tests for the curve, the state table, ice, cold and timing. Think a threshold is wrong for your state? Tell us, with the storm that proved it.