Central England Temperature

The longest single instrumental temperature record, one which has recently come under scrutiny in comments on this blog, is the Central England Temperature, or CET. The primary CET record consists of estimates of monthly average temperature from 1659 to the present. Let’s take a close look.

Here’s a graph of the CET data:

The first thing we notice is that the earliest data tend to fall along dense lines of whole-numbers values. This is because from 1659 through 1670, monthly estimates of CET are only given to the nearest whole degree C. From 1671 through October 1722 they’re only recorded to the nearest 0.5 deg.C, except for a brief interval from 1699 through 1706 when the data are to the nearest 0.1 deg.C. Only from November 1722 to the present do we have an uninterrupted record of monthly averages recorded to the neares 0.1 deg.C. In addition, there’s a record of daily temperature estimates from central England, but these don’t begin until 1772. Hence at the outset we can distinguish four episodes of varying precision in the data:

1659-1670	1 deg.C
1671-1722	0.5 deg.C
1723-1771	0.1 deg.C
1772-2008	daily data

Not only are the earlier episodes less precise by virtue of fewer significant digits, they were recorded when the science of thermometry was in its infancy. The alcohol thermometer was invented in 1709, and the mercury thermometer in 1714, so all thermometer records prior to that are due to different instruments whose accuracy and calibration are uncertain. Nor were these early measurements taken on modern scales; the Fahrenheit scale wasn’t invented until 1724, and the Celsius scale didn’t appear until 1742. Hence all thermometer measurements prior to the mid-18th century should be considered far less reliable than those taken later.

We can of course remove the annual cycle of the seasons by taking the difference between each month’s value, and the average for that month during a reference period. I’ll use the entire time span as the reference period, so that negative values are cooler-than-average months and positive values are warmer-than-average months:

Something caught my eye when viewing the anomalies, which is more evident when viewing a larger version of the graph (click to see the whole thing):

In the early years, there are dense areas with the same value. They’re highlighted in a closeup of the early years, here:

I haven’t read the documentation of the construction of the CET (if I were conducting a study for peer-reviewed publication, I surely would), but this indicates the possibility that during the early years, missing values were “filled in” with the average value over a longer time span. If true (and I emphasize the “if”), then those values aren’t direct measurements at all. This further emphasizes the reduced reliability of the earlier part of the record.

Using monthly values emphasizes the rapid fluctuations of temperature which are much more “noise” than “signal” in the long-term record. We can begin to reduce the noise by taking annual averages:

To make this easier to see, here’s a larger version (click to see the whole thing):

One thing that’s apparent is that most of the hottest years have happened recently; not to sound too much like a certain Nobel-peace-prize-winning former politician, but 9 of the 10 hottest years in the CET record have occured in the last two decades.

If we’re looking for climate rather than weather, even annual averages show more noise than signal. So I computed moving averages, using a 5-yr window, a 10-yr window, and a 30-yr window (the most common time period for computing climatological means):

As the averaging interval gets longer, the size of the “wiggles” in the smoothed data gets smaller. By the time we get to 30-year moving averages, the wiggles are greatly reduced but the signal still remains. This argues for the appropriateness of the “industry standard” 30-year window for climatological means. Hence I computed smoothed values on a 30-year time scale, using both moving averages and a Savitsky-Golay filter:

I’ve also plotted lines marking out the episodes of known changes in the precision of the data. It’s now clear that the recent rise in temperature is unlike anything seen in the more accurate part of the CET record. It’s also much warmer than anything seen in the entire CET record. If we plot only the most precise part of the data, from 1800 to the present, we get a very clear picture of the climatological changes in CET for the last two centuries:

In this limited view, the temperature increase during the modern global warming era is both quantitatively and qualitatively unlike any other part of the record. Even including all the earlier, less precise data, the recent warming has brought central England to a climate much warmer than previously measured.

The rate of warming in CET since 1980 is 0.05 +/- 0.02 deg.C/yr, or half a degree C per decade. If this trend continues, then by mid-century CET will have increased by a substantial amount, another 2 deg.C. This will bring CET to heights unknown for at least 350 years, probably several thousand years, and in all likelihood warmth not seen since humans inhabited the British Isles.