Guide To Ensembles

[Paul]

Ensembles - What are they?

All the major numerical models (GFS, ECMWF, UKMO etc) run more than once each time they are issued. The reason for this is that although they each get fed a lot of data into them for each run, the data can never be considered perfect. To counteract this the model is run again with slight changes to the original data. Using the GFS as an example, the model is run 10 more times, 5 times with the data positively changed and 5 times with the data negatively changed.

By following the ensembles you can get an idea of the likelihood of the forecast given by the main (control) run of the model being correct - if a large number of the ensemble runs are very similar to it, it is a good indication that confidence is reasonably high of the outcome. If very few of the ensemble runs show that outcome, then it is quite probable that it may well change.

What's an outlier?

An outlier is when the control run is different to the vast majority of the ensemble runs. A cold outlier is when the control run shows a colder outcome than majority of the others, and a warm outlier is when the control run shows a warmer scenario.

I hope that helps