How the snow day estimate works
Snow Day Call is an independent snow day calculator for parents and students in the United States and Canada. It is not a school district, a board, or a weather service. A percentage here is not a closure, a delay, or a bus cancellation.
Three numbers, not one
Each school morning gets three estimates, each capped at 97. Full closure is the chance school is cancelled. A 2-hour delay is the chance the start moves back because the morning is messy but not bad enough, or not early enough, to cancel. Early dismissal is the chance students go in and come home early, which rises when the snow is timed for late morning or afternoon. In Ontario and Quebec the closure line is labeled a bus cancellation, because that is the call families usually hear.
If tomorrow is a weekend or a widely observed public holiday, including a typical winter break from December 24 through January 1, the ticket leads with the next school morning instead of a zero. A short note still says that tomorrow itself is not a school day. We do not pretend to know your district calendar beyond those common days.
What goes into the closure number
Snow is weighted by when it falls, in local time. Evening before (6 p.m. to midnight) is multiplied by 0.85. Midnight to 7 a.m. is multiplied by 1, because that is when most districts decide. 7 a.m. to 10 a.m. is 0.45. 10 a.m. to 4 p.m. is 0.15. Later snow waits for the next morning. Those weighted inches become snow points.
| Effective inches | Points |
|---|---|
| 0 | 0 |
| 1 | 24 |
| 2 | 42 |
| 4 | 68 |
| 6 | 82 |
| 8 | 90 |
| 12 | 94 |
Snow falling above 2°C is multiplied by 0.6. Above 4°C it is multiplied by 0.35. Freezing rain of 2 mm or more adds 28 points. A lighter glaze adds 16. When ice is the main signal, the page says Ice day. Wind chill at or below −18°C adds cold points, and a cold morning with little snow is labeled a Cold day. Gusts with snow already falling add a smaller wind bump. Leftover snow depth adds a few points the morning after a storm. An active alert adds the highest matching value below, not a stack of every headline.
| Alert | Points |
|---|---|
| Ice Storm Warning | 18 |
| Blizzard Warning | 16 |
| Winter Storm Warning | 12 |
| Lake Effect Snow Warning | 12 |
| Wind Chill Warning | 8 |
| Winter Storm Watch | 7 |
| Winter Weather Advisory | 6 |
| Cold Weather Advisory | 5 |
| Freeze Warning | 3 |
| Snowfall warning | 10 |
Snowfall is a range
The likely amount is the middle of four models — ECMWF, GFS, ICON, and GEM — for the next 24 hours. The low and high are the smallest and largest of those four. “2+ / 4+ / 6+ inches” is how many of the four models reach that total. If they are 2 inches or more apart, the page says they disagree. That count is not a calibrated probability, and it is not an accuracy score.
What changed since yesterday
The change line compares this morning’s snowfall with the snowfall forecast issued about a day earlier for the same hours, from Open-Meteo’s previous-run feed. Temperature and alerts stay on the latest run. If yesterday’s file is missing, the page says so instead of inventing a trend.
How fresh the numbers are
City, ZIP, and postal pages keep the Open-Meteo forecast, National Weather Service alerts, and Environment Canada alerts for about an hour. After that the page asks for a new run. A failed request is kept only a few minutes, and a successful forecast is dropped once it is older than about seventy minutes, so a city page does not sit on yesterday’s storm. The line under the percentage is the capture time of that run (“Updated 18 minutes ago”) plus the next hour when the big global models are usually readable. Those models start at 00, 06, 12, and 18 UTC and typically land about four and a half hours later. The clock is the place’s local time. Check again then if a storm is still forming. The page does not invent a live countdown in the HTML. The browser computes “minutes ago” from the stored capture time.
How accurate is it?
We do not publish a hit rate yet, because we have not tracked a full winter of real closures against our numbers. Any site that quotes a precise accuracy figure without showing how it was measured should be read with care. What we can say is where the estimate is strong and where it is weak.
It is strongest the evening before and early in the morning, when the forecast for the next 12 hours is most reliable and the storm is already on radar. It is weakest two or more days out, for storms on the rain-snow line, and for freezing rain, where a small temperature change flips the outcome. It also cannot know local rules, such as a district that has used up its snow days or one that switches to remote learning. Read our guide on how accurate a snow day calculator really is for the longer answer.
What the model cannot know
It cannot read a district announcement, a bus-consortium post, or a text alert. It cannot tell whether plows have already cleared the route, whether the hill by your stop was treated, or how many snow days the district has used unless you say so in the optional school adjustment. That adjustment is a shift you asked for. It is not a more accurate forecast. The model also does not know if your school has a remote day. Days 3 to 5 of the outlook are marked lower confidence because a forecast that far out moves.
The home photograph is a snowy street with lamps, from Unsplash (photo 1477601263568). The logic lives in src/lib/snow-day.ts and is covered by unit tests. Background snowfall totals on city pages are rounded context, not official climate normals. Forecasts come from Open-Meteo. US alerts come from the National Weather Service. Canadian alerts come from Environment Canada’s public alerts feed when that feed responds. There are 183 city pages.
Read delay versus closure versus early dismissal, when schools announce, and why ice closes school faster than snow.