Saturday, October 3, 2009

Tree Ring Chronology Selection Problem

This article is an attempt to explain why tree ring temperature reconstructions have no value at all as they are currently done. I am not trying to make the point that they have no potential value; only the point that current methods of chronology selection destroy that value. To begin, I want to point out the method in question by using a quote from an Osborn/Briffa paper – 06 – SI.

“We remove any series that was not positively correlated with it “local” temperature observations [taken from the nearest grid box of the HadCRUT2 temperature data set (S9)]. The series used by (S3) were already screened for positive correlations against their local annual temperatures, at the decadal time scale (Table S1). We removed series from (S1) that did not correlate positively with their local annual or summer temperatures (Table S1), or which did not extend into the period with instrumental temperature to allow a correlation to be calculated. The series from south-west Canada (named Athabasca) used by (S1) did not correlate positively with local temperature observations, but has been replaced by a new, better-replicated series (S10) that does correlate very highly with summer temperature (Table S1) and has also been RCS-processed to retain all time scales of variability"

Now here is a quote from Briffa responding to Steve McIntyre’s criticism of his Yamal chronology.

"Chronologies are constructed independently and are subsequently compared with climate data to measure the association and quantify the reliability of using the tree-ring data as a proxy for temperature variations."

Now, one would think that this would be a good selection method if we wanted to use tree ring reconstructions to tell us 20th century temperature. But actually, it doesn’t even do that. The chronology simply mimics the surface temperature record; right or wrong.

But the real purpose of a reconstruction is to give us an idea of the relative temperatures of different periods of history. Especially those were we had no thermometers. And for that purpose, the selection method described by Osborn and Briffa absolutely destroys the value of the reconstruction.

It is clear from the descriptions of dendrochronologists themselves, as well as from what we know about tree growth, that trees do not all grow at the same rate, and the same tree grows at different rates throughout its life. One of the factors effecting its growth is temperature. But there are many others – sunlight, nutrient, water, CO2 availability etc. At any given point in time a tree may be able to respond to increases in temperature, or it may have it’s growth limited by a problem with the other factors. So, over the course of the life of a tree it may change it’s growth rate because it’s roots have reached a new source of nutrition; because the flow of water on the hillside where it is growing has slightly changed giving it more or less of the stuff; because a nearby tree has died giving it more sunlight; because a nearby tree has grown giving it less sunlight; because a nearby tree has died and it’s carcass is feeding the tree; because some bears have decided that it’s a good place to relieve themselves, etc. So let’s say that at any point in time a tree has a 60% chance of responding properly to temperature variation because it has the other resources that it needs. The number could be 20% or 80% - for purposes of this example it doesn’t matter.

So now, when we are in a period of increasing temperature, we select the series or the chronologies, as the case may be, that reflect that temperature rise. This means that 100%, or nearly 100% of the trees in the 20th century portion of the chronology are responding to temperature. If the number is much smaller, then only a fraction of the temperature rise will be reflected. But what happens to other historical periods within that chronology. The selection process did absolutely nothing to insure that those same trees were responding to temperature in say the Medieval Warming Period. So the likelihood that they are responding correctly to temperature tends to move towards our theoretical 60% average for those later periods. This means that by using Briffa and Osborn’s selection process, you are introducing a huge bias for having the 20th century show much more warming than the MWP. In fact, if the 20th century turns out to have anywhere close to the warming of the MWP, it will produce a chart where the 20th Century shows up as being much warmer than the MWP. And Al Gore can claim “unprecedented warming” for the 20th century.

I don’t know if Osborn’s and Briffa’s selection methods are unique to them. I rather doubt it. Most likely they are used by all of the dendrochronologists. If this is the case, then any reconstructions that use tree rings will tend to be worthless.

The effect of the selection method in question can be seen quite clearly in this IPCC spaghetti graph of temperature reconstructions. Notice how the separate reconstructions come tightly together in 20th century because their chronologies have been selected based upon their matching the surface temperature record. But then moving to later times, they spread out dramatically because their sensitivity to temperature is no longer assured. The percentage of trees that are sensitive to temperature has fallen, and the number of trees that are sensitive to temperature in any given reconstruction can vary widely. How useful are they when there is a .8C variation between them at some points.


Jeff said...

I agree with what you've written. Multi-proxy studies use a variety of methods to sort data. Rather than throw data out, Mann98 used decentered PCA to pull out the signal. There are also other math techniques which can highlight the signal you want, without actually removing the data. Multi-variate regression can get it done.

I've never seen a hockey stick that hasn't gone through this process.

Now in the Yamal case, Briffa says he didn't sort it. However, the original authors admit to sorting the data in the pre-measurement years and it is somewhat implied for the 12 series SteveM replaced.

From my experience, it is highly unusual for a set of treerings to have such a consistent upslope signal without being sorted as those 12 Yamal trees. Take any set of proxies anywhere and look at any point in their record and you won't find that. Of course it's impossible for me to prove that statement but when you spend some time with the data, that's what you find.

Nice job.

Tilo Reber said...

Thanks Jeff.

and then the_doctor said...

I've spent some time in forestry related jobs. The variance in microsites over time and their independence from temperature always seemed suspiciously absent from the certainty of establishing solid temperature data.

Interesting post.

Anonymous said...

My comment at Wikepedia protesting the listing of the Gavin - ran Real Climate website, not this present site where you are reading this comment ...

Below was posted at wikepedia in "Discussion" section at listing for Real Climate this week :

we are being faced with issues and people who feel they are right without the proper facts , and these new liberal thinkers feel that if they are right, well then, the facts will arrive eventually to back up their claims... At the Real Climate website are actual comments by some who wonder how the controllers of this RC site can be so self-righteous in their claims based on instrument readings from many devices placed all over the world...some have pointed out that some devices do not work properly or are giving data or temperatures that are sometimes influenced by improper performances or design of the instrument . ( Any machine can suffer from being in place in the environment ) Yet unlike our cars where a malfunction means a trip to the garage any slight mistakes in data are simply recorded as true data and this means, if the world is complex and big, instrumentation may not always give us the PROPER and ACCURATE data we need to make educated guesses of the future of Earth's climate...someone inform wikepedia that it is not proper to give those with biases that the Earth is warming a science website legitimacy that may or may not be well-deserved...Can we wait 30 years to see if the weather alters as warmers believe before we promote them to oracles and futurist geniuses who tell us what the future is ? Can we have some patience before we give legitimacy to those who FEEL they are right ? Thank You. I hope wikepedia has the moral courage to leave my comments up in Wikepedia.
- DerrUFO