Skip to content

New Attempt To Claim More Extreme Weather Not Supported By Data

September 9, 2014

By Paul Homewood  




The Telegraph carry a report on the latest piece of work from Phil Jones and co. I have not tracked down the original paper yet, but this is what the Telegraph have to say:


Britons should brace themselves for more extreme winters as weather conditions become more volatile, scientists have warned.

A study of seasonal records dating back to 1899 found that while most seasons have not changed dramatically, winter has become much more unpredictable.

The results suggest the idea of a typical British winter is increasingly becoming a myth, with wide swings from mild but stormy conditions like those which hit the UK this year to extremely cold temperatures and snow in another year becoming more common.

Researchers from the University of East Anglia (UEA), University of Sheffield and the Met Office found that seven out of the 10 most extreme winter conditions over the last 115 years have occurred in the last decade.

Professor Phil Jones, from the University of East Anglia’s climatic research unit, said: "This indicates that British winters have become increasingly unsettled.

If this trend continues, we can expect more volatile UK winter weather in decades to come.

Winter conditions are commonly defined using North Atlantic Oscillation (NAO) – a system of barometric pressure variations which indicates the strength of westerly winds approaching the UK.

When westerly winds are strong, Britain experiences mild, wet and often stormy weather – like last winter.

Weaker or reverse airflow typically brings cold, snowy weather, such as that experienced in 2009/10 and 2010/11.

The variations were particularly noticeable in early winter.

Mr Jones said: "When we look at the month of December in particular, our data shows that over the last 115 years, three out of five all-time record high NAO values and two out of five record lows took place in the last decade."

Professor Edward Hanna, from the University of Sheffield’s department of geography, said it was too soon to say whether the increased volatility is linked to global warming.

But the study, published in the International Journal Of Climatology, states that it was extreme unlikely the clustering of extreme conditions had happened by chance.

The trend could be due to random fluctuations in the climate system but could equally be due to factors including changing pressure and weather systems over the Arctic, especially Greenland, and changes in energy coming from the sun.

Mr Hanna said: "We cannot use these results directly to predict this winter’s weather but, according to the long-term NAO trend, we can say that the probability of getting extreme winter weather – either mild/stormy or cold/snowy – has significantly increased in the last few decades


The first thing to note is that they are actually trying to define the NAO, rather than looking at actual temperatures, etc. So, let’s see how the NAO has been behaving in winter.

The UK Climate Projection Report , last updated in 2012, includes the following graph of the NAO up to 2006, and comments:


There is considerable interest in possible trends in severe wind storms around the UK, but these are difficult to identify, due to low numbers of such storms, their decadal variability, and by the unreliability and lack of representativity of direct wind speed observations. In UKCIP02, we showed how the frequency of severe gales over the past century, although showing an increase over the past decade or so, did not support any relationship with man-made warming. Alexander et al. (2005) presented an analysis whereby a severe storm event is characterised by a rapid change in (MSLP) (specifically ±10hPa in a 3h period); this is different from the severe gales in UKCIP02. They found a significant increase in the number of severe storms over the UK as a whole since the 1950s. This analysis is being extended back in time using newly-digitised MSLP data from as many as possible long-period observing sites in the UK and Ireland, and some preliminary results are shown in Figure 1.14 (Allan et al. in preparation). It appears that an equally stormy period to those in the most recent full decade (1990s) was experienced in the 1920s. Similar conclusions are drawn in IPCC AR4 (Chapter 3, para and Fig. 3.41).

Whereas it is not our purpose here to discuss detailed links between the NAO and storminess, it will be immediately apparent that the two stormiest periods in Figure 1.14, in the 1920s and 1990s, coincide with decades of sustained positive NAO index, whereas the least stormy decade, the 1960s, is a time when the smoothed NAO index was most negative (see Figure 1.13). Although work by Gillett et al. (2003) has shown that man-made factors have had a detectable influence on sea-level pressure distributions (and hence atmospheric circulation patterns) over the second half of the 20th century, there continues to be little evidence that the recent increase in storminess over the UK is related to man-made climate change.




There is certainly no evidence here that the NAO is becoming more volatile, but what has happened since 2006. We can use NOAA to update the NAO.





The strongly negative NAO of 2010 stands out, but is certainly not unprecedented, and again there is no evidence of any unusual trends.


Cold Winters

Let’s now take a look at cold winters in the UK, and see if these are becoming more extreme. (I’ll take a look at mild winters tomorrow).

First, let’s consider the Met Office record since 1910. The cold winter of 2009/10, and the following winter appear to be nothing out of the ordinary, compared to earlier winters. Certainly, the variation from year to year is much less than several other periods of the 20thC.




We can also look back much further, using the CET records.





As winters nowadays are, on average, warmer than in earlier centuries, we cannot directly compare actual temperatures in order to determine which ones were extreme.

I have therefore calculated anomalies of winter mean temperatures against the 30-year average at the time. For instance, the anomaly for 2014 is set against the baseline of 1985-2014, while for 1690 the period 1661-90 is used. (The only exception here is the first 30 years of the record, 1660-89, which are all measured against 1661-90 as well).

Based on these anomalies, the chart below plots all winters when temperatures were at least 2C below the 30-year norm. As can be seen, there has been a notable absence of extreme cold winters, since the infamous winter of 1962/3.





 Final Thoughts

I notice that particular reference is made to the month of December only:


Mr Jones said: "When we look at the month of December in particular, our data shows that over the last 115 years, three out of five all-time record high NAO values and two out of five record lows took place in the last decade."


 It’s not difficult top find unusual trends when you pick one month out of, arguably, six. (Remember, NOAA take their NAO index up to March, and if you are looking at storminess you could certainly make a case for including October & November).

Is there any logical reason why December should exhibit more extreme trends, when the rest of winter shows the opposite? I don’t know, but it certainly does not justify the claim that British winters have become increasingly unsettled.

It is hard to avoid the conclusion that Jones has gone out of his way to look for examples of extreme weather, simply for headline making purposes, whether the facts justify it or not.


What Julia Slingo Said

I’ll leave the last words to Julia Slingo, who wrote this report for Sir John Beddington after the severe winter weather of December 2010. 




Slingo is clear – the cold winters of 2009 and 2010 were not unusual. To now attempt to claim that winters are becoming more extreme flies in the face of the facts. 

  1. September 9, 2014 10:31 pm

    When was winter ever “predictable”?

  2. Green Sand permalink
    September 9, 2014 10:41 pm

    Using data from 115 out of 4.54 billion years –

    “Researchers found from their statistical analysis that there was a less than 0.04 per cent probability of getting such a clustering of extreme NAO years by chance.”

    Inspired, brave or simply arrogant?

    When reading such “research” findings it is difficult not to come to the conclusion that the establishment’s commitment to dumbing down education has indeed been extremely successful.

    Aware questions – ignorance accepts

    • David permalink
      September 10, 2014 10:05 am

      Green Sand

      “Using data from 115 out of 4.54 billion years – …”

      Here’s the paper (pay walled):

      ‘Supporting information’ states the probability that 5 out of the 10 ‘most extreme’ (top and bottom 5) Decembers might occur in a 10-year period within a 115-year data set was calculated using 3 methods. All provided similar results, 0.03; 0.04; and 0.03%. Details are in the peer reviewed paper, so are available for scrutiny/rebuttal.

      I’ve been unable to find the 115 year data set used by the authors. It’s not the one used by NOAA, since only 3 of the 10 most extreme 10 events occur between 2004-2013 in that file.

  3. Richard111 permalink
    September 10, 2014 5:29 am

    I read somewhere that a cooling world causes an INCREASE in global wind movement due to the temperature differential between equator and the poles. A warming world would see a decrease in global wind. So far all weather/climate events indicate we live in a cooling world.

  4. igsy permalink
    September 10, 2014 11:35 am

    I sent the following to the Telegraph’s comments:

    Rather than using the actual NAO data, the authors choose to use an index derived from gridded sea surface pressure measurements by climate scientists using a multivariate statistical technique, the Empirical Orthogonal Function (EOF). They use the “leading” EOF, i.e. the major pattern in the data that has no commonality with other patterns. What percentage of the data’s variance is explained by this pattern, I haven’t had time to discover, but it is important: e.g. if the number is, say, 70%, then patterns representing 30% of the data have been left on the cutting room floor.

    They run some “what is the chance of x extremes in n years?” probability tests, one of which claims a 1 in 2500 chance of five extremes in the past ten years. They really are skating on thin statistical ice here. Obviously this statement is entirely dependent on an arbitrary definition of an “extreme”, effectively turning the problem into a subjective, discrete categorical one. There is no need to do this, we have continuous data that is N(0,1) with trivial autocorrelation. A standard chi-square test taking a ten-year sample suggests there is no evidence at standard statistical thresholds that the variance/stdev has changed. Suggesting a 1 in 2500 chance is just ridiculous, I’d give it 1 in 7 or 8 at most, and that’s just on the index data, the usage of which I will now contest.

    The comparison between the actual data and the 1st EOF index is instructive. The correlation is high, .82 from 1899 until the last five or ten years that excites Prof Jones so much, .77 for the full series. The chart of the differences between the datasets shows that they are nomally distributed about a mean of .5 until 2010 when we see a large outlier, followed by a smaller, but still big one in 2012.

    So the question is: why such a sudden in the relationship between this index and the actual data? I suspect (but don’t know) the answer lies in the second EOF, depending on how much of the variance the first EOF explains. However, there’s worse. December to March is winter? Nope, the meteorological winter runs from start-December to end-February. Using this definition (obviously the correct one), the 2010 difference between the datasets is a sole, humongous outlier. Unless this outlier can be explained, we cannot place any confidence on this analysis.

    Finally, the selected EOF index only starts in 1899, whereas the CRU’s NAO data goes back to 1821. Visual inspection of various rolling sub-period variances over all Dec-Feb periods show nothing exceptional, and if anything, exhibit a slight long-term downtrend over what is nearly two centuries, a result that would not have surprised the founder of the CRU, HH Lamb, who accepted that warmer higher latitudes reduce the temperature gradient and consequently the incidence of extreme weather.

    So, we have a result that is dependent on defining winter as December to March; on an erroneous probabilistic test; on using a derived index rather than directly measured data, allied with the undisclosed facts that 1) the derived index suddenly in 2010 exhibits a huge and unexplained departure from the directly measured data which inflates the recent variance, and 2) the derived index series starts in 1899, thus excising the pre-1899 data that would reinforce the interpretation that nothing unusual is happening.

    Sorry, this paper cannot possibly be used as evidence to support the claim there is anything particularly unusual in the UK’s winter weather experience.

  5. September 10, 2014 12:00 pm

    Reblogged this on the WeatherAction Blog and commented:
    A longer time frame, a few hundred years, would be needed before drawing any conclusions.

  6. September 10, 2014 12:12 pm

    Is there a standard definition for “extreme” or is this just a term like “cold” or “hot” that is interpreted differently by everyone?

Comments are closed.

%d bloggers like this: