Using the weather models
Computer models are the forecaster’s most powerful tool, and the easiest to misuse. Learn what Canada’s models are, how to compare them, and how to tell when to trust them.
In this lesson
- What a weather model is, in one paragraph
- Canada’s main models, and when to use each
- Ensembles: why forecasters run the same model many times
- Where models struggle in southern Alberta
- A smart way to read model output
What a model is
A weather model takes millions of observations from stations, balloons, aircraft, ships and satellites, places them on a three-dimensional grid around the planet, and uses the laws of physics to calculate how the atmosphere will change, step by step, hours and days ahead. The grid spacing matters: a model can’t “see” anything smaller than a few grid boxes, like a single valley or a small thunderstorm.
Canada’s main models
Environment and Climate Change Canada runs its models at the Canadian Meteorological Centre in Montreal:
- GDPS (Global): the whole planet at about 15 km, run twice a day, out to 10 days. Use it for the big picture, days 3 to 10.
- RDPS (Regional): North America in more detail. Use it for the next two to three days.
- HRDPS (High Resolution): a 2.5 km grid over Canada, run four times a day, out to 48 hours. Its fine grid handles terrain better, which matters near the mountains.
- Ensembles: Canada also runs ensemble systems, the same model many times with slightly different starting conditions.
Forecasters also look at other countries’ models, like the U.S. GFS and the European model, and compare them all.
Ensembles: measuring uncertainty
Tiny errors in the starting data grow over time. So forecasters run a model 20 or more times, each starting slightly differently.
- If all the runs agree, confidence is high.
- If they spread out in different directions, confidence is low, and the forecast should say so.
- The average of all the runs (the “ensemble mean”) is often better than any single run a few days out.
Where models struggle here
- Chinooks: the exact timing, and how far east the warm air gets, depend on mountain waves the models don’t always capture.
- Shallow Arctic air: very cold, dense air can hug the ground in a layer only a few hundred metres deep. Models often scour it out too fast, forecasting a warm-up that doesn’t come.
- The Arctic front’s position: a few tens of kilometres can mean the difference between -20°C (−4°F) and +5°C (41°F) in Lethbridge.
- Upslope snow amounts: small changes in wind direction change how much snow piles up against the foothills.
- Thunderstorms: models can say storms are likely; they can’t say which farm gets the hail.
Reading models the smart way
- Check how it’s doing now. Compare the model’s forecast for this hour with what the stations actually show. If it’s already wrong, be suspicious.
- Compare several models. If they agree, good. If not, look for why.
- Watch the trend. Is each new run moving the storm north, or making it stronger? The trend often tells you where the final answer lands.
- Know the biases. Keep notes. Does a model usually warm us up too fast after an Arctic outbreak?
- Use the right tool for the time. High resolution for today and tomorrow, global and ensembles beyond that.
Here at West Lethbridge Weather, the forecast is made by WXSIM, which starts from model data and real observations and simulates the weather above our one station. Its record is on the Forecast Scorecard.
Try it yourself
- On Environment Canada’s model charts, look at the forecast for southern Alberta 48 hours from now. Write down what it shows: precipitation, wind, and roughly how warm or cold.
- Look at the same time again tomorrow, on the newer run. Did it change? Which way?
- Two days later, compare both with what our station actually recorded.
Want the raw numbers? Our sister site explains how to download Canada’s models yourself: GRIB weather files on VE6DOK.
🏫 For teachers and parents
- Time: 45 minutes, plus a check two days later.
- Big idea: models are scientific predictions with known limits. A great discussion of uncertainty, error and why scientists compare different methods.
- Ensemble demonstration: have every student drop a paper “snowflake” from the same height and mark where it lands. The spread is your ensemble; the middle is the mean.
- Quick check: “Ten model runs agree, and another ten are spread out. Which forecast would you trust more, and why?”
Key terms
- Numerical weather prediction
- Using computers and physics to calculate future weather.
- Resolution
- The grid spacing of a model. Finer resolution shows smaller features.
- Ensemble
- Many runs of a model with slightly different starting conditions.
- Bias
- A model’s habit of being wrong in the same direction.
Check your understanding
1. Which Canadian model has the finest grid, 2.5 km?
The High Resolution Deterministic Prediction System runs on a 2.5 km grid over Canada.
2. The runs in an ensemble are widely spread out. What does that mean?
A wide spread means small differences grow into big ones: the outcome is uncertain.
3. Why do models often warm southern Alberta too quickly after an Arctic outbreak?
Dense Arctic air can be very shallow, and models tend to scour it out too quickly.
4. What should you check first when looking at a model?
If a model is already wrong about now, be careful trusting it about later.
Sources: Environment and Climate Change Canada, MSC Datamart documentation for the HRDPS, RDPS and GDPS; model forecast products.
Weather School is for learning and enjoyment. It is not official guidance and may contain errors. For watches and warnings, always use Environment Canada (WeatherCAN app or weather.gc.ca).