I received an email in my inbox that began:
Hi Lucia..
Has anyone checked the statistical significance of this out yet!
I think this story may get a LOT of coverage….
The Phil Jones interview, last year saying no statistically significant rises in temp since 1995
A new BBC Phil Jones article, says it is now significant…including 2010http://www.bbc.co.uk/news/science-environment-13719510
BBC: Climate warming since 1995 is now statistically significant, according to Phil Jones, the UK scientist targeted in the “ClimateGate” affair.
http://www.carbonbrief.org/blog/2011/06/global-warming-since-1995-statistically-significant.
Evidently, Phil Jones said this:
“The trend over the period 1995-2009 was significant at the 90% level, but wasn’t significant at the standard 95% level that people use,” Professor Jones told BBC News.
“Basically what’s changed is one more year [of data]. That period 1995-2009 was just 15 years – and because of the uncertainty in estimating trends over short periods, an extra year has made that trend significant at the 95% level which is the traditional threshold that statisticians have used for many years.
“It just shows the difficulty of achieving significance with a short time series, and that’s why longer series – 20 or 30 years – would be a much better way of estimating trends and getting significance on a consistent basis.”
Of course, lukewarmers– that is those who believe warming is real but likely to be on the lower end of projected values– do expect that the ‘trend since year X’ will become statistically significant as the number of years increases. So, it would hardly surprise me if 2010 happened to be the year when ‘the trend since 1995’ would become statistically significant at the end of 2010. I mean… why not 2010?
But since Barry asked, I thought I’d check to see what I get. Now, I think by now, everyone knows that whether the trend is deemed ‘significant’ can depend on the method used to estimate the uncertainty in trend based on analysis of the residuals. Do we use ARIMA? AR1? Fractional differencing? and so on.
In the first step: I decided to check if the trend is significant if the residuals are assumed to be white. It’s well known they aren’t white– the residuals exhibit positive autocorrelation. But if the trend isn’t significant when the residuals are treated as white, Phil Jones who claims it is significant would have some ‘splaining to do.
So, I wrote a simple R code.
Step1: Get the annual average temperatures from http://www.cru.uea.ac.uk/cru/data/temperature/hadcrut3gl.txt.
temperatures<- ts(c(0.275, 0.137, 0.352, 0.548, 0.297, 0.271, 0.408, 0.465, 0.475, 0.447, 0.482, 0.425, 0.402, 0.325, 0.443, 0.476),start=1995)
Create the time variable with names I like; center time because I like to do that:
# Fit assuming residuals are white noise:
years=c(1:length(temperatures)); norm_years=years-mean(years)
# Apply ols fit
olsFit<-lm(temperatures~years) # the fit.
output<-summary(olsFit) # nice summary.
tcrit_Student_OLS=qt(0.975,num_years-2) # find critical t for number of deg freedom. Note: 2*(1-0.975)=0.05. So this is the t for 95%, two tailed.
output; tcrit_Student_OLS # compare the 't' and the critical 't'
Call:
lm(formula = temperatures ~ years)
Residuals:
Min 1Q Median 3Q Max
-0.180941 -0.040926 0.000912 0.047301 0.208118
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.296000 0.049791 5.945 3.58e-05 ***
years 0.010971 0.005149 2.131 0.0513 .
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 0.09495 on 14 degrees of freedom
Multiple R-squared: 0.2448, Adjusted R-squared: 0.1909
F-statistic: 4.539 on 1 and 14 DF, p-value: 0.05134
[1] 2.144787
Note: 2.131 the ‘t’ associated with the trend is less than 2.144787, the critical value of ‘t’ required for statistical significance. That is: This isn’t even statistically significant if I assume the residuals are white! So, at least based on the data at http://www.cru.uea.ac.uk/cru/data/temperature/hadcrut3gl.txt, the trend since 1995 does not yet appear statistically significant.
By the way, you can do this in EXCEL. Just stuff in the data, do the fit, get the ‘t’ and compare to the student t value for 8 degrees of freedom.
Oh, and for those wondering: Is this ‘t’ stuff really right: Notice the 0.0513 highlighted in green? That value is greater than 5%. The 95% confidence intervals is exceeded when that value is less than 5%. So, if you use EXCEL, you don’t need to compute the ‘t’.
Ok…. but Hadcrut has another series. So, I tried this one:
http://www.cru.uea.ac.uk/cru/data/temperature/hadcrut3vgl.txt.
This one does show statistical significance if I treat the residuals as white noise. I bet Phil meant “If I use one of our series rather than the other.”
But now I need to do the next step: Is this still statistically significant if I use the quick correction for ‘red’ noise? That is, I assume the number of effective data point are reduced by a factor of (1-R)/(1+R) where R is the lag-1 correlation coefficient. (This assumes the residuals can be explained as being an AR1 process and tends to make uncertainty intervals a little too small for AR1 processes. But it’s a start, and it’s often used in climate science. For example, that multi-author Santer paper.)
So, I get a t. It’s t= 1.877057. I don’t even need to look up the student t, because the smallest possible value for the critical t is 1.96. So, no. This is also not statistically significant.
On what does Phil Jones base his claim?
I don’t know what Phil Jones did nor do I know what data he used or what method he used. But if I claimed statistical significance for tests I do without accounting for serial autocorrelation or using a one sided test, people would scream bloody murder or patiently explain that I am daft.
Maybe Richard Black should ask Phil which data he used and how he did his calculation. ‘Cuz I can’t guess. Maybe he used GISTemp? I’m not inclined to go on a hunt to guessing what Jones may have done.
If we use HadCrut, trend since 1995 does not yet look statistically signficant to me. I don’t think this lack of significance has great scientific importance– I’m sure it will eventually be significant, but Phil Jones needs to wait a few more years.
Jones: Significance R script.
I think that Lucia gives an example of why the consensus appeal of climate scientists is poorly understood and suffers from a lack of detail in matters such as this one. Phil Jones is obviously making a point with the media, and given the lack of detail that normally involves, we get the answer of a scientist/advocate.
What becomes apparent from Lucia’s analysis is that with little effort (if you know what the issues in the measurement involve) one can present a more detailed analysis than a keeper of a commonly used temperature data set evidently did. It raises questions in my mind of Jones’ motives and/or understanding of time series. Jones in the mean time can be quoted widely as an expert in the matter and without defending and showing details of the evidence.
I am not at all sure what a trend in a short time period means for global temperatures long term, but surely Jones’ updating his year ago statement indicates that it means something to him – for purposes that might be more advocacy than science.
That’s what I suppose. But as advocacy, this is going to backfire because no one has to do anything “fancy” to show that using HadCrut and the statistical model the IPCC used to demonstrate significance over the 20th century, the trend since 1995 is not statistically significant. I would be rather astonished if this embarrassing to Jones story doesn’t get aired rather widely. To the extent that anyone unquestioningly accepted what Jones claimed and posted, they too will be criticized for either lack of understanding or diligence.
So, this is bound to be a low-light in the PR-advocacy campaign Jones might wish to advance. At a minimum, he should have pro-actively explained his assumption.
Did he use NOAA/NCDC? GISTemp? Neglect serial autocorrelation etc. I don’t know. But I think his statement ought to have given readers a hint and if he knows the trend with HadCrut is not significant, he would have been wise to mention that the trend based on HadCrut remains ‘not statistically significant’.
Just out of curiosity is the values closer to being significant than if you use the 95-09 range he admitted didn’t reach significance?
Oh, come on. The guy gets interviewed a year ago, gets asked “Do you agree that from 1995 to the present there has been no statistically-significant global warming”, says “Yes, but…” and gets quoted all over the world as having said “there has been no warming since 1995”, completely ignoring the fact that the 2nd half of his statement was that there was a trend which was “positive, but not significant at the 95% significance level”.
So, he comes back a year later, does a quick, simple calculation just looking at the HadCRUT temperature trend which most HadCRUT people use (eg, the variance adjusted one), sees significance, and sees an opportunity to maybe start correcting a widespread misconception regarding his original statement.
Now, maybe the trend still isn’t significant using more sophisticated methods, but given the typical BBC audience, which is closer to the truth: “the Earth has warmed since 1995” or “the Earth has not warmed since 1995”? I would say its the former, for most purposes, even if “the warming since 1995 has been statistically significant” is not correct when using realistic assumptions about self-correlation.
Basically, he’s in a no-win situation: make a scientifically correct statement, and see his statement used to obfuscate and confuse and leave people with a really wrong misconception. Or, make a statement which is correct using inferior methodology, and just barely incorrect using superior methodology (and I’d argue that, like horseshoes and hand grenades, “close” does have some value in statistics, though I know some disagree), and get jumped on. Science communication is not easy, and there are times when the 100% correct statement actually communicates reality less well than the 98% correct statement.
I wish someone would explain to anyone that there isn’t a whole lot of use in trying to figure out the statistical significance in the temperatures since 1998 anyway.
Does it make a difference if one uses annual values vs. monthly values or a monthly 12 month running average?
Jones must have some sort of basis to make the statement, even if it overestimates the significance by ignoring the autocorrelation.
Your Jones.txt script uses a variable num_years without defining it.
Adding num_years=length(temperatures); seems to do the trick.
Weird– my home version has num_years=length(years) in it!
I usually try to remember to remove all defined variables and rerun.
From the article linked above we get this:
“Furthermore, a short-term hiatus in warming does not detract from a longer warming trend. In fact such pauses are only to be expected, as US scientists demonstrated in 2009:”
The US scientists they are talking about are Easterling and Wehner. And their link takes you to the abstract of their paper. I read their paper and noted why the points they were making were invalid more than a year ago, here:
http://reallyrealclimate.blogspot.com/
It’s humorous how junk science is used to support other junk science in the global warming community.
M
It would be more correct to say “incorrectly diagnoses significance”.
Whatever opportunity he thought he saw, his blunder will only reinforce the misconceptions about the importance of the lack of statistical significance since 1995.
“More sophisticated” would be the easy, peasy to apply approximate method, which everyone in blog-dom knows is the minimum threshold required.
I’d sympathize with your argument except that he is not required to contact the BBC the moment he thinks he might have detected statistical significance. He would have been wiser to shut up for the time being, make sure he didn’t make a mistake. Ask a few people (including bloggers) and do other things to avoid egg on his face. If his calculation turns out to be based on monthly data (which doesn’t seem to match the wording of the BBC article) he should have said he was using monthly data and also explained the assumptions he made when identifying statistical significance.
Do you not understand that the statement he made is incorrect? It’s not ‘inferior’ to neglect serial autocorrelation when it exists. It’s incorrect.
In this case, his statement mis-communicate reality. Reality is that — based on HadCrut3 or HadCrut3v annual average data, the trend since 1995 is not statistically inconsistent. The answer may be different based on monthly, there is a good chance the answer will be different in 7 months. Jones should have shut up for 7 months and waited to see what he could say in Jan 2012.
Tilo–
Yes. Very, very bad misleading paper. Very.
Charlie A
It looks like JeffId used monthly data from HadCrut3V and found that Jones’s announcement came during the first month when significance was reached using a) monthly data and b) correcting for auto-correlation using the simplified method.
http://noconsensus.wordpress.com/2011/06/11/experts/#comment-51233
So, significance is just reached in April. Since La Nina is over, this will probably stick for a while– likely forever. If this is how Jones did his calculation, he would have been wise to write a discussion that doesn’t give the impression he used annual data– but at least it’s not as embarassing as having on absolutely, positively no basis to claim significance!
Oops, sorry, got the link wrong. It should be this.
http://reallyrealclimate.blogspot.com/2010/01/another-inconvient-truth-for-agw.html
I was wondering about “the standard 95% level that people use”. How common (and in which other sciences than Climate Science) is this? I do some “value at risk” calculations for my employer (a bank), and I never use a significance level below 99.5% (or above 0.005, depending on your nomenclature).
“Science communication is not easy, and there are times when the 100% correct statement actually communicates reality less well than the 98% correct statement”
Significance, like virginity, is absolute. A data set is either statistically significant or not. If you have to transform the data, jump on it, stretch it, pad it, prod it, ‘improve’ it or reinvent the meaning of the ‘t-test’ to attain significance; then it is bollocks.
All it means is that the variation we see is not odd when compared to what would happen in 20 sample-length times.
Would you be happy to give your child pharmaceutic products that had been tested, in-house, using data sets that were ‘improved’ and ‘homogenized and using statistical tests that give ‘added value’.
In drug testing we have to state how the study will be run, how and what data will be collected AND how the data will be analyzed BEFORE beginning the study.
It is now customary to run the experimental design past a statistician prior to beginning work.
Oh, we tend to work blind.
Tomorrow I have 8 96-well plates to analyze; I know that I have four Aut’s and four Twins. I don’t know which of the plates is an Aut’s or twin. I don’t know which half of each plate contains stripped and unstripped media, nor do I know which contains DDE (as in DDT).
I will not know the identity of each quadrant or the cell ID’s until I have done all the three assay methods, done statistical analysis on 24 well quadrants and deposited the data as a time-stamped data file.
Even then I will only know then as Fam1, Fam2, Fam3, e.t.c. I will not know who is who, until I have done all 11. Then and only then can I compare the cells to people.
Espen
It’s fairly common as a criterion for reportable vs. not reportable in science. However, before an interpretation based on observations is taken as being well supported, one expects some level of replication from similar experiments done by other groups. This type of replication is difficult to obtain for some climate science questions because there is only 1 earth so we can’t get replicate samples of global mean surface temperature for the period 1995- now.
Well, I think this demonstrates selectivity of desired result. Please, can we separate the data collectoras from the data analyzers; and also go easy on the “corrections” of the data. Let’s just look at the raw data
Lucia,
Check out the update at tAV.
“lucia (Comment #77041) June 11th, 2011 at 1:46 pm
This type of replication is difficult to obtain for some climate science questions because there is only 1 earth so we can’t get replicate samples of global mean surface temperature for the period 1995- now”
With the greatest of respect, garbage. One does not only examine one outcome of a hypothesis. IF CO2 causes an increase in the rate at which photons in the IR are re-radiated from the atmosphere and back to the Earth then there should be a huge range of testable hypotheses.
For instance, the RATE at which the Antarctic cools and reaches steady state during the polar night should have slowed since the 1960’s and the RATE at which it warms during the polar day should increase.
This should also be the case for sand desert regions.
If one is trying to look at a very small signal in spectroscopy then one uses difference or double difference spectroscopy.
Why is the AGW signal not paired? There should be an anisotropic in the amount of CO2 induced re-radiance as one moves from pole to equator.
The essence of experimental design is know that one can theoretically deliver data that should have the ability to destroy a testable hypothesis.
Stating that we have only one Earth does not cut it; big claims should have a huge weight of evidence to support them, from a huge range of statistically tested hypotheses.
Doc–
Of course there are many different testable hypothesis. But it is also true that it is sometimes not possible to tests certain specific hypothesis against replicate realizations because there is only 1 earth.
Testing the rate at which temperatures in Antarctica vary is not the same as testing the rate of temperature variation over the entire globe. (Worse, the tests aren’t even independent as one is a subset of the other.)
Sure. If you wanted the ideal set of experiment to test the response of the earth’s surface temperature to greenhouse gases, you’d do one set controlling forcings one — and repeat samples. You’d do another set another way– controlling forcings at some other level and have multiple earths’ to have replicate samples. But you can’t do this. The equivalent is possible in fields where laboratory experiments are possible.
This doesn’t mean there are no experiments that can be done.
Sure. But that doesn’t mean that sometimes it’s difficult to actually perform the experiments required to obtain evidence. It can be easier in disciplines where people can quickly and cheaply repeat experiments in a lab. It can be more difficult in disciplines where some experiments requires using the entire planet and take a long time.
lucia This is not a pop at you, but you state
“Testing the rate at which temperatures in Antarctica vary is not the same as testing the rate of temperature variation over the entire globe. (Worse, the tests aren’t even independent as one is a subset of the other”
and respectfully disagree, as my mentor says; “there are only two things, Thermodynamics and Kinetics”.
Concentrating on looking changes in temperature, which is only a proxy for steady state thermodynamics, is only a part of the story; especially on a rotating planet which is in an elliptical orbit.
The rates that temperatures change should bear the finger-prints of AGW.
Of course temperature is only part of the story. Of course temperature changes should bear the fingerprints of AGW. But I have to admit I’m a bit befuddled because you seem to have gone on to some subject you want to discuss but thinking it has some connection to my earlier comment about making replicate samples.
I just loaded in the monthly data from HADCRUT3v, and ran your code over that, and for the monthly data from 1995, and using your red noise scaling, the significance factor were:
> t; tcrit; tcrit_Student
[1] 2.184812
[1] 1.959964
[1] 2.04356
Re: M (Comment #77028),
I’m guessing that the commenters that have responded to you haven’t caused you to revise your opinion. People, myself included, don’t change their minds easily on topics that they think are important.
Still, it might be worth noting that a number of (seemingly) sensible people are critical of Dr Jones for his not-really-correct assertion to the BBC, all for more-or-less the same reason: that commonly-used data and methods don’t support his claim.
Your position defending Jones is rather lonely, at least at this lukewarmer-friendly site.
I have no doubt you’ll find yourself in plentiful company at pro-AGW Consensus advocacy sites.
Which brings me to my point, akin to what Lucia said. From a public relations perspective, Jones’ public-relations-centric action will tend to increase doubts about the reliability of scientist/advocates such as him. Among “regular” scientifically-literate laypeople.
If the story develops true to form, his statement will be feted within the AGW advocacy community. Because, as you explain, “science communication isn’t easy…”
If the point is to educate more and more people about the strong science behind the claim of alarming warming trends and alarming climate sensitivity — this isn’t the way to go about it.
But perhaps that isn’t the most important objective.
“GregH (Comment #77051) June 11th, 2011 at 3:19 pm
I just loaded in the monthly data from HADCRUT3v, and ran your code over that, and for the monthly data from 1995, and using your red noise scaling, the significance factor were:
> t; tcrit; tcrit_Student
[1] 2.184812
[1] 1.959964
[1] 2.04356”
What was 1995 to 2008 and 1995 to 2008?
GregH–
Yep. Jeff-Id did that. See my comment. lucia (Comment #77037)
June 11th, 2011 at 1:06 pm followed by back and forth conversation. I initially thought that finding meant Phil Jones has some basis for saying “statistically significant since 1995”. But Doug Keenan pointed out a big problem.
Believe it or not, it turns out that Phil Jones paper on HadCrut specifically says we shouldn’t use Hadcrut3v for time series. It says
That is: Phil Jones own documentation on the data sets says doin’t use HadCRUT3v when analysing time series. Use HadCRUT3 for that. So…
Jones really should have waited. Significance will come– but he really should have waited. (7 months will probably do it!)
I mean version 3, I acually used the file hadcrut3gl.txt, which is the version Jones says to use.
Of course I’m of the opinion on one with any common sense actually cares if the warming since 1995 is significant or not.
Re M;
The BBC is the UK state broadcaster and had a reputation for being impartial, and offering up the truth. It is supposedly the upholder of the Reithian ideals of educating, informing and entertaining. The BBC may have failed at this, and Lucia succeeded because the BBC has entertained.
It’s uncritically taken a statement from a perceived authority and made it public. Lucia has educated and informed by explaining why there are questions around Jones’s press release and Black’s regurgitation of the same. There is no space at the BBC to comment or discuss Black’s article, so the BBC has to an extent already lost control of the story. Why it ran the story in the first place without checking might be a more interesting question, and might relate to the questions being raised about current UK policy and spending on preventing climate change that doesn’t exist.
Lucia,
Phil Jones says: “It just shows the difficulty of achieving significance with a short time series, and that’s why longer series – 20 or 30 years – would be a much better way of estimating trends and getting significance on a consistent basis.”
.
You say: “Of course, lukewarmers– that is those who believe warming is real but likely to be on the lower end of projected values– do expect that the ‘trend since year X’ will become statistically significant as the number of years increases.”
.
As you have HADCrut data available since 1850, why don’t you plot the start date of 1850 and see if you get a significant trend? Just a suggestion…
Fascinating analysis Lucia! Thanks for doing that.
I’m rather sad to see Phil so hastily (and sloppily) updating his statement of last year. There were a series of interviews he did soon after Climategate where he really seemed to be making an attempt to be more accurate and forthcoming. Climategate seemed to have caused him to experience one of those ‘long dark tea times of the soul” that give perspective and cause introspection. It appeared as though he emerged having thought, “maybe I lost my way a bit, became more of an advocate than a scientist”, perhaps concluding “this isn’t worth sacrificing my integrity. I’m just going to stick to the facts from here on out”.
That’s what I imagined when reading those interviews. It’s not that everything he said was straight-up, it was still colored by his beliefs, but I really thought he was trying (from his perspective) to be relatively neutral. Now of course, Climategate has faded away with far less impact than Phil or anyone else ever thought possible. It proves that when both the majority of the press and most bureaucrats will close ranks to support you by selectively underplaying, reinterpreting or ignoring awkward facts, stunning things are possible.
However, there was still the problem of some things that Phil said in those particular moments of candor. Like the MWP, they needed to be got rid of because skeptics were making effective use of Phil’s statement. I admit having done so in certain forum discussions. Warmists tend to adore their appeals to authority and citing *their own authority* back at them (with link to a MSM interview) was remarkably effective in shortening the interminable back and forth over recent temps.
I imagine that there were many such citings of Phil’s words and someone on the warmist side saw it as a PR problem and starting playing with the numbers. They conjured up some statistical approach that appeared to them to pass muster by some criteria and sent an email to Phil encouraging him to correct the record at his next opportunity.
What I find amazing in all this is that it doesn’t matter. The vast majority of skeptics and all lukewarmers are in agreement that the earth’s climate has been warming slightly since the little ice age. Whether any recent warming trend is not quite statistically significant or very slightly statistically significant, it’s still far below the model projections and warmist predictions. That’s all that matters. Current temps still falsify previous warmist theories of CAGW.
Lucia,
Thx for doing this. Facts over message.
Reality is, Phil Jones has been in a significant [self-inflicted] credibility pickle since Climategate broke, in spite of the subsequent “inquiries” into CRU activities and policy which have been shown up as classic whitewashes.
As Atomic Hairdryer points out, the BBC is supposed to be impartial. Unfortunately we have reams of evidence over a long period of time that this has not been the case on a number of topics, AGW/ACC being one of the most blatant examples of BBC bias on record. So Jones did not choose the BBC by chance as an “outlet” for his new findings, knowing from experience that he would get a sympathetic ear.
The root of his travails remains: he has a fundamental credibility problem, one he just compounded by this “advocacy” driven flip flop which is supported by the data.
Lucia,
I think that we discussed what test Phil Jones used in a thread a while back. Based on what he said, all he did was:
1.) take the yearly values from here:
http://www.metoffice.gov.uk/hadobs/hadcrut3/diagnostics/global/nh+sh/annual
2.) work out the r value
3.) use a two-tailed test in a calculator like this:
http://faculty.vassar.edu/lowry/tabs.html#r
4.) and find that the 1995-2009 value is not significant at the 95 per cent confidence level (two tailed value of 0.0726) and that the 1995-2010 is significant at that level (two tailed value of 0.0328)
He did not change his test. The autocorrelation for the 1995-2010 values (assuming that I have done the calculation correctly, which is a big assumption, I must admit) is effectively non-existent. The residuals offset by one have an R^2 value of .003.
So assuming no autocorrelation is not a horrible crime in this case.
Looking at the values that they have used, they are not the ones that I believe that Phil Jones used. The values from the met office link above are:
1995 0.275
1996 0.124
1997 0.356
1998 0.517
1999 0.263
2000 0.239
2001 0.399
2002 0.456
2003 0.459
2004 0.431
2005 0.474
2006 0.427
2007 0.402
2008 0.312
2009 0.439
2010 0.496
2011 0.277
So that difference is likely the reason for the difference in the outcome of the significance tests.
Re: lucia (Comment #77037)
Somehow I find this hard to believe. Monthly anomalies were a lot higher during the first half of 2010 than they are now. If he was going to use monthly data, it seems that he could have claimed significance a year ago. Am I missing something?
GregH,
CAGW is based on a warming trend that started in the 70s, right? So, about 35 years of warming is being interpreted as cause for great alarm. In that context lack of significant warming for the last 15 years of that 35 year period seems interesting to me. If I had all my eggs in the CAGW basket, it would be enough to at least get me a little worried, I think.
As further evidence that the series that I posted is the one Phil Jones used, in 2010 he stated that the non-statistically significant 1995-2009 trend was .12 degrees per decade. Neither of the two datasets that were used in the OP have such a trend – both are .108, which rounds to .11 degrees per decade.
David Gould:
And you know this…how?
David Gould:
Do you have any idea why the “annual” file has a different value than the annual average in the “monthly” file?
Weighting the monthly values according to the length of the month generates a slightly different value compared to the unweighted average, but only by about 0.001 K. Nothing nearly as “big” as, say, the .020 K difference of 2010 (.496 in annual file, .476 in monthly file). So that can’t be the reason.
Carrick,
I did the calculation. I may have done it wrong or interpreted it wrong (as I said, I am not sure if I did it right). The r^2 value from the residuals at a one-year lag was .003, which indicates a very low level of autocorrelation. If I have this wrong, that would be good to know.
HaroldW,
I am not sure what relevance that has. I am simply pointing out what values Phil Jones used in his calculations – both the 1995-2009 calculation and the 1995-2010 calculation.
And I had done the calculation incorrectly: I foolishly included the residual for the 2011 value. As such, the r^2 value is .01, so there is a lot more autocorrelation than I suspected. This reduces the value of N, and thus tips it out of statistical significance.
David, the reason I’m relatively certain something is fishy with your autocorrelation calculation is that there is a well-known 1/f characteristic to climate noise. (Very none-white.) In addition there are pretty well defined atmospheric-ocean oscillations, such as the 4-year ENSO, the 9-year AMO, and so forth.
Zero autocorrelation pretty much assumes the spectrum is white (constant amplitude with frequency). What assuming white noise does is to underestimate the uncertainty in the trend.
I prefer a Monte Carlo approach to estimate the uncertainty in the trend induced by the observed time series (with the assumption of homoscedasticity)… other methods work equally well.
I think the problem is that you are looking at the 1-year lag, seeing that it is small, and assuming from that, that there is no issue with autocorrelation.
Carrick,
How would you implement a monte carlo approach in this instance? While I have done a little bit on monte carlo in examining data series with noise and trends (I did this to look at the chances of getting flat decades, given certain assumptions), I would be at a bit of a loss in this instance. Can you give me a hint or two?
If I examine other lags, there are indeed some with higher levels of autocorrelation (and some with negative slopes).
If the extra year made a difference, why not start in 1994 and end in 2009? Then compare with 1995 – 2010.
David, what you do is estimate the power spectral density of the detrended time series. You get something like this: PSD for GISTEMP.
(Do the same thing for hadcrut3gl of course.)
Then what you do is generate a random series of “noise” based the assumption of homoscedasticity: You do this by taking the Fourier amplitudes, assigning each of them a random phase, then inverse transforming. Choose a 16-year period (equivalent to 1995-2010 inclusive), do a OLS fit to determine the trend associated with this instance of noise. Do this N times to get an estimate of the standard deviation associated with with this natural variability on temperature trend estimation.
Use that temperature trend uncertainty together with your fitted value for the temperature trend from hadcrut3gl (0.11 °C/decade in this case) to perform a test of the null hypothesis (the probability that this particular fitted trend could be this distance from a 0-trend by accident).
This involves having access to software that can compute Fourier transforms and knowing something about how to compute spectral periodograms
A simplified alternative is to download my pink_noise.m program, run it in octave (freely downloadable version of Matlab), fix the variance of the output of this program to that of the detrended, measured data, then use that to estimate the uncertainty in the OLS trend.
It wouldn’t be terribly difficult to adapt the pink_noise.m matlab/octave code to use an estimate of the fourier amplitude to generate multiple instances of short-period climate fluctuation. I did a quick and dirty calculation, and I got 0.11±0.07 °C/decade. If I did it right, this gives a p value of 0.11 (two sided distribution).
One thing to note is if you can control for some of the natural variability–extracting it using MEI as Lucia does for example—this will tighten the uncertainty bounds on the resulting OLS trend estimation.
David, if you are using R you could fit an ARIMA model to detrended HadCrut using auto.arima. Use the modelled coefficients to run your 15(?) year monte carlo simulations. Scale the variance of the simulations to equate to HadCrut. Fit ols to the simulations and extract the trend coefficients. Plot the histogram of this trend range along with 2 sigma confidence intervals to demonstrate the likely limitations of the effects of noise on HadCrut’s trend.
If, as seems likely, the 2011 temperature is ‘significantly’ lower than 2010, is Phil Jones going to give another statement next year that the temperature rise is insignificant again?.
.
The thing is, no-one is arguing about whether there has been warming since 1850, or that there have been periods since that year of both higher trends or lower trends. The question is why has there been warming and, specifically, how much is due to anthropogenic GHG emissions.
.
It seems that Phil Jones is just trying to make up for his earlier statement of ‘insignificant warming’. If so, he is playing a dangerous game to shore up any credibility he may still have, as the difference between significant and insignificant could be a couple of hundredths of a degree C. He could be changing his statement every year for the next decade!
.
ps, the trend since 1850 is around 0.06 C per decade. Is that significant?
Thanks to Lucia for taking a look and publishing the email I sent to the Blackboard..
The problem is, the intent of this headline is to be used to prove the ‘sceptics’/’lukewarmers’ wrong somehow. It HAS BEEN sent around the world’s media, and Jeff’Id’s, Keenans, and the Blackboards fact checking will be never be communicated to the masses.
What also gets lost, is not whether the world is warming, plateauing, cooling, etc, or the rate of which over what timescale, but none of this Proves the actual cause of temperature changes
The cause AGW or natural or realisticall what % of both (and nobody knows what % of either) which is of course the whole point. The argument has descended to a media propaganda soundbite, it’s warming again, proof of AGW?!?!?
(what happened to nature, has every warming, cooing process stopped, or all cancelled each other out in a bizarre equilibrium, where only AGW makes a difference?)
The very simplistic message that the ‘sceptics’ are wrong, because the world is warming, is what this Phiul Jones staement and BBC article was intended to convey, not least repeated by the Carbon Brief, who twittered and spun this article to the world’s media (ie their twitter followeres is a who’s who of the AGW media,NGO extablishment, including the Committee on Climate Change and UEA, Climate progress, Guardian environment, washington POst, Time, Independent, BBC, etc)
Carbon Brief:
“The claim that global warming has stopped – one of the most overused and deeply flawed climate sceptic arguments – can finally be laid to rest today, following the publication of new data analysis by one of the country’s leading climatologists”
http://www.carbonbrief.org/blog/2011/06/global-warming-since-1995-statistically-significant
Comments are open (slow moderation at the weekend – ie paid staff) but no one that it will have been distributed to will ever read the comments…
Carbon Brief Background (sceptical scientists, LIndzen, Mckitrick, etc) smeared by the CB.
http://wattsupwiththat.com/2011/05/14/smear-job-by-the-carbon-brief/
http://wattsupwiththat.com/2011/02/18/the-carbon-brief-the-european-rapid-response-team/
Did anyone notice Paul Dennis’ comment at Bishop Hill about the BBC article(a UEA colleague of Jones)
“I’m rather bemused by the article. 1995-2009, no significant warming, 1995-2010 significant warming and perhaps 1995-2011 no significant warming depending on this years temperature. Who knows! Adding a year to the trend and suddenly claiming significance as the headline asserts (‘Global warming since 1995 ‘now significant’) really shows a complete lack of understanding of linear regression, let alone the nature of the data.
Jun 10, 2011 at 6:32 PM | Paul Dennis ”
http://bishophill.squarespace.com/blog/2011/6/10/jones-post-1995-warming-significant.html#comments
The point of Phil Jones’ statement is that you can’t expect to estimate significance reliably with short series (read it again). So I think that he and this Paul Dennis fellow are in violent agreement here.
Hi Lucia
Using monthly data and arma(1,1) correction for autocorrelation I find that the trend in Hadcrut3 is statistically signifikant since 1993-1994:
See this plot
If you read the BBC article (Richard Black) and The Carbon Brief, the point(tone) of the articles are clear… not debate about why the world is warming, but a pr/media piece to put a solitary point of view to discourage debate/questions.
At the end of the BBC piece look how Black spins the Berkley Best project for example (on 2% of data, uncorrected)
does anyone know why the SD’s for the monthly data from 1850-2010 HADCRUT3 decline?
The trend from 1940-2010 is quite noticeable, so the implication is that temperature distributions 50 years ago are different from those taken now.
I really do hope that Phil Jones did not use the variance adjusted version.
It would open up the whole can of worms about endpoint smoothing assumptions.
From the FAQs at this site:
http://www.cru.uea.ac.uk/cru/data/temperature/#faq
“In addition to this the method of variance adjustment (used for CRUTEM3v and HadCRUT3v) works on the anomalous temperatures relative to the underlying trend on an approximate 30-year timescale. Estimating this trend requires estimation of grid-box temperatures for years before the start of each record and after the end. With the addition of subsequent years, the underlying trend will alter slightly, changing the variance-adjusted values. Effects will be greatest on the last year of the record, but an influence can be evident for the last three to four years. Full details of the variance adjustment procedure are given in Jones et al. (2001).”
It would be devastating to discover that the statistical significance for 1995-2010 depended on estimations of temperature measurements that have yet to be taken.
Jorge–
That’s interesting.
If the final temperatures in the series are based on estimated temperature, that might also reduce the standard devaition of residuals, there by reducing the estimate of variability of 15 year trends. So, even if the mean temperature was estimated close to correctly, it would tend to overstate significance.
Doc Martyn
Possible valid reasons I can think of: The increase in measurement stations would result in a reduction in variance. Better measurement equipment could result in reduction in variance. Periods with minimal volcanic eruption should show less year to year variability that other periods.
You can show significance using the method of Quenouille. (That is assuming residuals are AR1. ) Of course one can debate whether than’s right– but you also get significance with a variety of ARIMA choices that describe the data fairly well. (Of course you can debate ARIMA.)
But the fact is, there is evidence warming is significant. I’ll have to dig up the Zorita paper and describe that.
BTW: The Keenan method has been discussed here in comments. His analysis assumes natural variability can be described with a model that is unphysical. So, in my opinion, he’s shown pretty much nothing.
Jorge,
See http://www.woodfortrees.org/plot/hadcrut3vgl/from:1995/plot/hadcrut3gl/from:1995
The variance-adjusted and unadjusted versions are virtually identical from 1995 to present
David
Yes. And if Jones did that, he might be able to claim significance. But… he doesn’t seem to do it.
Yes. But they differ just enough to affect the test for statistical significance using simple tests.
“The variance-adjusted and unadjusted versions are virtually identical from 1995 to present
Yes. But they differ just enough to affect the test for statistical significance using simple tests.”
That would seem to defy common sense.
We can ask all sorts of short-term questions such as the significance of the warming between 1995-present. With a small but persistent signal that is obscured by both random and non-random but cyclic noise signals that are often far larger than the forced signal, such short term analyses are of little value.
The simplest technique of all, that of moving-boxcar smoothing with a window wide enough to remove most of the non-random but cyclic effects (ENSO) strips the data of much of its noise and shows a clear, persistent, unabated warming in the past century. See http://www.woodfortrees.org/plot/hadcrut3gl/from:1880/mean:180/plot/gistemp/from:1880/mean:180/offset:-.09 which uses a 15-year smoothing window applied to both HADCRUT and GISS since 1880. The warming signal could not be clearer and its significance not lost on any of us.
Using AR(1) for the residuals is inappropriate – they don’t really follow an AR(1) so it is not really relevant whether that model is or is not significant. A model with AR(2) residuals fits the data better than a linear trend, using GLS, and the trend is significant there. I wrote an illustration analysing the HadCRUT3vgl data set using R that you and your readers might be interested in: http://ucfagls.wordpress.com/2011/06/11/global-warming-since-1995-now-significant/
You are right to point out that the result depends on the data one use and I too have been unable to ascertain, as yet, to what data Phil Jones was referring.
Gavin–
The reason AR1 gets attention is that it is — for some reason– very commonly used in climate science papers, and was used by the IPCC to test significance of 20th century warming in the AR4. (It doesn’t fit that data well either.)
That said: I get what you get on AR2. I used the ‘forecast’ library and auto.arima
auto.arima(temperatures, xreg=norm_years, stationary=TRUE,stepwise=FALSE)
tcrit_Student_OLS=qt(0.975,num_years-4);tcrit_Student_OLS
0.0083/0.0031
This gives a significant trend on for both HadCrut3 and HadCrut3v.
Gavin–
I should add: I’ve done numerical simulations on synthetic data and I’ve found that if you find the best fit ARIMA to a short data set, and the data are AR(1), you will systematically under-estimate the uncertainty even more than if you use Quennouille. (I think the difficulty is that it’s very difficult to pick out the correct ARIMA form with short data and the AICc criteria is strongly influenced by the magnitude of the residuals. So, when it make mistakes, the direction is toward under-estimating uncertainty intervals and so over stating significance. So this is a problem with using anything more sophisticated than Quennouille’s (1+R)/(1-R) to adjust number of degrees of freedom. (It also means no reviewer should let you get away with using something else– but likely some would because it does sound reasonable to use AICc to pick out the best fit to the residuals.)
Lucia, I did some downloading of temperature data series from KNMI and looking at them for statistically significant trends @ 5% probability and adjusted for AR1 using the Nychka method.
Annual with CRU3 agrees with your result of not having a significant trend. There is very little AR1 to adjust for.
CRU3 using monthly data is different: The difference between the trend slope and 1.96 times the adjusted (for AR1) SE becomes smaller as one proceeds from Dec 2010 to April 2011. The significance could well go not significant in another month if the anomaly for that month is in the range of the first 4 months of 2011.
Instead of April being the first month of significance it is more like April might be the last month of significance (not saying for how long).
I looked at monthly GISS250, GISS1200 and GHCN for surface temperature measurements and the trends there are larger and more significant than for CRU3. UAH and RSS (RSS projected based on 1995-2010 available data) the satellite measurements do not show statistically significant trends over the period of interest.
These calculations may have already been presented as I have not read completely here or at TAV.
Kenneth–
I think you are the first. I’m waiting to read whether Jones elaborates since the response to the basis for his claim will depend on what the basis for his claim actually is!
Re: Owen (Jun 12 06:32),
yes if you look into the math of the variance adjustments you will see that its tied to the the number of stations in a grid cell. as the number of stations stabilized you dont see much difference.
Owen
It’s frequent when the result is just on the border of significance. After a while the differences won’t matter any longer. My guess is “a while” is in about 6 months or less.
Owen:
Long term trend, no doubt (one that starts well before AGW forcing became important, which the AR4 suggests started circa 1975).
But the problem with your figure is your individual data points aren’t statistically independent (only one out of 15 points are independent.)
Seen that way, this isn’t a dodge for the question of what happened to warming from 2002 to 2010?
Owen (Comment #77096) June 12th, 2011 at 7:21 am
“We can ask all sorts of short-term questions such as the significance of the warming between 1995-present. With a small but persistent signal that is obscured by both random and non-random but cyclic noise signals that are often far larger than the forced signal, such short term analyses are of little value.
The simplest technique of all, that of moving-boxcar smoothing with a window wide enough to remove most of the non-random but cyclic effects (ENSO) strips the data of much of its noise and shows a clear, persistent, unabated warming in the past century. See http://www.woodfortrees.org/pl…..ffset:-.09 which uses a 15-year smoothing window applied to both HADCRUT and GISS since 1880. The warming signal could not be clearer and its significance not lost on any of us.”
.
A few points…
1. I agree that short term trends are misleading, that’s why the only useful trend is the overall trend. I suggest using the IPCC chosen date of 1850 (as the ‘start’ of an anthropogenic effect) is a good year to start the overall trend.
2. Just prior to your chosen start point of 1880, there was a significant (?) warming between 1870 and 1878. The HADCrut data goes back to 1850 and the overall temperature rise – although it looks mighty steep in your chosen WFT graph – is 0.8 deg C in 161 years. This equates to 0.057 deg C per decade as an overall trend. Or, if you prefer, 0.005 C per year. Maybe the Y axis should use whole degrees instead of decimals?
3. I have no argument with the statement that warming has occurred. What is not proven, however, is WHY it has occurred.
4. Your final sentence is, at least in part, subjective. The significance of a warming signal of 0.005C per year is certainly lost on me.
5. What significance are you referring to?
Carrick (Comment #77111)
June 12th, 2011 at 12:36 pm
“Seen that way, this isn’t a dodge for the question of what happened to warming from 2002 to 2010?”
——————————–
Well, I didn’t think so. 10-15 years is just too short a time to make any conclusions about this small, persistent, and cumulative long-term change in temperature. I have pointed out in the past that multi-year windows can produce short-term cooling in the midst of long-term warming ( see http://www.woodfortrees.org/plot/wti/from:1980/to:1987/plot/wti/from:1980/to:1987/trend ). In this latter case, we could say that there was no warming (in fact cooling) from 1980-1987, or we could say that the warming signal in that interval was not discernible from the substantial noise during that period. Now, if I go from 1987 to 1997 (http://www.woodfortrees.org/plot/wti/from:1987/to:1997/plot/wti/from:1987/to:1997/trend ), we see more short-term “cooling.” My two sequential short term analyses tell me that we had cooling from 1980-1997. Let’s quit acting like warming has stopped in the first decade of the 21st century. Nothing could be further from the truth.
Finally, look at the ENSO behaviour in your 2002 to present time period (http://www.weatherzone.com.au/climate/indicator_enso.jsp?c=nino34&p=monthly) – how would the Nino and Nina patterns be predicted to affect the slope of the temperature line in this time interval?
Arfur Bryant (Comment #77112)
June 12th, 2011 at 1:03 pm
“4. Your final sentence is, at least in part, subjective. The significance of a warming signal of 0.005C per year is certainly lost on me.
5. What significance are you referring to?”
You did your best to pick the smallest per-year temp change (from 1850 to present), rounding to 0.005C per year from 0.057C per decade, and assuming, I guess, that the temp change has been linear in the chosen interval and that it will extrapolate into the future at that same rate (even though the rate is substantially higher from 1970 to present).
I try to not belittle or minimize the impact of a 1-3 C change in global average temperature over the next couple centuries. The effect on world agriculture could be devastating, even at the low end. We do not know what impact release of carbon, largely CH4, from the warming tundra will have on global temperatures – it could be very large (http://nsidc.org/news/press/20110216_permafrost.html).
We are seeing a significant and increasing loss of northern ice albedo in the summers, providing many hours of direct absorption of sunlight energy by open waters in the Arctic Ocean and Hudson Bay. The loss of albedo drives further loss of albedo in a cooperative process.
lucia (Comment #77110)
June 12th, 2011 at 12:09 pm
“It’s frequent when the result is just on the border of significance.”
—————————-
Biologists often use the 90% confidence level to test for significance (you know, messy complex systems affected by a multitude of variables), while chemists generally use the 95% CL. The whole process shouts ARBITRARY so loud it hurts my ears. But I know, we need to have an objective standard to make decisions, and 95% is the law of the land.
“The effect on world agriculture could be devastating, even at the low end.”
Owen, The moral at the end of the Greek version is that ‘the story shows that this is how liars are rewarded: even if they tell the truth, no one believes them.
http://en.wikipedia.org/wiki/The_Boy_Who_Cried_Wolf
Andrew
Owen:
It does tell us something about the scale of AGW forcing compared to natural variation. It’s also always been interesting to me that the IPCC suggests that anthropogenic warming didn’t initiate until circa 1975, but we’ve had warming since the late 1800s.
In the future AGW could be off-scale compared to natural climate variation, but it hasn’t happened yet. But the real take home message is AGW forcing isn’t dominating climatic variation to this point…that is mostly an inconvenience to the people trying to argue for drastic measures now as well as those who try and attribute every blip on the weather map to AGW.
For the AGW smitten, prominence of short-term climate fluctuations is not really good news, as is transparently obvious to their responses to these sorts of observations.
Owen (Comment #77114)
“You did your best to pick the smallest per-year temp change (from 1850 to present), rounding to 0.005C per year from 0.057C per decade, and assuming, I guess, that the temp change has been linear in the chosen interval and that it will extrapolate into the future at that same rate (even though the rate is substantially higher from 1970 to present).”
.
No, Owen, I didn’t. I chose the 1850 ‘start year of the anthropogenic effect’ as chosen by the IPCC. I then chose the last full year of data with the only set that encompasses both years (HADCrut). Its not my fault the numbers don’t agree with your belief in a potential threat from CO2. I have not at any time extrapolated the linear trend into the future. I only deal in the facts as observed. I’ll leave doomsaying predictions to you…
Arfur:
1850 isn’t the “start” year according to the IPCC. It’s the 1970s. Note that prior to the 1970s you didn’t need to invoke AGW forcings to explain the temperature trend…according to the IPCC.
The year 1850 is when we first start having reliable surface temperature records, again according to the IPCC.
Owen says:
Not much. The trend goes from slightly negative (-0.026C/decade) to slightly less negative (-0.012/decade). I ran multiple regression with detrended HadCrut using ssn for solar, sato aerosols, and Nino3.4 for ENSO and then removed their modelled effects from HadCrut for an adjusted time series.
I blew the link to my plot. Let’s try it again.
Layman, thanks for the plot.
The take home message is it is OK to exploit upward swings in short-period variability as evidence of AGW , and it is also an imperative that downward swings that don’t fit the AGW narrative be written off as “natural variation.
(I’m not suggesting Owen is doing this. I happen to agree with him about the dangers of trying to interpret temperature trends over short time periods.)
Owen
Selecting a specific ‘p’ level to decree statistical significance is arbitrary. No one denies this. Nevertheless, 95% is used in many fields as a threshold. Having a threshold is someone useful to avoid everyone wasting time trying to explain observations that could just arise due to random chance. By having some traditional threshold, people who design experiments know they will likely need to collect a enough data to get a significant result before people will pay much attention to their results.
In the case of Jones press announcement– he would have been well advised to just wait before announcing statistical significance since 1995. Is 95% arbitrary? Yes. But it is a common standard and he said the data achieved it.
Agreed on Owen – he’s a good guy. As long as anyone discusses rather than use snark and dismissiveness then everything is cool in my books.
I dug up some scripts I was playing with a few months ago to generate the plot. It is all very interesting. For instance, when fitting an autocorrelation model to the “adjusted” and detrended series, the noise is still obviously much more complicated then AR1. In fact, I think removing these factors result in much more interesting residual patterns then similar detrending of the full series. Here are the residuals from detrending the “adjusted” HadCrut. I used a 3 month lag for both ssn and enso.
Lucia,
What if we provided for any fit the degree of statistical certainty for which the fit (or hypothesis test) is significant. For example, we are 83% confident that this data follows a linear trend, or 99.9% certain in this second case, etc. We don’t say right or wrong, yes or no, based on an arbitrary cut-off, but provide instead a level of confidence in our fit.
What do you think?
Arfur Bryant (Comment #77118)
June 12th, 2011 at 3:54 pm
“No, Owen, I didn’t. I chose the 1850 ‘start year of the anthropogenic effect’ as chosen by the IPCC. I then chose the last full year of data with the only set that encompasses both years (HADCrut).”
———
When you selected an 1850 anomaly and then a 2011 anomaly, and calculated a slope between the two points over the 161 year interval, that was the lowest slope you could produce. I used the 15-year smoothed graph ( http://www.woodfortrees.org/plot/hadcrut3gl/from:1850/to:2011/mean:180 ) to estimate a slope from 1911 to 2011 with your two point method and got ca. 0.085 C per decade. Then I used your method for the period from 1971 to 2011 and got ca. 0.16 C/decade. The slope is increasing and is currently nowhere near 0.057 C/decade.
I have a couple of questions just on the actual data-set its self.
1) Why is there a massive step change between 1944 and 1946. This is best viewed by plotting the yearly max anomaly vs. year?
2) Why does the SD fall in a linear fashion? What happens in 2130 when the SD of the monthly anomalies falls to zero?
3) Why does yearly max minus yearly min fall in a linear fashion? What happens in 2136 when the yearly max minus yearly min fall of the monthly anomalies falls to zero?
4) How can we do a simple statistical test to compare means of a 12 point yearly sample when we know that the anomaly distribution is different in different decades?
Owen
No. We aren’t 83% confident the data follows a linear trend or anythign like that. What confidence intervals mean is ” What is the probability of seeing an event like the one we saw if the null hypothesis is true.” When we are arguing about models for the stochastic part, it also means “under the assumption the model for the stochastic part is ‘x'” (with ‘x’ white noise, AR1, ARIMA, Fractional diff or whatever the person used to create an estimate.)
But you never become ‘x’ confident the alternative is true. Rejecting the null at a high confidence only means that if the null is right, the data we actually saw would have happened very rarely.
Papers generally state the ‘p’ value and also state whether the null was rejected at some particular confidence level. No individual person is required to be governed by a particular confidence interval. Saying 95% is common in a field only means that if you do experiments to reject something most people consider a plausible null then it’s highly unlikely reviewers are going to let you write a paper describing a result where you rejected it at 70% or something like that.
If your notion is correct, it is generally the case that if you take more data you can eventually get the 95% confidence interval. So…. go back to the lab and take enough data. Or in the case of global surface temperatures, just wait a few months!
People in most fields find it more productive to insist on some arbitrary cut-off and send people back to the lab rather than have loads of experiment results of experiments based on too little data being written up discussing results at low confidence intervals. That’s why the arbitrary cutoff exists.
Different fields have different balances for a variety of reasons– but 95% is pretty common. My undergrad stats book suggested 90%, 95% and 99% are all common — but in different fields. In fields where data mining is comment, they might insist on 99.9% for a result obtained by data mining.
DocMartyn–
I’m not sure I undestand all your questions, but I’ll try to answer:
1) There is a sudden step change in global step temperatures sometime during the 40s. It is widly attributed to a change in measurement technique for sea surface temperartures.
2) By “the SD”– could you clarity “the SD” in what? Do you mean in the errors discussed by hadcrut? Why do you think the errors would all go to zero in 2130? I’m mystified by the question.
3)Dif in yearly max or min in what? And how do you know what happens in 2136?
4) How do you know the anomaly distribution is different in different decades? (And what do you mean by it being different?)
Why am I reminded of this from xkcd?
http://xkcd.com/882/
Oops. I messed up the solar lag in my code when I posted the enso, solar, aerosol adjusted HadCrut since 2002. The correction now results in a significant trend of 0.095 per decade for the adjusted series.
I think one should merely report the observed significance and leave the decision up to the reader. That’s who decides anyway.
Lucia,
As I understand it, Phil Jones one year ago said that there was no significant warming since 1995 at the 95% confidence level, but there was was significant warming at the 90% CL. What if he had determined that significant warming still occurred up to the, say, 92.7% CL (ie, this is the maximal confidence level at which we can still reject the null hypothesis based on available data)? Then Jones could have stated that we are 92.7% confident (but not higher!)that warming has been significant (i.e., > 0) since 1995.
No black and white, yes or no, simply a level of confidence. People could then decide for themselves whether or not they would accept the findings.
Owen
He could have said any number of things– but he didn’t. In his recent press release, he could have said any number of things, but chose to say warming was now significant at 95%. People want to know if what he said was true or not and they also want to know the basis for his claim.
That he might have said something different or that he might not have said something to the press when he thought the 95% confidence level was passed etc is rather irrelevant to the answer one supplies people who ask “Is what Jones claimed correct”.
People can and do anyway. But when people ask if 95% is commonly used: It is. If you do an experiment and want to publish, it’s quite likely reviewers will want to see 95% confidence levels. If you don’t achieve that you go get more data. (There can be rules in certain fields saying you have to start from scratch. These apply because you don’t want to have people deciding to be monitoring results as they come in and the ‘stop’ the experiment when they get significance. “Two out of three flips win!”. Ok… let’s make that 3 out of 5! Ok,let’s make that 5 out of 7 and so on until they get the results they “like”. But in other fields, the data keeps rolling in whether people want it to or not, so that rule is suspended because even if you publish your paper, data will still come trickling in and other can see if the results reversed. In contrast, in a pharmaceutical experiment, if a group stops the experiment, it’s just stopped. No more data comes in.)
lucia, I just had a look at the data in the hadcrut3gl.txt. file.
Firstly I looked to see if it was a continuous data series. The huge step change between 1944 and 1946 indicates that it is not.
So the data falls into two distinct blocks 1850-1944 and 1946-2010.
Then I looked at the anomaly distribution for each year, looking at the standard distribution, the max, min and max minus min.
The standard distribution of the 12 months, from which the mean is derived, indicates that the distribution of the temperature anomaly gets smaller each passing year.
The max minus the min temperature anomaly of the 12 months also gets smaller each passing year.
Plot either the standard distribution or the max-min vs time and the intercept is around 2130.
So either CO2 alters temperature so that the variance of each month gets smaller, with respect to some nominal value, or the data has be ‘enhanced’ so as to arrive at this position.
What is clear is that the variance of each monthly anomaly shrinks, with respect to time. Which is a little odd to say the least.
Doc
This might be arising due to the combination of increased number of stations and better measurement equipment. These are also why Hadley’s estimate of uncertainty is greater for past years.
Well…. if the weather/climate process is stationary and the decrease in sd is due to improvements in equipment of a larger number of stations, the sd would be expected to approach an assymptotic value determined by the properties of weather.
Yes. But I think the more common notion is the decrease in sd is due to increasing number of stations which permit better resolution of the measurement and/or better equipment.
Now that I know what you are asking: At least if you go back in time all the way to the 1900, this isn’t odd. It’s expected because the surface coverage was once so poor.
The question of whether we can treat the data series as homoskedastic is a good one– but it’s not really that surprising if it’s not.
Lucia,
”
People could then decide for themselves whether or not they would accept the findings.
People can and do anyway.
——————————————-
Actually, when Jones is asked if there has been significant warning and he says NO, even if people know he is using the 95% CL they don’t know whether the actual level of confidence is 5% or 94% (to reject null hypothesis) – they just know that we fail to reject the null hypothesis at 95%. The common everyday reader assumes therefore that the results are conclusive and settled – there has been no warming, period (easily misleading to the public).
Owen– If your point is that people don’t understand what “statistically significant” means, I agree.
But be careful what you recommend. Telling people that the warming is significant at 94% vs 95% really doesn’t resolve the problem. After all: statistically significant at 94% does not mean the probability of warming is 94%. If you start advocating people only reporting the p value and dispensing with arbitrary cut-offs for what a field calls “statistically significant”, you’ll observe other problems that are worse. For example: How would you feel a bout someone reporting that cooling since 200x has a ‘p’ value of something like 60 or 70% (whatever it is right now) and having people think the probability of cooling was 70% or so? (If the trend since ‘x’ shows cooling at all the ‘p’ value coming out of these tests will always be cooling with p greater than 50%. )
People should report both the p-value and whether or not that p is considered ‘statistically significant’ in a particular field. The declaration of ‘statistically significant’ is especially important for people who know nothing about statistics. If scientists do a test using frequentist statistics and decide to stop saying things like “it’s statistically significant at p=95%”, I’m pretty sure mis-conceptions will be worse not better.
“People should report both the p-value and whether or not that p is considered ‘statistically significant’ in a particular field.”
————————————
Such a practice would be a great improvement.
Owen, I always report the p-value. I don’t always make a “judgement” as to whether I consider that significant, but it’s usually pretty obvious.
(With experimental measurements, you can get p< 1e-11. 😉
Carrick
No post on the Lindzen and Choi fiasco, what a travesty. I find it incredibly amusing that I tell you again and again, the radiative forcing value for co2 is wrong and that there is more radiation escaping into space than is calculated as well. Also, Dr. Happer again is completely smeared and is not deemed credible enough to review the paper. I am wondering what personal grudge you hold against him?
I recommend every1 here to go check out masterresource to see how the global warming team simply cheats to block papers they don’t like.
@Owen
You know what else is misleading? Claiming the world is warming while its below historic GAT. Furthermore, it is also misleading to claim a warming trend because the previous cooling trend lasted far longer than the supposed warming trend. I say supposed because I think some of the calculations are wrong, I think the surface stations are useless and the adjustments to previous temperature records have all left me highly suspicous.
Carrick (Comment #77119
.
“Arfur:
chose the 1850 ‘start year of the anthropogenic effect’ as chosen by the IPCC.
1850 isn’t the “start†year according to the IPCC. It’s the 1970s. Note that prior to the 1970s you didn’t need to invoke AGW forcings to explain the temperature trend…according to the IPCC.
The year 1850 is when we first start having reliable surface temperature records, again according to the IPCC.”
.
Carrick, you are wrong, wrong and right (in that order). You are correct in that 1850 was the year when reliable data was first recorded, which is why the IPCC uses 1850 for the start of many of the x-axis on its graphs. This is why I referred to it as the ‘start’, although the IPCC quotes the start as the mid-eighteenth century. Do you seriously want to prove your point by using a series of graphs of models? Here are a few quotes from the IPCC AR4 and FAR:
.
http://www.ipcc.ch/publications_and_data/ar4/wg1/en/spmsspm-human-and.html
.
“Global atmospheric concentrations of carbon dioxide, methane and nitrous oxide have increased markedly as a result of human activities since 1750…†and
“The understanding of anthropogenic warming and cooling influences on climate has improved since the TAR, leading to very high confidence[7] that the global average net effect of human activities since 1750 has been one of warming, with a radiative forcing of +1.6 [+0.6 to +2.4] W m–2 (see Figure SPM.2). {2.3, 6.5, 2.9}…â€
.
http://www.ipcc.ch/ipccreports/far/wg_I/ipcc_far_wg_I_spm.pdf
.
Fig 8: “Simulation of the increase in global mean temperature from 1850-1990 due to observed increases in greenhouse gases, and predictions of the rise between 1990 and 2100 resulting from the Business-as-Usual emissions…â€
.
http://www.ipcc.ch/ipccreports/far/wg_I/ipcc_far_wg_I_annex.pdf
.
A.2: “The year 1765 was chosen as the pre-industrial baseline for greenhouse gas concentrations (and, hence, for man made forcing and global mean temperature)…â€
.
In summary, the IPCC makes it quite clear that anthropogenic effects (in terms of GHGs) started well before 1970!
.
For your second error: you don’t need to invoke AGW after the 1970s, let alone before them! The trend in the period 1970 to 1998 is matched quite easily in the periods 1872-1878 and 1910-1945. If AGW didn’t start until the 1970s, how do you explain those warming periods? Natural variability works both ways…
Owen (Comment #77126
.
“When you selected an 1850 anomaly and then a 2011 anomaly, and calculated a slope between the two points over the 161 year interval, that was the lowest slope you could produce. I used the 15-year smoothed graph ( http://www.woodfortrees.org/pl…..1/mean:180 ) to estimate a slope from 1911 to 2011 with your two point method and got ca. 0.085 C per decade. Then I used your method for the period from 1971 to 2011 and got ca. 0.16 C/decade. The slope is increasing and is currently nowhere near 0.057 C/decade.â€
.
Owen, your cherry-picking of start dates is unimpressive. Why use 1911 and why use 1971? These are random dates without meaning, carefully chosen by you to try to give some credibility to your belief in cAGW. Short-term, interim trends are meaningless compared to the overall trend. Your argument about me picking end-points is silly. The total global warming since the start of the HADCrut dataset is 0.8 deg C. That is in 16.1 decades. Do the math. If you want to use smoothing, that’s fine by me. Go ahead – the result will be similar. If you use 1850-2010 (or 2011 if you like) with your 15-year smoothing (Wow! That’s some smoothing…), you will get a 0.059 C per decade trend. The long term trend is about 0.06 per decade and no amount of number-massaging will change that. Of course there are steeper short-term trends as well as negative ones, but they don’t count. You need to be consistent.
.
Picking any trend shorter than the overall trend will be a weak argument. If the AGW signal is evident, the overall trend must become steeper. Any straight line drawn from the origin of (whatever) start date must (repeat, must) eventually be steeper. This is what was implied in the MBH98 graph and this is what has most certainly not (repeat, not) happened.
.
Therefore my question still stands – to what significance are you referring when you say the significance is there for all of us to see?
.
Oh, and by the way, as the temperature has been relatively flat for over 12 years, the slope cannot (repeat, cannot) be increasing.
Just for kicks I looked at some past statistical significances of trends for CRU3 using monthly and annual data starting from Jan 1995 and going to the indicated end month or year as shown in the tables linked below. I computed the trend slope (TS); the standard error of the slope, SE, and the first order autocorrelations of the regressed residuals, AR1. I then subtract from TS the value 1.96*adjusted SE where SE has been adjusted for AR1 using the Nychka method. If the difference is positive the trend is significantly different than zero with a p-Value = 0.05 and alternatively if the value is negative the trend is not significantly different than zero.
Using annual data it can be seen that the trend is not different than zero for the periods 1995-2008, 1995-2009 and 1995-2010. Using monthly data the trend was not different than zero for the all ending months Dec-08 to Oct-09 and then it becomes an increasingly more significant trend from Nov-09 to Nov-10 and then decreasingly significant to Apr-11, where it is just barely significant.
These historical trends from 1995 would seem to go against what Phil Jones said about 2009 and 2010 (or April 2011) whether you use annual or monthly data.
I would judge that a lot of this is dithering with the data and does not have a lot of importance and probably Jones would agree and is why you probably will not hear from him explaining how he made his calculations. The world now knows we are back on track for short term current global temperature trends that are positive and different than zero and that is what is important – or at least for an advocate.
http://img716.imageshack.us/img716/844/cru3trends.png
Cadbury, Happer is about as well qualified to review climate papers as he is to review string theory ones. In other words not at all.
The problem is Eli you dont need to be a physics expert to see the problem with the Lindzen paper. You dont need to be a dendro expert to see the problems with some of mann’s papers. Happer is definitely qualified to see the problems, whether he would point them out is another matter. Clearly the reviewers of mann were qualified to see the shambles he made of locating proxies correctly, but they did not.
Being “qualified” in the field and being ABLE and WILLING to find errors are two different things
I shoud have further explained in my previous post that the difference btween the trend slope and the 1.96*adjusted SE for the trend slope is in degrees C per decade. Also the 1.96 multiplier while near correct for the monthly data should have been increased to 2.13, 2.15 and 2.16, respectively, for the yearly data from 1995-2010 , 1995-2009 and 1995-2008. This would make these differences in the direction of the trend being different than zero even more unlikely and not changing the conclusions one might want to draw from it.
Mosher, you are joking of course. If you read the reviews of L&CII you see that they rely on previous work, e.g. the reviewers knew the field, they rely on observed data and they rely on a knowledge of the field, all of which Happer lacks. So of course, in your view, he could see all the problems, which kind of tells everyone where you are viewing from.
For some examples look at the most positive (but still negative ) review, #4
They rely on previous work:
They rely on their knowledge of the data:
They rely on their knowledge of the field:
I find some irony in the selected passages above provided by Eli Rabett criticizing L&C. The three selected criticisms boil down to (a) charges of cherry-picking; (b) overbroad/ungrounded assumptions and (c) failure to understand moisture-related feedbacks.
I thought that those three alleged flaws (cherry-picking, shaky assumptions and simplistic water vapor feedback modeling) are the basis for all policy-conscious climate science.
People in glass houses and all that…
I like how Jones screws up and its his fault that dishonest people will now use his screw-up to double down on their dishonesty. If people were really interested in the truth, they might do posts on how 15 years without statistically significant warming doesn’t mean no warming at all.
But this is another cautionary tale of playing by skeptic rules. The proper answer is that by using the same data analysis, the last fifteen years cannot statistically rule out trends one and a half times what the IPCC predicts. Oh, well, liars gonna lie.
“Oh, well, liars gonna lie.”
Thanks, “Boris”. Tell “Natasha” I said hello.
Andrew
Boris–
I wrapped up my post in 1995 with
Huh? Using red noise we get 0.22C for the max trend. This is 10% higher than the nominal trend, but 1995 is shortly after Pinatubo. You have to compute multi-model mean trend starting in 1995 to compare. Have you?
Apparently Jones still insists he’s right, see this incoherent article at “New Scientist”.
“By releasing the latest analysis,”
Huh? What “analysis” has he released?
“Jones insists that his latest analysis is correct because it is a combination of land and marine temperatures.”
Double Huh??
I noticed an exchange upthread in which Arfur Bryant and Carrick disagreed about when “global warming started,” and I thought perhaps I could clarify the issue. Carrick’s argument about global warming starting in the mid-20th century stems from conflating two issues. He is not referring to a start of warming. Instead, he is referring to when that warming was distinguishable in the record.
There is noise in data. This means when a change happens in the data, it is not immediately noticeable. However, it is still there. In the same way, human induced warming began well before the 20th century, but confirmation for it exists only in the 20th century.
Re: Brandon Shollenberger (Jun 14 12:35),
If you believe Ruddiman, AGW began ~8,000 years ago with the invention of agriculture and the associated land use/land cover changes plus methane production. CO2 induced warming is thought to have become significant when the rate of change of CO2 became high enough. 1970 is not an unreasonable estimate of when that could have happened.
Brandon Shollenberger (Comment #77199)
.
Brandon,
Thanks for trying to clarify the exchange between Carrick and me but your post raises a few dubious points:
Firstly, I assume that when you say ‘global warming’ in your first paragraph you are referring to ‘anthropogenic global warming’, yes?
.
Secondly, how do you know that confirmation of AGW exists in the 20th century – or any other century for that matter? In other words, how do you know ‘that (anthropogenic) warming’ was distinguishable in the record?
.
To clarify that second point, I accept that it is theoretically likely that anthro-CO2 has made a contribution (albeit very small) to global temperature, I’m just not sure if any evidence exists to confirm that idea.
.
The warming of the late 20th century is significant but it is not necessarily confirmed as having been caused by anthropogenic factors.
DeWitt Payne, I don’t think that requires much “belief” in anyone (unless you want an estimate of impact). It seems obvious to me humans have impacted the planet’s climate for thousands of years. I just didn’t see it as relevant to the disagreement I was discussing. I’ve actually been attempting to get more information about how historic methane levels are determined in part because of my interest in that very topic.
Arfur Bryant, your assumption is correct. As for your questions, I wasn’t meaning to take a side. I just wanted to explain what the sides were. Unfortunately, as I see now, my second paragraph implies otherwise.
I’m not inclined to argue the position I unintentionally adopted as I have issues with the “consensus” position. I apologize for wording that second paragraph poorly.
Brandon,
No apology necessary, old chap. Your comments are appreciated.
Regards,
Arfur
DeWitt Payne (Comment #77200)
“CO2 induced warming is thought to have become significant when the rate of change of CO2 became high enough. 1970 is not an unreasonable estimate of when that could have happened.”
.
Ok DeWitt, I’ll bite…
.
The warming trend from 1910 to 1945 was appx 0.24 C per decade. This was before the time “CO2 is thought to become significant…” (do you have a reference for that thought, by the way?).
.
The warming trend from 1970 to 1998 (when the temperature stopped increasing) was appx 0.23 C per decade.
.
Please explain why the trend after the ‘significant rate of CO2 increase’ is equal to or less than the rate before…
Brandon:
I’m afraid you misunderstood what I said.
I never said warming started in the mid-20th century.
What I said was according to the IPCC, we can explain all of the warming through natural forcings until circa 1975. This comment was in reference to one made by Arfur:
That statement is false. it’s not the “start year chosen by the IPCC” for the anthropogenic effect.’
Since the IPCC graphic backs up my statement, I’m not sure what there is to really argue with.
Arfur:
Since it’s a series of graphs of models chosen by the IPCC to represent their view on climate change, yes, of course.
Otherwise we can just make up stuff.
I think lukewarming or whatever is no longer relevant please read its really over for the AGW camp it is
is no longer relevant
http://news.google.com.au/news/more?pz=1&cf=all&cf=all&ncl=d-UvhPYxunMkIUM5hc6ayHMqN1V7M
for different reasons both skeptical and pro. Mainstream media are onto this big time 3 independent studies. The fact it just MAY cool dramatically will be enough to scare the S### out of both warmistas and deniers.
Carrick (Comment #77207)
“Otherwise we can just make up stuff.”
Carrick,
LOL! SO you’re saying that the IPCC graphic of model predictions re-inforces your quote:
“…1850 isn’t the “start†year according to the IPCC. It’s the 1970s.” Yet you completely ignore the three references I have provided which are actual statements from the IPCC FAR and AR4 which quite clearly state that anthropogenic warming (in the opinion of the IPCC) started well before 1970!
.
I suggest you are the one who is making stuff up. Choosing a random model graphic does not constitute a ‘statement by the IPCC. Have a serious think about what you’re saying.
.
And if you and the IPCC can explain ‘all of the warming due to natural forcings’ before 1975′, perhaps you’d like to address the question I put to DeWitt above about why the warming from 1910 to 1945 is indistinguishable when compared to the warming from 1970 to 1988? According to you and the IPCC, one is natural and one is anthropogenic. How do you know which is which?
.
I will have to leave any further response until tomorrow evening…
Well, I did a fit of 1910 to 2010 to a fifth order polynomial, and from that got the rate.
The residuals show the fit is nice, although with a fifth order polynomial it damn well should be.
The rate shows a linear rate 0.0095 degree per year, unchanged since 1910, and a cyclical rate change that is +/- 0.0125 degree per year at peaks 1917-1955-1992.
So the global heating will be very slow until about 2029.
[IMG]http://i179.photobucket.com/albums/w318/DocMartyn/hadcrut3fit.jpg[/IMG]
Arfur, can you read graphs?
It’s from their report. I’m pretty sure it didn’t get inserted in there randomly.
(As to the other, I was under the impression we were discussing what the IPCC claims, not what DeWitt or I believe to be true.)
Re: DocMartyn (Jun 14 16:56),
Exponential plus sine fits have been explored here. Quasi-periodic oscillations like the AMO may be involved. But that’s starting to get far OT for this thread.
Just playing around here a little, but if you go back to 1700 and use the IPCC AR4 Model’s CO2 warming response rates versus the actual Hadcrut3 temperatures, it is not off by alot.
Still less than expected, but it does start to make one think about the aerosols offset and the 1970 warming start date. By 1890, for example, temps should have been up by 0.2C already.
http://img818.imageshack.us/img818/119/ipcc1700tempstheory.png
Dewit–
I don’t think it’s off topic. Or… if it’s off topic, here it’s on topic on the climacogram posts! I’m trying to figure out how to create decent climacograms for vostok data. (I made some without converting depth to time.)
Bill-
Maybe the were. 🙂
Carrick, perhaps I did misunderstand you. You claim you “never said warming started in the mid-20th century.” This is true. However, as Arfur Bryant pointed out, my comment was about anthropogenic warming, and you did say that started in the 1970s:
I assumed this was poor wording, as that is not what the IPCC says. The IPCC says anthropogenic warming started before the seventies, but it isn’t until the seventies the anthropogenic nature of it can be confirmed. This interpretation requires I assume poor wording on your part.
Now then, I could be wrong. You could have literally meant what you said. In that case, you’d be making things up about what the IPCC said, and you’d be wrong in a rather dumb manner. But hey, at least your wording would have been clear.
I apologize if assuming your wording was wrong was a bad idea. I just didn’t think you could actually mean anything as absurd as what you said. I guess having faith in people can be a bad idea.
Brandon, I’m really puzzled what you are getting confused about.
According to the analysis presented in the IPCC AR4 it is unnecessary to invoke any anthropogenic forcings to explain the observed variation in global mean temperature until the 1970s. That’s the implication of the figure, assuming you know how to interpret it. So unless you’re suggesting I made up the figure, there isn’t much to argue with here.
As a footnote, I choose the words I use carefully, and I mean them to be used in the technical, rather than lay sense.
To disambiguate, of course there were anthropogenic forcings befor 1975… for at least for the last 8000 years, as DeWitt observed above. But “warming” implies a measured increase in temperature. “Nonmeasureable anthropogenic warming” is the same thing as saying “not anthropogenic warming” and “anthropogenic forcing” means something different than “anthropogenic warming”. (The latter means you’ve observed a warming, and you have attributed it to anthropogenic causes.)
If you meant “forcings” when you said “warming”, there wouldn’t be anything left to disagree about.
There… all that said without suggesting you are saying dumb or absurd things (which was the passive aggressive subtext of your own comment). Was that difficult?
Carrick, you say:
Imagine you have a temperature sensor in a closed room which only measures in multiples of five degrees. A heating element in the room is turned on, and a minute later, a person says, “That heating element has warmed up the room some, huh?” You haven’t measured the temperature, but he is obviously right.
Moreover, if I’ve measured warming, but I can’t attribute it to anthropogenic causes (you incorrectly name this “nonmeasureable anthropogenic warming”), that does not mean it is “not anthropogenic warming.” As with my example, knowledge about something does not dictate it’s nature.
You claimed the IPCC said anthropogenic warming started in the seventies. It said no such thing. It said anthropogenic warming could be distinguished only after that point (given current approaches). You have taken this clear point and claimed it means the IPCC says anthropogenic warming started in the seventies. When this error was pointed out, you said nonsensical things to defend your ridiculous conflation. So with no passive-aggressive subtext, let me say just this.
What you are saying is incorrect, misrepresents the IPCC, causes pointless confusion, and quite frankly, is idiotic.
Brandon, sorry but it is an abuse of language to intermix “forcing” with “warming”.
Warming is an empirical statement regardless of whether you have a precision of 5°C, 1°C, or 0.1°C (or whether it is measure sensually as in your example). If you don’t understand scientific terminology, perhaps you could spend a bit more time brushing up on it and a bit less time trying to lecture other people on the meaning of words…
Beyond not understanding the difference between forcing and warming, you apparently have trouble with that word “claim”. I linked a graph from the IPCC AR4, which has a clear interpretation in this case. That’s not a “claim”, that’s called a reference.
I see you’ve chosen to mix rudeness with your puerile arguments. Have the floor.
DeWitt Payne, the only reason one ever uses a polynomial is that you get the fit in a form that allows very integration. All I did was fit the data and get the rate. The numbers fell where they did.
Brandon Shollenberger (Comment #77258)
[“What you are saying is incorrect, misrepresents the IPCC, causes pointless confusion, and quite frankly, is idiotic.”]
.
Brandon, welcome to the Twilight Zone…
…where the interpretation of a graph of observed data and model simulations is magically transformed into said graphs becoming a reference for a CLAIM made by Carrick that anthropogenic warming started in the 1970s:
[“Carrick (Comment #77119) June 12th, 2011 at 4:00 pm
Arfur:
…1850 isn’t the “start†year according to the IPCC. It’s the 1970s.”]
.
Do-dee-do-dee-do-dee-do-dee… “You’re traveling through another dimension — a dimension not only of sight and sound but of mind. “
Carrick (Comment #77211) June 14th, 2011 at 4:58 pm
.
Arfur, can you read graphs?
Choosing a random model graphic does not constitute a ‘statement by the IPCC.
It’s from their report. I’m pretty sure it didn’t get inserted in there randomly.
(As to the other, I was under the impression we were discussing what the IPCC claims, not what DeWitt or I believe to be true.)
.
Carrick, can you read English? You still ignore the three references I have provided you in which the IPCC states (not interpretations from a series of graphs) clearly:
.
“The understanding of anthropogenic warming and cooling influences on climate has improved since the TAR, leading to very high confidence[7] that the global average net effect of human activities since 1750 has been one of warming,”
.
And: “The year 1765 was chosen as the pre-industrial baseline for greenhouse gas concentrations (and, hence, for man made forcing and global mean temperature)…”
.
How much clearer do I have to make this?
.
According to the IPCC’s explicit statements, they consider that AGW started in the mid-eighteenth century. However, they consider that reliable observed data started in 1850.
@Eli
Eli you and I know both know full well that Happer has an expert understanding of the absorption bandwiths of co2. He has studied the greenhouse theory in detail, although his official title may not be a greenhouseologist. Just to compare though, look at Heidi Cullen. Do you honestly think she knows even half of what Happer knows about the system? There are several other alarmists who are also woefully inadequate to publish anything. I’ll give you the point that his specialty is quantum mechanics though, not climatology. However, Robert Laughlin is also a physicist and if you read What the Earth Knows, you can see that he has a master’s grasp on these concepts and beyond what many of the mainstream scientists know.
@Eli
Schmidt, Mann and Trenberth do not come off as very smart because instead of speaking about historic climate, they play pretend and act like the earth formed in 1950, it is absurd. The best and most credible researcher that believes in global warming, in my opinion, is Isaac Held. The reason I say this is because he has actually admitted that there is a lot of uncertainty regarding global warming theory because of the massive amounts of co2 the earth has had before.
Arfur, it truly is mind-boggling. He seems adamant about conflating “we don’t know” with “it is not,” yet he suggests I don’t understand what words mean.
I genuinely thought I was helping solve a simple misunderstanding. I never thought he could have actually meant that.
Carrick is correct on this one, the AR4 argued that anthropogenic forcings only became statistically distinguishable from natural forcings post-1970s. That doesn’t mean that there were no anthropogenic forcings pre-1970s, just that their magnitude was small enough that they were not distinguishable from a world with no anthropogenic forcings, at least from a GCM standpoint.
Zeke, that isn’t the point in contention. It is actually the very interpretation I initially gave and was told was wrong. He has clearly shown that is not his position.
Brandon Shollenberger (Comment #77331)
June 15th, 2011 at 3:17 pm
I also have been following along on this topic. I had understood Carrick’s point to be exactly the one that Zeke elaborated. Prior to 1970, the inherent noise in the climate system was too large to allow the smaller warming signal to be decisively assigned to forcing. Warming did occur from, say, 1910 to the present, but after 1970 could clearly be assigned to greenhouse gas forcing. At least that’s the way I understand the situation.
Owen, he clearly said anthropogenic warming started in the seventies, not that it became distinguishable in the seventies.
I agree your interpretation is more sensible, and it is basically the same one I suggested, but Carrick has clearly rejected it.
Sounds like folks are talking in circles past each-other. Lets just let Carrick clarify what he meant when he gets back, as I’m rather sure that the overall picture here (various influences pre-1970 like GHGs and aerosols, but only differentiable from background natural variability post 1970) is not too controversial.
Zeke and Owen,
While we’re waiting for Carrick to clarify what he meant when he said [“…’start year of the anthropogenic effect’ as chosen by the IPCC. 1850 isn’t the “start†year according to the IPCC. It’s the 1970s.”]…
.
How can you attribute the warming from 1970 to 1998 to anthropogenic GHGs and yet attribute the warming from 1910 to 1945 to natural factors? The trends are exactly the same.
.
What evidence do you have that the later warming can clearly be assigned to GHG forcing?
Zeke, as it stands, he’s made it quite clear. I’m happy to let him post more about his position, but in the mean time, I see no reason to ignore what he has clearly stated. He indicated he intentionally said “the IPCC suggests that anthropogenic warming didn’t initiate until circa 1975” the way he said it, and he stands by it. This is clearly not the position you suggested.
Seeing as he has made his position clear, why should we wait for him to clarify it?
Arfur Bryant (Comment #77344)
June 15th, 2011 at 4:28 pm
“How can you attribute the warming from 1970 to 1998 to anthropogenic GHGs and yet attribute the warming from 1910 to 1945 to natural factors? The trends are exactly the same.:
——
I look at it this way: if you consider the HADCRUT temperature series from 1850 onward (http://www.woodfortrees.org/plot/hadcrut3gl), the warming signal (which began in the late 1700’s due to the industrial revolution) was quite small as the CO2 produced by a smaller world population with less overall industrialization was far lower than today. That smaller signal is lost in the considerable noise of the climate system. I therefore consider the period on the HADCRUT graph from 1850 to, say, 1920 (somewhat arbitrary) to be largely noise. As CO2 emissions have increased and the cumulative increase in the atmospheric CO2 (and other greenhouse gases) has risen, the warming signal has risen out of and above the background noise. Only when rises above the noise can it be considered significant. The warming signal, however, has been present all the time since 1850, and is increasing, but it is not increasing necessarily in a linear fashion.
Owen (Comment #77352)
[“I therefore consider the period on the HADCRUT graph from 1850 to, say, 1920 (somewhat arbitrary) to be largely noise. As CO2 emissions have increased and the cumulative increase in the atmospheric CO2 (and other greenhouse gases) has risen, the warming signal has risen out of and above the background noise. Only when rises above the noise can it be considered significant. The warming signal, however, has been present all the time since 1850, and is increasing, but it is not increasing necessarily in a linear fashion.”]
.
That is a change from your support of the year 1970 as suggested by Carrick. You originally stated that [“Prior to 1970, the inherent noise in the climate system was too large to allow the smaller warming signal to be decisively assigned to forcing. Warming did occur from, say, 1910 to the present, but after 1970 could clearly be assigned to greenhouse gas forcing.”].
.
Now you say that the noise was mostly before 1920. Are you now saying that the warming period 1910-1945 is assigned to GHG forcing, or not? If not, why did you support Carrick’s assertion in the first place? If so, how do you rationalise the similarity between the two trends (1910-1945 and 1970-1998)?
.
Furthermore, the observed temperature has stopped increasing since 1998, and yet atmospheric CO2 has continued to increase at a more-or-less linear rate. Logic, therefore, would suggest that your statement [“As CO2 emissions have increased and the cumulative increase in the atmospheric CO2 (and other greenhouse gases) has risen, the warming signal has risen out of and above the background noise.”] is demonstrably incorrect. The ‘cumulative’ effect should be increasing the warming, not decreasing it. The absence of further warming therefore suggests that the warming of the late 20th century was not assigned to CO2 – as you postulate.
.
Unfortunately, there is just no ‘real world’ evidence to suggest that the CO2 signal can have anything other than a marginal or insignificant effect on global temperature.
Arfur,
“Logic, therefore, would suggest that your statement [“As CO2 emissions have increased and the cumulative increase in the atmospheric CO2 (and other greenhouse gases) has risen, the warming signal has risen out of and above the background noise.”] is demonstrably incorrect.”
————-
Your logic maybe. Please take a look at decadal average global temperature anomalies since 1880, relative to a 1951-1980 baseline (http://en.wikipedia.org/wiki/Instrumental_temperature_record):
2000-2009 = +0.535C
1990-1999 = +0.313C
1980-1989 = +0.176C
1970-1979 = -0.001C
1960-1969 = -0.014C
1950-1959 = -0.020C
1940-1949 = +0.035C
1930-1939 = -0.043C
1920-1929 = – 0.175C
1910-1919 = -0.276C
1900-1909 = -0.259C
1890-1899 = -0.254C
1880-1889 = -0.274C
The decadal averages, which smooth out much of the climate system noise, show that the global temperature was fairly flat from 1880-1919. But, boy has the temperature shot up in the past 4 decades! No one could possibly miss that – not even you.
Owen,
What a pointless post. There is no need to produce a tome from wikipedia when I have never tried to ‘deny’ that warming has taken place. Where have I done that? I have consistently queried the assignment (as you put it) you give to the CAUSE of the warming being CO2. The signal we were discussing was the CO2 signal, not the warming itself. Stop trying to change the subject.
.
Your earlier graph of HADCrut temperature from woodfortrees was quite sufficient for everyone to see the data. Everyone can also see that the temperature has not risen above the 1998 figure.
Re: Owen (Jun 16 15:00),
Even better, take the 66 year moving average (792 samples) to remove the quasi-periodic signal and you get something that looks a lot like the ghg concentration.
Arfur,
Please feel free to keep telling yourself whatever stories you’d most like to believe. Just ignore that decadal data from Wikipedia and keep telling your self that there is no evidence that CO2 is a forcing agent. Keep repeating the mantra that there has been no warming since 1998 (hurry, it won’t be true much longer) and ignore the fact the the first decade of the 21st century has been, by far, the warmest on the instrumental record. Keep shouting it from the rooftops – maybe some day you’ll actually believe it.
Owen,
Ok, you don’t want to address the subject we were discussing and you want to talk nonsense. You keep avoiding actually producing the evidence for the forcing agent which you believe in. You keep talking about ‘the warmest decade’ and keep forgetting that when you’re at the top of a hill you are higher than you were on the way up but you’re not getting any higher. But you know – for a fact – that the no-warming period “won’t be around for much longer”. Very objective of you.
.
If you don’t want to answer my perfectly reasonable question about the two trends either side of the 1970 ‘anthro warming divide’, just say so. Or, like Carrick, just ignore it and pretend you didn’t engage. Trying to deflect difficult questions is a typical warmist tactic. Have a nice day.
“there is no evidence that CO2 is a forcing agent”
I’ve never seen any. Owen, what *evidence* would you put forth that C02 is a forcing agent?
Andrew
DeWitt,
.
Why not go all the way and use a 150-year smoothing? You get a very frightening straight line!
.
Trouble is, the trend is still only 0.06 C per decade!
.
http://www.woodfortrees.org/plot/hadcrut3gl/mean:1800
.
Quote from Owen:
[“The slope is increasing and is currently nowhere near 0.057 C/decade.”]
.
Priceless…
Arfur,
Since you are still contesting the that CO2 and (by extension) other infrared absorbing greenhouse gases are forcing agents for climate change, we really don’t have much more to say. The evidence has been laid out in great detail, for example, at the Science of Doom web site. The radiative transfer equations are admittedly formidable, but are based on the known physics of absorption and emission and are used in explaining properties of planetary atmospheres. I have not fully made my way though the mathematics of the theoretical equations as they are difficult for me, even as a chemist. But I understand the concepts of long wave thermal radiation emitted from the earth’s surface, absorption of such radiation by CO2, collisional energy transfer to N2 and O2, and emission of radiation by CO2 at a new temperature regime. I accept that the question of sensitivity has not been thoroughly worked out and the level of predicted warming must have a range associated with it.
You need to move off of step one and quit belligerently fighting something that I do not believe you have really tried to understand.
Owen (Comment #77425)
Well, for someone who doesn’t have a lot to say, you’re saying quite a lot. What you’re not doing, is either answering the questions I asked previously or providing any evidence that CO2 is capable of causing significant anthropogenic warming! I have no problem with accepting each individual CO2 molecule is capable of absorbing and emitting radiation, and please rest assured that I have read some of the absolute nonsense on the science(sic)ofdoom site. It is not evidence. It is opinion. Evidence – and, just to be sure, I mean evidence that CO2 can cause or has caused significant global warming – is something you are demonstrably unable to produce.
.
[“But I understand the concepts of long wave thermal radiation emitted from the earth’s surface, absorption of such radiation by CO2, collisional energy transfer to N2 and O2, and emission of radiation by CO2 at a new temperature regime.”]
.
As a non-chemist, I too understand the concept of long-term radiation and ‘collisional energy transfer’ to N2 and O2… However, would you, as a chemist, like to clarify how many N2 and O2 molecules, on average, surround each CO2 molecule in the atmosphere (well mixed)? How about, say, over 2000? Ok, how many of those molecules will gain energy from collisions with the CO2 molecule?
.
And, if the forcing of CO2 is so – according to you – powerful, where the F**K is all the warming? 0.8 deg C in 161 years? By your own arguments, you have managed to condense the total cAGW hype/alarmism to a 28 year period between 1970 and 1998. And then, you ASSUME that all the warming is due to CO2 because you can’t face the fact that natural warming has taken place before 1970 and you just want to brush that under the carpet.
.
If you’re a chemist, stop confusing radiation with heat. If you want to go down the heat transfer by conduction, or low energy collision route, then please explain how 0.028% of the atmosphere that was GHGs (that’s all of them, not just CO2) in 1850 can produce a significant portion of the 33 deg C Greenhouse Effect, yet a massive 40% increase to 0.04% of the atmosphere can only produce a PORTION of 0.8 deg C over one and a half centuries later – and then miraculously stop having an effect!
.
Jeesh… and you say I’m the one who hasn’t tried to understand?
.
I’m off to bed.
Zeke and Brandon, what I trying to say was that the IPCC analysis suggests that we need not invoke anthropogenic forcings to explain the observed warming until the 1970s.
This was in response to Arfur’s comment that:
And what I said pretty much jibes with Zeke’s suggested choice of language.
However, I was also referring to the fact the models used by the IPCC in making their analysis are assuming net anthropogenic forcings were either not large enough to cause a measurable amount of warming, or were actually negative (they produced a net cooling due to aerosol effects).
As an example, here are is a plot of the GISS model net anthropogenic forcings over time.
If you want to quibble there is a small, probably measurable, anthropogenic cooling from 1890 to 1970 in this particular assumed model forcings, but since I thought we were discussing anthropogenically forced warming, hopefully I’m to be forgiven for the oversight in mentioning that the explanation for the observed cooling from 1950-1970 in the instrumentation record in particular is explained in the models in terms of indirect effects of rapid industrialization (pollution) swamping the smaller effects expected from AGW CO2 emissions.
So while I agree with Zeke (his is probably a “safer” claim to make), the IPCC model assumptions and results actually suggest something a bit stronger.
In either case, it is safe to say that the IPCC analysis suggests that AGW only initiated circa 1970.
(I apologize for the delay in the response…we had a large-scale sensor deployment at the end of last week, and I’ve been preoccupied with seeing that go off well.)
(Second update: The embedded link in my PDF file isn’t functioning correctly so here is the link to GISS forcings.)
Carrick, that doesn’t resolve anything. You say anthropogenic forcings aren’t needed to explain the observed warming prior to 1970. So what? That they aren’t necessary prior to 1970 in no way means they didn’t exist prior to 1970. It just means the record doesn’t prove their existence until 1970. Prior to 1970, anthropogenic warming could (and did) exist. It’s existence just can’t be proven through an examination of the temperature record.
You claimed the IPCC report says 1970 is when anthropogenic warming started. It did no such thing. Unless you retract that claim, there is no way you are going to get an agreement. Until then, your position is nothing more than saying a lack of proof equals disproof.
Carrick:
http://dl.dropbox.com/u/4520911/giss_net_forcings.pdf
What an absolute load of crock. What are the units of the y-axis? Could they be decimals of degrees Celsius? Really? If so, the logical conclusion is that the global temperature rose about 0.8 deg C between 1910 and 1945 WITHOUT anthropogenic forcings, and yet the temperature after 1998 has not risen when there has been a constant increase in anthro-forcings? How does this make any sense from a warmist point of view?
.
You have effectively condensed the entire cAGW hype to a period of 28 years between 1970 and 1998 with an assumption about forcings when you have already admitted/conceded that a similar warming took place 30 years before 1970 which could have had no anthropogenic factor.
.
I suggest you stop digging.
Carrick,
.
Furthermore, you keep talking about the ‘IPCC analysis’ whilst ignoring the references I have provided where the IPCC have clearly stated that AGW started well before 1970.
.
I repeat:
According to the IPCC’s explicit statements, they consider that AGW started in the mid-eighteenth century. However, they consider that reliable observed data started in 1850.
.
See Arfur Bryant (Comment #77164) for references.
Brandon:
Actually I never said positive forcings didn’t exist before 1970 (in fact I said the opposite), what I said AGW warming didn’t exist before 1970 according to the analyses included in IPCC AR4.
In fact, as you can see from the figure I included above, GISS assumes that net anthropogenic forcings are negative prior to 1970. And since AR4 figure is a composite over many models, and reflects this same picture, it is clear this is a prevalent feature of the models.
Sorry that you have such trouble with basic logic, it must be tough.
Arfur, you can find people who will say whatever you’d like them to say. The question is whether their rhetoric is backed up by any analysis…and it’s not.
Address the fact that the figure I linked to came straight from the AR4 analysis and what its implications are, or just drop it. It’s obvious you don’t really understand how to read and process technical literature.
Oh as to this:
They are forcings and in MKS units (of course). That makes it Watts per meter squared (W/m2). You could have learned this by following the link that I posted.
The models generally assume that GHG forcings are balanced by the cooling effects from aerosols released through pollution (mainly sulfates) until roughly 1970. That’s generally accepted and well understood by people within the modeling community. I don’t necessarily endorse this position, but I believe this is an accurate description of their viewpoint.
The cooling from 1998-now from a warmingist viewpoint would be due to short-period climate forcings from atmospheric-ocean oscillations. According to that viewpoint, which I do endorse here, you need at least 15-years, and I think at least 30-years, to separate out the secular forcing from CO2 emissions from short-period climate fluctuations.
Arfur:
I believe this is an accurate statement of the findings of the IPCC.
What do you think I’m digging for? Do you think I endorse CAGW?
Carrick, you just responded to one part of my post which did not use the same terms as you while ignoring the part of my post which did. Immediately following the part you quoted as:
I first addressed forcings then I addressed warming. You selectively quoted me in order to cut out the part of my post which discussed warming, and then you insulted me based upon your misrepresentation. I’m not going to continue this exchange after such nonsense.
But before I quit, I have to point out you are flagrantly misrepresenting the GISS forcings you refer to. Anyone can look at the bottom figure on that page and see net forcings are positive for most of the time after 1920. Even if you remove the solar forcings (leaving only anthropogenic forcings), the net forcings are still predominantly positive prior to 1970. It is easy to confirm this by examining the records in this file. As one example, 1940 has .8759W/m2 in positive anthropogenic forcings. Negative forcings in 1940 are only .7213W/m2. Clearly, anthropogenic forcings are considered positive in 1940, despite you directly stating otherwise.
You’ve misrepresented the evidence you reference, and you’ve misrepresented me. Seeing as the entire base for your position and for insulting me is these misrepresentations, I’m quite content to leave things at that.
Actually, I’m not ready to stop there. Starting with 1900, there are 25 years with net negative anthropogenic forcings. Of those, four are after 1970. This means 21 out of seventy years prior to 1970 have (net) negative anthropogenic forcings, and 49 have (net) positive anthropogenic forcings. If you instead start at 1910, there are only 16 years (out or sixty) with (net) negative anthropogenic forcings.
Carrick has simply made a claim up while pointing at the very source which shows him wrong. I have no words to describe that behavior, or at least not ones I’d be comfortable saying out loud. On a final note, I seem to have messed up the link in my last comment, so here is one that will work:
http://data.giss.nasa.gov/modelforce/RadF.txt
Re: Brandon Shollenberger (Jun 20 08:14),
Look at the magnitudes. Before 1960, the sum of AG forcings (everything except Stratospheric Aerosols and Solar ) barely exceeds 0.2 W/m². After 1960, the forcings increase rapidly to a value of ~1.7 W/m². While you may be technically correct, Carrick has the better argument.
See graph here.
Thanks for the plot DeWitt, I was left guessing which forcings they considered anthropogenic versus natural. (Some of them seem mixed to me, for example, atmospheric aerosols have a component that relates to climate…)
It is obvious Brandon is pretty confused here. He says:
Of course I never claimed that net forcings were zero.
Solar forcing is not the only natural forcing.
Thirdly, if I am flagrantly misrepresenting GISS and the IPCC report, then why does this figure, which is included in both the technical report and the summary for policy makers, suggest that one need only consider natural forcings to explain climate change until circa 1970?
I also concede that DeWitt’s figure is a more accurate representation of the GISS modeler’s view (it is generated by them), but 0.2 W/m^2 is in the noise level compared to the uncertainties involved in the forcings themselves.
(SteveF has gone as far as to suggest that the aerosol histories are invented.)
Finally, it was not my original contention that the AGW forcings used in the climate models were negative until 1970 (though they may be) but rather that climate change according to the models was predominantly due to natural forcings until the 1970s.
Here is the original statement that I made that led you/Arfur to go into attack dawg mode:
I believe this statement is completely accurate.
Your original comment on the other hand was this:
Here is what I think the IPCC report suggests: Global warming started circa 1850, but most of it was natural until circa 1970. (I also accept that the globe has been warming since around 1850, but I also don’t think the data are reliable until circa 1950 for their use in this debate.)
If you had said “anthropogenic global warming” instead of warming, I would have pretty much agreed with the statement.
And unlike your claim that you started by discussing forcings, you didn’t first address forcings, you discussed warming and you mistakenly conflated global warming with anthropogenic global warming.
You said later:
This is nearly a paraphrase of what I said, but it has one crucial error in it: it isn’t until the seventies the anthropogenic nature of it can be confirmed.
Had you said “it isn’t until the seventies [that an] anthropogenic component of it can be confirmed,” we would have been in agreement.
If it was your intention from the start to say, the IPCC AR4 suggests that:
1) there has been global warming since 1850,
2) there are natural and anthropogenic components to this warming,
3) it wasn’t until the 1970s that the influence of an anthropogenic component can be confirmed,
then we were always in agreement. Based on your other language, I suspect this is not the case.
Finally, the IPCC figure that I have linked ad nausea illustrates and confirms points 1, 2 and 3 above.
DeWitt Payne, I’m obviously aware of the magnitudes involved seeing as I had to calculate them in order to get the information I put in my comments. However, the magnitudes have no relevance when the comment in concern was:
I’m at a loss as to how you conclude Carrick has a better argument. He claimed net anthropogenic forcings were (assumed to be) negative prior to 1970 and pointed to a source which contradicted his claim. I said net anthropogenic forcings were positive at points prior to 1970 and demonstrated his source shows this. He was obviously wrong, and I was obviously right. Despite this, you claim he has a better argument…
Ignoring the strangeness of your response, I have to ask what the source for the values in that graph as. It shows significantly higher forcings than the source Carrick provided (and it has far less variability). I’m not sure how that graph could be relevant to the discussion at hand, but if you’re going to bring it up, you need to give some information about it.
By the way, it’s very hard to believe Carrick “was left guessing which forcings they considered anthropogenic versus natural.” I had no trouble reading the paragraph below the first figures to see the distinction:
Brandon:
Actually it confirms my claim, as I’ve reiterated above. As well it provides recognition to everybody following this thread that your technical skills are somewhat lacking in this arena. 1) You were unable to interpret the original graph that I linked, 2) you have trouble with orders of magnitude discussions like this one.
Obviously I didn’t see that paragraph (I’ve had that particular data set on my computer for years and was just going by the header in the first line of the file).
It’s interesting in spite of your apparently more thorough reading, that you came away with the conclusion that the only natural forcing was solar.
To illustrate this isn’t quite as straightforward to interpret as Brandon suggests, here is a list of the columns in the file:
Year W-M_GHGs O3 StratH2O Solar LandUse SnowAlb StratAer BC ReflAer AIE
GHGs includes the influence of “carbon dioxide (CO2), methane (CH4), nitrous oxide (N20) and chlorofluorocarbons (CFCs).”
I don’t know how you break down O3. It is of course predominantly produced in the stratosphere by ultraviolet radiation, soI would take this one to be a natural forcing, though modulated by anthropogenic forcings. Technically I guess you would fit natural O3 level prior to the introduction of CFCS, then assume the loss in O3 is related entirely to CFCs (I believe however this point is in contention, as atmospheric circulation changes related to global warming can affect how much O3 is retained in the stratosphere).
StratH2O-Stratospheric H2O…it is predicted to increase with increasing temperature regardless of why the temperature increases (so it is really a feedback that is not directly modeled rather than a true forcing). I’m treating this as a natural forcing.
Solar…obviously natural.
Land usage…mostly anthropogenic (but there is a global feedback on changes in surface albedo e.g. of forests that is a response to climate change, natural or anthropogenic, and this can be very large).
Snow albedo…technically it is a feedback not a forcing, but since it’s not modeled directly, it gets treated as a forcing. SA is affected by both temperature change and by heavy, dark particulates that accumulate on its surface. In turn, there are both natural and anthropogenic sources for these (volcanos produce ash and forest fires produce soot…in really bad years, there is a substantial contribution to atmospheric opacity from natural forest and range fires). Again this is heavily intermixed….because we tamper with the natural cycle of forest fires, thereby leading to heavier conflagrations, how much of that is natural versus man made? At zeroth order though—it’s a natural forcing.
Stratospheric aerosols—dominantly these are volcanic eruptions, but large convective storms also inject aerosols into the stratosphere. If large convective storms are “on the rise” then this is a feedback in addition to a forcing. At zeroth order another natural forcing.
BC–black carbon. This is the one I missed. It is dominantly man-made (though forest and range fires produce it too as I mentioned, and to complicate things we affect the number of forest and range fires through water usage changes and forest fighting policies). I’ll rate this one as anthropogenic, but that’s a guess. I’ll update my graph to include this as a forcing. I’ll admit I missed this one (sorry yesterday was father’s day and I had other things to do besides play with numbers).
ReflAer–reflective aerosol, as I understand it, this is dominantly from anthropogenic sulfate emissions. I included this in my original graph.
AIE-aerosol indirect effect, I believe this is generally assumed to be anthropogenic and I have treated it as such.
Now my question is…for all of Brandon’s snark over this…is this really that obvious and easy to interpret based on the paragraph that he included?
Here is an updated figure showing natural, anthropogenic and total based upon the choices I’ve made.
GISS radiative forcings
Here is the data I used for the graph.
(Note for Brandon, the reason that there is more variability in the natural forcings is due to volcanic eruptions).
Re: Brandon Shollenberger (Jun 20 11:10),
No, that’s what you believe he claimed. That isn’t a direct quote and it isn’t how I interpreted what he wrote.
I second Carrick on the AG forcings prior to 1960 being at the noise level. The IPCC estimates of uncertainty for anthropogenic forcings other than WM ghg’s are very high. Even ghg’s are +/- 10%. I consider it very likely that if error bars were attached to those forcing numbers, they would include zero until after 1960. I’d do it myself, but it’s not worth the effort.
DeWitt:
It’s always easier to argue against a position that you’ve constructed using paraphrase rather than the real one you to disagree with. This is why I generally try to include the actual language from the person with whose argument I am disagreeing. In the computer age…how hard is it to cut and paste?
DeWitt is right though…whether anthropogenic forcings were positive or negative are immaterial to the argument I was trying to advance. Prior to 1960, you can safely ignore the anthropogenic forcings: as is clear in this figure. In other words, exactly what I stated in my initial comment.
(As a side comment, from 1950 to 1970, it is generally assumed by the modelers that anthropogenic sulfate emissions are responsible for the cooling/plateau seen over that period. It is likely that you will find models in which anthropogenic forcings decreased over that period…even if they didn’t go negative…to explain that cooling/plateau in global mean temperature.)
I can admit error when I make it, in fact I have no problem at all doing that. Does Brandon ever make an error and can he admit it when/if he does?
DeWitt Payne, your response is dumbfounding. You say:
I actually provided a direct quote. It said:
I described it as:
Exactly how is my description inaccurate? How do you interpret that direct quote in any way other than how I interpreted it?
Also, while we’re on the subject of dumbounding responses, could you tell me what the point of you providing that graph was? Neither it, nor any of your comments about uncertainties change the fact Carrick (incorrectly) said:
Carrick and DeWitt Payne,
Right, lets stop all this deflection and talk about integrity.
.
Here is what Carrick said in post #77119:
[“1850 isn’t the “start†year according to the IPCC. It’s the 1970s.”]. That was referring to the start of anthropogenic warming.
.
In comment #77207, Carrick said:
[“What I said was according to the IPCC, we can explain all of the warming through natural forcings until circa 1975.”].
.
I then provided Carrick with ABSOLUTELY CLEAR statements by the IPCC in which they clearly stated that they (the IPCC) consider that anthropogenic warming started in the mid-eighteenth century and that reliable data started in 1850.
.
I repeat, for the third time:
“A.2: “The year 1765 was chosen as the pre-industrial baseline for greenhouse gas concentrations (and, hence, for man made forcing and global mean temperature)…â€
.
And “Global atmospheric concentrations of carbon dioxide, methane and nitrous oxide have increased markedly as a result of human activities since 1750…†and
“The understanding of anthropogenic warming and cooling influences on climate has improved since the TAR, leading to very high confidence[7] that the global average net effect of human activities since 1750 has been one of warming, with a radiative forcing of +1.6 [+0.6 to +2.4] W m–2 (see Figure SPM.2). {2.3, 6.5, 2.9}…â€â€¨
.
Reference in my post #77164.
.
Back to integrity.
Integrity is having the guts to admit you are wrong when it is honest to do so. So, as an example:
I was wrong when I implied that the units on the y-axis of Carricks forcing graph were degrees Celsius. They were, in fact, W/M2 as Carrick pointed out. I had missed those units written in the second link from Carrick. See what I did there? I was in error and I have admitted such. That is because it is honest and morally correct for me to do so.
.
Carrick, you have on at least three occasions now ignored the references that I have provided which CLEARLY STATE the FACT that the IPCC consider anthropogenic factors to in global warming to have started in the mid-eighteenth century. That you insist the IPCC analysis is otherwise – starting in the 1970s – is appallingly poor form.
.
You are using model projections to make your point through interpretations of model graphs, and yet you CONTINUOUSLY ignore the clear and unambiguous statements made by the IPCC in black and white which totally refute your earlier statement.
.
To say now that you believe the IPCC intended to condense AGW into a 28-year period from 1970 to 1998 is sheer cheek.
.
I have pointed out your inability to argue the points that I have made – and I note Brandon is also having the same problem with your selective answering – on several occasions.
.
You need to man up and admit when you are wrong. No amount of deflection and sanctimonious belittling of my ability to read will change the fact that you refuse to address the points.
.
I expect a certain amount of groupthink bunker mentality from the lukewarmists on this forum but I didn’t expect the level of bald-faced deviousness that I am witnessing.
.
The warming from 1910-1945 was of equal magnitude to the warming of 1970-1998, yet you allocate totally different causes to those warmings. You do this without any proof and consistent use of model projections over real data. There is no logic to your arguments, you merely select the particular model you wish without any real evidence to back those models up.
.
Now to try to dismiss the lack of warming since 1998 – when, as you posted, your CO2 forcings have been increasing steadily – as being too short to be of significance (“The cooling from 1998-now from a warmingist viewpoint would be due to short-period climate forcings from atmospheric-ocean oscillations. According to that viewpoint, which I do endorse here, you need at least 15-years, and I think at least 30-years, to separate out the secular forcing from CO2 emissions from short-period climate fluctuations.”) is a completely illogical u-turn when your previous argument was that the warming of the late-20th century was due to the increasing forcing of CO2. You can’t have it both ways.
.
What you have here is an AGW sandwich between overpowering natural effects. Your logical fallacy is that you argue CO2 can overpower natural factors when the temperature goes up, yet apparently any lack of ‘temperature going up’ is due to to natural factors. Natural factors work both ways, Carrick. Make your mind up.
Carrick (Comment #77661),
IMO, the GISS forcing history in the graph you linked to is a kludge to make the GISS model work. The trends in methane, N2O, and CO2 (ice core data) show a clear and substantial increase in GHG forcing through the entire period, starting at least in the mid 1800’s. The volcanic and man-made aerosol effects and solar effects, which so perfectly offset rising GHG levels in the GISS graph, are speculative (wild-ass guesses!) at best. It would be nice if we had real historical aerosol data (even recently!). We don’t. We do have lots of speculation which is substituted for data.
Brandon:
This was a result based on the analysis I provided (which left out the black carbon, which I believe they would take as an anthropogenic forcing). As DeWitt and I have separately pointed out, this was never central to my argument, which relates to the magnitude not the sign of the forcing.
SteveF:
This may well be, but my only point is you need to at least get right what the models are actually assuming before you criticize them.
Arfur, this is wasted energy but I’ll try one last time. What the IPCC actually is saying is that they take the zero year for anthropogenic forcings to be e.g., 1765 (in fact as DeWitt points out, it has been going on for about 8000 years).
Anthropogenic warming (by definition) starts at the point where a) you measure a net warming (sensibly or via instrumentation) and b) you can attribute that warming to anthropogenic causes. According to the IPCC AR4 that starts around 1970. Why 1970 and not say 1950? Because temperatures were remarkably stable from 1950 to 1970 (it only started really warming again around 1975…. you can’t have AGW without the “W”.)
Getting pretty dense aren’t we? I’m not claiming the models are right, merely trying to correct your misunderstands about what those models (and climate science by extension) claims to be the truth.
You can…
(Think about the effects of a fitting a sine wave + straight line to a straight line… )
Carrick
.
Sure, and I don’t criticize the modelers for making assumptions. I do criticize for using those assumptions to fit known data and then proclaiming (based on the fidelity to historical data) that the models are capable of calculating the true climate sensitivity and of projecting future warming.
.
Models are only validated when they make predictions that are accurate about the future (the real future, the one that nobody has any data for). I criticize the entire effort because the individual models substantially disagree with each other and with the (measured) behavior of the Earth, and yet no model is simply dumped in the trash bin; the entire ensemble is “averaged” as if averaging unvalidated (and mostly wrong) models somehow is “better” than any single model. (This is something so bizarre that I blink in disbelief each time I think about it.) But most of all I criticize because no modeler says “Gee, these models are all pretty crappy in a lot of ways; they don’t do lots of things that are observed on Earth, they don’t seem to match credible data, like the projection of amplified tropospheric warming, and they are all kludged up with arbitrary aerosol forcings to match the historical temperature record. Maybe we should not claim to be able to calculate climate sensitivity using models that all have serious deficiencies.” Climate modeling is the only scientific field I have ever heard of where there seems to be no real progress, yet nobody is held to account; it most certainly is in need of strong critique… or perhaps more effectively, future funding based on accuracy of projections.
.
Lucia is doing it right: show how the model projections are consistently higher than the observed warming; at some point the the discrepancy can no longer be ignored. AR5 will generate a whole new set of “projections” based on “model improvements” since 2000…. and (of course) the knowledge that the measured trend since 2000 was lower than projected in AR4. I’ll make a rash prediction: the new AR5 “projections” for 2001 to present will (surprise!) much more closely match the measured temperature history. By the time many of those new projections (following AR5) are shown to be again incorrect, it will be time for AR6, and the cycle will begin again. It is an intellectually corrupt enterprise.
SteveF:
I believe this is an unattainable goal. In the absence of knowledge of future forcings (which are inherently unpredictable), it is impossible to predict the warming that will be expected.
I think a more fruitful approach would be to demand greater fidelity from the models: They needn’t be including ice albedo as a forcing, that should be a result, anthropogenic forcings of clouds… similarly. Same goes with the frequency content of short-period climate fluctuations.
Arfur Bryant, you say:
I set my RSS reader to filter out Carrick’s comments after I said I was done responding to him, but from the little I’ve caught skimming past his recent remarks, it has gotten absurd. He has contradicted his own comments, flagrantly misrepresented my comments, and yet came up with a correct position. It is depressing to think this could have been resolved with no trouble had he simply behaved differently.
However, his position (assuming it doesn’t change again) is correct.
Brandon:
And oddly…it was the same position I started with.
It’s amusing that Brandon thinks that I “flagrantly” misrepresented his views, given that, unlike him, I provide pretty detailed context in the form of direct quotes from him. Apparently, neither logic, intellectual honesty nor civility are among Brandon’s strong suits.
Good riddance.
Carrick (Comment #77686):
[“Arfur, this is wasted energy but I’ll try one last time. What the IPCC actually is saying is that they take the zero year for anthropogenic forcings to be e.g., 1765 (in fact as DeWitt points out, it has been going on for about 8000 years).
Anthropogenic warming (by definition) starts at the point where a) you measure a net warming (sensibly or via instrumentation) and b) you can attribute that warming to anthropogenic causes. According to the IPCC AR4 that starts around 1970. Why 1970 and not say 1950? Because temperatures were remarkably stable from 1950 to 1970 (it only started really warming again around 1975…. you can’t have AGW without the “Wâ€.)”]
.
Carrick, watching you wriggle like a suffocating fish because you have dug yourself a massive logical hole in the form of twisted words is NOT a waste of time! Its actually quite fun…
.
Here is precisely why what you are saying now and what you said at the start of our little contretemps are different, and why you are displaying the most appalling lack of maturity:
1. For about the fourth time, here is a direct quote from the IPCC: [““The understanding of anthropogenic warming and cooling influences on climate has improved since the TAR, leading to very high confidence[7] that the global average net effect of human activities since 1750 has been one of warming, with a radiative forcing of +1.6 [+0.6 to +2.4] W m–2 (see Figure SPM.2). {2.3, 6.5, 2.9}…â€
.
Here, again, is the reference:
http://www.ipcc.ch/publications_and_data/ar4/wg1/en/spmsspm-human-and.html
.
I refer you most politely to the words …the global average net effect of human activities SINCE 1750 has been one of warming…. By the way, ‘net effect of human activities is the ‘A’ in AGW. Please note the use of the word ‘warming’, as opposed to ‘forcings’. This refutes your point that the IPCC consider (your opinion only) anthropogenic ‘warming’ started in 1970. Maybe you should write to them and point out their error…
.
You have persisted in ignoring this clear and unambiguous statement by the IPCC, in favour of your penchant with using your interpretation of models (which, as SteveF points out, don’t actually project into the future…) whereas I am providing you with actual written statements by the IPCC. How much clearer do I have to be? And you infer I am dense?
.
2. The ‘anthropogenic warming’ of which you speak – and claim that it is proof of your point – is INDISTINGUISHABLE from the 0.8 deg C warming that took place a few decades prior to 1970. This, therefore, is a warming – your own definition – which has taken place after the IPCC start of 1750 (and after the IPCC start of accurate data in 1850. And yet you conclude that it is the later warming that is proof of your point. This is why I say you can’t have it both ways; you have no way of judging (from real-world data) whether the 1910 warming was not due to anthropogenic and whether the 1970 warming was not due to natural effects. Your entire argument is based on interpretation of models which, as Lucia is proving, are of highly dubious efficacy.
.
This discussion is about WHAT the IPCC says about Anthropogenic Warming. I have provided you with clear proof that your assertion is wrong, yet you choose to ignore this clear proof in favour of your own interpretation.
.
The fact that you will not even refer to – or even acknowledge – my referenced proof in any of your posts is indicative of your lack of ability to carry on a reasoned and honest debate. As I pointed out in my last post, sometimes we make mistakes and, if we are man enough, we admit them. Its as simple as that.
Arfur, this is pointless. Have a good life.
Brandon Shollenberger (Comment #77694)
[“However, his position (assuming it doesn’t change again) is correct.”]
.
Brandon, I totally understand why you would not wish to continue a discussion with Carrick in the face of such bold impudence. However, here is why Carrick’s position is NOT correct:
.
He stated:
[“Arfur:
chose the 1850 ‘start year of the anthropogenic effect’ as chosen by the IPCC.
1850 isn’t the “start†year according to the IPCC. It’s the 1970s. Note that prior to the 1970s you didn’t need to invoke AGW forcings to explain the temperature trend…according to the IPCC.
.
I have pointed out on several occasions now that the IPCC explicitly considers that ‘anthropogenic warming’ started in the mid-eighteenth century:
[““The understanding of anthropogenic warming and cooling influences on climate has improved since the TAR, leading to very high confidence[7] that the global average net effect of human activities since 1750 has been one of warming…“]
.
As you can see, his statement is flatly contradicted by the IPCC. In black and white.
.
This is why his position is incorrect. The discussion (if you can call it that!) was always about what date the IPCC considered AGW started. Any comments about 8000 years, or interpretations of model graphs are a sideshow. He is using his opinion to try to shield his lack of real evidence. The man won’t even acknowledge my IPCC references! He is wriggling. A little humility on his behalf would go a long way to earning some respect.
Sock puppet much, Brandon/Arfur?
😉
Greg,
If you’re referring to this:
.
http://en.wikipedia.org/wiki/Sockpuppet_(Internet)
.
…then, no. Absolutely not as far as I am concerned. Just as I will admit when I’m wrong, I will also admit to being nonplussed (in the literal sense!) when I genuinely don’t understand something. Your post leaves me nonplussed…!
.
Regards, Arfur
Arfur Bryant, sorry about the slow response, but the problem is Carrick’s position isn’t as you (or even he) claim. He actually agrees with the point you raise. He just refuses to admit he has misspoken on multiple occasions, so he can’t retract his comments which contradict you. Both he and you hold the same position, but it’s understandably hard to tell given the schizophrenic nature of his comments.
Greg, I have no involvement with Arfur Bryant, and I don’t much care to be accused of intellectual dishonesty.
Brandon,
Just in case you check this (the thread is old, I know…), it has been a pleasure conversing with you.
You, SIr, are a gentleman.
Regards,
Oddly enough, just this morning I found a monocle while looking for a cane (I had twisted my knee). I kind of feel like a true (British) gentleman.
Granted, I wear glasses so I couldn’t actually use the monocle, but still…