But I saw ealier in this blog – as far as I understand – that you would consider +0,3K for UAH ok. And as I have 5 other trend types around +0,2K the + 0,3K is not that “far out”.

So lets move on :-) (administrator would love that!)

Then you write:

“Whether your peaks are 0.3 or 0.4K above current temperatures, you have to demonstrate that these peaks are likely, even when they contradict all other proxy reconstructions.
Unless you can do that, I’m afraid you are wasting your time.”

Tom, I have really done EVERYTHING you could possibly demand to illustrate how much and how come tree rings are not usefull.

And you know very weil (?) that leaves only Loehle´s research that “contradicts” my results! (Moberg uses partly tree rings)

So your argument is: Frank, your results does not match Loehle, therefore its wrong. But if Loehles results are so correct that anything not matching it is wrong, then I hope that Loehles and my COMMON conclusion that tree rings are NOT usefull will be accepted by you.
Loehle is right, you know :-)

Craig Loehle did a SUPER good anf important reasearch!! He was definetly one of the reasons i did this larger investigation. Loehle used 18 datasets including some sets also used by Moberg.

I have for weeks and weeks searched after a bigger basis of data to if possible get more precise result. A MWP around 0,3-0,4 K above todays average is supported by 55 independent datasets. Loehle stopped at 18 datasets.

Craig, if you are out there: “I have 55 datasets, that does in average show a 0,3-4 +K MWP. If you have the time, i would like to hand over all data to you and see what came out of it. We would win Tom P, how about it??”

In addition i have assembled a BIIIIIG pile of similar research papers, where direct temperature curves are not avaulable, by datasets like O18 C14 etc IS.
These clearly comfirms a substantial warmer MWP than current temperatures.

Tom, with all the datasets I have seen, i find it VERY hard to believe you could possibly find 55 datasets that told a quite different story.

Tom, Im not Loehle, so lets just agree that you dont think ive got the correct answer, and end this dialog.

K.R. Frank

“Notice that both after Pinatubo and El Chicon it does take years after the direct effect before the temperature levels are up on the “before-vulcano” trend. You have thus only removed a small part of whats needed if you really want to use flat trend UAH.”

Please look more carefully at:

http://img366.imageshack.us/img366/6383/uahexvolcwtrans.png

I have removed all the temperature data where the transmission coefficient measured at Mauna Loa Observatory is outside of normal variation. Is there any contribution outside of that? Yes, but it’s less than 1% of the transmission coefficient, which itself is varying by 2%.

A more valid criticism is that I’m actually taking out too much data, as the eruptions are only partially responsible for cooling in that period – for instance there was a La Niña episode in 1984

” do you honestly think, that the low temperatures in 1985 just after El Chicon Has nothing to do with El Chicon??”

Yes. This is more than two years after the eruption, the transmission coefficient has almost completely recovered – the cause was another La Niña event that year.

“AND that the low temperatures in 1993 just after Pinatubo has nothing to do with Pinatubo??”

I’ve already removed the 1993 data from the plot!

“And as i said, if “flat trend” is your preference, no matter that if for sure gives an outlier trens result, then you should use a just as faulty flat trend for the historic proxies.”

I prefer a linear fit when a more complex fit gives no significant improvement in the correlation. This will not be the case historic proxies, so a more complex fit can be justified.

“I then used 0,3 K, and showed that we still have over 0,4 K MWP.”

Whether your peaks are 0.3 or 0.4K above current temperatures, you have to demonstrate that these peaks are likely, even when they contradict all other proxy reconstructions.

Unless you can do that, I’m afraid you are wasting your time.

I said: “Tom: Studying the effects of El chicon and Pinatubo more:
http://atmoz.org/img/global_temperature_index_1.png

Notice that both after Pinatubo and El Chicon it does take years after the direct effect before the temperature levels are up on the “before-vulcano” trend. You have thus only removed a small part of whats needed if you really want to use flat trend UAH.
”

And you just go on showing this:http://img366.imageshack.us/img366/6383/uahexvolcwtrans.png

etc etc. This is not dialog, sadly.

http://atmoz.org/img/global_temperature_index_1.png
Tom , do you honestly think, that the low temperatures in 1985 just after El Chicon Has nothing to do with El Chicon??

AND that the low temperatures in 1993 just after Pinatubo has nothing to do with Pinatubo??

Coincidence coincidence?
Tom, then we just dont agree even in these simple matters, and can agree that we dont agree….

And as i said, if “flat trend” is your preference, no matter that if for sure gives an outlier trens result, then you should use a just as faulty flat trend for the historic proxies.
This also gives an outlier result, + 1,05 K for the MWP.

Then i suggested +0,3K as a compromise to be able to move on in the dialog.
You did not comment.

I then used 0,3 K, and showed that we still have over 0,4 K MWP.
You did not comment.

“Have you not heard of validation in engineering studies?”

Heard of? I’ve had great enjoyment over the years in developing and using validation and verification matrices!

As for “robustness” in wikipedia, try

http://en.wikipedia.org/wiki/Robust_statistics

“Robust statistics provides an alternative approach to classical statistical methods. The motivation is to produce estimators that are not unduly affected by small departures from model assumptions.”

I think you might enjoy the example concerning the determination of the speed of light.

“I believe you have not corrected for Elchicon.”

If you use the Mauna Loa Observatory data to remove all temperatures when the atmospheric transmission was outside of normal variation, the trend actually increases from 0.120 to 0.122 degC/decade:

http://img366.imageshack.us/img366/6383/uahexvolcwtrans.png

“Lets look at a handfull approaches to UAH temp rise 1980- today, and get a realistic picture, right?”

As I explained before, there is no justification for using anything other than a linear trend to fit the UAH data – the correlation coefficients show no significant improvement as a result of putting in the additional fitting parameters for a polynomial trend. To use such fits has no mathematical basis.

“Finaly DO NOT DO THE MISTAKE, that you blindly want to compare PEAK temperatures of today with AVERAGE temperature of the MWP!!!! Did the MWP also have EL NINO Peak with far above average temps?? YES!!”

I am not combining peak temperatures of the UAH record with filtered averages from the past – I agree it would be incorrect to use the 1998 El Niño maximum to make my point. Any filtered average of a linear trend is the same as the trend itself, so the comparison I’m making is absolutely proper.

“So to declare MWP “dead” is mostly a kind of wishful thinking.”

I have never declared it dead – there’s certainly evidence of a warm period at that time. But there’s little evidence that it was noticeably warmer than the climate today, as indeed Loehle states.

Your proxy reconstruction does not amount to evidence as it stands. At a minimum you first have to follow anna v’s advice:

“Frank has to take care in the correct statistical method the addition of errors, statistical ( which are added by squaring and square rooting) and systematic ( which are added straight, no ice) and display it for his final curve. It may be that that after that the whole significance of any bumps becomes moot.”

Notice that both after Pinatubo and El Chicon it does take years after the direct effect before the temperature levels are up on the “before-vulcano” trend. You have thus only removed a small part of whats needed if you really want to use flat trend UAH.

So to declare MWP “dead” is mostly a kind of wishful thinking.

I believe you have not corrected for Elchicon.
Here you see indications of atmosphere disturbing afer Elchicon and Pinatubo, Stratosphere and troposphere UAH.
http://atmoz.org/img/global_temperature_index_1.png
I cannot say exactly when the direct effect of El chicon ends, perhaps around 1985. A few years after El chicon temperatures seems to recover.
So if you want a more total flat trend curve (for some reason) then remove ALL the El chicon effect.
The MWP magnitude should obviously not depend on the random timing of recent volcanoes.

After doing so, you will have a UAH trend curve under 0,3K.

This demonstrates the random-sensibility of using a flat trend curve.

Lets look at a handfull approaches to UAH temp rise 1980- today, and get a realistic picture, right?

Temperature difference 1980 to 2000 and 2009 :

UAH actual vals, 5 year smooth. 2000: 0,10K 2009: 0,20K
http://www.climate-movie.com/wordpress/wp-content/uploads/2008/10/uah_global.gif

hadcrut, actual vals. 2000: 0,1K 2009: 0,3K
http://img201.imageshack.us/img201/5254/tropicshadcrut3rb1.png

polynom UAH trend. 2000: 0,18K 2009: 0,25K
http://www.holtlane.plus.com/images/uah_anomaly.jpg

polynom 2 UAH trend. 2000: 0,2K 2009: 0,15K
http://www.greenbuildingforum.co.uk/newforum/extensions/InlineImages/image.php?AttachmentID=631

polynom 5th order trend. 2000: 0,1K 2009: 0,0K
http://wattsupwiththat.files.wordpress.com/2008/07/uah_june08_curvefits.png
(- linear trend also shown to compare)

Linear trend , Tom P. UAH. 2000: 0,24K 2009: 0,36K

I suggest: 2000: 0,2K 2009: 0,24K

Now Tom you want me to use your 0,36K, but in order to make me accept this outlier value, you need to do some serious argumenting.

OK, lets make a compromise, we say UAH 1980-2000 shows a difference of 0,3 K, ok?

Fine. So we “know” that UAH contributes with 0,3K to the level at 1980.

Take a look at fig 2 in the article. Here you see not the “drawn” tree rings, but the actual, more precise average of 11 multiproxies, mostly trees.

The value here in 1980 id -0,18 K .

UAH temp rise of 0,3K from 1980 to 2009 would then result in + 0,12 K

BUT!

The tree ring average IS 50 YEAR AVERAGED (!!!) which gives a FAR higher 1980 value than original tree ring data says!
So we end up considerably LOWER than the + 0,12 K after inserting UAH +0,3K

These numbers are NOT absolute Kelvin values (please understand this), but Kelvin values that is then callibrated to the Proxi data in order to compare (!)

Summa: Incl + 0,3 K from UAH, we today have most like a value of +0 to +0,1 K which can be compared with the +0,4 to +0,6K in the long MWP.

This resembles the situation on my graphs.

On top of this, before the MWP it was even warmer, 1-1,5 K! For 6 many thousand years.
Probably sod cant find evidence for this either…! But others can.

Finaly DO NOT DO THE MISTAKE, that you blindly want to compare PEAK temperatures of today with AVERAGE temperature of the MWP!!!!
Obviously, if you somehow extract the absolute PEAK temperature after an El Nino, YOU MUST COMPARE WITH PEAKS OF THE MWP!!!
Did the MWP also have EL NINO Peak with far above average temps?? YES!!

So what we have done here is UNDERESTIMATING the MWP, because we have had a positive PDO + som some strong EL Nino data that we compare with MWP.

Tom. Now I have argued 100% truth seeking, please show that you do the same, and that you dont have som kind of agenda/religion. I dont want to waste my time on you if thats the case. Let me see some honest scientific thinking from you.

This thread is fast getting out of front page. I suppose we can pick up arguing in a newer thread :).

I was talking about IPCC models when I was talking of chi**2 fits.
Robustness is good for gymnastics and it seems genetics. http://en.wikipedia.org/wiki/Robustness
Have you not heard of validation in engineering studies?

The world is asked to commit trillions and destroy its way of life on studies that are “robust” instead of validated, as should any engineering product be before use?

Maybe you should edit the wikipedia entry to include what robustness means for climatology.

For pure data:

Frank has to take care in the correct statistical method the addition of errors , statistical ( which are added by squaring and square rooting) and systematic ( which are added straight, no ice) and display it for his final curve. It may be that that after that the whole significance of any bumps becomes moot. This means that the significance of the present “hot period” also becomes moot and a chasing of video game visions.

no strawman at all. if he can pick his datasets, so can i.

i am sorry, if you don t like the data. but you will have to struggle hard to find scientific support for the claim, that the 30s were NOT 0.5°C below the temperature of the last decade.

Are you suggesting NOAA and AMSU2 databases are not correct?

The peak value of the MWP is 0.526 Deg C above the mean over the period (again as a 29 year mean, not annual, value). This is 0.412 Deg C above the last reported value at 1935 (which includes data through 1949) of 0.114 Deg C. The standard error of the difference is 0.224 Deg C, so that the difference is significantly non-zero at the 10% level (t = 1.84). While instrumental data are not strictly comparable, the rise in 29 year-smoothed global data from NASA GISS (http://data.giss.nasa.gov/gistemp) from 1935 to 1992 (with data from 1978 to 2006) is 0.34 Deg C. Even adding this rise to the 1935 reconstructed value, the MWP peak remains 0.07 Deg C above the end of the 20th Century values, though the difference is not significant.

Scientific articles must not be read the same as you read the Bible, i.e. taking verses out from the context:

While instrumental data are not strictly comparable, the rise in 29 year-smoothed global data from NASA GISS (http://data.giss.nasa.gov/gistemp) from 1935 to 1992 (with data from 1978 to 2006) is 0.34 Deg C. Even adding this rise to the 1935 reconstructed value, the MWP peak remains 0.07 Deg C above the end of the 20th Century values, though the difference is not significant.

“Tom, I have worked for over 35 years in fitting models to data in the field of particle physics. All these “sensitivity”, “robustness” and “estimates” would be laughed out of the conference room had one dared to use them.

When we came out with a number from the fit, there were chi square per degrees of freedom and error bars +/- statistical, +/- systematic.”

My background is in physics as well, as is also the case for a lot climate scientists. The techniques physicists generally use are to fit theory to data, but this is not what a proxy reconstruction is trying to do. Such reconstructions are trying to derive the underlying time series behind a set of derived measurements. There is mostly no actual past temperature series to compare, and so the techniques you cite are of little relevance – what model is Frank supposed to do a chi-square test against?

Hence the tests of the “skill” of a reconstruction are to show that it is robust in that it is insensitive to choice of proxies, that the proxy temperatures are derived correctly, and the weightings are reasonably applied given the uncertainties in the derived values. I was also unfamiliar with these techniques, but that is no reason to dismiss them.

Climate models have the same problem in that again there is no future data to model against. My subsequent engineering practice, though, suggests a way of analysing the robustness of such models. If I want to be sure that my finite element analysis (FEA) of a mechanical system, for instance, is giving a reasonable answer, I first check it is consistent with a simplified analytical approximation and then change the analysis parameters, such as cell number and shape, to see that my FEA solution doesn’t change drastically. If I want an estimate of the accuracy of the solution, I tweak the input parameters.

This is the day-to-day experience of modellers in engineering, and I’m not at all surprised that climate scientists use similar approaches in their GCM modelling, which is just time-dependent finite element analysis.

no strawman at all. if he can pick his datasets, so can i.

i chose the two most common used sets of surface temperature.

http://www.woodfortrees.org/plot/hadcrut3vgl/from:1935/plot/hadcrut3vgl/from:1935/trend

i am sorry, if you don t like the data. but you will have to struggle hard to find scientific support for the claim, that the 30s were NOT 0.5°C below the temperature of the last decade.

Really, can you make ANY argument without some logical fallacy? Just curious, I’m waiting to see one.

using a different dataset is not a logical fallacy. you might want to google that term, before you use it.

“Tom, you can see as well as I that if not El chicon and Pinatubo where placed in 1985 / 1993 there would have been a linear UAH trend of 0,2K 1980-today, approx. Therefore your 0,4K is complete random and for no scientific use.”

Maybe some data would help clarify. I’ve removed all below-trend data for temperatures after El Chicon and Pinatubo. The UAH series gives:

http://img11.imageshack.us/img11/6061/uahexvolc.png

and a trend of 0.120 rather than 0.133 degC/decade, so a 0.36 rather than 0.40 degC change from 1979 to date. The removal of these lower temperatures means the current anomaly has actually slightly increased and is now +0.28 degC.

I’m sure you will amend your analysis accordingly.