Christopher Monckton and the Great Pause

Christopher Monckton of Brenchly posts a monthly article on the Watts Up With That website updating the extent of what he calls The Great Pause. Each month he gives a very specific time frame, saying exactly when it started and how long it has been going to the nearest month.

His most recent post (based on the January 2016 figures) is called The Pause hangs on by its fingernails, archived here. In this he proclaims that the Great Pause is 18 years and 8 months old, having started in June 1997. He also warns that the Pause will end next month. I think he's being a little pessimistic here, as it will require a record breaking February for this to happen. We will see in a few days time.

Here's what the Great Pause looks like:

How is Monckton able to get such a precise time for the start of the Great Pause? He can do this because he has his own definition of a Pause, which makes little sense statistically. What he says is:

The hiatus period of 18 years 8 months is the farthest back one can go in the RSS satellite temperature record and still show a sub-zero trend. The start date is not cherry-picked: it is calculated. And the graph does not mean there is no such thing as global warming. Going back further shows a small warming rate. And yes, the start-date for the Pause has been inching forward, though just a little more slowly than the end-date, which is why the Pause has continued on average to lengthen.

As he says, his starting point is calculated each month - that calculation is to look at every month from the start of the RSS data and check the trend, from there to the current month. As soon as he finds a month that gives a negative trend, he declares that that is the start of the Great Pause.

There are many problems with this definition. For a start it assumes some magic property relating to the difference between a negative and a positive trend, even if they differ by just a few hundredths of a degree per century. Then there's the fact that it is only defined up to the current date, which means its start date can change each month, and as he says may disappear entirely. Think about that. One month there is a Great Pause lasting exactly 18 years and 8 months- the next month it has softly and suddenly vanished away. This is not a robust definition.

What I find interesting about Monkton's Pause is that many of the problems with it are things that Monckton himself has warned about them in the past, often quite vociferously.

Cherry Picking the Start Date

The IPCC 4th assessment report contains this graph.

FAQ 3.1 How are Temperatures on Earth Changing?

Monckton does not like the graph at all. He's still complaining about it in this current article - referring to it as ... one of the most mendacious graphs in the IPCC reports.

I'm not particularly keen on it either - it's not so much the graph but the gloss the IPCC put in the description that I find problematic (my emphasis):

Linear trend fits to the last 25 (yellow), 50 (orange), 100 (purple) and 150 years (red) are shown, and correspond to 1981 to 2005, 1956 to 2005, 1906 to 2005, and 1856 to 2005, respectively. Note that for shorter recent periods, the slope is greater, indicating accelerated warming.

The problem here is using different length trends to infer acceleration. In general you would expect longer trends to show less variation than shorter ones. You cannot necessarily conclude that a greater short trend proves acceleration, anymore than you can infer that a lesser trend proves a slowdown.

Monckton has much stronger objections to the graph, and focuses on what he calls the endpoint fallacy, which he says is a statistical lie.

Dr. Pachauri's mistake, which was politely unchallenged by his learned audience, is an elementary instance of the insidious but well-known statistical fallacy known as the endpoint fallacy, by which inappropriate conclusions are drawn through the careful selection of the start-point and end-point of a graph.

SPPI Monthly CO2 report - February 2009

It's clear from Monckton's description that by endpoint fallacy he's talking about cherry picking. He is saying that by choosing the endpoints of a trend line carefully you can get any trend you want.

But isn't Monckton's technique for finding the Pause specifically based on carefully selecting the start point that gives the longest negative trend? As Monckton was answering questions in the comments section, I thought I'd ask him about this:

Monckton of Brenchley,

A few years back you made several references to the endpoint fallacy, where you said:

This is a statistical lie known as the start point or endpoint fallacy. Where you take a jiggly up and downy dataset like temperature, where you don't know which way it's going to go next, a stochastic dataset. If you choose your start point and your endpoints carefully enough you can make it look as though any trend you want is happening. Here they've tried to show a rising trend.

Could you explain how your pause, defined as

The hiatus period of 18 years 8 months is the farthest back one can go in the RSS satellite temperature record and still show a sub-zero trend.

is not an example of the endpoint fallacy?

Here is his response in full:

Bellman appears poorly schooled in logic. The very sentence he cites explains that the start-point of my monthly Pause graphs is calculated: it is not plucked arbitrarily from the ether.

The central point is very simple. One-third of all anthropogenic forcings since 1750 have arisen during the period of the Pause. And that period is now of sufficient length to pass the NOAA test: i.e., 15 years or more without any global warming indicates a discrepancy between the predictions and the observations.

What is more, NOAA were talking of ENSO-adjusted data. After adjustment for ENSO, the discrepancy between prediction and observation is actually wider than before making the adjustment.

There is, therefore, a problem with the official global warming theory. The problem is that the warming is either not happening at all, as the satellites suggest, or not happening at anything like the predicted rate, as the terrestrial tamperature datasets suggest.

Given that the head posting considers two satellite and three terrestrial datasets over periods commencing with three distinct starting dates, each chosen because it was the year in which IPCC had made another set of predictions against which the observed trend could be tested, the suggestion that the RSS graph in the head posting is an instance of the endpoint fallacy is feeble-minded.

Only the first paragraph has any relevance to my question, and all it seems to say is that it's not an example of the endpoint fallacy because the start-point is calculated. This is a very odd distinction, as calculating a start point to give you the desired result is, self evidently, more likely to give you the desired result than plucking a point arbitrarily from the ether.

I did ask for clarification but Monckton didn't reply. Quite a few other commentators did try to explain why this is not an example of the endpoint fallacy. Mostly this meant repeating what the selection process was, and in some cases demonstrating they didn't understand how Monckton was calculating it themselves. A few extracts (all emphasis are from the original comments):

RACookPE1978:

We (the skeptical community of science observers and realists) do NOT pick a start date for this trend. Granted, you have been schooled (propagandandized) into believing the start date is picked but that is exactly opposite of the process, and your choice of believing their lie does fit YOUR mindset and prejudices. (I will rant your prejudices appear to control your thought processes, but do not (yet) accept your conjecture that you are feeble minded. Firm-minded, prejudiced, incapable of independent thought based on the facts presented? True. But feeble minded? Probably not.)

What actually happens each month is the following, and it was described in the original text as well. The global average satellite temperature is read from those instruments, and a flat line is projected BACKWARDS as far as that trend line remains indistinguishable from zero. We DO NOT pick a point and draw a line, we pick this month's global average temperature (whatever it actually is) and go BACKWARDS until the trend line rises. Whatever date that point is, we report. Same process every month, regardless of whether the results confirm or contradict your assumptions. The process is neutral and breathe-tauntingly honest. Indeed, the essence of this month's report is that February's report may actually show a near-continuous rise!

richardscourtney:

The answer to your question is NO!

The 'endpoint fallacy' is when a researcher chooses the ends of an analysed period to indicate a desired result.

The analysis of 'Pause' length does not involve choosing an analysed period. In this case, the start point is now (i.e. the most recent month for which data is available). Each previous successive month is then assessed as being the other end of a time-series to determine the longest period back from 'now' that shows a sub-zero trend.

In other words, the endpoints are determined by the data - they are NOT chosen by the researcher - and, therefore, the 'endpoint fallacy' is not possible in this case.

Quite a few mention that the start point is actually the most recent date, but I failed to get an explanation as to what difference that made. None of this really goes beyond claiming that because the start date of the Great Pause is calculated, it cannot be a cherry pick.

I did also try asking whether it would have been acceptable for the IPCC to have chosen their dates proving accelerated warming by calculation, rather than choosing arbitrary dates such as 25 years? To which richardscourtney replied:

THE IPPC METHOD IS PLAIN WRONG. I REFUSE TO CONSIDER WHETHER OTHER END POINTS WOULD MAKE IT MORE OR LESS WRONG.

Cherry Picking the Data

There are multiple data sets of global temperature, but for the current Pause calculation Monckton only uses the RSS satellite data. He does also mention the UAH satellite data, but only to say the Pause is slightly shorter in that data. Indeed this month he claims the Pause has gone completely in the UAH data, but he's wrong - because of a minor update to the UAH data the pause reappeared, starting in November 1997.

The current non-beta version of UAH (5.6) is not mentioned by Monckton. That version hasn't shown a long Pause since 2009.

Of the non-satellite data, Monckton only says After many unconvincing alterations to all of the principal global surface tamperature datasets over the two years leading up to the Paris climate conference, the Pause all the datasets once showed had been erased. He makes no mention of the larger alterations to UAH, presumably because he finds them more convincing.

For comparison, here are the different satellite trends over the length of the Great Pause, along with a sample surface set from NOAA. (Anomalies based on the years 1981 - 2010.)

It's interesting to note that the older UAH data was very similar to the altered surface data set, whilst the adjusted UAH data set is in much better agreement with RSS.

So, what does Monckton think is the best data set to use for establishing a global warming trend? Well quite a few times, Monckton has insisted in using averages of multiple data sets (both satellite and land based). In a long Letter to Representatives Ed Markey & Joe Barton he says (my emphasis):

The advantage of a composite global-temperature index, however, is that the satellite datasets for the lower troposphere (not, as Mr. Karl implied, for the troposphere as a whole) are to some extent less prone to heat-island distortions arising from progressive urbanization than the terrestrial datasets on their own.

The composite index is accordingly more reliable than any individual dataset , particularly since there is evidence that at least one of the terrestrial datasets has been tampered with by its administrators to create a false impression that global temperature in the late 20th century rose more sharply than it did in reality - a point to which I shall return infra.

Then there's the question of why use satellite sets which are measuring temperatures in the troposphere, but ignore warming on the ground? In 2012 Monckton wrote an open letter to Skeptical Science, which contained this interesting point (my emphasis):

It is claimed that we were wrong to say there has been no statistically-significant global warming because the oceans have warmed. However, the standard definition of global warming is warming of the near-surface atmosphere . Also, measurements to date are inadequate to tell us reliably how much - if at all - the oceans have warmed in recent years.

So why is is it now appropriate to refer to a pause in the lower troposphere as being a pause in global warming?

Statistical Significance

Even if a trend line indicates some sort of pause, there is still the question - is the pause statistically significant? Lack of significance does not mean there is no pause, just that so far there is not enough evidence to know if the pause is a real phenomenon, or just something that has occurred by chance.

You could rewrite the previous paragraph and substitute the word warming for pause, and the statement would be just as correct. Which is where I have a problem with some pause advocates, they are quite happy to claim a period of no statistically significant warming as meaning no warming has occurred, but won't apply the same logic to any pause, let alone the Great Pause.

As an aside, the main difference between two is that whilst you can easily find periods of no significant warming, you can always resolve the question by adding more data, that is looking back a few more years to find the warming has become significant. By contrast if you do that to the claimed pauses you just find the no warming turns into some warming.

Monckton himself has argued that statistical significance is required to confirm warming. Here, for example is a typically pretentious piece, where he tells a journalist how to work out significance using Excel:

Step 4. Check whether the warming (which is the difference between the first and last value on the trend-line) is greater or smaller than the measurement uncertainty. If it is smaller, falling within the error-bars, the trend is statistically indistinguishable from zero. There has been no warming - or, to be mathematically nerdy, there has been no statistically-significant warming .

Aside from the fact that his description of how to calculate statistical significance is completely wrong, he does say that no statistically-significant warming means there has been no warming.

By the same logic, if his Great Pause is not significantly different to the warming trend, then it is statistically indistinguishable from the warming trend, and hence there has been no pause.

In order to get some idea of the problem here's a graph of the pause for RSS, using Dr Kevin Cowtan's online trend calculator. I'm using this rather than generating my own here, simply because it shows confidence intervals that take into account the all important corrections for autocorrelation.

As can be seen from this, the trend from June 1997 - Jan 2016 was effectively zero, but the confidence interval was a whopping ±1.69 °C / Century. In other words, we can only be about 95% confident that the RSS trend between those dates was somewhere between -1.7 and + 1.68 °C / Century. We cannot say the Pause was significantly different from the RSS trend before the Pause, or over the whole of the satellite era.

Here's what the various trends look like for RSS data, along with their 2σ confidence intervals.

In all cases the trends are within each other's confidence intervals, and so are not significantly different from each other.

The Problem With Trends

When people apply linear trends to a time series, they usually only mention the slope of the line - that shows the rate of warming. But there's also the intercept that determines how warm the trend was to start with. Just referring to the slope of a trend over part of a series can give a very misleading impression.

In Monckton's narrative he mentions the slow rate of warming during the first 18 years of RSS data before the start of the Pause, followed by the 18 plus years of no warming, but what does this look like if we show these two trends in context (anomalies in °C):

What Monckton's narrative doesn't mention is the big break between the two trends. If these trend lines are a correct interpretation of the data, we have to assume there was global warming of around 0.2C all occurring in June 1997. This shows how much warmer Monckton's Great Pause was compared to the pre-pause era.

You can give a very different impression by splitting the trends at different points. Monckton himself mentions that the fastest warming period ended in 2006. This alone seems to be a problem with the Pause narrative as it means the first nine years of the Great Pause were also the last nine years of the Great Warming.

The narrative of this warming scenario is that RSS data shows a pretty fast period of warming, around 1.5 C° a century, lasting 27 years, followed by a 9 year period of only slightly slower warming.

Here's the same data, but with trend lines before and after the end of 2006:

This narrative is also misleading. What it misses is there's a bit of a drop at the end of 2006. Not as big as the rise in the previous graph, but still misleading.

Finally, here's what the trend for the entire RSS data looks like.

To me the single linear trend seems as convincing as either of the split trends. If a case can be made for a pause more likely to have started after 2000 and then only in the satellite data, but for now I think there is insufficient evidence.