Interesting Independent analysis of IPCC and AGW data

[Guest someguy]

came across this analysis by this Doctor from Australia . he does what I think is a decent review of the IPCC data assmuption and forecasts and comes away from an interesting contrary theory.

Dont know if he is right or wrong .... but its worth a read....

from what I can see this guy does NOT appear to aasociated with any "group " in the AGW debate

Girma J Orssengo

Master of Applied Science, University of British Columbia, Vancouver, Canada

Doctor of Philosophy, University of New South Wales, Sydney, Australia

http://wattsupwiththat.files.wordpress.com/2010/04/predictions-of-gmt.pdf

[k***]

quick google search shows this guy loves ayn rand, portrays AGW supporters as being "anti-plant food", and is an optometrist.

like DT, i'm not sure if he's right or wrong, but i have my suspicions...

[Kelvin Wave]

This appears to be a huge oversimplification of the issue, regardless of what you believe. His method for fitting the cosine curve to the data doesn't make a whole lot of sense. Using a combination of sines and cosines (Fourier), one could perfectly fit a sinusoidal wave to the data - why settle for a fit that only explains about 3/4 of the overall variance?

This statistical model that he has created is nowhere near sophisticated enough to properly capture, nevermind predict, global temperatures. His model only works so well because he created to work well. So his argument that because the model captures the major changes in the original data, it can be used to predict the future, is ridiculous.

Simple statistical models like his only have a prayer of predicting the future when the data are stationary - that is to say that the major attributes (mean, variance) upon which the model was created do not change in the future. Statistical models can only predict patterns (or combinations of patterns) with which they have been trained. That's why we have dynamical models for this stuff instead.

He also does other shoddy things... one in particular is that using the PDO as the main oceanic force behind global warming is far too simple. Much of the effect from oceans has to do with them being a CO2 sink hole.

I'll stop here, but clearly statistics is not his strong suit.

[BethesdaWX]

This appears to be a huge oversimplification of the issue, regardless of what you believe. His method for fitting the cosine curve to the data doesn't make a whole lot of sense. Using a combination of sines and cosines (Fourier), one could perfectly fit a sinusoidal wave to the data - why settle for a fit that only explains about 3/4 of the overall variance?

This statistical model that he has created is nowhere near sophisticated enough to properly capture, nevermind predict, global temperatures. His model only works so well because he created to work well. So his argument that because the model captures the major changes in the original data, it can be used to predict the future, is ridiculous.

Simple statistical models like his only have a prayer of predicting the future when the data are stationary - that is to say that the major attributes (mean, variance) upon which the model was created do not change in the future. Statistical models can only predict patterns (or combinations of patterns) with which they have been trained. That's why we have dynamical models for this stuff instead.

He also does other shoddy things... one in particular is that using the PDO as the main oceanic force behind global warming is far too simple. Much of the effect from oceans has to do with them being a CO2 sink hole.

I'll stop here, but clearly statistics is not his strong suit.

Please back youself up instead of spitting out critisism with no basis.

Thankyou

[skierinvermont]

Basically agree with what the poster two above me said.

He's created a statistical model and we all know how well statistical models work in meteorology. It's analogous to the XTRP line for hurricane tracks. Extrapolating past trends is not a strong predictor of future trends. Moreover, he could have created a much stronger statistical model of past trends. The curve would fit better if the amplitude of the sin function were not as large. But he obviously chose to use an inferior function because that would predict less warming in the future.

That being said, I think some of the oscillatory nature of 20th century temperatures, is related to the PDO, and I think this is one reason we have not seen temperatures rise very much at all since the late 90s. Skeptics, and this model in particular, tend to exaggerate the physical importance of the PDO on GMTA.

[Kelvin Wave]

Please back youself up instead of spitting out critisism with no basis.

Thankyou

I shall do so. I have put the commentary inline...

This appears to be a huge oversimplification of the issue, regardless of what you believe. His method for fitting the cosine curve to the data doesn't make a whole lot of sense. Using a combination of sines and cosines (Fourier), one could perfectly fit a sinusoidal wave to the data - why settle for a fit that only explains about 3/4 of the overall variance?

The fact that any curve can be fit perfectly using either a Fourier type fit or a polynomial fit is a continuation of the logic one uses to explore Taylor series in a Calculus class. Secondly, he shows that his model correlates with observations with a correlation coefficient of 0.88 (this is R). The coefficient of determination, which is the square of R, says how much of the variance is explained by the model. This is about 0.77 which I rounded to 3/4. If you don't believe me, check here: http://en.wikipedia.org/wiki/Coefficient_of_determination .

This statistical model that he has created is nowhere near sophisticated enough to properly capture, nevermind predict, global temperatures. His model only works so well because he created to work well. So his argument that because the model captures the major changes in the original data, it can be used to predict the future, is ridiculous.

Simple linear models that are trained to a dataset cannot predict inherent changes in the dataset that may occur over time. Let me elaborate... this man creates a statistical model that is purposely fit to the original data. He then says that because his model is able to predict the major regime shifts in the original data, then it is useful to predict the future. The reason it predicts the original regime shifts is because it was TRAINED to do so... that is, the model was created using the original regime shifts, so it must be able to predict them!

The fact that it fits the original data well does not imply that it will fit future data well. The model is only useful for predictions if the mean, variance, trend, etc. in the future are the same... clearly with global warming this wouldn't be the case. The model cannot capture the nonlinear effects that are inherent to the climate system.

Simple statistical models like his only have a prayer of predicting the future when the data are stationary - that is to say that the major attributes (mean, variance) upon which the model was created do not change in the future. Statistical models can only predict patterns (or combinations of patterns) with which they have been trained. That's why we have dynamical models for this stuff instead.

Tied in with above comment...

He also does other shoddy things... one in particular is that using the PDO as the main oceanic force behind global warming is far too simple. Much of the effect from oceans has to do with them being a CO2 sink hole.

The PDO is defined by SST anomalies, and therefore doesn't take into account anything regarding CO2... this site has a lot of good PDO info... http://jisao.washington.edu/pdo/

I'll stop here, but clearly statistics is not his strong suit.

[Clifford]

Here is another model similar to this one above.

http://www.climate-s...-important.html

Here is another one (or bunch of notes by the same author) about long-term periodicity in solar cycles.

http://www.vukcevic.co.uk/

http://xxx.lanl.gov/...401/0401107.pdf

http://www.vukcevic....olarcurrent.pdf

One can argue whether the last decade showed any actual warming.

There certainly has been flattening of the trend since 1998 (or 2002).

And, with the first 2 months hitting close to the 30 year average.

http://www.coaps.fsu...e_anomalies.jpg

There are indications that this year will also show some "flattening". Is an "average year" an anomaly?

It is much harder to prove a trend with 60 year cyclic events.

However, one of the big complaints of the so called "sceptics" is that the "alarmists" have chosen the maximum slope for their projections. Yet, they recognise that there have always been long-term climate cycles in the past... just not NOW!

We also know there are longer term periodicities in the climate, perhaps a few hundred, or a few thousand year periodicities.

And...

Sometime we'll get hit with a forcing agent equivalent to an 8°C temperature drop. It could hit tomorrow, it could hit in a thousand years, or it could hit in ten thousand years. But, humanity should have no doubt that the pressures that caused previous glacial periods will return. Concerned about shrinking glaciers? What will we do when the glaciers start growing again?

[WeatherRusty]

Here is another model similar to this one above.

http://www.climate-s...-important.html

Here is another one (or bunch of notes by the same author) about long-term periodicity in solar cycles.

http://www.vukcevic.co.uk/

http://xxx.lanl.gov/...401/0401107.pdf

http://www.vukcevic....olarcurrent.pdf

One can argue whether the last decade showed any actual warming.

There certainly has been flattening of the trend since 1998 (or 2002).

And, with the first 2 months hitting close to the 30 year average.

http://www.coaps.fsu...e_anomalies.jpg

There are indications that this year will also show some "flattening". Is an "average year" an anomaly?

It is much harder to prove a trend with 60 year cyclic events.

However, one of the big complaints of the so called "sceptics" is that the "alarmists" have chosen the maximum slope for their projections. Yet, they recognise that there have always been long-term climate cycles in the past... just not NOW!

We also know there are longer term periodicities in the climate, perhaps a few hundred, or a few thousand year periodicities.

And...

Sometime we'll get hit with a forcing agent equivalent to an 8°C temperature drop. It could hit tomorrow, it could hit in a thousand years, or it could hit in ten thousand years. But, humanity should have no doubt that the pressures that caused previous glacial periods will return. Concerned about shrinking glaciers? What will we do when the glaciers start growing again?

Cycles in Earth's orbital configuration involving 22,000, 41,000 and 100,000 periods are quite well established as the "pressures" that modulate ice age/interglacial periods. It will be many thousands of years before the phasing of these orbital cycles supports a recurrence of an ice age.....which will take thousands of years to develop.....you don't grow ice sheets a mile deep and thousands of miles across in a few decades or centuries.