HadCRUt4: revision or revisionism?

Guest essay by Christopher Monckton of Brenchley

As Bob Tisdale has recently pointed out, the monthly temperature anomaly series published by the Hadley Centre and the Climatic Research Unit at the University of East Anglia has just been extensively revised.

A commenter on a previous posting about the temperature record asked whether the entire 163-year HadCRUt4 dataset could be displayed graphically, with the uncertainties (red and green) and trends illustrated. Here is the requested graph, based on the revised data:

The least-squares linear-regression trend across the entire 1959-month period shows warming equivalent to 0.47 Cº/century (0.9 Fº/century), well within the natural variability of the climate. The 95% (2 σ) confidence interval falls between 0.33 and 0.6 Cº/century, as shown by the trend-lines flanking the central trend.

The result should be adjusted to allow for the finding in Michaels & McKitrick (2007) that urban heat island effects and other extraneous influences over the past 30 years have led to overestimation of the warming rate over land by as much as double. On the assumption that this bias may have existed since 1850, the true warming rate since then is equivalent to just 0.4 Cº (0.7 Fº) per century.

The bounds of the 2 σ confidence interval converge because measurement uncertainties are thought to have diminished with improvements in methods and reliability.

On the other side of the account, terrestrial coverage has declined sharply over recent decades. For this and other reasons, the HadCRUT4 record takes explicit account of three distinct species of uncertainty: measurement and sampling, bias, and coverage. Combining the effects of these three, the 2 σ uncertainty bounds today are approximately one-sixth of a Celsius degree either side of the central estimate:

In March 2013, for instance, the central estimate of the anomaly was +0.412 Cº. However, the lower and upper bounds of the 2 σ interval were given as 0.249 Cº and 0.569 Cº, respectively 0.163 Cº and 0.157 Cº distant from the central estimate.

The discrepancy between the projections in the IPCC’s Fifth Assessment Report (red central projection bounded by orange region) and the outturn in HadCRUT4 (bright blue trend-line) is startling. The difference between the observed cooling at 0.86 Cº/century and the predicted warming of 2.33 Cº/century is equivalent to a hefty 3.2 Cº/century.

However, the period of record since the backcast “predictions” began in 2005 is short. And one does not want to suggest that part of the reason why the HadCRUt4 projections show steeper cooling than the satellite datasets is that the terrestrial results are now being tuned to bring them into correspondence with the more accurate and complete satellite results, and that, since the 20^th-century warming has been – as it were – enhanced, there now has to be a corresponding disenhancement of the 21^st-century warming.

The head of the World Meteorological Organization tried to have me thrown out of the Durban climate conference in December 2011 because I had dared to question his assertion that one should study temperatures over 30-year periods. I had politely pointed out that the warming and cooling phases of the Pacific Decadal Oscillation each last 30 years, so that to cancel them out one should either study periods that are multiples of 60 years or study periods centered on the year of a warming-to-cooling or cooling-to-warming phase transition.

“Are you presuming to lecture Me?”, he bellowed. The event was captured on film by one of my CFACT colleagues, and the WMO’s flimsy pretext for having me banned was that it had not given permission for the filming. So I skydived into South Africa from a great height and told the organizers that they had made the mistake of acting unreasonably within a formerly British jurisdiction, so that their actions would be subject to judicial review.

They said they were immune from judicial review (as they are about to find out, this is not the case). But, since the local papers had had enormous fun at their expense because they had tried to ban me, they let me back into the conference.

With that background, here is the graph of temperature change in the 60 years since 1953:

The warming rate throughout the period during which we could in theory have influenced global temperatures has been equivalent to just 1.14 Cº/century.

The period since 1990, the year of the first IPCC Assessment Report, falls more or less neatly either side of the warming-to-cooling phase transition in the Pacific Decadal Oscillation in late 2001, broadly canceling any bias from this and the related ocean oscillations (see Tsonis, 2006, for an interesting discussion). Here is the graph:

The warming rate is not much distinguishable from that since 1950: it is up by less than 0.3 Cº/century. Besides, as the IPCC is now learning to its cost, one cannot draw any conclusion from the relative slopes of multiple arbitrarily-chosen trend-lines on the same data about the rate at which global warming is accelerating.

There have been some significant changes between HadCRUt3 and HadCRUt4. In the earlier dataset, the warming rates from 1860-1880, 1910-1940 and 1976-1998 – the most rapid supra-decadal warming rates in the entire record – were statistically identical at 1.6 Cº century, as Lord Leach of Fairford discovered on St. George’s Day, 2009, when he asked Her Majesty’s Government –

“… whether the rate of increase in global mean surface temperatures between 1975 and 1998 was similar to the rates of increase observed between 1860 and 1880 and between 1910 and 1940 …”

Lord Hunt of King’s Heath replied –

“Observations collated at the Met Office Hadley Centre and the University of East Anglia Climate Research Unit indicate that the rate of increase in global average surface temperature between 1975 and 1998 was similar to the rates of increase observed between 1860 and 1880 and between 1910 and 1940 (approximately 0.16 C° per decade) [1.6 Cº/century]. …”

There is a tendency towards progressive revisionism in the terrestrial temperature datasets, in a manner calculated to depress temperatures substantially in the earlier part of the record and to nudge them upward a little in more recent times (but not by too much: the satellites are watching).

The effect of this tampering has been inexorably to steepen the apparent rate of warming since 1850, so as to sex up the dossier and make global warming seem to be a bigger problem than it is. This tendency, long evident in the untrustworthy NASA GISS dataset (J. Hansen, prop.), is also detectable in the HadCRUt series.

A version of HadCRUt3 more recent that that used in Lord Hunt’s answer gives warming rates equivalent to 1.0, 1.6 and 1.8 Cº/century respectively over the three periods. HadCRUt4 continues the tampering trend. It gives warming rates equivalent to 1.0, 1.4 and 1.9 Cº/century respectively. The following revealing table summarizes the changes:

It is almost as though Lord Hunt’s answer, demonstrating the inconvenient truth that the three most rapid supra-decadal warming rates in the global instrumental record were statistically identical, and implying that there has been little or no man-made acceleration in global warming, caused such concern among the usual suspects that they dropped it down the memory-hole, replacing it with successive new and more politically-correct versions of the Party Line.

Yet even the highest of the exciting, freshly-minted supra-decadal warming rates in HadCRUt4, at 1.9 Cº/century, is less than half the warming rate equivalent to 3.9 Cº/century that prevailed for 40 years in Central England (and, inferentially, globally) from 1695 to 1735. Those four decades entirely preceded the onset of the Industrial Revolution in 1750. We could not have been to blame. It was (and largely still is) the Sun, stupid.

Accordingly, even on the much-altered HadCRUt record, the anthropogenic component is in the warming observed since 1850 seems rather small and difficult to detect. As Bob Tisdale has pointed out (and if you have not seen his slide-show on el Niño events and global temperature change, you should), even the historically warm temperatures since 1998 were elevated to their new plateau in a step-change by the naturally-occurring Great El Niño of that year.

Fred Singer has been pointing this out for years, but the usual suspects have kept their fingers in their ears, preferring to conduct bogus and insulting pseudo-psychological surveys about why skeptics are skeptical rather than looking at the most obvious reason why skeptics are skeptical: the science itself.

Professor Ian Plimer says that there is some evidence to suggest that undersea volcanic tremors occur in the equatorial eastern Pacific for six months before each el Niño begins there. If so, the el Niños, including the Great el Niño of 1998 and its two predecessors over the past 300 years, are of volcanic and not of anthropogenic origin.

For how long has there been no statistically-significant warming? In the following graph, the zone of statistical insignificance is shown in pale blue. Since the entire central trend-line falls within that zone, the warming since February 1996, more than 17 years ago, is not statistically distinguishable from zero.

On the evidence summarized in these graphs, there is no rational case for taking drastic and costly steps to mitigate global warming. It will be cheaper, and wiser, to wait and see.

More detailed and larger graphs are available in this PowerPoint file: wuwt-hadcrut4