forecaster

Doesn't seem to be an amazingly strong correlation over the British Isles. Also....it looks opposite to what might be expected. Negative correlation would imply that as QBO goes up, surface temperature goes down.

Amusingly, where I live the winter max temperatures are almost always in that range (more like 9C-14C). Houses have thin wooden walls, single glazing, and no central heating. Can be dismal. We have 4 dehumidifiers in this house and I usually wear a down jacket indoors all winter.

To touch very briefly on this. Two images from work by Anders Persson (previously SMHI and ECMWF) regarding forecast consistency:

Australia's problems are less. Most of the population is on the east coast, which has climatologically much wetter summers than California, though as you say it is dependent on the ENSO cycle. It also has a smaller population than California. This Californian drought is deeply concerning. It might take until April or May, when the chance of rains are over, before it hits home properly.

Pretty much. The most robust highs on the planet are the subtropical ridges (think Azores High, North Pacific High, South Pacific High) rarely go above 1030hPa. A 1050-1060hPa high could occur from purely low level conditions (cold dense air near the surface), with little upper support. Hence, could be barrelled out of the way fairly easily. Or, it could be persistent. Hard to tell from looking purely at a surface pressure field.

MJO active in the W Pacific during northern winter increases chances of negative NAO occurring 10-20 days later when compared with climatology.

No firm definition. We often call 500hPa upper air, but mid levels would be perhaps more accurate. It also depends on the seasons. The troposphere is deeper in summer.

This type of thing was seen in model runs in early November too.

To open GRIB files (meteorological standard for model data) you can use: http://www.unidata.ucar.edu/software/idv/

Yes, just a good example I found from an old model thread, not a dig at the ECMWF specifically.

Just a quick look back in time. Older EC run valid for Tuesday: FAX: This, for Thursday, shows the EC perhaps 10C out in 850hPa temperatures:

Seems to have done well in picking the general character of December, especially given he must have known it would not have been a popular forecast! I thought it was risky for him to mention that Christmas Day specifically had a good signal for being mild. That is very high precision forecast which turned out to be inaccurate. It wasn't really mild on the day or the days either side.

Models really seem to progress that fairly promising chart quickly. At 144 the EC jet profile looks good. Very meridional out to our west, diving into North Africa. However, it seems like a split in that jet south of Greenland allows a depression to form there (with double jet structure - i.e. entrance and exit regions of separate streaks), which could be the spanner in the works.

I remember Derek on BBC Wales hinting at the possibility of colder conditions to come in a piece at the end of November. He was referring to the Dec ~6th cold snap, which barely materialised for most of us. The models really led us astray on that event.

I am not sure if balloons are as massively important as they once were, due to the quality of satellite soundings available nowadays. A real sounding (balloon) is better, but it's only for one point (or assumed to be for one point, despite drift during flight), but inferred soundings from satellite can cover the entire globe.

There is a little value in forecasting by persistence and the pros will use it to some extent. One measure of the quality of a forecast is whether it can beat both persistence and climatology. If both are better, the method tested obviously does not look too clever!

When you talk about phasing and phased energies what exactly do you mean?

A quick glance at the 2nd Jan EC32 update does not suggest any very big changes during that period from this model.

Yes, and referring to the original post, the model is not really an "algorithm".

You need some evidence to back that up.

Greenland is mostly elevated. Have to be careful with what you look at. Nwp shows pressure reduce d to sea level. Station might be reporting station level pressure which would be relatively low. Reduced to sea level pressure is called Qnh, that's what you want to look at

LEWP?

Not convinced. Xmas 04 had mid Atlantic ridge and scandi low with relatively deep n to nw flow. Current modelling has no substantial ridge in mid Atlantic. What was interesting in 2004...not super cold airmass. 850hpa temps above -6 everywhere I think. Dewpoints in s Wales sub zero....not sure we can get that with what you are showing.

Those aren't thicknesses. Just 500hpa heights it looks like.

The short answer is that they're rigidly programmed. However, members here tend to anthromorphise the models somewhat, or at least treat them differently to how they really are. My experience is that forecasters also don't think of models in their truest sense. This is all understandable, because at their core, it is just a set of equations which are integrated forwards in time. Models are pretty much just a massive equation solver. That is hard to picture. It's also pretty dull (unless you're really into that sort of thing). Decades ago, the equations used were the primitive equations: http://en.wikipedia.org/wiki/Primitive_equations Most (or all?) of these do not have "analytic solutions". That's a way of saying that they don't have solutions that can be precisely written by another formula. For an example of the converse: That's the form of a quadratic equation, familiar to anyone who's done GCSE maths. What is x? The answer is always: That's an analytic solution. None of the primitive equations have solutions that can be written in such a way. All the solutions are numerically produced approximations (though they are very good approximations). If anyone has ever done "trial and improvement" solutions at school, it's a similar thing. This is part of the reason we need supercomputers. The reality is that the primitive equations are only approximations themselves. I don't know what goes into modern models, but the equations will be more complicated than that. Other issues that arise are how to deal with things that cannot be explicitly modelled. These are many. Convection being the classic example, and all the components of that. Models don't go around thinking "I'll have a high there, and a jet there, and a nice low there, and we'll stick a front there" even though that's the output we see. The gritty reality is that it's an unfathomably vast solver of ugly looking equations. Everything we see is produced through masses of equation solving; it's just that these equations are rather good at approximating the state of the atmosphere, and how the parts link together. So when we see a low, everything within the model has contributed towards making that low, and there is a lot of stuff going on there. I'm no modeller, but hopefully the above is approximately true enough.