Monday, 3 October 2016

An examination of CET versus global temperature anomalies: Part I – overall and July to December

The Central England Temperature (CET) series goes back to 1659 and is the longest temperature record in the world. Like the global temperature record, CET shows a major increase since the 1970s, which I have documented in the first chart from an earlier post here but will reproduce for your interest:
Mean Central England Temperature for each fiscal year from 1659/1660 to 2014/2015, plus mean and percentiles for period from 1765/1766 to 1973/1974, reproduced from ‘Two “little ice ages” revealed by CET summer data’
Despite the similarity between CET and global temperatures over the long term – both are now about 1˚C or 1.8˚F above the mean from 1880 to 1974 – at shorter time scales there does not superficially appear to be any natural relationship between the two. The largest CET anomalies during the very cold months of February 1947 and January 1963 do not show a corresponding global decline, and what I previously observed as the coolest period globally around 1910 was not abnormally cool in the CET series, especially vis-à-vis an exceptionally cool period (taking the CET series as a whole) from March 1885 to January 1893.

However, as the map below shows, the period from 1903 to 1912 during which all but one of the “record cool” months at a global level occurred was actually hotter than 1885 to 1892 over the northern hemisphere’s land, but cooler over the tropics and oceans.
Moreover, the areas without data seem to have been hotter during the later period, which covered only a minor proportion of the great pluvial over Central Chile from 1898 to 1905. Wetness over Central Chile is known to be correlated with blocking, stronger anticyclones and colder conditions over West Antarctica, as shown below:

Mean 500 millibar (50 kilopascal) May to August geopotential height for 1898 to 1905 vis-à-vis the period since the second “magic gate” identified by Tim Flannery in ‘The Weather Makers’
Whilst we can see the anomalous wetness over Central Chile due to stronger onshore flow very easily from this chart, one cannot see the cold over West Antarctica that would be expected on the eastern side of the block. However, this is not very clearly seen from the temperature graph below:
The diversion of the Central Chilean pluvial put aside, in order to really look at the correlations between CET and global temperature anomaly over the period from 1880 to 1974, I have taken the following steps:
  1. Compiled an unrounded monthly mean CET for every month from 1880 to 1974 via the daily CET data, which can be downloaded online here
  2. Averaged the CET for each calendar month using only years from 1880 to 1974 inclusive
  3. Using an Excel file and my dowloaded GISS global temperature maps (which have an anomaly vis-à-vis 1880 to 1974 means to two decimal places), I compiled all the global means as independent variable and plotted CET against it as a dependent variable
  4. As a final step, I plotted scatter plots of CET anomaly against global temperature anomaly (versus 1880 to 1974 mean in both cases) and calculated the Pearson product-moment correlation.
There were too many data for a Spearman correlation and I lack any means of separating large number of equal ranks in the global temperature database, whose anomalies are naturally within a much narrower range than CET anomalies covering only about 0.1 percent of the planet’s ice-free land. 1880 to 1974 is chosen as the review period for two reasons: first, 1880 is when the earliest maps of global temperature begin, and secondly, 1974 is when, as shown by the sudden fall in runoff into Perth’s dams, the first “magic gate” caused by greenhouse emissions from the oil- and mineral-producing nations of Australia, South Africa and the Gulf States began controlling the climate. Because GISS cannot in its latest form provide fiscal year temperature means (which have the advantage in the northern hemisphere of not dividing abnormally cold or warm winters which provide most of the annual temperature variability) I have done only monthly data for the moment at least.

All Months:

Looking at our scatter plot of CET versus global temperature anomaly between 1880 and 1974, we do see some positive correlation both at a glance and with the correlation coefficient of +0.13730109. However, this coefficient is sufficiently small that variation in global temperature can explain a mere two percent of the variation in CET, since percentage explained equals the square of the variation.

It is of course possible that different seasons behave differently. Because the sun is weak or nonexistent, high-latitude winter temperatures (including those of the UK) are controlled almost entirely by air mass. Variations in wind vector could thus overwhelm global temperature trends due to variations in greenhouse gas concentrations or dimming from volcanic aerosols. In the hotter months, however, air masses possess less influence and sunlight of cloudiness has a more critical role: thus, one might expect a better correlation between CET and global temperature anomaly here.

It is also plausible that because the influence of greenhouse gas levels upon CET was shown in ‘Two “little ice ages” revealed by CET summer data’ to peak in the autumn and trough out from late winter to early summer, that we would see a peak in correlation in the autumn. We will look month by month through the fiscal year to see if either thesis can be supported, and I will divide the fiscal year into two halves to avoid having too much memory in one post.


July CET anomaly versus global temperature anomaly in ˚C, 1880 to 1974
The results here do not really support our prediction that a much better correlation between CET and global temperature anomaly: the virgin mean correlation is actually less than that for all months, when we would expect high values in mid-summer. The smallness of our sample size (95 months vis-à-vis 1,140) means that strong outliers are very easy to see, notably the famously hot and dry July 1911:
As we can see, there are, apart form the generally cool Southern Hemisphere three very substantial cool regions during the UK’s hottest month since July 1852, and Europe’s hottest since July 1859 – a belt from Finland to the Black Sea, China and eastern Russia, and most of the US and Canada outside of Québéc. Newfoundland and New England, where Vermont, New Hampshire and Maine experienced records for heat that still stand today. On the fifth of July, Nashua recorded 106˚F or 41.1˚C.


August CET anomaly versus global temperature anomaly in ˚C, 1880 to 1974
In contrast to July, August’s results do suggest that the summer months with more and less indirect sunlight should provide a better correlation between CET and global temperature anomalies: the Pearson correlation coefficient is almost twice that for all months. The result does however support the thesis of increased correlation in the autumn months which are being approached in August.

However, global temperature anomalies for August still can explain only about one-seventeenth of the variance of CET between 1880 and 1974. 1911 is again a huge outlier:
Global temperature anomalies for August 1911. Note the alternating patterns of heat and cool across the Northern Hemisphere, and the consistent cool in South America and over the oceans


September CET anomaly versus global temperature anomaly in ˚C, 1880 to 1974
As the sun moves southwards we would expect the correlation between CET anomaly and global temperature anomaly to weaken. This is observed vis-à-vis August, suggesting that the July sample may be too small at ninety-five years to eliminate the effect of outliers. Septembers of 1944 (the hottest year globally between 1880 and 1974) and the chilly autumn of 1952 are the most substantial outliers:
September 1944 global temperature anomaly – note the cool area centred around the UK amongst an otherwise circumlatitudinal hot anomaly
September’s poor correlation coefficient also opposes the thesis of highest correlation in the autumn months: the fall from August would not be expected if greenhouse gas concentrations have c=most consistent effect in the autumn as observed in recent CET records.


October CET anomaly versus global temperature anomaly in ˚C, 1880 to 1974
October seems to contradict our first thesis: the correlation between CET and global temperatures is greater than for July or September, although between 1880 and 1974 global temperature variation can explain only 4 percent of variation in CET. It does support the thesis of an autumn maximum in correlation, although the Pearson is less than for August. The major outlier of October 1896 – Central England’s coolest since 1817 but 0.13˚C hotter than average globally – shows the classic alternating “sinusoidal” pattern of hot and cool anomalies, with a very cool band centred upon Iceland and Western Europe and a hot band on Eastern Europe and Russia away from the Pacific:
October 1896 global temperature anomalies. Note the cool over Western Europe and the Great Lakes region and heat over Eastern Europe, northern Russia and Australia


November CET anomaly versus global temperature anomaly in ˚C, 1880 to 1974
As we move towards winter, both theses outlined at the beginning of this post would predict a decline in the correlation coefficient between global mean temperature and CET. Vis-à-vis what we saw for October, November does seem to suggest this: the correlation coefficient is down to 9 percent and global temperature anomalies can predict less than one percent of those for CET. Striking outliers are completely absent despite the low correlation coefficient.


December CET anomaly versus global temperature anomaly in ˚C, 1880 to 1974
The two theories noted at the start of this post give different predictions of how the correlation coefficient between CET and global temperature anomalies would change from November to December: an autumn maximum (October) would predict a rise, whereas a winter maximum would predict a continuing fall in the correlation coefficient.

What is actually observed is almost exactly the same correlation coefficient between CET anomaly and global temperature anomaly as for November. There are two major outliers from the general pattern:
December 1910 global temperature anomalies. Note the hot area of northern Africa and Europe, and the very cold area over Siberia and China
December 1910 is interesting not so much because of the contrast between Europe and the rest of the world, but because unlike in a conventional positive NAO situation, hot conditions extended through the Sahara, which is normally cool during positive NAO winters. As can be seen from the 500 millibar anomaly chart below, there was unlike typical positive NAO winter months much blocking over Greenland, but not a classic low-latitude block that would produce cold northwesterly flow, nor a high-latitude block in the correct place to produce a very cold easterly flow:
Departures from 1880 to 1974 mean of 500 millibar heights (in metres) over the northern hemisphere for December 1910. The block over the Black Sea and Ukraine is decisive in shaping the observed temperature pattern.
December 1939, in contrast, is notable as relative to the 1880 to 1974 mean globally the second-hottest month between 1880 and 1978 (behind January 1944 by 0.02˚C) and the relatively hottest over the Northern Hemisphere. Over the contiguous United States it was the hottest December between 1890 and 2014, although December 1889 was certainly hotter there due to remarkable heat in the East and Plains.
December 1939 global temperature anomalies vis-à-vis 1880 to 1974 averages. Note the unusual warmth over the interiors of Asia and North America – this was the latter continent’s warmest December of the century.
Whereas the colder weather of the Chukotka Peninsula relates to the fact that the anomalous high-index westerly airflow was offshore, Europe’s failure to participate in the Northern Hemisphere’s anomalous warmth was because a block over Greenland and Canada, which drove warm air into northern North America throughout this month (and the four cold “War Januaries” of 1940, 1941, 1942 and 1945), drove colder air into Europe. Unlike the “War Januaries” this block was not circumpolar and the easterly flow into the UK much weaker:
December 1939 500 millibar height anomalies in metres. Note the blocks over Canada and Central Asia driving warm Atlantic and Pacific air into Asia and North America.
In Part II of this post, we will look at January to June and make an overall examination of the correlation trends. So far, evidence of seasonality in correlation between CET anomaly and global temperature anomaly is quite indecisive and there remains the possibility that no differences exist between seasons.

No comments: