Snow Day

How it works

How Snow Day predicts a snow day

Snow Day estimates the probability that your school board cancels buses or closes schools on the next school morning, using tonight's hourly forecast, official alerts, and a model of how boards actually make the call. This page explains every piece, so you can decide how much to trust the number.

Which morning?

Before 9 AM on a weekday, the app is talking about this morning, because the decision is being made (or has just been made). From 9 AM on, it's the next school day. Friday evening and all weekend, it looks at Monday.

The window that matters

Boards and bus consortia decide between 4:30 and 6 AM. What they look at is the road at that moment and the next three hours: is it still snowing, how hard, is there ice, can the route supervisors see. So Snow Day measures the hours from 6 PM the evening before through 9 AM, and it weighs them differently:

Fifteen centimetres that ends at midnight often means a normal day. Eight centimetres between 5 and 8 AM often does not. Snow Day scores those two nights the way a bus operator would, not by the total.

Corrected snowfall

Forecast models report snowfall at a fixed ratio of ten millimetres of water to seven centimetres of snow. Real snow at −12 °C is far fluffier than snow at 0 °C. Snow Day converts the forecast back to water and applies a temperature-dependent ratio, from 8:1 near freezing to 20:1 below −15 °C. On a cold night the raw forecast can understate the snow on your driveway by half; the app corrects for it.

The drivers

Five things can cause a snow day on their own. The strongest one sets the score.

The conditions

Two modifiers scale the driver between 70% and 130% of its value: the feels-like temperature in the morning, and blowing snow, which needs both wind (gusts of 40 km/h and up) and loose snow on the ground. Reported visibility under 800 metres for two hours of the run is treated as conclusive.

Alerts

Snow Day reads active alerts from Environment Canada and the US National Weather Service and keeps only winter products: Blizzard, Winter Storm, Ice Storm, Freezing Rain, Snow Squall, Snowfall, Winter Weather Advisory, and the cold products. Each sets a floor for the score (a warning counts fully, a watch counts half) and only while it is in force during the decision window. A frost advisory in October or a heat warning in July does nothing.

Your board

Cancellation culture is local. Loudoun County cancels on a forecast of five centimetres; Buffalo shrugs at twenty. The snow tolerance setting moves the curve: "Cancels easily" makes four centimetres a coin flip, "Typical" twelve, "Rarely cancels" eighteen. A cold policy adds extreme cold as a driver. If you like tinkering, every factor's weight is adjustable too.

Confidence

Snow Day fetches two independent weather models (in North America, the American GFS and the Canadian GEM) and blends them. Confidence is how much they agree about the morning that matters: high when their snow totals are within three centimetres and on the same side of your tolerance, low when they disagree about whether it's snow or freezing rain.

What it can't know

Whether the plows got ahead of it. What the route supervisor saw at 5 AM. Whether the board has a policy about the third snow day in a week. Snow Day is a forecast-based estimate, and a good one, but the final call always belongs to your school board. Use it the way you'd use a rain forecast: to decide whether to set the alarm.

Data sources

Hourly forecasts from Open-Meteo (GFS and GEM in North America). Alerts from Environment and Climate Change Canada and the US National Weather Service. Your location is sent to those services to fetch the forecast and to nobody else; read the privacy policy.