Model Blind SpotsAdded : Friday at 9:34 We had an observation from one of the Metcheck team the other day when discussing the potential for colder weather next week. We noticed how the GFS ► would fluctuate quite wildly in the intensity and timing of any colder air coming from the North. Granted, this is still in the medium range of the models and to take model output at face value at T+210 is asking for trouble.
But.. "Why did the GFS ► do so well with the remnants of Hurricane Ophelia a couple of weeks ago, but struggles so much with developments around Greenland?"
To understand this, take a look at this image :-
This shows a rather crude picture of the global weather observations which go into numerical weather models. It's this ingest as we call it which is so crucial to every numerical weather model which you view. Observations from satellites, radars, met stations, balloons, airplanes, ships, buoys, sondes and a plethora of other instruments all observe the atmosphere and then put these into BUFR files which numerical models like the GFS, ECMWF, UKMET etc all use as starting conditions for a model.
The problem is that we never get a 100% pure view of the atmosphere at any one time. There are places which may have no observations at all. What happens then? The models will use pre-forecast data for that location as a starting condition, but you can see where the problem with that lies.
The other issue with Greenland is the physical characteristics of the place. A cold and barren plateau which surrounded by a mixture of water, ice and snow. In short, this is an awful lot of energy and physics required to model how the weather behaves around Greenland. Add in a lack of observations in the area and you can begin to see why changes in model output fluctuate so wildly in that particular area.
The latest overnight run from the GFS ► takes the plunge to the East of the UK later this week :-
But it may return for the 6z run and then disappear for the 12z run... who knows?
The catalyst for the whole thing is a tropical disturbance which is currently a trough of low pressure over the northwestern Caribbean Sea :-
Later this weekend, the system drifts Northeast and merges with a mid latitude low and heads North :-
This is where the models go a little hazy. Before, the GFS ► was taking this system up the West coast of Greenland which in turn promotes high pressure to build across Greenland which in turn throws a Northerly wind down the East coast of Greenland and sends colder air our way.
Until the models can resolve this system they simply will not be able to get the timing or intensity of the cold plunge across the UK at the end of next week, in fact, if the system drifts to the South of Greenland then the plunge won't happen.
So, why was Ophelia so well forecast? The reason was thanks to some ultra high resolution satellites which were observing and monitoring the system constantly from NOAA and NASA. Accurate and high quality data into the models shows what they are capable of. Also, Ophelia had no land interaction as she headed towards the UK.
Conversely, the system which could bring a colder phase next week are remnants of a tropical system which has to complete post tropical transition and merge with a mid latitude low and then have to interact with the land mass of Greenland. That's an awful lot of physics and plenty of potential for errors, errors which grow the further you head down the line.
The way this uncertainty is dealt with is by using ensembles which perturb the starting conditions using what we call Stochastic noise. It basically means that it changes the data for each starting location slightly in the knowledge that the actual starting data might be wrong, but that's another discussion for another day.
But when you see low pressure running up the West coast of Greenland at T+240 and then a cold plunge down the East at T+276 then enjoy looking at the physics in action, but at the same time have that salt shaker to hand as that scenario at that range is something which should never be taken too seriously.
METEOROLOGIST : MARSH |