# the Air Vent

## Because the world needs another opinion

• JeffId1 at gmail dot com

Everything you need to know about Climategate. Now available at Amazon.com

• ## Follow Blog via Email

Join 140 other followers

## Temperature Scale Distortion in Hockey Sticks

Posted by Jeff Id on September 24, 2008

Just a short post today. I am tired from working on this for so long. I have been analysing and trying different things so I get a good feel for how the historical data in from what I know, every hockey stick temperature graph ever produced is actually on a different scale from the “CALIBRATION PERIOD”.

The Calibration period is the secton of every hockey stick which shoots upward. This is becasue measured temperatures have risen for the last 100 years or so.

Scientists compare the shape of temperature curves to temperature proxies which are things like tree ring width graphs, mollusk shell isotopes and other obscure objects to try and determine historic temperature. If the shape of these proxies matches measured temp they consider the proxy good.

In comparing shape one thing scientists do is find the absolute best fit, magnifying and offsetting the data to give the best possible correlation. This is where they go wrong.

If you are new to building hockey sticks read

Ten Things Everyone Should Know About the Global Warming Hockey Stick

If you already know paleoclimatology but are unfamilliar with this reconstruction read:

How Come So Many Independant Papers Claim Hockey Sticks

and

The Flaw in the Math Behind Every Hockey Stick

Ok, the same process as before – I used random Red noise (the integral of white noise or noise with a trend)

I applied a fake temperature signal to the noise.

My fake temperature has the same 1 degree in recent times (2000) as it does in history (1000 AD)

I then went to look for the temperature using analogous methods to hockey stick papers.

Assume – linear temp rise from 1900 – 2000 of -0.2 to 0.6C

step 1 – correlate high r value data >0.8

step 2 – remove any negative slopes

step 3 – magnify individual data series to match my temperature trend

step 4 – average data

This produced the following graph. I looked for data with a high correlation to my intended temperature. I scaled the data to match the 0 – 1 C rise between 1900 and 2000. I found the signal which I had inserted in the data by adding the above graph to the data. But its shape changed.

We noted the hump in history is diminished compared to recent times.

I made the data so we know the peak at 1250 was really 1C and the value at 1300 AD was 0. What happened to temperature.

It is as I have said, the scale of temperature in the past is distorted by sorting for preferential data.

Today I have extended the graphs to show what actually happened to the temperature by sorting the data according to these methods.

The dark blue line is the temperature curve from the second graph up from here. The rest of the lines are actual temperature as should be measured on the graph. The top line represents actual 1 degree C, the yellow line represents a -1 degree C. The entire graph is not on the normal Y axis scale except for in the 1900 to 2000 calibration range. Temperature was compressed in this graph by 0.48 times original scale simply by looking for a trend in recent times!!!!!

There is something else which I can show that demonstrates how prevalent this false temperature shape is in paleoclimatology. In the graph above section A is rising behind the 1900-2000 correlation period. This feauture is prevalent in every fake reconstruction I have done. Now paleoclimatology data is much noisier but if I am right and we look close, you should see the same distortion as I have at A in the graph above.

Look at the top picture In this graph the red line is the calibration range but beneath it (a lot cooler) and barely visible is the actual proxy data. From right to left note the steep drop in the top graph followed by a gradual rise. The next graph is from climate audit. Again you see the drop from the right end to left followed by a gradual rise.

Again this graph is thanks to climate audit. The recent temperatures show a drop followed by a gradual rise into history.

These graphs have clearly been according to the same mathematical properties I have worked to show you. From messing around with the data. Temperature in these graphs is on a distorted scale. I have determined a method for calculation of the magnitude of the distortion but I am still working on the offset.

1. ### StuartRsaid

Thanks for another illuminating post

So if I have this right in my layman mind, if the real world paleoclimatology data shown above is basically using the same match and select method previously outlined in these simplified models Not only have you ensured that the current temperature record matches the shape of the selected proxies but it would also, as a by-product of the method, tend to confirm/support the anecdotal evidence for a Little Ice Age . Even if the proxies, in reality, are not very good and could be as useless as the random red noise data, you are still going to get something that looks like a good correlation for at least a couple of hundred years in the past?

In that case it looks to me like this method could be very self-deceptive and would need some further safe guards upon the method of data selection.

I guess that’s why there isn’t much argument about the existence of the Little Ice Age but the dispute gets more heated about the level of the Medieval Warm period?

As I said, I am a total layman in this area and so I feel sure that there is more to it than that, but I’m still very interested thanks again…

2. ### jeff idsaid

The graph has a single vertical compression factor which applies throughout its length. I want to discuss that in my next post. I spent a bunch of time trying to learn another software language yesterday.

Since the graph creates a compression and a variable offset to the actual temperature graph, interesting things will happen when you correct for it. My data had a constant scale multiplier of 0.48 throughout its length. In the hockey stick blade range of he graph the true temperatrue lines are shifted very strongly, this makes the recent time of the graph have a full scale temperature value of 1 degree. In the historic part of the graph the offset of the true temperature curves becomes more constant and the amplitude of he signal is much smaller than reality.

So the question becomes, can we calculate the shape of real temperature for proxy reconstructions and what would it do to the final outcome of the graph.

The big thing I see is that reconstructions which show a medieval warm period which is slightly smaller than today could get a correction of as much as 2X plus an offset just based on the math!

3. ### Chris Hsaid

I see that you have scaled all your pictures (graphs) in this latest post, which is good as in previous posts Firefox (on a 1024×768 display) was cutting the right-hand half of every picture off (so I had to right-click then select View Image to see the whole picture).

4. ### Chris Hsaid

Actually, the 4th picture (temperature-scale-distortion-0-1-deg-with-a.jpg?w=628&h=417) was still slightly truncated for me.

5. ### Jeff Idsaid

I scaled it down now.

Thanks.

6. ### rob rsaid

I am waiting to see what happens when the most prominent proxy series like MBH98; Moberg etc are re-“adjusted” to account for the issues you are discovering. Will we start to see multiple proxy temperature records that look like the “non tree-ring” record produced by Loehle & McCulloch (2008)? Is this finally going to resurrect the MWP? We live in hope.

The next question is – how can you get these findings out to a vastly wider audiance?

7. ### Micksaid

Just a word of thanks for your efforts in bringing the detail to the layman, that is, myself. Have been lurking around ClimateAudit and WattsUp and find much of interest, your efforts adds a further dimension to my understanding of the roles and unique application of statistics in Climate Science. In short, there appears to be a need for an extensive audit along the lines of the Wegman Panel review to other aspects of climate science.

Regards Mick.

8. ### JamesGsaid

Playing devils advocate here, if we were certain that the 20th century instrumental data was correct then this data mining for slopes would successfully replicate it but it would also show any true MWP signal (assuming sufficient data) because the noise is still averaged out in that period. So even if most proxies are no better than noise, isn’t the technique semi-valid because it still gets a MWP signal if there is one to detect?

It seems to me to always come back to the validity of the 20th century temperature plots. If they are correct (which with all those upward adjustments I highly doubt) then ok we can’t say which period was warmer (due to scaling and dearth of data) but the striking aspect is not the height but the rapid slope. Most intelligent warmers say that it is the slope, not the amplitude, that indicates AGW. So while I can see you’ve invalidated the MBH98 procedure I’m still not sure you are invalidating Mann’s 2008 effort, which in any event shows MWP proxies higher than the early 20th century proxies. It is still only the instrumental record that gives that rapid rise at the end.

9. ### JamesGsaid

Now if we were to find out that current temperatures free from spurious adjustments (especially the TOBS adjustment) are no higher now than the 30’s/40’s then that represents the death blow to the AGW hypothesis and invalidates virtually every scientific paper written on it. Of course, I hastily add, we only want to find out the the real truth: ie are we experiencing unusual warmth or is it more likely natural?

Of course then people would be saying – what happens to all the CO2 then? My own feeling is that it isn’t even well proven that CO2 is well-mixed. I’m pretty sure it varies hugely with altitude and most of fossil-fuel-produced CO2 is quickly gobbled up by nature. Truth will out…eventually.

10. ### Jeff Idsaid

JamesG

Thanks for the comments. I am a bit new to climate science so, I am still working out all the details. I got a bit mad when I found out they weren’t calibrating trees to temperature and then averaging the data (three weeks ago!). The correlation and elimination process obviously distorts the scale. Some of these climatologists are good enough at math to know it yet they present the papers anyway. There are a few things you said which my recent intense study might clear up (maybe not).

The tree ring proxies used which represent the majority of the reconstruction data are clearly non-linear. I want to do a post on Craig Loehle’s latest paper but I need permission yet. When I think about a tree growing ever faster the warmer it gets the whole thing stops making sense. Sure there is some response but it can’t be infinite and it won’t be linear. Therefore treating tree rings like thermometers is borderline insane but again what do I know.

Tomorrow, time permitting I will show a graph which demonstrates the origin of the MWP in Mann 08. I have been living in the data and I can say without looking at it the MWP by the proxies is generated from very few measurements. The Mann 08 paper is a bit of practice for me, I want to expand my temperature corrections across all hockey sticks sorted by correlation. From what I can tell that includes every single one!

The correlation process is the problem and it can be mathematically demonstrated. If I am right, the rescaled graphs won’t make any sense to people so they will either accept the math and find new proxies or ignore my posts and any of my future papers which might get published.

If I were them, I would choose — Ignore.

11. ### John Nicklinsaid

I have been curious, for a long time now, about why Mann, et al have to graft-on the instrument readings. It goes way beyonf apples and oranges in my opinion. If the reconstructions show that it is warmer now, why bother with the instrument data. Trying to peer through that thick red instrument plot to see what the proxies look like is difficult, perhaps intentionally so, otherwise they could have put the fat red line behind the other series.

At any rate, in my humble opinion, the proxies are approximations which cannot be directly compared to precision instrument readings. More like grapes and pineapples, at least apples and oranges are about the same size.

Thanks Jeff for taking this on, I can only imagine the brain power required for all the heavy lifting. Your insight is very much appreciated.