Numerical weather models are the staple diet of many meteorologists. In order to look into how the weather will develop in the future we use numerical weather models to see how features suck as lows, highs and frontal systems will be handled based on each models own set of physics packages.
One thing though which all models require is quality input data. This is the initial or starting conditions which show the atmospheric state at one moment in time.
When you see numerical models, they will have an initialisation time, you can see this on the bottom left of all models :-
This, in theory is the time which observations were taken in Zulu time (i.e UTC or GMT) around the world and sent to the forecasting agencies so that they can run their models. However, in recent years many numerical models use 4dVAR technology which allows observations to be taken either before or shortly after the initialisation time.
Many of the forecasting agencies around the world such as NCEP, UK MET, ECMWF, METEO FRANCE etc run numerical models and are also responsible for collecting observation data and sharing this with other agencies.
So, what data is collected and sent into the models?
- Observation data such as METAR or SYNOPs are recorded by either humans or automated stations and encoded to WMO standard.
- AMDAR and ACARS data comes from airlines around the world which record and send back basic outside weather information from upper level flights, especially across oceans which have a lack of observational surface data.
- Weather buoys which are stationed or floating in the oceans and beam back weather/sea information
- Weather balloons or Rawinsondes are released from locations twice a data and send back a profile of the atmosphere.
- Weather satellites and radars can collect a plethora or weather information on the atmosphere below and this data is also collected and fed into weather models.
There are also many other types of observations which are used which include ships, profilers, dropwinsonde, tide gauges etc
Normal amount of hourly observations used in models such as the GFS ► :-
METAR - 45,000
SYNOPS - 20,000
AMDAR - 20,000
ACAR - 150,000
Buoys - 10,000
Weather Balloons - 1,000
Whilst satellites and radar technology will also provide millions of observations for different layers in the atmosphere and different atmospheric sensors.
Once the data is collected it is put into BUFR files which numerical models use. Many agencies also use pre-processors which analyse the BUFR files to ensure any incorrect or erroneous data is purged before the numerical models are run.