2011-01-14 10:56:30Monckton Myth #2: Temperature records, trends, ENSO
Robert Way


be brutal :P and John is there any way I can pretty up the post a bit?
2011-01-14 12:40:49Prettied post
John Cook


Firstly, your tables are too wide - the maximum width is 570 pixels. Any wider and the header graphic goes funky.

Did you want to sexy up your tables? There's some sexy looking tables at http://designshack.co.uk/articles/css/15-tips-for-designing-terrific-tables to copy from.

Add some more visual structure by turning "Then which year is it?" into a subheading (h3). Perhaps add a few other subheadings to separate the different arguments you rebut.

Put blockquotes around Monckton's quotes for more visual structure - but you need to edit the raw HTML to do this.

Maybe at the end, use a graph of ENSO rather than two links (but link to the sources in the caption)

2011-01-14 12:55:46Question - is this post too long?
John Cook


Not sure of the answer myself but should this post be broken up into several posts? One on hottest year on record, one on warming trends, one on 1998, one on ENSO? Or because they all overlap, better to keep them in the one post? I'm easy either way, just thought I'd raise the idea (aka muddy the waters)

UPDATE: probably better as one post - we don't want to drag this out any longer than we have to.

2011-01-14 16:21:35comments
Dana Nuccitelli

"I think it is first important to begin by settling the issue as to whether 2010 was the warmest year on record?"  Question mark should be a period.

I'd change "Figure 1" to "Table 1".  Same for 3, 4, and 5.

I think you need to define "Global coverage" or at least say something about it right before or after Table 1.  The reader should be able to figure it out based on the Arctic warming discussion, but it would help to clarify it earlier.

I agree with John on making "Then Which Year is it?" a Heading 3.  I'm big on using headings because it lets the reader know that you're starting a new discussion on this new topic in the next section.  It would be good to add a few section headings, actually.

"Each temperature dataset has their own individual caveats so it is difficult to assess which is the most reliable" - add a comma here.

You may want to calculate a standard deviation and throw in a +/- 2 sigma in Figure/Table 3 trends.  Otherwise Monckton could criticize that you're not including the margin of error.

"Secondly, if he was comparing the three periods in Figure 3 [Table 2] then the only index which covers that period is Hadley, [what about SSTI?] making his 1.6°C per century number wrong for 1975-2001 as it is actually much higher at 1.78°C per century."

It's a bit on the long side, but it seems okay as one post to me.  Could be 2 or 3 too though.

2011-01-14 21:15:32
Rob Painting
Love that All Series Temperature Index, certainly a powerful image isn't it?. Long post, but probably necessary, Monckton can insert a whole bunch of insinuations in just one sentence it's hard to let them just slide by. He's a master debater!.
2011-01-14 21:24:11

First paragraph
"Yesterday" -> might not be "yesterday", just a previous post

Second paragraph
I'd explicitly say that as far as climate change is concerned it does not really matter if 2010 is the warmest or not.

Table 1/figure 1
clearly separate surface temperature datasets from lower troposphere datasets.
I find the discussion of datasets not including 2010 a bit convoluted. You should think to drop these datasets and the relative discussion. You're talking about if 2010 is a record year or not, afterall.

Paragraph before fig.2
put Skeptical Science Temperature Index in double quotes
extend the baseline periods to more decades than just 1990-2000, if it's not too much work. Not that it will change anything, but it's more appropiate.

Paragraph before fig. 3
C° -> °C

Table 2/fig. 3
use the same °C/decade units that Monkton used both in the table and the discussion
use the same names/acronyms as in table 1. Should you decide to use the shorter version as in table 2, expand the acronyms in the caption. The same applies to the other tables.
I'd also add a few words on the fallacy of his reasoning. If due to a hypothetical rapid and strong increase in TSI the planet had warmed faster than now, can we conclude that we cannot possibly warm the planet at all?

table 2/fig. 4
add the statistical uncertainties in the table. You'll probably not get statistical significant trends in any of the dataset. So, after the discussion of the most probable trends you may conclude that anyway the period 2001-2010 is too short to have any climatic meaning.

Missing link to the rebound from LIA

Last paragraph
the trends alone do not point to "a human induced impact on climate", they just tell you that the planet is indeed warming. It's just one piece of the puzzle that goes in the right place or, better, it was already there despite Monkton's claims.

Finallly, I don't think this post should be splitted, maybe only drop the part on El Nino/La Nina.

2011-01-15 01:06:11
Ari Jokimäki


You know, that is not the all series temperature index. You are missing at least these:

- NOAA STAR satellite analysis:


- Global surface temperature analysis by Lugina et al. (2006):


- Global surface temperature analysis by JMA:


Since you include satellite analyses of troposphere temperature, you might also include the radiosonde analyses:

- HadAT:






- IUK:


Well, I'm not saying that you should include these all, but perhaps the name all series temperature index should be reconsidered. ;)

2011-01-15 03:03:31
Mark Richardson

Needs some words culling IMO.


Uncertainties in your trend calculations too. And I think sometimes you state things with a bit too much uncertainty: after all, if you've just taken a mean of the sets then this definitely is unscientific (though still useful & worth keeping!)

 On the 'warmest year' stuff, I'd include something on ocean heat and/or sea levels since 1998. 2010 was unambiguously far ahead of 1998.


I'd be careful with the ENSO stuff at the end. There is a well known lag between ENSO and temperatures (particularly satellites). But importantly the '97/'98 El Nino was stronger and longer lasting (perhaps also worth including the ONI? MEI is only one way of quantifying an El Nino) AND, whilst '98 was around a solar minimum the recent one has been spectacularly weak, so we know that the effect of the Sun since 1998 has, if anything, been one of slight net cooling.



So it's not the Sun, and we know it's not ENSO because a) this one was weaker and b) there's more heat in the oceans. Now how did that happen?

2011-01-15 03:40:20
Mark Richardson

Over 1500 words is quite chunky, example shortening of wordcount:


"Yesterday John addressed some of the issues with Christopher Monckton’s paper targeting Michael Steketee’s column in The Australian, where Monckton asserted that the oceans are not accumulating heat. John showed that when you consider the full body of evidence it seems that they are, which is contributing to global sea level rise. Now I will address Monckton's claims regarding temperature records, temperature trends and the intensity of 2010’s El Nino.

Mr Steketee noted that 2010 had the warmest January to September recorded, which Monckton highlighted as cherry-picking. Now the data is in for most indices - what does it show? 

I've brought together the thermometer records from Hadley, NASA & NOAA plus the satellite measurements from UAH and RSS plus 5 'reanalyses' which include data from thermometers, buoys, weather balloons and satellites. The 10 sources I could find with their 2010 and 1998 ranks are shown in Table 1:"



It's readily apparent that 2010 was a warm year: every data source that includes it concludes that it was either the warmest or 2nd warmest year on record. However it seems that the devil of a year in 1998 still tops the global temperature records for HadCRUT, RSS and UAH (although NASA, NOAA and UAH are all statistically tied). There is one important caveat: column three shows that all datasets which include up to date Global Coverage have 2010 as the warmest year. Amongst datasets with global coverage reaching 2005 or beyond there is only one dataset which has 1998 in its top 2 and this record does not include 2009 or 2010. The global coverage identifier is key because it is well known that the Arctic is the reason that Hadley has been undersampling the warming, and that the satellite records do not include the poles. The Arctic climate system is one of the most dynamic of any places on the planet and it has experienced the fastest rate of warming out of anywhere in the planet. Neglecting a region such as this will undoubtedly force 1998 into a position that it should not be in, as the warmest year in an incomplete record. 




This is 149 words compared with 285 originally. If you acn cut the entire thing down from over 1500 to under 1200 then I think it would be a lot easier to read!

2011-01-15 03:58:44Comment
Robert Way

Hey Ari,
Thanks for the links. I'm only going to incorporate lugina et al and the JMA however I can't figure out the JMA formatting. Anyone know how to?
2011-01-15 06:44:20
Ari Jokimäki


It might be better to ignore the JMA for now. They seem to be giving only the data for the individual points in their grid of the Earth but not the computed averages for the global anomaly. If you want to add two series, it might be easier to add the NOAA STAR so that you would have three satellite analyses.

2011-01-15 07:04:15Comment
Robert Way

Hey Ari.
I will be adding Lugina et al. 2006 but I won't be adding in star because STAR measures the mid troposphere whereas the RSS and UAH measure the lower troposphere. The JMA would be nice to have but I can't for the life of me figure out how to make sense of their data.



Lugina et al. is land-only so it will be excluded. As it stands I will not be changing any of the 10 datasets used because the balloon measurements are too sparse (only 85 stations) and are incorporated into the reanalysis datasets already used, because the STAR is for mid troposphere rather than lower troposphere, because lugina et al is land only and because JMA has a data issue that I have yet to resolve.

Hope this clarifies things.
2011-01-15 07:14:32Trend error calculations
Robert Way

Hey all, was wondering how I would calculate the + or - in the trend estimates.
2011-01-15 07:56:27responses to MarkR and Riccardo
Robert Way



extending the baselines cannot be done unless we want to lose data. We cannot compare ERA-Interm (1989-present) if we do not put it on a 1990-2000. How can I determine the statistical uncertainties and whether the trends are statistically significant?

I agree regarding the lag in the ENSO but Monckton stated explicitly that we had an El Nino for the majority of the year which is not true. With the lag we could have possibly had El Nino conditions but that is not what he said. I was going to do a 12 month running average of elnino/la nina intensity but my problem is that I can’t seem to get monthly matrices of temp data into a format where it can be used as monthly data. For example turning:
Jan Feb Mar
1     4      6

into Jan 1
       Feb 4
       Mar 6

I agree with the shortening and I will try to do so.

2011-01-15 08:39:27

Referring to JMA annual gridded data, in each line there are the 72 5°x5° boxes at equal latitude starting from 87.5 N; in this way you have 36 line per year. You should average them considering the different area of the boxes. Not straightforward but manageble.

The statistical uncertaintainty can be calculated but any fitting program I know return the best estimate of the parameters together with the 1 sigma uncertainty (error). It's customary to use the 2 sigma uncertainty. If the program you use does not give you the uncertainties, send me the time series you used and I'll give you the uncertainties in the trends.
2011-01-15 09:20:38Response to Riccardo
Robert Way

Hey Riccardo

I just used excel's slope function for determining the rate of warming so all I have is an excel file. You don't by any chance know how to implement the uncertainty using excel do you?

Regarding the JMA annual gridded data. I do find it hard to deal with to be honest. I can't really seem to get it into an excel compatible format. Any ideas?
2011-01-15 10:01:31stdev
Dana Nuccitelli
Can't you get the margin of error by just calculating the standard deviation (stdev formula in Excel), then multiplying by 2 to get 2 sigma?
2011-01-15 11:01:30response to dana
Robert Way

Hey Dana,

But what would I be calculating the stdev of? Like just the one value? Cause all I do is get the slope of the trendline for a given x value ( a year) and then multiply by 100 to get my degrees per century value so what would I take the stdev of?
2011-01-15 11:10:34hmm
Dana Nuccitelli

I was thinking of the standard deviation of the data, not the slope.  My bad.  I'll have to think about that one....

2011-01-15 18:20:23


I can't help with excel, we scientists tend to use complicated things to do simple tasks :)

I found googling that the function LINEST should do.

2011-01-15 19:13:10LINEST
John Cook


Yes, I believe the LINEST function in Excel will get you the standard deviation and you can calculate the 95% uncertainty from that.

Rob, can I suggest you change the title of your blog post to  "Monckton Myth #2: Temperature records, trends and El Nino" so we have a consistent style across all the MM posts?

Similarly, what do you think of changing the Filename to something like Monckton-Myth-2-Temperature-records

2011-01-15 21:52:12
Mark Richardson

LINEST function in excel is simply wonderful :P  it got me through a lot of my degree when I was on holiday and only had access to MS Office...



You can do it by typing quite easily - it shows you the formula on the second page of the .pdf John linked to. And rather than type ",TRUE,TRUE)" you actually just need "1,1)".


So it's =LINEST(y1:y2,x1:x2,1,1) and then CTRL+SHIFT+ENTER. Make sure you have enough cells highlighted when you type it in et voila.


2011-01-16 00:30:20Feedback in my most brutal mode
James Wight


Robert, I think some of your arguments are a bit weak. The post starts off well with your combination of the various temperature records and your comparison of the historical trends in Table 2. However, I think some of your rebuttals to Monckton’s arguments miss the point, and there are stronger counterarguments you could make.

Most importantly, I think your rebuttal of “virtually no warming since 2001” wrongly accepts Monckton’s unspoken premise that it matters what the trend was from 2001-2010. Instead I think you should challenge that assumption. Ten-year temperature trends fluctuate wildly: from the last 30 ten-year periods you can cherry-pick some that are basically flat (eg. 1977-1986 in NCDC), and some that are double the long-term warming rate (eg. 1974-1983 in NCDC). The data are just too noisy on that timescale to draw any conclusions. When you step back and look at the last 25-30 years, what you see is a clear warming trend of 0.17-0.18°C/decade. 10-year moving averages show continued warming up to the present.

MarkR also has a point: ocean heat content is still rising; sea level is still rising. Arctic sea ice, ice sheets, and glaciers are melting faster than ever. These are all signs global warming is still happening.

Making arguments based on ten-year trends only sets us up for failure when they turn against us, as short-term variability always will. The 2001-2010 trends in NOAA and RSS, though positive, could reasonably be described as “virtually no warming” compared to the long-term rate, and even in UAH the 2001-2010 trend is only half the long-term rate.

Also, I think it is more meaningful to express trends in degrees per decade rather than per century, because nobody expects the trend to remain the same over a century.

In the discussion of the warmest year, you should also state that it doesn’t matter much what the individual warmest year is, whether 2010 is first or second. Even Monckton himself comes close to saying this in his point #2.

I think the language “overwhelming majority of evidence” (about 2010 being the warmest year) is too strong. There are legitimate arguments to be made for and against each method, so it could reasonably be the case that the warmest year was actually 1998 or 2005, with 2010 coming second. I suggest you leave it at “majority of evidence”. Let’s save the word “overwhelming” for when it actually applies – if a word is overused it ends up not meaning anything anymore.

I also have a few minor suggestions/corrections:

  • “Each temperature dataset has their own individual caveats” should be “Each temperature dataset has its own individual caveats”.
  • I would change “create the All Method Temperature Index” to “create an All Method Temperature Index”.
  • Why is part of Monckton’s claim #2 in italics?
  • In Tables 2-4, “SSTI” should read “AMTI”.
  • I would change “within 0.25°C per century” to “within 0.25°C/century of 1975-2001” for clarity.
Sorry to be so negative, but you did say to be brutal!
2011-01-16 04:20:07good points James
Dana Nuccitelli

That wasn't so brutal - it just involves adding a discussion about how looking at short 10-year trends or individual record years is not informative when trying to assess global warming (plus a few other small edits).  Good examples with the NOAA data.  A reference to Mark R's 1998 DIY Statistics post could be helpful here.

It's a good point - not only is Monckton wrong in his claims about the data, but he's also focusing on too short of a timeframe.

2011-01-18 03:52:37need more time?
Dana Nuccitelli

If you need more time to work on this Robert, we can switch Myths #2 and #3.

*edit* whoops, I didn't notice that this one had already been published!  I guess I'll go live with #3 pretty soon.  John, the next one is supposed to be the climate sensitivity post, if we're staying in order.

2011-01-18 08:17:43#4
John Cook

I know, I've been meaning to get to it but was distracted by the ABC piece. Will post a draft in the forum today.