DaleC, could you ask your graphics colleague to elaborate on why he think this picture is a fake?

I never bothered checking before; I do some of this kind of work. At first I quickly passed it off as a pasteover job (see below for some of the things I saw.) Now I think it likely it’s just a low quality photo, saved for the web by “save at low resolution” software that managed to make the tree look very different due to its different level of detail. A good copy/paste job would look the same; can’t tell the difference.

What I can say with assurance is it is not an original digital photo. But that’s no big deal; we know it was saved into a screen presentation anyway!

Almost anyone can do the following ‘telling’ test. Open up (free) Picasa, click on a picture to edit, go to the ‘effects’ tab and bump up the ‘sharpen’ level all the way. An original digital photo or scan will look much different from one that has been edited, either through low res save and/or copy/paste work.

Here are two screen shots of tests I did.

First, look closely. Here’s a 200% view:

Note how blocky the tree portion is, in stark contrast to the background. Those blocks are characteristic of a low-resolution JPG file.

Then I looked using a more professional tool, NeatImage.

NI allows you to view a photo in Y/Cr/Cb space, which is best for noise analysis. The tree-based ‘blocks’ are very obvious when blown up large. The details on the tree are clearly preserved at a different resolution than details of the mountains or nearby rocks. But low res JPG saves can do this just as much as a copy/paste job.

While the Cr/Cb noise is distinctly different over the tree, I wouldn’t claim anything from that. A tree is distinctly different from sand, rocks and mountains.

(By the way, there’s a free version of NeatImage available. It’s unique in its ability to clean up digital pictures, eliminating all kinds of grain and other noise! I use the pro version for bigger projects)

Cum Hoc, Ergo Propter Hoc

I have a strong sense that this is a large gulf between twq (and many others) and the CA group. We repeatedly see evidence that various AGW promoters are expending great energy to prove “reasonable” correlations, and use that to convince the statistically gullible of their stance.

I found this to be a nice, if dry, summary of the fallacy.

But there’s another explanation I appreciate much more. Unfortunately, it is cultural and perhaps will not communicate well to twq.

The Texas Sharpshooter is famous in some circles, because he’s soooo good at hitting the correlation bullseye. Only problem is, he draws the bullseye around the shots that landed in a tight cluster, and you never get to test his ability to shoot at a fresh target.

I’ll spell it out clearly to ensure language and culture don’t get in the way. Twq, it’s great to find a correlation in a data set. But you’ve learned nothing at all until you go much further:
* Apply your methods and hypothesis to different data (in this case from the same areas since you’re trying to find climate info from a specific area.) It’s not valid to throw out the data that “doesn’t fit.”
* Apply your methods to appropriate kinds of random data, to ensure your method isn’t automagically producing the interesting result.

CA has had many good conversations on these topics before, such as here. The questions I raised before have never been addressed. The more I’ve looked, the more I see work that would never pass muster in my world.

I’ve mentioned before that in the business world, we rather directly feel the pain of poor analysis. Accounts are quickly lost, stocks drop precipitously. Auditors help make sure that’s the case: invalid claims get knocked out ASAP. Sadly, too much of climate science today is distanced and insulated from the impact of bad work.

It’s very very much like the Y2K days. I’ll be glad when this is all over.

Would you please re-read all my comments?

Uh, pass.

Isn’t the issue in reconstruction about finding a measurable parameter which correlates with temperature and not proving that temperature cause tree ring widths.

IMO no, but other opinions are welcome This is a clear problem of observational studies. What if the correlation applies only for a short time? What if the correlation is completely spurious? Like bender said, the list of response variables seems very large.. And in the next post we have a 40-year moving average, please save me from those ..