Weather is an extremely complicated phenomenon made up of hundreds of different factors and variables, and simply tracking it can be a rather daunting task. Making a correct forecast is a totally different beast. These days, weather forecasts are made by supercomputers solving insanely difficult and complex mathematical equations. Those equations can be drastically different depending on the region, its terrain, its man-made objects and many other inputs.
Moreover, forecasting requires a vast array of meteorological data, collected by satellites, observation systems, automatic and manned stations, aircraft, ships, weather balloons and so on. That is why there are dozens of forecast models currently in existence, each tailored for a specific task.
There are two main types of forecast models: global ones, covering the entire planet, and local ones, covering specific areas, such as continents, countries, mountain ranges and so on. Both global and local models also vary in their resolutions, which is the distance between two grid 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.
Resolution – various (11km)
The ECMWF is a European global forecast seamless model and it is widely regarded as the best and most reliable model currently in existence. It uses a concept called 4D, which is an assimilation that allows the model to be constantly updated as new satellite or other input data becomes available. It is a well-known fact that the ECMWF was the only model that accurately predicted where Hurricane Sandy was moving.
Mean sea level pressure | ECMWF
ECMWF layer on the map in Windy.app
Resolution – 27 km.
The GFS is the most well-known global weather model and it’s updated every six hours by the American meteorological service. It is actually made up of 4 separate models which work together to paint an accurate picture of weather conditions: atmospheric, ocean, land/soil and sea ice models. However, it doesn’t take topography and shapes of coastlines into account, so it isn’t very accurate for places next to bodies of water. Good for oceans.
The GFS+ is another version of the same model. While the standard GFS27 interpolates data for each point of the 27 km x 27 km square, the GFS+ version always shows the maximum value in each square.
Precipitation map | GFS
GFS wind forecast on the map in Windy.app
Resolution – various (Europe 7km, ICON7), (global 13km, ICON13)
Created by the German Meteorological Service (Deutscher Wetterdienst), the ICON is generally considered to be even more accurate than the ECMWF due to the better resolution, albeit only in Europe. The most important variables of the ICON are considered to be air density and virtual potential temperature, horizontal and vertical wind speed, humidity, cloud water, cloud ice, rain and snow. Its small-scale part includes the COSMO model, which will be fully integrated into the ICON by 2020.
Precipitation map | ICON
ICON wind forecast on the map in Windy.app
Resolution – 1.5 km (UK), 10 km (Global)
The UM (Unified Model), also often referred to as the UKMO, is a global model developed in the UK. This model runs every 12 hours, with its output running out to 3 days. Due to its resolution, the UKMET is the most reliable model for the United Kingdom. The global model is considered reliable, and is a base for some regional small-scale models (e.g. New Zealand).
Relative Humidity | UM
The CFS is a global numerical model. It is produced by the US NOAA NCEP (National Centers for Environmental Prediction). The model is based on 11 years of weather observations. On each calendar date, the straight average value, determined from the available 11 values, is in general composed of the following components: the true climatological annual cycle, meteorological noise, climatological noise, model noise. The model forecast period is 9 months. It may have poor prediction value, but looks useful for long-term planning.
Precipitation map | CFS
Resolution – down to 500 m
The WRF was a result of a collaborative effort of several agencies and laboratories across the globe in the 1980s. It is a code base for further forecast model processing. It is applicable globally and can take local geography and topography into account. It has a wide range of different physical parameters, and demands vast resources to process. Many institutes use the WRF as a base to develop their own regional models.
Temperature valid | WRF
WRF8 wind forecast on the map in Windy.app
Resolution – various
The ALADIN is a consortium sponsored by Météo-France with some countries of Central and Eastern Europe. In 2015, the consortium initiated a partnership with the HIRLAM, with a goal of an integrated management of the two models. Besides the so-called ALADIN model, it has developed the global ARPEGE (55km grid) and the regional AROME (1.25km) for France and domains. These two are considered as good ‘chemical’ models, as they take into account chemical additives. The AROME has a very good resolution, thus taking into account many weather parameters, which makes it reliable in Europe.
Domain and orography of the ALADIN model
Resolution – 12 km.
The NAM is a regional model for North America produced by the American Weather Service. It has a much better resolution than global models such as the GFS and takes small-scale weather phenomena into account, which makes it the most effective model in that region.
Total surface precipitation | NAM
NAM forecast in Windy.app (open any North American spot and tap NAM to see forecast)
Resolution – 3 km.
The HRRR is a real-time regional (USA) atmospheric model that is updated every hour and has been running since April 2017. It is a cloud-resolving, convection-allowing atmospheric model, initialized by 3km grids with a 3km radar assimilation.
Composite reflectivity of HRRR
Resolution – 2.5 km (Canada), 25 km (Global)
In North America, this model is often referred to as the CMC (Canadian Meteorological Centre) model. Due to it’s resolution and specialisation, it is the most accurate model for forecasts in Canada.
Temperature map | GEM
Resolution – 5 km.
A regional high-resolution weather model developed by the University of Athens on the base of the WRF. It is generally regarded as one of the most reliable models for the Mediterranean.
Accum. precipitation | Skiron
O-SKIRON forecast in Windy.app (open any Mediterranean spot and tap O-SKIRON to see forecast)
Resolution – 2.5 km. (7.5km)
The regional model covers Northern Europe. This model is a result of cooperation between 10 European meteorological institutes, created with a goal of setting up a numerical short-range weather forecasting system to be operated by all the participating institutes.
Surface pressure | HIRLAM
Resolution – 25 km.
This global model was developed and tested by the Australian Government’s Bureau of Meteorology based on the UK Met Office’s Unified Model, which is a weather prediction and climate modelling software developed by the Met Office and used by many forecasting agencies around the world.
Position of cold front | ACCESS-G
Resolution – 9 km.
The RTOFS is a forecasting model of the Atlantic ocean currents run by the US National Weather Service.
Oceanic model | RTOFS
Currents' forecast by RTOFS in Windy.app (open any spot and tap SAIL to see currents' forecast)
Resolution – 12 km.
This regional weather model covers certain regions bordering the United States and it is run by the American Navy.
Surface pressure | COAMPS