March CET

[Stratos Ferric]

There are always arguments about which CET period to use and by using the 61-90 seems to be skewing things, IMO this is used by some to back theories of global warming as its obvious we will always find it more difficult to get below average months using that comparison.

At what point will the met office finally ditch that comparison? the year 2100 . Why dont they just combine both to take account of both periods and give a truer representation of recent years, it seems the use of one or the other will always cause argument.

Nick, comparative assessments of one year against reference period are an irrelevance in making any case for general change in the climate. The (sensible) arguments made are based on comparison of ABSOLUTE temperatures, not the temperature for any given year against the current reference period. Only absolutes provide a definite indication of change. The reference period comparison is useful for the layman in gauging any particular month in its current context only, which for many purposes is fair enough.

For precisely this reason many on here are describing this winter as cold - it is by recent standards, but by absolute standards it has not been.

[Stratos Ferric]

I may have been slightly over the top about the 100 year average being closer to 1971-2000 than 1961-1990. Here are the figures:

The 3 figures are, in turn, 1961-1990 then 1971-2000 and finally the 100 year rolling 1906-2005 average

January 3.8 4.2 3.9

February 3.8 4.2 4.1

March 5.7 6.3 5.9

April 7.9 8.1 8.1

May 11.2 11.3 11.4

June 14.1 14.1 14.2

July 16.1 16.5 16.1

August 15.8 16.2 15.9

Sept 13.6 13.7 13.6

Oct 10.6 10.4 10.2

Nov 6.5 6.9 6.6

Dec 4.6 5.1 4.7

The annual CET from the 3 means is as follows:

1961-1990: 9.5

1971-2000: 9.75

1906-2005: 9.6

I cannot really see much justification for now using the 1961-1990 figures. Someone has had a familiar nip at the Met Office (!) but in fact they are moving to the more recent 30 year average. It may be true that some of those who use the 1961-1990 average do so to impress the global warming argument. On the other hand, the same might be true the other way of those using the 1971-2000 figure? I do think the 100 year rolling average lessens complaints from all quarters. It sits between the two 30 year means and that is probably a fair indication that the 1961-1990 figure is skewed to cold, and the 1971-2000 is skewed to warm. I'll start a campaign for the 100 year rollking average! Great to see John's comments, and look forward to Philip's in due course ... and others of course.

The problem with a long rolling period is that changes are too small to be perceived year by year.

In constructing a sensible time series the following things matter:

1 - smoothing out short-term variation

2 - not smoothing out long term variations

3 - providing a sample size large enough to have statistical robustness

No. 3 is established by basic maths: as a rule sample sizes of 30+ are the most robust. Smaller than this can be used but the attriubution of statistical (i.e. calculated) confidence limits becomes less robust at a rapid rate.

No.s 1 and 2 are actually dependent upon the cyclicality in the data. For example, we measure temperature at least daily because this aligns with the normal period of variation.

On the basis of the above 30-40 years is actually a good basis for placing any given year / period in its current context. The longer series simply serves to tell us what we all know, particularly when viewed on a rolling basis, i.e. the climate at present is warming.

It's rather like measuring our cash wealth through our lives on the day we die as an average of our total banked cash averaged across our lives. All you end up with is a ranking of what we know i.e. who was richer, who poorer. However, day by day, the actual situation would have varied, and would have been experienced very differently by each of us.

[Stratos Ferric]

Does your last sentence really make total sense Nick? One of the reasons I do like the CET is that when tested against a long mean it is possible to discern patterns. The problems only emerge when people read into the statistics 'meaning' based on a week, a fortnight or even an isolated month's data. So I would say the March 2006 CET is a very fair reflection of, well, March 2006.

The really interesting aspect is that March 2006 continues a trend begun last November. We have had 5 months of average to below-average months and that begins to look like a pattern.

You're quite right, and Nick is making a slight mistake in inferring that it is easier to get a different net (to periodic mean) outcome for any period other than a calendar month. If you were to take, say, Feb 20-Mar 10 in a given year where this was a cold period, sandwiched between warm periods, then he is right to suggest that smoothing in the calendar month would mask what actually happened, but the fact would remain that however you slice and dice the 59 day period, the overall average for the period as a whole is unchanged. Can the statistic therefore "lie": no, it's just that the way the data is clustered can tell different stories.

If you wanted to compare the Feb20-Mar 10 period you would, in any case, have to compare it with itself down the years - otherwise you're comparing apples and elephants: on this basis there might also be years when the cold period ran Feb 5-Feb 20, say, and our reference period of Feb20-Mar 10 would again "be in the wrong place" to be a fair reflection of "what actually happened".

If you're not careful you end up in a place where the goalposts are continually changed to try to tell the story you want to believe, rather than what actually happened.

Another example to make the point. European Cup Final a few years back. For the bulk of the match Bayern Munich were winning 1-0; in the last two minutes of play Man U scored 2 goals. Man U walk off with the cup despite the fact that they were only in the lead for around 1% of the match.

It may seem like a silly example, but my point is that we can't moan that the CET should be lower because March felt cold - which is one way of reading Nick's complaint: that said, if we had twenty-five days where the temperature was 1C below the norm, and five when it was 5C above, we would have a month that averaged out around normal, yet which was cold for the bulk of the time. Viewed in this way the extension of Nick's point actually makes an interesting case about the problem of using a single data point to describe what has been a dynamic variable across a longer period. For that reason, the use of commentaries in support of the numbers, the likes of which Philip provides on his site, are fundamental, and there is also an argument for an added measure of cold / warmth which is not just the measured arithmetic mean, but also the proportion of individual days that were warmer / cold.

[drfeelgood]

I may have been slightly over the top about the 100 year average being closer to 1971-2000 than 1961-1990. Here are the figures:

The 3 figures are, in turn, 1961-1990 then 1971-2000 and finally the 100 year rolling 1906-2005 average

January 3.8 4.2 3.9

February 3.8 4.2 4.1

March 5.7 6.3 5.9

April 7.9 8.1 8.1

May 11.2 11.3 11.4

June 14.1 14.1 14.2

July 16.1 16.5 16.1

August 15.8 16.2 15.9

Sept 13.6 13.7 13.6

Oct 10.6 10.4 10.2

Nov 6.5 6.9 6.6

Dec 4.6 5.1 4.7

It's interesting to note that although most months have indeed warmed up, a few haven't or have even experienced slight cooling! I must admit I wasn't expecting this.

[acbrixton]

yes and June is esp interesting as the other 2 summer months (esp August) are noticeably warmer drier and sunnier when comparing the 1961-1990 and 1971-2000 averages.

I set out below the stats for London for avge max temp, avge min temp, rainfall and sunshine for the 3 summer months. 1961-1990 first and 1971-2000 following :

June:

20.2 20.2; 11.0 11.1; 49.3 53.0; 189 180.6

July:

22.2 22.8; 13.2 13.6; 44.0 38.3; 182.3 190.3

August:

21.8 22.6; 12.9 13.3; 49.8 47.3; 179,8 194.4

Note the increase in August's sunshine: although it has only the 4th most daylight it is now the sunniest month and June is only the third sunniest month...

Regards

ACB

[The PIT]

Overall from our weather data from 1955 the biggest changes have been in the Winter Months while Summer has hardly changed. Thats temperature wise.

Rainfall wise bar nothing has really changed although day to day perception is that it is dryer.

[Thundery wintry showers]

These stats do, however, hide the fact that the 1990s were exceptionally warm in most months.

For example, it may well be that August hasn't warmed significantly over the reference periods, but the 1990s average was a whopping 16.9C, and the 2000s have exceeded an average of 17C.

[drfeelgood]

yes and June is esp interesting as the other 2 summer months (esp August) are noticeably warmer drier and sunnier when comparing the 1961-1990 and 1971-2000 averages.

I set out below the stats for London for avge max temp, avge min temp, rainfall and sunshine for the 3 summer months. 1961-1990 first and 1971-2000 following :

June:

20.2 20.2; 11.0 11.1; 49.3 53.0; 189 180.6

July:

22.2 22.8; 13.2 13.6; 44.0 38.3; 182.3 190.3

August:

21.8 22.6; 12.9 13.3; 49.8 47.3; 179,8 194.4

Note the increase in August's sunshine: although it has only the 4th most daylight it is now the sunniest month and June is only the third sunniest month...

Regards

ACB

That's very interesting, ACB. So basically, in terms of sunshine distribution, it has been "displaced" from early to late summer. Temperature-wise, a similar displacement can also be seen, with August being almost as warm as July. In fact, if we only take into account the last 10 years, August is now probably the warmest month.

[Optimus Prime]

Does anyone find it Ironic that everytime we get a below average month (apart from October 2003) the BBC mess it up;

March 2006 "0.6 °C above the 1961-1990 average, which is in the below average category. Coldest since 1996."

November 2005 "0.1 °C above the 1961-1990 average, , which is in the close to average category."

November was 0.1c below average

March was 0.6c below not above.

I've seen them do it before but never when we get an above average month. Yet the metoffice site never do it.

This couldn't be due to their mostly over the top Global Warming scare could it?

They've also done it for Northern Ireland;

March 2006 0.6 °C above the 1961-1990 average, which is in the above average category."

Oh, dear.

Edited April 9, 2006 by Optimus Prime

[drfeelgood]

Does anyone find it Ironic that everytime we get a below average month (apart from October 2003) the BBC mess it up;

March 2006 "0.6 °C above the 1961-1990 average, which is in the below average category. Coldest since 1996."

November 2005 "0.1 °C above the 1961-1990 average, , which is in the close to average category."

November was 0.1c below average

March was 0.6c below not above.

I've seen them do it before but never when we get an above average month. Yet the metoffice site never do it.

This couldn't be due to their mostly over the top Global Warming scare could it?

Oh, dear.

Write to them Optimus and ask them to use a calculator if they can't count !! Go on!

Edited April 9, 2006 by drfeelgood

[Metomania]

The problem with a long rolling period is that changes are too small to be perceived year by year.

In constructing a sensible time series the following things matter:

1 - smoothing out short-term variation

2 - not smoothing out long term variations

3 - providing a sample size large enough to have statistical robustness

No. 3 is established by basic maths: as a rule sample sizes of 30+ are the most robust. Smaller than this can be used but the attriubution of statistical (i.e. calculated) confidence limits becomes less robust at a rapid rate.

No.s 1 and 2 are actually dependent upon the cyclicality in the data. For example, we measure temperature at least daily because this aligns with the normal period of variation.

On the basis of the above 30-40 years is actually a good basis for placing any given year / period in its current context. The longer series simply serves to tell us what we all know, particularly when viewed on a rolling basis, i.e. the climate at present is warming.

It's rather like measuring our cash wealth through our lives on the day we die as an average of our total banked cash averaged across our lives. All you end up with is a ranking of what we know i.e. who was richer, who poorer. However, day by day, the actual situation would have varied, and would have been experienced very differently by each of us.

Sorry, I missed this excellent post and never responded.

Whilst I would often agree with this in statistics I am not sure climatology makes it quite so clear-cut. The issue revolves around what 'short-term' means in metereology doesn't it? Except for the exceptional warm up at the end of the C20th patterns in climatology are surely measured over decades? There is little doubt that the 1961-1990 mean is 'unfairly' cold, but that the 1971-2000 is also rather warm. It's true these might be indications of patterns, in which case climate change is measurable over what would be considered in metereological terms incredibly short time-spans. But I do take your point.

On balance I rather like Philip Eden's suggestion of using the 1971-2000 mean, but also referencing how this compares to the last 100 years.

[Metomania]

The Met Office have released both the February CET and the March CET. You can see the Hadley figures on their site, here:

http://www.metoffice.com/research/hadleyce.../HadCET_act.txt

February is officially 3.7C and March is officially 4.9C.

[Weather-history]

The Met Office have released both the February CET and the March CET. You can see the Hadley figures on their site, here:

http://www.metoffice.com/research/hadleyce.../HadCET_act.txt

February is officially 3.7C and March is officially 4.9C.

2nd coldest November-March period since 1987 with only November 1995 - March 1996 colder.

[reef]

The Met Office have released both the February CET and the March CET. You can see the Hadley figures on their site, here:

http://www.metoffice.com/research/hadleyce.../HadCET_act.txt

February is officially 3.7C and March is officially 4.9C.

It looks like Phillip Eden wasnt far off when he predicted a Hadley CET of 4.96°C. Interesting that both February and March were the coldest since 1996, that along with 4 out of the last 5 months being below the 1971-2000 average makes for quite an impressive cold blip. The second coldest November - March period since 1987 is equally astonishing.

It looks like April will end up above average for sure, however even this month is not looking to be too far above average as the warm spell of the next few days has been downgraded somewhat. Then again, even an April CET of 8.8°C would make it the "coldest" since 2001.

Oh, and it looks like Shuggee was spot on aswell in this thread, with his prediction of 4.9°C

Edited April 22, 2006 by reef

[Weather-history]

. The second coldest November - March period since 1987 is equally astonishing.

The period November 05 -March 06 was 0.6C below the 1971-2000 norm or 0.2 below the 1961-1990 norm

[Metomania]

It looks like Phillip Eden wasnt far off when he predicted a Hadley CET of 4.96°C. Interesting that both February and March were the coldest since 1996, that along with 4 out of the last 5 months being below the 1971-2000 average makes for quite an impressive cold blip. The second coldest November - March period since 1987 is equally astonishing.

The six months from October 1st to March 31st was the coldest since 2000-1. I'm not too sure about the 'astonishing' point really Reef - it all depends on when you pick your reference point. We need to see this in a long-term context. The average (or slightly below) months apart from January will need to be viewed over the following years. If the general trend in the CET remains upward then the 3 months below average 4 if you include February which was average to the 1961-1990 mean) will be an interesting blip. Those who believe in cycles will be hoping the next few winters show the trend to cooler conditions was no one-off, but something more significant. But I do think we need to reserve words like 'astonishing' for now.

The period November 05 -March 06 was 0.6C below the 1971-2000 norm or 0.2 below the 1961-1990 norm

How did it fare for the 6 months from October 1st to March 31st?

[Metomania]

Just done the figures, and the six month period from 01st October to 31st March was:

0.3C above the 1961-1990 average

0.1C below the 1971-2000 average and

0.1C above the 100 rolling average 1906-2005.

So although there were colder/cooler months, the last six months have been average (the 1961-1990 mean is undoubtedly cold-biased). To have an average 6 months is significant after a large number of months with an upward trend. It will be very interesting to see where things head from here!

Edited April 22, 2006 by Metomania

[Weather-history]

How did it fare for the 6 months from October 1st to March 31st?

It was about the 1971-2000 norm and 0.3 above the 1961-1990 norm.

The period October 1995- March 1996 was only 0.1 below the 1961-90 norm

[reef]

The six months from October 1st to March 31st was the coldest since 2000-1. I'm not too sure about the 'astonishing' point really Reef - it all depends on when you pick your reference point. We need to see this in a long-term context. The average (or slightly below) months apart from January will need to be viewed over the following years. If the general trend in the CET remains upward then the 3 months below average 4 if you include February which was average to the 1961-1990 mean) will be an interesting blip. Those who believe in cycles will be hoping the next few winters show the trend to cooler conditions was no one-off, but something more significant. But I do think we need to reserve words like 'astonishing' for now.

How did it fare for the 6 months from October 1st to March 31st?

The period October 1st to March 31st isnt anything special I agree, but then again it is offset by an inflated October figure. Although in longer term trends the November - March period isnt anything particularly special either, in the context of the mild trend post-1988 it is rather astonishing. Its very rare post-1988 to get ANY below average months, so to get 4 out of 5 below the 1971-2000 average in a bunch is an interesting development. I do think its only a blip, but time will tell.

[Guest Mike W]

Not sure if this has been mentioned probably has but last winter was 4.1 average which is the coldest since 1996/97 winter at 4.0.

Edited April 22, 2006 by Mike W

[North-Easterly Blast]

No, winter 2005-06 was the coldest since 1995-96 overall. The overall winter Dec/Jan/Feb CET was 4.13*C, but was followed by a cold March at 4.9*C, so it was the coldest winter for ten years. Dec/Jan/Feb 1996-97 may have been 0.1*C colder but it is often forgotten that the March that followed was very warm at 8.4*C, so 1996-97 would seem a somewhat milder winter overall than 2005-06. Since 1987 only 1990-91 and 1995-96 were colder than 2005-06 and 96-97 was only marginally colder and only 1995-96 was followed by a cold March, so in general it could be looked as the second coldest winter since 1987.

[Weather-history]

No, winter 2005-06 was the coldest since 1995-96 overall. The overall winter Dec/Jan/Feb CET was 4.13*C, but was followed by a cold March at 4.9*C, so it was the coldest winter for ten years. Dec/Jan/Feb 1996-97 may have been 0.1*C colder but it is often forgotten that the March that followed was very warm at 8.4*C, so 1996-97 would seem a somewhat milder winter overall than 2005-06. Since 1987 only 1990-91 and 1995-96 were colder than 2005-06 and 96-97 was only marginally colder and only 1995-96 was followed by a cold March, so in general it could be looked as the second coldest winter since 1987.

Winter CETs are measured from the 1st of December to the last day of February. March is not included in that value.

:o

[Methuselah]

Tbh, I think that we can split as many hairs as we like over the details...But, judging from others' comments, personal experience and - of course - the available data, I'd be surprised if winter 2005/6 was not the coldest (least mild?) since 1995/6??? :o

That would be nicely in-line with the Met Office's early expectations? :o

[Nick F]

Tbh, I think that we can split as many hairs as we like over the details...But, judging from others' comments, personal experience and - of course - the available data, I'd be surprised if winter 2005/6 was not the coldest (least mild?) since 1995/6??? :o

That would be nicely in-line with the Met Office's early expectations? :o

Apparently it's been the coldest winter for 9 years over the Southern quarter according to Philip Eden in his Sunday Telegraph column, but he suggest that it's been a warmer than average winter over Northern areas - locally the warmest for 14 years over Scotland. Though, March ended up 1C below the Scottish Mean Temperature Series according to the Met Office - though not officially a winter month!

Anyway, this is a CET thread, and yes the Met Office were pretty spot on last winter for the South, and i'll be interesting to see what they'll rustle up forecast wise for next winter.

[Methuselah]

Apparently it's been the coldest winter for 9 years over the Southern quarter according to Philip Eden in his Sunday Telegraph column, but he suggest that it's been a warmer than average winter over Northern areas - locally the warmest for 14 years over Scotland. Though, March ended up 1C below the Scottish Mean Temperature Series according to the Met Office - though not officially a winter month!

Anyway, this is a CET thread, and yes the Met Office were pretty spot on last winter for the South, and i'll be interesting to see what they'll rustle up forecast wise for next winter.

Thanks Nick...I think that orographic factors must have played a significant part in making some areas of Scotty that warm - that and a lack of widespread severe frosts at night?? :o Either that, or it's been so damned bland that one has failed to notice the temperature! :o

Whatever the outcome - I'll not be quibbling over the data...