What Weather Model Does Canada Use? The GEM Model Explained

Canada's national forecasts are built on GEM — the Global Environmental Multiscale model — developed and run by Environment and Climate Change Canada (ECCC). GEM powers a family of systems: the global GDPS at about 15 km, the regional RDPS at about 10 km over North America, and the high-resolution HRDPS at about 2.5 km for local detail. Together they underpin nearly everything you read on weather.gc.ca.

What is the GEM model?

GEM stands for the Global Environmental Multiscale model. It is a numerical weather prediction (NWP) model — a physics-based simulation of the atmosphere that solves the equations governing air motion, heat, moisture and pressure on a three-dimensional grid, then steps them forward in time to produce a forecast. GEM was developed in-house by ECCC's meteorological research and operations teams and has been Canada's operational workhorse for decades, continuously upgraded.

The word "multiscale" is the point: the same core model can be configured to run over the whole planet or zoomed in over a small region at high resolution. That flexibility is why one model family can serve everything from a ten-day national outlook to a two-day, street-scale snowfall forecast.

The GEM family: GDPS, RDPS and HRDPS

ECCC does not run a single forecast — it runs a tiered system, each tuned to a different range and resolution.

The logic is simple. Farther out in time, small-scale detail is unpredictable anyway, so the global GDPS at coarser resolution is the right tool. Closer in, the high-resolution HRDPS can resolve terrain, coastlines and lake effects — the very features that drive Canadian weather — giving sharper, more realistic short-range forecasts.

What about ensembles?

Alongside the single "deterministic" runs, ECCC runs ensemble systems: the Global Ensemble Prediction System (GEPS) and a regional counterpart. Instead of one forecast, an ensemble runs the model many times with slightly different starting conditions and physics. The spread between those runs is a direct measure of uncertainty. When the ensemble members agree, confidence is high; when they scatter, the forecast is genuinely uncertain. This is the scientific basis for any honest "forecast confidence" score.

How does GEM become the forecast on weather.gc.ca?

Raw model output is not the public forecast. ECCC meteorologists take GEM's guidance — often alongside ECMWF, the American GFS and other international models — and apply human judgement, statistical post-processing and local knowledge before issuing the forecast and any warnings. So when you read Environment Canada, you are getting GEM filtered through experienced forecasters, not the bare numbers. This blend is a major reason the official forecast often beats any single raw model.

GEM also feeds Canada's apparent-temperature indices: the summer Humidex, a made-in-Canada measure of how hot it feels with humidity, and the winter wind chill index, the national standard for how cold the wind makes it feel. Both appear in ECCC forecasts and warnings.

Is GEM any good compared with other models?

GEM is a credible, respected global model, but it is worth being honest: in most independent verification, the European ECMWF model is the world's most skilful global system, with the American GFS and Germany's ICON also strong. GEM is competitive rather than the outright leader at the global scale. Its standout strength is the high-resolution HRDPS over Canada, which is purpose-built for Canadian terrain and lake effects that global models smooth away. For local Canadian weather, that home-field advantage matters.

How often does GEM run, and what does it take?

Running GEM is a serious computational undertaking. The global GDPS is typically produced twice a day, while the shorter-range regional RDPS and high-resolution HRDPS run more frequently through the day to fold in the newest observations. Each run begins with data assimilation — ECCC ingests millions of measurements from satellites, weather balloons, aircraft, buoys, radar and surface stations to build the best possible snapshot of the current atmosphere — before the model steps that state forward hour by hour on one of Canada's supercomputers.

The higher the resolution, the more computing each hour of forecast demands, which is exactly why HRDPS is limited to short ranges and a Canadian domain rather than the whole globe. Understanding this trade-off explains a lot about forecasts: you can have global coverage and long range at coarse resolution (GDPS), or fine local detail over a short window (HRDPS), but not both at once from a single run. ECCC's tiered design squeezes the most useful forecast out of finite computing power.

Why does the model behind your forecast matter?

Because every model has different strengths, and knowing which one you are reading tells you how much to trust it. A ten-day outlook is really the global GDPS or ECMWF; a sharp two-day snowfall map is really HRDPS. This is exactly what checkweather.io surfaces: it lines GEM up against ECMWF, GFS and ICON, converts their agreement into a plain-English forecast confidence score, and keeps an accuracy scoreboard of how each model actually performed. No black boxes — you see whether today's GEM forecast is a lone outlier or backed by the pack.

The bottom line

Canada uses GEM — the Global Environmental Multiscale model — run by Environment and Climate Change Canada in three tiers: the global GDPS, the regional RDPS, and the high-resolution HRDPS. Global models like ECMWF may edge it at long range, but GEM's high-resolution Canadian configuration is its real strength, and it is the engine behind virtually every official forecast and warning you read at weather.gc.ca.