2011-01-16 00:08:10Response to Soares et al
Mark Richardson



What do you think?



Took me ages, I originally wrote about 2k words and had to cut it. There was so much wrong with the paper's conclusions that it's unbelievable that this was actually reviewed.

2011-01-16 02:06:40Nice title
Daniel Bailey
Daniel Bailey

You need quotes around your Soares quote at the top.

It gets a little confusing switching between your Figures and Soares' Figures (ex: the first Figure used is Soares' Figure 8, which you call Figure 1; later you refer to it as Figure 8).  Perhaps you could just use your verbiage to keep it simple?

Nice summary close.

2011-01-16 03:11:07
Mark Richardson
Changed on both points.
2011-01-16 06:21:51comments
Dana Nuccitelli

"... theory and measurements show increased CO2  heating?" - I would re-word the end of this sentence for clarity.  Something like "...theory and measurements show warming resulting from increasing CO2?"

"...a climate model estimate of how much global warming was expect from greenhouse gases..."  I believe should read "was expected"

The following sentence: " tens of times smaller than the noise which effectively hides the incline."  Hides it over what timeframe?

I'd add an attribution to whoever created that Meehl graph (globalwarmingart perhaps?).

"So if you plot a slope you expect it to be positive – going from 0.01 on the left up to 0.02 on the right but practically impossible to find amongst so much noise..."

Again, over what timeframe?  The trend is pretty obvious in the long-term, but I presume you're arguing that if you look at short timeframes, the trend is hidden by noise.  This isn't quite clear.

When you talk about aerosols causing mid-century cooling, it would be good to link to the rebuttal on the subject.

I'm glad you had time to do a write-up on this paper.  Mainly you just need to hammer home the timeframe noise vs. signal issue clearer.  Might even be worth referencing your 1998 DIY statistics post.

2011-01-16 08:58:39Soares
Dana Nuccitelli

By the way, Soares is apparently a retired former researcher and professor.


2011-01-16 09:03:39
Rob Painting
The title is a hoot. I like it, both informative & humorous. 
2011-01-16 21:10:46Thanks, great you wrote this
John Cook


Am getting sick of people quoting this paper, without even reading it but just glad to have a skeptic paper in "peer review". So future quotes can now get referred to your rebuttal.

For your theory link, could you link to the abstract rather than the full paper - non-subscribers can't even see what the paper is.

2011-01-16 23:55:46
Mark Richardson

dana; I've changed teh wording, linked to the rebuttal.



Not qualified on the noise because I didn't have the time to take the data and analyse it properly myself when it's already largely irrelevant. I looked at GISTemp using temperature change and found the standard deviation to be 0.1 C. So, assuming all noise was random and that you're looking to determine with 95% confidence that a positive value exists, and the positive value you're looking for is 0.01 C average, then you need over 350 measurements. GISTemp has 130 years, so...

But the noise isn't random, and the signal size changes too, making it more complicated. If my rule of thumb test is right then I've shown you won't detect the signal, and all these effects do is obscure it even more. So I'm pretty confident...



Although someone could do with checking my standard error: I just assumed central limit theorem so you can get 95% confidence from 2 * STDEV / SQRT(N).

2011-01-17 11:18:31Thanks Mark and question for Baerbel
John Cook


I've added your blog post into the rebuttal database - our 142nd rebuttal:


BTW, that headline and concluding paragraph are both brilliant writing :-)

Quick question - do you still have your 2000 word original draft? Would it be suitable as the advanced rebuttal? If so, can you add it and then we can add an update to the blog post, linking to it. Note - I added a green box announcement to the bottom of the blog post, linking to the rebuttal just to let people know it's now part of the "canon".

Also, note to Baerbel - I'm told the Soares paper is getting lots of mention in German denier sites. Any chance of your translation team having a look at Mark's rebuttal?

2011-01-19 19:49:46German version now online


We've translated the Soares' blog-post and rebuttal into German:

Blog: http://www.skepticalscience.com/translationblog.php?n=522&l=6

Argument: http://www.skepticalscience.com/arg_Soares-Zusammenhang-CO2-Temperatur.htm

Unfortunately, the original heading for the blog-post didn't lend itself to a similarly witty German translation....


2011-11-29 07:03:59
Klaus Flemløse


Discussion of the methods by Soares

I am trying to argue for that the findings of Soares are correct and expected. Any two time series with at linear trend will show the same results as indicated by Soares.  This means  a correlation between the two time series and no correlation between the delta time series.

If the CO2 concentrations equals C(t)  and the Global temperature equals T(t) and you want to investigate if there is a linear relation between these function, it as good idea to plot  this to function (C(t),T(t)).                                                                             

Let us assume that such a relation exists. This means that C(t)=f(t) and T(t)=beta*f(t) + alfa.

Then  the graph of (C(t), T(t)) will be a subset of  the line Y=beta*X+alfa.

If you differentiate C(t) and T(t) we  will that have C’(t)=f(’t) and  T(‘t)=a*f’(t).

 The graph of (C’(t),T’(t))  will be a subset of the line Y=beta*X.

What do we learn from this?

If f(t)= alfa+beta*t,  we will have that f’(t)=beta and the graph of (C’(t),T’(t)) will be the point (beta, b*beta).

If the C(t) and T(t) are two time series with a positive linear trend in year and  with random variation, then the  points of (C(t),T(t)) will be a scattered around a straight line.  In most cases a correlation will be found.

Remembering that differentiation in respect to time correspond to taking year to year differences. Then points of (ΔC(t), ΔT(t)) will therefore be scattered around  one point and no correlation is expected.

If you have linear trend in CO2 and in the global temperature you must expect

1)  the plot of the points  (CO2,Temperature) will be scattered around a straight line

2)  the plot of the points  (ΔCO2,ΔTemperature) will be scattered around one points  with no correlation

If we assume as an approximation, that  there is a linear relation between  CO2 and the Global temperature  then, the findings of Soares are 100% correct and in line with expectations.

The following graph shows a simulation of the global temperature and global CO2 concentration for the years 1880-2010, where  the CO2 curve and the temperature curves  both have the same form: Y=alfa+beta*exp(gamma*year).

The parameter (alfa, beta, gamma) are estimated by a graphical methods. Due to the random variation, the Soares method produces the same results. One can't se the exponential form of the curves.

 “ width=

The following graphs show the plot of (CO2,Temperature) and (ΔCO2,ΔTemperature).

 “ width=


 “ width=

To my understanding Mr  Soares does not know what he is doing, and if he knows what he is doing, then it is fraught.

This point of view could be included in blog post.














2011-11-29 07:17:19
Klaus Flemløse