How Forecasts Are Made

Step one: take the atmosphere’s pulse

Every forecast starts with a snapshot of what the atmosphere is doing right now. A forecast can only be as good as that starting picture, so the world’s weather services gather an enormous amount of data, around the clock:

• Surface stations at airports and automatic sites, plus ships and ocean buoys, report temperature, pressure, wind and humidity.
• Weather balloons are launched at hundreds of sites around the world at the same two times every day, measuring the air from the ground to well above the jet stream.
• Satellites supply by far the largest share of data, measuring temperature, moisture and winds across oceans and remote areas where nobody lives.
• Radar shows precipitation, and commercial aircraft automatically report temperature and wind as they fly.

Step two: the computer model

A weather model, or numerical weather prediction, chops the atmosphere into a three-dimensional grid of boxes, like a giant stack of ice-cube trays wrapped around the Earth. Each box holds numbers for temperature, pressure, wind and moisture. The computer then applies the laws of physics (how air moves, how heat flows, how water evaporates and condenses) to work out what each box will look like a few minutes later. Then it does it again, and again, millions of times, stepping days into the future.

First, though, the model needs its starting picture. A step called data assimilation blends all those new observations with the previous forecast to fill in the gaps, much like a detective combining witness statements into one best account of what happened.

The big names

• GDPS: Environment Canada’s global model, covering the whole planet.
• RDPS: Environment Canada’s regional model, with finer detail over North America.
• HRDPS: Environment Canada’s high-resolution model, with boxes just a few kilometres across, useful for thunderstorms and mountain weather.
• ECMWF: the European Centre’s global model, widely regarded as one of the most skilful in the world.
• GFS: the United States’ global model, run by NOAA.

Step three: ensembles and confidence

The atmosphere is chaotic. A tiny error in the starting picture, a slightly wrong wind here or temperature there, can grow over several days into a completely different forecast. This is the famous “butterfly effect,” first described by meteorologist Edward Lorenz.

Since the starting picture can never be perfect, forecasters run the model many times, each with slightly different starting conditions. This bundle of runs is called an ensemble. When the runs agree, forecasters can be confident. When they fan out in different directions, the forecast is uncertain, and an honest forecast says so.

This is also why forecasts for a week or more ahead are given in broader terms. Day-to-day detail becomes very hard to predict beyond about a week to ten days, though the general pattern (warmer or colder than normal, for example) can sometimes be seen further out.

Step four: the human forecaster

Environment Canada’s meteorologists, at storm prediction centres such as the one covering the Prairies, do not just copy the computer. They compare models, know each model’s habits (some tend to mishandle the shallow Arctic air that hugs the ground east of the Rockies, for example), watch radar and satellite minute by minute, and decide when to issue watches and warnings.

Many phone apps, on the other hand, show raw model output with no human checking it. That is why two apps can disagree, and why an app may show a precise number when the honest answer is “it depends where the Arctic front stalls.”

What “40 percent chance of rain” really means

Probability of precipitation, or POP, is one of the most misunderstood numbers in weather. In Environment Canada forecasts, it is the chance that measurable precipitation will fall at any given spot in the forecast region during the forecast period.

So a 40 percent chance of showers does not mean it will rain on 40 percent of the city, or for 40 percent of the day. It means that on days like this one, your spot would get measurable rain about four times out of ten. A forecaster can reach 40 percent in different ways: very sure that scattered showers will hit some places but not others, or only moderately sure that a band of rain will reach the whole area at all.

The West Lethbridge forecast: WXSIM

The 5-day forecast on West Lethbridge Weather comes from WXSIM, a local forecasting program. Rather than modelling the whole planet, WXSIM simulates the column of air above one location. It starts from the station’s own live readings and is guided by data from the large models, then works through the physics of daytime heating, overnight cooling, cloud and wind for that single spot.

Think of the big models as a regional road map and WXSIM as a detailed street map of one neighbourhood. Both are useful, and they complement each other. For watches and warnings, though, always rely on Environment Canada.

Why forecasts change

New observations arrive constantly, and the major models run several times a day. When tomorrow’s forecast shifts, it is not that yesterday’s forecaster got it “wrong”; the picture simply got sharper with fresh information, much like a GPS updating your arrival time as traffic changes.