Meanwhile I’m still enjoying Jeff’s amazing work, and hope to read even more eventually.

Cheers, Robin

I think the important thing about Loelhe’s data is that the proxies have been validated as actual temperature proxies. Tree rings are uncertain to be temperature proxies as there is a band where temperature is optimum above which there is slow growth.

Peer review sucks. It’s initial motivation was to keep nutcases out of scientific journals but it has now focused on supporting popular paradigms. Of course, Einstein would be rejected out of hand these days.

I’ve often contemplated writing something about the tendency of climate to change in steps from one regime to another. I did try once, with “Weather” the journal of the Royal Meteorological Society – must have been about 8 years ago, I suppose, but the referee clearly didn’t like my conclusions, though the editor did! I don’t think the ref understood what I was doing. Must use smaller words next time.

Cheers, Robin

BTW: I just did a post which I used Mann’s favorite CPS method to extract positive and negative hockey sticks from his groups own proxies.

I also demonstrated temp rises and falls in historic points as well as sine waves just for fun. All of it used the M08 group math and data sets.

http://noconsensus.wordpress.com/2008/10/11/will-the-real-hockey-stick-please-stand-up/

No doubt the true main signal is coming through after smoothing, tho’ I’d still really like to see the untouched data!

Cheers, Robin

I also have checked some of Loehles work out. The difference between his papers on climate and others is that he isn’t doing any post data sorting. Therefore smoothing just makes it easier for your eyes to see trends.

Not that smoothing is correct, all I’m saying is that smoothing hasn’t affected the overall trend as in a Mann paper. The data were pre-calibrated and therefore scale was unaffected by smoothing.

His approach is pretty minimalist as far as typical data molestation in this field. Craig is a forester by trade so he has an excellent handle on tree growth. He has also recently issued papers on the non-linearity of tree response to temp.

As far as my two cents, these are the best representations of climate I can find. The number of proxies is not complete enough to draw conclusions from and the calibration was done by other scientists so I couldn’t review that.

Well, I’ve downloaded it all, thanks to your link, and have to say I’m a mite disappointed! I can of course reproduce his conclusions about the MWP ad LIA by simply looking at the plot of his data. They look quite reasonably spiky, as one expects from anything related to climate, but when I looked at the cusum plot it was /instantly/ obvious that the data were not “as found”, but heavily smoothed. This pretreatment is always recognisable from cusum plots merely from /their/ lack of spikiness. Then I read the summary of his paper – can’t get at the whole thing – and saw that the data had all been treated by using a 30 year smoother. Unfortunately, this ruins what can be deduced from cusums about sudden change. That of course is what smoothing of any sort is all about! It is used to disguise unpalatable amounts of noise, disregarding completely the fact that the most interesting signal may be in the noise that the smoothing system takes out, and that we cannot know what the cause of the noise is. It could be old-fashioned iid “gaussian” with mean zero, but it could be a reflection of the vicissitudes of the complex and unforecastable things that are caused by the drivers of climate. Further, the more you smooth the greater are the problems with the start and end of the series. Inevitably you have to resort to guessing.

Another problem is “outliers”. No doubt you’ve addressed this one, probably without reaching an acceptable conclusion. Again, in my software I’ve implemented a couple of procedures, but in reality the only reason for discarding data is knowing that it is in error from sources other than the data values themselves. A typical scenario is the transposing of digits, which happens to result in a “strange value”, say 28 instead of a more reasonable 82. This stranger may indeed by picked up by suitable outlier detection techniques but if you don’t contact the owner of the data with your suspicions you really have no proper grounds for rejecting it.

You’ll guess that smoothing frustrates me. If I need it I can do it myself, but with a heavy heart. Simplification of real data for the convenience of those who choose not to try to understand that nature is not all smooth curves or linear models is an unwelcome aspect of current popular science, in my opinion. It tries to gear things to the level of science correspondents of newspapers, whose editors further manipulate them to form headlines, or to make them “understandable” to politicians, who are very seldom proper scientists.

So, where are the original data, I wonder. Those are what I’d really like to see.

Cheers, Robin