Guide to weather forecast models

Guide to weather forecast models

Learn about weather forecast models used in

What is weather forecast model?

The weather depends on lots of different factors: wind speed, cloudiness, pressure, humidity, etc.

All of them can be measured and represented in meters per second, millimeters of mercury, etc. A mathematical weather model is a combination of all these factors. Such models tend to be very complex, powerful computers calculate them.

Usually, weather models are provided by national weather services, but sometimes they can also be produced by groups of enthusiasts. In this section, we want to introduce you to our models, from which we get forecasts. has two types of forecast predictive models: global — on the map, and local — for a spot:

Model Region Grid, km Forecast depth, days Step, hours Daily updates, time/day
ECMWF Global 14 10 3 2
ECMWF-ENS Global 27 6 1 1
GFS27 Global 27 10 1 4
GFS+ Global 27 10 1 4
ICON 13 Global 13 5.1 1 4
ICON 7 Europe 7 4.7  1 4
NAM North America 12 2.5 1 2
HRRR North America 3 2 1 24
WRF8 Europe 8 3 3 1
Open Skiron Mediterranean 12 5 3 1
Open WRF Mediterranean 12 4 3 4
AROME France 1.25 1.7 1 4


Other data in the table:

  • The grind or resolution is the distance between two points. Bigger resolutions of 50 to 10 km in size are usually deployed in relatively flat terrains, while mountain ranges require the nodes to be a lot closer to each other, usually 5, 2, or 1 km.
  • The depth of the weather forecast is the number of hours or days for which the forecast is made. As a rule, the lower the depth, the more accurate the forecast.
  • The step shows how long each next change in the atmosphere will be recorded. After each step, the supercomputer calculates the rate of weather change and makes a forecast one step into the future. These measurements are repeated until the solution reaches the desired forecast time, i.e. the meaning of its depth. Time steps for global models are about tens of minutes, and for regional models — from one to four minutes. The larger the step, the faster the model is computed.
  • The update frequency of the forecast is the regular time interval after which new forecast data is received from the supercomputers. For example, the forecast can be updated once a day, 2 times a day, or every 6 hours.

ECMWF — Global

A global model by the European Centre for Medium-Range Weather Forecasts. Considered the best for precipitation and cloudiness. Resolution — 14 km. Updates every 12 hours, providing forecasts for 10 days.

ECMWF-ENS (Ensemble Model) — Global

A global model by the European Centre for Medium-Range Weather Forecasts. It is precise data for more than a 3-day forecast. Resolution is 27 km. Updates one a day, forecast for 10 days.

Learn more about ut in an article "How to Use Ensemble Models for Weather Forecast".

GFS27 — Global

The US Global Forecast System, originally created for aviation. Updated 4 times a day, providing forecasts for 10 days.

GFS+ — Global

While the GFS27 interpolates data for every dot within a 27 km x 27 km square, the GFS+ always shows only the maximum result in each square. Updates 4 times a day, forecasting for 10 days.

ICON13 — Global, ICON7 — Europe

A global model by Deutscher Wetterdienst (German Meteorological Service).

For Europe, higher resolution (7 km) is available for the ICON model, making it even more precise than the ECMWF. The ICON with a 13 km resolution is the global version available worldwide.

  • ICON7 — updates 4 times a day, every 6 hours. Provides forecasts for the next 5 days.
  • ICON13 — updates 8 times a day, every 3 hours. Provides forecasts for the next 5 days.

NAM — North America

Developed specifically for North America, provides the most precise forecast for the region. Resolution — 12 km. The forecast is updated 2 times a day. Provides forecasts for the next 61 hours.

HRRR — North America

HRRR is High Resolution Rapid Refresh. The model was developed by the National Oceanic and Atmospheric Administration (NOAA) of the US.

It provides a great forecast for the continental US with real-time 3 km (1.9 m) grid resolution, hourly updates and radars' data — the unique advantage of HRRR.

It is assimilated every 15 min over a 1 h period. It gives short-range weather forecasts (forecast depth is 18 hours) with pretty good accuracy. So the forecast from HRRR is definitely worth looking at.

In HRRR you find temperature (2 m), gusts (10 m), wind speed and direction (10 m), total cloud cover on your spot forecast screen.

WRF8 — Europe

The WRF is a model for weather forecast calculation developed by several scientific laboratories in different countries.

Using it, calculates a hyperlocal forecast for Europe and the Mediterranean. Updates once a day, giving forecasts for the next 3 days.

OpenSkiron — Mediterranean

A local model developed at the University of Athens. Considered one of the best for the Mediterranean. Resolution — 12 km. Updated once a day, providing forecasts for the next 5 days.

OpenWRF — Mediterranean

A local model for the Mediterranean. Based on the WRF technology, adapted by the University of Athens and a group of enthusiasts. Resolution — 12 km. Updated 4 times a day, giving forecasts for the next 4 days.

AROME — France and the surrounding territories

A local model developed by Meteo France (French meteorological service). Covers France and some adjacent regions. Resolution — 1.25 km. It is updated 4 times a day and provides forecasts for the next 42 hours.


Read about the advantages of each model and lots of other useful info in the guide on forecast models in blog.

Cover photo © markusspiske / Unsplash


Share:   WINDY.APP Facebook   WINDY.APP Twitter
This website uses cookies to improve your experience. If you continue to browse this site, you are agreeing to our Privacy Policy and Terms of Use.