Thanks Gavin, a bit too early for me to worry though.
Gavin, do you remember what the NAO forecasts for winters 2009-10 and 2010-11 were indicating back in May 2009 and May 2010 respectively? Thanks.
SST's in May 2009 pointed towards a positive NAO for the winter of 2009/2010;
This was one of things that encouraged the Met Office to forecast a mild winter for 09/10.
SST's in May 2010 pointed towards a negative NAO for winter 2010/2011 with the tripole quite clearly in evidence;
Edited by user 06 June 2012 16:17:46(UTC)
| Reason: Not specified
SST's in May 2010 pointed towards a negative NAO for winter 2009/2010 with the tripolequite clearly in evidence;
So basically, SSTs in May are more or less useless for forecasting NAOs the following winter?
The method is not without its uses but I believe it's just one factor of a bunch; a factor that can be overridden also IMO.
Yes, the MetO got it horribly wrong for Winter 2009/10 on the back of the May SST so to be honest I wouldnt hang too much on this.
More important IMO is solar activity which seems to show a better correllation with the NAO.
Odds are on a wet/windy/mild winter in 2012/13 but thats just down to the dice effect after 3 cold one 2009-2011.
The higher solar activity will not help either.
Either way I think I'll find this coming Winter hard to beat the record -17.7C that was recorded here in Jan 2010 under clear skies and over a foot of snow cover, now that was what you call proper Winter weather.
As I say in the video, its one thing to look at. Its one element, but there are many other things that are as, or maybe even more important than Atlantic SST's in May.
I think that what you have mentioned is a good example of what Crewewcold said earlier in this thread about the NAO prediction in May each year being rather hit and miss. It seems that in 2010 the MO were correct with their negative NAO forecast for the 10/11 winter, but the previous year they were very much wrong.
Personally, I take the view that the weather patterns that develop during the preceding autumn, especially during the latter part in November, go a long way to determining the type of winter we will get. I seem to recall Brian saying one year (think it was 2009) that if he had issued his 2009/10 winter forecast at the start of November '09 instead of at the end, eh would probably have called a mild winter instead of a cold one. It seems to me therefore that even though the UK' weather during that November was very mild and wet, there were changes taking place elsewhere which began to have a noticeable effect on us a few weeks later.
I don't think last winter could be described as ever being particularly cold, with the exception of the first half of Feb. It was a much milder winter overall compared to 2009/10 and 2010/11.
Actually, their model was also pointing towards a mild winter for 2009/2010 as well. It wasn't just the Atlantic SST's.
Thanks guys. Good discussion.
I just put the evidence out there for you. Its up to you guys whether you want to take it seriously or not. Ultimately its for fun and for enjoyment but its a small part of the overall picture.
Yes that sounds about right.
Yes. Here is the 2/3 figure quoted from their 2005/6 forecast:
A tripole predictor pattern of warm SST anomalies at low latitudes and to the south ofGreenland and cold anomalies off the eastcoast of North America extending into thecentral Atlantic leads, in two cases out ofthree, to negative winter NAO (see Part 2 andRodwell and Folland 2002, hereafter RF02)
My first thought was how many years' data were they able to consult before obtaining this 2/3 probability.
Happily the Rodwell and Folland paper from which they took this percentage is available:
We find observational evidence from the last half of the twentieth century for SST forcing of the atmospheric circulation in the North Atlantic region.
Hmm ... so about 50 years of data. Personally I would not be putting my money down on the basis of a 2/3 correlation from just 50 samples.
Were the data available to investigate any further back in time? It turns out they were:
Analysis of observational and atmospheric model data from the first half of the twentieth century identifies very little predictability of the winter NAO.
Oh dear. Increase the sample size to 100 and the correlation disappears. So how did they, and subsequently the Met Office, justify their faith in the 2/3 probability?
We suggest that the change in seasonal predictability is genuine and may be related to the strength of decadal oscillations.
Which sounds like suspiciously circular reasoning to me: if there is a visible correlation in recent years, it must mean that conditions have changed to allow a visible correlation.
My conclusion would be: "Interesting. Give it another century and let's have another look." I certainly wouldn't have made it the central plank of a published forecast, even an "experimental" one.
For a bit of fun for the last two months I have been looking at the cfs daily nine month runs on Meteociel and specifically checking out December 2012 and January and February 2013. bearing in mind all the caveats which come with the tool.It should be remembered they are run forward from genuine starting data rather than assumptions. I have made notes of what the runs have suggested. So far they run around 70/30 in favour of a winter with substantial cold or very cold periods. In some weeks the percentage has been as high as 80/20 in favour of a winter with substantial cold or very cold periods.Any runs showing a genuinely mild winter have been very very few and far between.Please feel free to shoot me down in flames ( as I say this has been for a bit of fun)However at this stage of the year I would venture to suggest that this method is as accurate as any other for trying to get a handle on trend for the coming winter.Purely based on what I have seen gleaned from the above and nothing else The suggestion is that that tendency would be for a colder winter overall than last year with the coldest spells coming in December and January and possibly very late feb.
Edited by user 09 June 2012 12:14:20(UTC)
| Reason: Not specified
For a bit of fun for the last two months I have been looking at the cfs daily nine month runs on Meteociel and specifically checking out December 2012 and January and February 2013. bearing in mind all the caveats which come with the tool.It should be remembered they are run forward from genuine starting data rather than assumptions. I have made notes of what the runs have suggested. So far they run around 70/30 in favour of a winter with substantial cold or very cold periods. In some weeks the percentage has been as high as 80/20 in favour of a winter with substantial cold or very cold periods.Any runs showing a genuinely mild winter have been very very few and far between.
This I can very much believe and I base much of this on the basis of the summer we are having - Some indeed, quite a few, may disagree with me here - but this summer is now very much resembling the very bad ones we had in the 1960s and 1950s and many of those delivered cold or even very cold winters.