6 min read

Understanding Forecast Models

A forecast model is a simulation of the atmosphere, run several times a day by a national weather service or research centre. SmartMixin serves the output of many of them, so you can see not just what the forecast says but who is saying it. This article explains what the models are, how they differ, and how to read their specifications.

Seeing which models are available

Not every model reaches every point on Earth. Regional models cover one area — one country, one continent, one overseas territory — so the list you are offered depends on where your station is.

  1. Open a station and click the Forecast tab
  2. The models available at that station are listed as chips, with a Forecast model comparison button beside them
  3. On the dashboard, the chip floating in the bottom-right corner opens the model picker for that station

Reading a model's details

Open the picker from the dashboard chip and click the ⓘ on any model row. The chips on the Forecast tab have no ⓘ of their own.

The detail view has two sections.

Coverage

A model is labelled one of two ways:

  • Global — This model covers the entire planet.
  • Regional — This model only covers a specific region.

Regional models also show a map of the area they cover.

Global models are available everywhere but are usually relatively coarse. Regional models are usually finer, but only inside their domain, and they generally look a shorter distance ahead.

"Regional" says where a model applies, not how fine it is: NAM North America is regional at 32 km, coarser than every global model in the table below.

Specifications

FieldWhat it means
EditorThe organisation that runs the model
Grid sizeHow far apart the model's calculation points are. Smaller means more detail — a 1.3 km model resolves an individual valley or shower; a 25 km model averages it away
HorizonHow far ahead the model looks, in hours
FrequencyWhen a new run starts — usually a list of UTC times, or simply how often for models that re-run continuously (HRRR is hourly)

Above those sections you may also see a Run: line — for example Run: 2026-09-10/12 · 42 minutes ago. That is the model's most recent finished run: the date, the hour it started (in UTC), and how long ago it became available. A few models are fed by a live service rather than by scheduled runs, and show no run line at all.

Tip: Grid size is the single most useful number when choosing a model. A fine grid is what puts a thunderstorm in the right valley.

The models SmartMixin offers

These are the ones you are most likely to meet. The exact list at your station depends on where it is.

ModelRun byCoverageGrid size
GFSNOAAGlobal25 km
ECMWF HRESECMWFGlobal25 km
ARPEGE GlobalMétéo FranceGlobal25 km
GDPS (Global)Environment CanadaGlobal15 km
ARPEGE EuropeMétéo FranceEurope10 km
RDPS (Regional)Environment CanadaNorth America10 km
ICON-EUDWDEurope7 km
NAM CONUSNOAAUnited States5 km
NAM North AmericaNOAANorth America32 km
HRRRNOAAUnited States3 km
ICON-D2DWDGermany and neighbours2.2 km
AROMEMétéo FranceFrance and neighbours1.3 km
MET NorwayMeteorologisk instituttGlobal

Météo France also runs AROME at 2.5 km over the French overseas territories: AROME Antilles, AROME Guyane, AROME Indien, AROME Polynésie and AROME Nouvelle Calédonie.

Grid sizes are shown in the distance unit you chose in your settings — see Setting Your Unit Preferences.

SmartMixin multi-model

SmartMixin multi-model is not a weather centre's model but our own blend. For each hour at your location it picks the best model available — typically a high-resolution regional model where one exists, falling back to a global model elsewhere — and fills any gaps from neighbouring sources.

The result is a single seamless forecast, anywhere in the world, refreshed whenever any of the underlying models updates. If you want one good answer rather than a choice to make, this is it. It is also what the SmartMixin blend option gives you in your preferences — see Choosing Which Forecast Models You Get.

Which models you can use

GFS is available on every account. Every other model, including the SmartMixin blend, is part of Premium.

A free account is served GFS everywhere, whatever else it can see in the list. See Premium Features for what a subscription includes, and How to Subscribe to start one.

Why two models disagree

Different models start from slightly different pictures of the current atmosphere and simulate it with different physics on different grids. Small differences at the start grow, which is why forecasts for the same afternoon diverge more the further ahead you look.

There is no single best model. A high-resolution regional model is usually better for tomorrow's showers; a global model is what you have for next week. Comparing them is often more informative than picking one — see Comparing Forecast Models Side by Side.

Was this article helpful?


Still need help?

If this article didn't answer your question, I'm happy to help you directly.