Temperature Trends from 1659 to 2017 (Central England Temperature series)

M.T. Cranium

This is a detailed report on my research into trends in Central England temperatures as provided by the UK Met Office on the websites indicated below. I also provide an excel workbook with the data being discussed here.

To see the data, go to

http://hadobs.metoffice.com/hadcet/mly_cet_mean_sort.txt

which sorts the monthly and annual data by rank. You can navigate within that link if you prefer to see the data in chronological order. However, that presentation is also available in my excel file. You can download that from the link at the end of this post:

At this point I am going to post this short summary to make sure the excel file is embedded, and go on to a discussion of what is shown in the analysis of 30-year running means of each month and the year as a whole.

First, a brief overview of major trends in the data.

The "Maunder minimum" was a period from about 1660 to 1710 when solar activity was very low compared to most decades since then. This was a colder time in all available European data and the deepest cold was around 1684 to 1700. The year 1686 broke with this trend and was warmer than some years in the recent past. Also, it is evident that the cold was developing through the first decade of the series, the 1660s were not as cold as the 1670s. At the deepest point of this cold spell, most monthly means and the annual means were depressed by 1 to 2 deg compared to the long-term or more recent averages.

The climate of central England (which is probably well correlated with most locations within 1,000 kms at least) then warmed slowly through the first two decades of the 18th century and reached a temporary peak of warmth in the period 1729 to 1738 when some years were as warm as the modern averages. During that warm spell, it is particularly noteworthy that September became considerably warmer than in the Maunder, and reached modern values around 14.0 C for several decades.

The winter of 1739-40 was harsh and marked a sudden downturn in average temperatures that was maintained to some extent into the 1760s. It did not become as cold as the Maunder, but settled into an intermediate zone about a degree colder than the peak warmth earlier.

The 1770s to 1790s were a period of increased variability without much of an overall warming from the 1760s, although some is evident. The frequency of both warm and cold daily records (which begin in 1772) is higher than average in these decades. Most of that period was a time of high solar activity combined with occasional large volcanic events globally. Those conflicting drivers might explain why the circulation appeared to become rather chaotic for some period.

The "Dalton minimum" in solar activity lasted from about 1795 to 1835 and the coldest signal from that appeared around the mid-1810 decade. Although there was a major volcanic dust veil from Tamboro 1815-17, this would not explain why 1814 and early 1815 were very cold, so a combination of natural signals must be the cause. Despite no upturn in solar activity, the 1820s became a considerably warmer decade, while the 1830s were notable for high variability.

The mid-19th century was a period of relatively strong solar activity but this was not necessarily reflected in temperature trends. Some months show a colder trend at this time. April, May and October show the greatest downturn relative to trends of all months. This might show that two different scenarios were competing and forms a theme for further research since the outcome is rather complex. Winters in the mid-19th century showed a tendency to high variability, some were very cold and some very mild. Summers were not as different from long-term trends as the other seasons proved to be.

In the 1880s there was another downturn in solar activity combined with a large global dust veil event (Krakatoa) and the signal from these drivers appears to have peaked around 1887 to 1892, a generally cold period.

The climate erratically warmed after about 1893, a trend also visible in North American data. There were some clusters of colder years around 1905 and 1915 (this period continued to have rather low solar activity) but in general the "modern" climate was developing with milder winters becoming normal after 1895. Winters were particularly mild in the 1920s and 1930s. An interesting trend visible in monthly data concerns the coldest part of winter. This had normally been January and was sometimes shifted to a December-January overlap (most noticeable around 1790s). However, February began to provide as many cases of coldest winter month by the early 20th century, and the mean temperature for January and February have never been very far apart since about 1900. The year 1921 was a high water mark in the evolving warming trend, not again reached until 1949.

The early 1940s were rather cold and of course there was a notably cold winter in 1946-47, followed by about three years of generally warm years.

The 1950s and 1960s were generally rather cool decades against the modern background although about 1.5 deg warmer than the Maunder minimum. The mid-1980s were also quite cold compared with most recent data. Otherwise, from about 1970 to 2017, the general trend has been warmer than average and this has of course been increasing since about 1988, although appears to have flattened out since about 2006. The warmest 3 and 4 year periods in the data were back around 2003 to 2007. The year 2010 was anomalously cold and is near the average for the first six decades of the data, mainly due to very cold months at each end of the year.

So far, the year 2014 ranks as warmest (10.95 C) but this year has been running about on par with 2014 to mid-July and would become the warmest year if the July to December average is more than 0.2 higher than in 2014.

Anyway, have a look at the data file and in particular the graph at the bottom of the data section, that graph shows the trends of 30-year variations from start to finish, for each month as well as the year as a whole. There are some interesting features such as how September has gradually been catching up to June over 360 years of data, and how March and November have varied, with March sometimes catching up to November then falling back. In these monthly variables, there may be more useful information available to flesh out the bare bones of the climate change theorem (whatever you may think of it), because there is no implied reason for either adherents or critics to suppose that months will vary at different rates and time scales. It may be just random noise but I suspect there are implications for teleconnections buried in some of this second-order variability.

---

Data in this file from Hadley Centre, UK Met Office

M.T. Cranium

Looking at winter data in more detail ...

January and February had similar mean temperatures (over 30-year intervals) during the Maunder minimum. January begins to become colder than February in increasing differentials after 1700 and through the milder period of the 1710-1739 interval. At this point, February starts to edge closer to December which is making less progress upward and briefly catches up in the 1713-42 interval (both at 4.0 if we use 1712-41 for December (same 30 winters) while January is at 3.29 C at this time.

January then becomes increasingly separated from both February and December through a colder interval in the mid to late 18th century. The greatest separation achieved is around the time of January's minimum 30-year average, 2.09 in 1766 to 1795. Here, February is 3.83 and December 3.6 (for 1765-94). A few years earlier, December had become second coldest month, a position it then occupied until the mid-19th century (1842-71 is the exact interval when February resumes second place). While that transposition was in place (about 60 years so that for two generations this was the normal state of affairs in winter) January continued to run much colder than in the present or even most of the 20th century. A second minimum of 2.30 was recorded in the Dalton minimum (for 1802-31) followed by a third of 2.25 (1809-38).

By about the 1851-80 period, January was steadily warming while February and December held more steady around 4 C. After the interval of 1849-78 December resumed second coldest place and became barely warmer than January, the closest they came was around 1865 to 1900 (any 30 year interval) when December was generally just 0.2 milder than January (which was then up to 3.3 C). This coincided with a time when February was considerably milder than both (4.45 in 1865-94).

Although March has never been colder than any winter month (over a 30-year average) it became quite close with February at the end of the 19th century; winters in that period could be described as starting early, finishing early but with a very strong chance of a second wintry episode. I suppose that 1917 might represent an extreme case of this tendency, but in the period 1865-1894, February's mean of 4.45 was followed by a March of only 5.0 (it had been in the 4.9 range slightly earlier but February was relatively colder then).

The period in which December was then second coldest (we noted that it started with 1849-78) ended with 1882-1911 whether we consider the same calendar year or the same winters (Dec 1881-1910 that is). So that was another 33-winter period of a colder December, so that from about 1750 to 1910, most intervals experienced a colder December than February.

Since about 1910, the three months have continued rather closely bunched, and January finally catches up to February in the interval 1901-30 (Jan 4.23, Feb 4.22) and for about ten further 30-year intervals, the difference remains very small. January begins to resume its status as coldest month with the biggest separation coming in the 1938-67 interval (3.15 to 3.87). These differences were maintained through all 30 intervals that included the very cold Feb 1947 except for the last one, 1947-76 when the months reversed for a few intervals off and on through the end of the 1980s. February has remained marginally colder than January in most but not all 30-year intervals since 1957-86, the exceptions were between 1971-2000 and 1980-2009 where January was colder. All through the period discussed in this paragraph, December was rather mild with averages usually around 4.4 to 4.6 C and edging up towards 5.0 until the very cold December of 2010 entered the intervals and pushed December back down towards the close-run January and February. The December average for 1981-2010 dropped 0.2 but has since recovered (only 1981-2010 had both 1981 and 2010 included).

The most recent averages for the three months show that February may become milder than Deccember for the first time in over a century. The table lists a lower value now for December but that's a 29-year average for 1988-2016 until we can enter December 2017 in the table. So at the moment, with February at 4.96 (January at 4.74) December will need to average 5.0 or lower to get the current December average below 4.96.

The next post will examine differences between March and November through the period.

M.T. Cranium

March and November represent the occasional extremes of the winter period in some years; November has sometimes been the coldest month of winters and provided the coldest day of winters, and so has March (less frequently). On the other hand, the normal state of affairs especially in recent decades is for November to be very much of a continuation of the mild autumnal pattern (an exception being late November 2010), and for March to see only modified cold spells if any (one exception being 2013).

March has always averaged colder than November. I looked at the 358 years of monthly temperatures (excluding 2017) -- about one third had a colder November than March, and about one tenth had the same value. The overall margin is 0.8 C in favour of November (6.1 to 5.3).

The difference has varied considerably over the period of record.

During the Maunder and first half of the 18th century, it remained fairly static at about 1 C deg through the cold and milder interludes. The separation began to narrow after about 1750. By the interval 1750 to 1779, the difference was only 0.22 C deg (March 5.34, Nov 5.56). There were runs of several years when March was milder than November (for example, eight of ten years 1773-1782). This smaller difference was maintained generally into the early 19th century but then November widened the gap again, hitting a peak of 6.34 for 1817-46. March had a much flatter peak a few years later and stayed around 5.5 C through the mid-19th century. Averages were coping with the very cold March of 1845 so perhaps there was generally more equal warming of these two months than the numbers reveal.

After the 1840s, November means began to tumble while March just got slightly colder, so the difference narrowed until it reached only 0.16 in the 1853-82 interval (March 5.29, November 5.45). This is as close as they ever came, the gap slowly increased again later in the 19th century and by 1888 to 1917, it was over 1.1 degrees (6.34 to 5.22). That separation slowly narrowed again into the early 20th century. The next minimum separation was about 0.6 from 1909 to 1938.

Since around 1940 the two months have continued to warm with the separation increasing to almost one degree for most of the past half century, but recently the gap is closing again, and in 2016 (1987-2016) it stood at 0.56 deg (7.24 in Nov and 6.68 in March). A cold November this year could reduce the ongoing trend to 0.4, a mild one would retain the 0.6 difference and an average November around 0.5 for the 1988 to 2017 period.

The next post will look at the April-October, March-April and October-November patterns.

M.T. Cranium

When comparing April and October in the graph, note two things, they happen to have the same coloured lines (labelled to left) and October (the higher of them) has the annual mean tagging along just below it.

The normal difference is 1.8 degrees (9.7 for October to 7.9 for April) and this separation has seldom varied all that much, except for a period in the late 19th century into the early 20th century. Here, October became rather cold while April stayed closer to its long-term average. The smallest separation of only 0.87 deg takes place in the interval 1867 to 1896 (Oct 8.92, Apr 8.05). You have to wonder if coal smoke might be a factor in low-sun October showing this temporary disadvantage; it also dropped below the annual mean briefly. That period starts with 1866 to 1895 and ends with 1873 to 1902. The October-April difference was back to about 1.5 degrees by 1881 to 1910 and continued to widen after that, reaching a maximum of 2.12 degrees for 1901 to 1930 (9.77 to 7.65). The separation then reduced slightly before expanding to as much as 2.71 degrees for the period 1957 to 1986 (10.65 to 7.94). This has been decreasing steadily and now stands at 2.14 degrees (10.92 to 8.78) for 1987 to 2016.

Now, how about March and April as springs warm up?

The average warming is 2.6 degrees, but there are two intervals when March was getting colder while April was getting warmer. The first of those was around the late 18th century into the early 19th century (around the Dalton minimum). For 1778 to 1807, the separation increased to 3.27 degrees (8.11 to 4.84). Things got back to more average warming for the mid 19th century, then another peak of spring warming was reached in the mid-19th century. In the interval 1840 to 1869 the average warming was 3.13 deg C. This gradually reduced into the 2.5 range later in the century. It reached a minimum of 2.06 deg C for 1910 to 1939 (5.70 to 7.76). In the past eight decades the separation was slowly increasing (springs warming faster) until a peak of 2.8 around 1942 to 1971; since then the rate has dramatically slowed, and stands at only 1.9 C now (it was as low as 1.8 about ten years ago, i.e., for 1976 to 2005 when it averaged 1.78).

Then, how about the pace of autumnal cooling from October to November? That is on average about 3.6 C deg (9.7 to 6.1).

This shows a tendency to pulse in periods of about 50-60 years. The peaks in cooling can be found around these intervals:

At the start of the period of record, cooling was almost 4 deg for 1659 to 1688 and it steadily decreased to 3.2 deg around the end of the Maunder period (intervals spanning years around 1685 to 1714). The cooling then steadily increased for much of the 18th century.

The next peak of cooling rate was in the late 18th century, 1787 to 1816 had an average Oct to Nov cooling of 4.36 deg.

That rather quickly reduced to 3.33 deg for 1821-50. The next peak of cooling rate was established in the interval 1851 to 1880, also 4.36 C deg tying the previous mark. These two have never since been duplicated.

Cooling rates dramatically decreased until they reached only 2.78 deg for 1879 to 1908.

A fourth peak of cooling, 3.93 C deg was observed in 1901 to 1930.

That slowly reduced down to 3.19 C deg 1926 to 1955, and has since reached another peak of 4.03 C deg for 1949 to 1978.

Assuming there may have been a peak just before the period of record around 1651 to 1680 in arbitary terms, the separations of the five peaks are 136 years, 64 years, 50 years, and 48 years. Another peak of late autumn cooling may be just ahead at this rate (the first period was a lot longer than the next three cycles)

M.T. Cranium

One of the most interesting comparisons is September and June CET values.

The basic overview is that both have shown similar tendencies over the period of record, with one notable exception, but also September has been slowly catching up to June. The difference between them overall is 1.0 deg (June 14.3, September 13.3). But by 1987-2016 this had narrowed to 0.33 degrees (June 14.48, September 14.15).

The two months were about 1.5 degrees apart in the Maunder minimum period. June was not that chilly in the Maunder, managing to stay just below 14 C, but September was a lot cooler then with means in the mid-12 C range, considered quite cold nowadays. This difference narrowed to less than 1 degree around the mid-18th century. At this time June was about as warm as nowadays, or even slightly warmer (the intervals from 1726 to 1755 had an average of 14.64), and September was close to 14.0 C peaking in that same interval at 13.97 C.

Both months began to cool slightly later in the 18th century but June recovered while September continued to cool into the Dalton minimum period. June managed another peak value, 14.70 for 1772 to 1801, just after September had bottomed out at 13.19 C for 1764 to 1793 and 1765 to 1794. June then cooled slightly and September remained rather cool reaching 13.12 several times around 1786 to 1815.

The two months had their one instance of different trends around the mid-19th century. The coldest period for September after the Maunder was around 1812 to 1841 and also 1813 to 1842 at 12.99 C but slightly colder readings followed in the mid-19th century, the lowest being 12.88 for 1835 to 1864. But while September was in this chilly mode, June hit an all-time peak of 14.72 for 1822 to 1851 and remained above 14.5 until past 1860.

After about 1880 the two months resumed their parallel trends with some gradual warming for both into the mid-19th century. June, unlike most other months, has shown no great tendency to modern warming. The highest value it displayed after the 19th century maximum was 14.57 C for 1931 to 1960. June then cooled a little and reached 14.11 by 1962 to 1991. September, on the other hand, has begun to close in on June quite noticeably since about 1950.

This differential must imply that the summer season is shifting later in time, with little change in its June onset but a gradual tendency for summer conditions to persist well into September. The "old" climate especially that of the Maunder and some colder parts of the 19th century often saw summers that began to fade even in August and little if any summer warmth in September. June, however, was always about the way it is nowadays.

July and August have shown a marked tendency to keep pace, but the cooler climate intervals, notably the Maunder and most of the 19th century, produce larger differences between these peak summer months than other, warmer periods. In warmer intervals, August stays within 0.1 of July, whereas in colder periods, they exhibit larger separations of about 0.3 to 0.6 degrees. For example, at the coldest point in the Maunder, 1671 to 1700 July had an average of 15.47 and August 14.90. And in the coldest part of the 19th century, 1836 to 1865, July was at 15.56, August 15.25.

The most recent difference was only about 0.2 degrees, but August has on three occasions come very close to overtaking July and for just one interval it briefly did so; these were observed in the following three 30-year intervals (some years near the 1702 to 1731 peak had smaller differences than 1787 to 1816 and some years near that one, smaller than the most recent event).

INTERVAL ____ JULY __ AUG ___ diff

1702 to 1731 __ 15.84 _ 15.85__ --0.01

1787 to 1816 __ 15.86 _ 15.76___ 0.10

1975 to 2004 __ 16.59 _ 16.46 ___ 0.13

The greatest differences were all around the same values:

1672 to 1701 __ 15.55 _ 14.92 ___ 0.63

1808 to 1837 __ 15.87 _ 15.24 ___ 0.63

1900 to 1929 __ 15.85 _ 15.21 ___ 0.64

M.T. Cranium

September was the warmest month of only one year in the period, namely 1890. It came very close in 1865 as well.

June has been the warmest month fewer than thirty times, the most recent being 1970.

Of the remaining years, it is more often July than August (about 55 to 38 per cent).

There are various tied scenarios, most often July tied with August, and one year (1662) with June, July and August all tied.

The coldest month stats are just about a mirror image of summer from November to March although more than one March has been coldest and November does better than May which never qualified even in 1833 with the exceptionally warm May (15.1). January is somewhat more frequent than February and December makes an occasional appearance, last time being 2010 (for winters or for years, about the same profiles).

The difference between winters and summers is interesting to track. I have just added that stat to the table you can download, so that if you download it now, you'll see the graph being discussed (it does not appear in the earlier version).

This gives some indication of the continental influence on British climate.

The difference in the Maunder minimum and for much of the 18th century was static around 12 degrees.

The largest differentials and therefore the most continental form of the climate came in the intervals 1757 to 1786 (12.36 degrees) and 1771 to 1800 (12.37 degrees).

The smallest differential, likely indicating the strongest zonal flow period, came in the early 20th century. The lowest value is 10.55 degrees in the interval 1903 to 1932. Mild winter and cool summers were common in this interval.

The index value then increased again and the highest value in modern times was 11.67 deg for 1940 to 1969. You could interpret this as a peak of high latitude blocking which gives the result of cold winters and warm summers.

The index is now on the way down generally and stands at 11.04 degrees.

The average for all intervals was 11.57 degrees. Only sixteen consecutive values (those ending 1955 to 1970) were higher in the 20th century. The second half of the long period of record averages about 11 degrees.

The year 1947 averaged a winter to summer difference of 15.9 degrees. That establishes an extreme value for this particular statistic at least in recent times. There may have been slightly higher values in the 18th century in particular years.