The Power Of EnsemblesAdded : Tuesday at 14:05 A good example today of the power of ensembles. If you saw the discussion on Sunday, then we talked about the GFS ► and how it was trying to build in a ridge from the Southwest for later next week.
Yesterday, we showed how the ensembles simply didn't agree with this and kept the pattern far more zonal.
Today, the GFS ► has backtracked somewhat and here is the chart for later next week :-
Sure, the ridge is there, but nothing like as settled as previous runs had shown. So, why don't we just use ensembles? The reason, is that the ensemble output suite is the GFS ► run forty odd times with some slightly different starting conditions and also some slightly different physics packages. Ensembles are also run at lower resolution due to the time taken to generate each run.
If you look at the GHX Risk model which is based on ensembles for this afternoon then you can see the high confidence (red) in rain being expected across our part of the world :-
But, by the end of the run you can see the confidence dropping remarkably and a risk of rain in most places by the end :-
Deterministic models are incredibly important to show the higher resolution output, but you'll notice they chop and change from run to run after around 10 days or so.
We counter this by adding rainrisk and confidence in the forecasts based on the ensemble spread. If the main model says 22C as the expected maximum temperature in two weeks, but the ensembles range from 16C to 27C then the confidence is low, but if the ensembles are well clustered then confidence is higher.
Deterministic model runs should always be used alongside ensembles to give a region of confidence in the forecast.
So, whilst both types of models have their own strengths, combined is what makes them the invaluable tool to meteorologists.
METEOROLOGIST : MARSH |