## Forward Modeling of Ring Widths

bender drew our attention to the following interesting paper from Evans et al (including Heghes and Vaganov) attempting the salutory exercise of forward modeling tree ring site chronologies from climate data. bender quoted from the intro:

Two major uncertainties lie in the statistical development, analysis and interpretation of tree-ring data for paleoclimate studies. First, there are nonclimatic influences on tree-ring records, including tree biology, size, age and the effects of localized forest dynamics [Cook and Kairiukstis, 1990]…. Perhaps of more concern is that tree ring data reflect a nonlinear response to multivariate climate forcings.

and reported:

Read the whole paper for context. This shows the dendros are working on the problem.

I don’t have time to analyze the paper and merely bring it to people’s attention. For reference, the sites in question were said to be 190 MBH98 sites and 8 Russian sites.

Of these, 190 data series for North America are from the Mann et al. [1998] data set. These data were screened a priori for several quality control variables [Mann et al., 2000] to produce a data set most conducive to paleoclimate reconstruction, and represent an excellent target for this study. Data from eight sites in Russia are from published or unpublished data sets developed by Vaganov et al. [1999, 2006, submitted manuscript, 2006].

Although this study is published in an AGU journal requiring data citation, the 190 MBH98 sites are not listed in an SI (nor the 8 Russian sites). There were 232 sites listed in the original SI, of which only 212 sites were actually used. In the location map, I noticed that there were no Alaska sites, so MBH98 Alaska sites appear to have been removed from the population for some reason. In the early screening which reduced the population from 232 sites to 212 sites, sites that were removed in order to “produce a data set most conducive to paleoclimate reconstruction” included the Hart’s Pass WA site (together with other Peterson sites in Washington). I mention this because the Hart’s Pass WA site is the only ITRDB site used by Rob Wilson in Wilson et al 2003 to supplement his own collection – so it would be interesting to know why this site is rejected by one dendroclimatological group and accepted by another. It would also be interesting to know why the Alaska sites were removed, as some of these were presumably temperature proxies.

Their Figure 2 is for the Ulan-Ude, Buryatia region, southern Siberia: 51.8N, 107.6E; 510 m elevation. It’s in the same general location as Barabinsk (57,5N, 97.5E), where I experimented with information on Russian stations, I’ll try to see what station data is available in Ulan-Ude at some point.

In terms of recent debate, the issue is not whether dendroclimatologists are or are not working on their problems, but whether the IPCC is over-selling.

Update: Here are a couple of graphics showing gridded and station data for Ulan-Ude and its gridcell. Several versions of station data are shown here. The station data shows a rather remarkable increase since the 19th century in the GHCN adjusted version (which does not incorporate recent Russian data.)

Figure 1 – Several versions of Ulan-Ude station data

Figure 1 – Several versions of Ulan-Ude gridded data

Reference:
Evans, M. N., B. K. Reichert, A. Kaplan, K. J. Anchukaitis, E. A. Vaganov, M. K. Hughes, and M. A. Cane, A forward modeling approach to paleoclimatic interpretation of tree-ring data, J. Geophys. Res., 111, G03008, doi:10.1029/2006JG000166, 1-13, 2006
url

1. bender
Posted Apr 3, 2007 at 9:20 AM | Permalink

the issue is not whether dendroclimatologists are or are not working on their problems, but whether the IPCC is over-selling

And, if so, “what can dendroclimatologists do to ensure their work is not oversold in glossy color cartoons?”

2. bender
Posted Apr 3, 2007 at 9:23 AM | Permalink

‘welikerocks’ was actually the first to bring our attention to that paper, back when it was an unpublished manuscript.

3. EW
Posted Apr 3, 2007 at 9:25 AM | Permalink

4. Bob Koss
Posted Apr 3, 2007 at 1:09 PM | Permalink

Giss has raw station data on Ulan-Ude for 1886 to 1989. Homogeneity adjusted 1913-1989.
Put this station number 308230000 into the search box found here.

Their annual figures run from Dec of the prior year to Nov of the year they display it.

As usual, the raw data is on average about 0.22C colder than the data adjusted for homogeneity. Wish I knew how they made the adjustment. I’ve compared several sites and it seems the adjustment is always greater further back in time and slightly less in the later years.

5. Craig Hamilton
Posted Apr 3, 2007 at 4:52 PM | Permalink

Have there been any studies done on how tree ring size varies with CO2 levels?

This seems like a fairly obvious question, since these levels have been rising over the past century, which we seem to be using as the calibration period for the climate studies. Since trees absorb the carbon from air to form the rings, I would think that even temperature-stressed trees would have thicker rings in high-carbon atmospheres. How do climatologists compensate for this when trying to predict temperatures from times where CO2 concentration was unknown, but assumed lower than today?

My apology if this has been answered before.

6. bender
Posted Apr 3, 2007 at 4:57 PM | Permalink

C uptake has been discussed before, but may be worth discussing again. My advice is to use the search tool to scan the blog, then ask your question in a thread where the fit is good.

Posted Apr 3, 2007 at 5:33 PM | Permalink

RE: #4 – Application of the HS transform …. 😉

8. John Baltutis
Posted Apr 3, 2007 at 6:49 PM | Permalink

Apparently, my previous post was lost in the shuffle to consolidate threads WRT to dendroclimatologists, so I repost.

In regards to the Evans et al paper, I asked Bender for his take on this:

5.3. Implications for Statistical Paleoclimatology
…in a majority of the tree-ring width chronologies studied here, the decadal variability was not skillfully resolved at or above the 90% level of significance…by either modeling approach. This suggests that much of the decadal-scale variability evident in the tree-ring width data may not be directly related to decadal-scale climate variations.

My take, which may be erroneous, is that both the forward model and linear regression techniques are unskillful and neither approach can separate tree-ring width data into the component decadal-scale climate variables. As cited by numerous others here and elsewhere, until someone can establish a causal relationship between tree-ring widths and the many sources of decadal-scale variation, tree-ring widths cannot be used as proxies for any climate variables, including local temperature.

9. Steve McIntyre
Posted Apr 3, 2007 at 10:59 PM | Permalink

I’ve collated a number of versions of Ulan-Ude station data and a number of versions of gridded data. There are some odd features that I’ll try to analyze on another occasion. There is a HUGE temperature increase in Ulan-Ude temperature data – about 3.7 deg C since 1886 in the GHCN adjusted version. The gridded versions aren’t as big an increase. I’ve plotted an older Gridded version from ndp020 (Jones 1990) and it is out of phase with later versions. It would be a very big job to figure out all the things that they’ve done.

10. Ken Fritsch
Posted Apr 4, 2007 at 9:23 AM | Permalink

Re: #6

Would an analysis of the differences in adjusted station temperatures over time and as compiled by the various “official” groups start with the uncertainities that these various groups claim for their measurements and adjustments and determining how frequently the overlap between groups is zero or by a lower than expected amount?

Perhaps this has already been done and published and there has been a significantly probable overlap or the results have shown a problem with no substantial reaction from the groups or such a study has not been undertaken.

If there are a number of cases of no or little overlap should not the various offical groups be looking to evaluate those differences amongst themselves or perhaps increasing their uncertainty ranges? May be we are just looking at the very worst cases of disagreements?

11. Andrey Levin
Posted Apr 6, 2007 at 5:30 AM | Permalink

Re:5

Craig, apparently your question is not going to be answered by professionals. I asked same question couple of weeks before, and couple of posters asked it since. The answer we got is:
“tree ring temperature proxies DOES NOT account for CO2 fertilization effect AT ALL”

See for reference:

12. bender
Posted Apr 6, 2007 at 7:39 AM | Permalink

Re #11 Your professional answer, which I gave in #6, is that C uptake has been discussed before*. Read the blog. Use the search tool. Temperature reconstruction models do not include effects of C. And this is only natural. It is something that they have only recently begun to examine in the detail required, through the FACE experiments. In terms of statistical apporaches to reconstructing temperature, how the heck are you going to disentangle the independent effects of two inter-correlated upward-trending variables (CO2 & Temp)? You aren’t, that’s how. And that’s why a second-order effect like C is not included in models designed to reconstruct first-order inputs**. And that’s why process simulation models are growing in importance relative to the statistical models.

* See comment #33 at http://www.climateaudit.org/?p=610 (among others).

** Don’t bother pointing out that the argument is circular (“how do you know which effect is first-order and which is second-order?”). Google FACE and read.

13. Mark T.
Posted Apr 6, 2007 at 9:55 AM | Permalink

Good description, bender. I’ve tried to make this point several times since it confounds the use of component analysis methods which typically start with the assumption of at least uncorrelated sources. They aren’t, so those that use these have to assume one has a much larger impact than the other. The assumption to date has been temp >> CO2, without justification.

Mark

14. Paul Maynard
Posted Apr 6, 2007 at 10:05 AM | Permalink

DENDROCHRONOLOGY 101

I’ve swallowed my pride and decided to show my ignorance in public. I’m a regular reader and occasional poster on CA. The threads on the use of tree rings for paleoclimate reconstruction have got to the point where I would appreciate, off thread maybe, some answers to some basic questions. I doubt that I am alone.

1. Mann et al propose that tree rings are good proxies for temperature but there appears to be plenty of work that contradicts this. Yes tree growth is affected by temperature but also and more so by other climatic forces such as rain. So are there some trees that are better temperature proxies than others or do all trees exhibit similar characteristics?

2. Analysis of tree ring growth is I presume based upon cores extracted on site. Are measurements calculated on site or are the cores removed to a lab for analysis?

3. How do the scientists allow for simple measurement error? How do they make the measurements?

4. Is tree ring growth affected by the age of the tree? If yes, how is this adjusted for?

5. Presumably, cores from trees of different ages but in a geographically close group are compared so that the pattern’ is aligned?

6. What are the criteria for the selection of the locations? Does proximity to the tree line influence the sensitivity to temperature? Are tree lines considered to give the best proxies?

7. Has anyone performed a tree ring analysis for Central England where we have a long running set of measured temperatures, whatever their weaknesses?

8. Are the reconstructions that go back to say the 1600’s based upon living trees that old ‘€” I thought most trees lived about 200 years of are they based upon other wood samples?

9. How is the tree rings’ sensitivity to temperature calibrated? Since I presume that the subject trees are in areas where they are no other direct temperature measurements, do they calibrate tree ring growth against some more modern temperature record and then extrapolate backwards?

10. And now my really stupid question. Why is it necessary to perform the statistical contortions to make sense of or extrapolate the meaning from the tree rings? Why can’t the measurements simply be displayed graphically for each tree and group of trees. Is one of the problems that the resolution’ by which I take to mean representative accuracy of temperature change so poor that it can only be extracted by statistical jiggery pokery?

11. And what is the noise’ red or otherwise, that is referred to frequently. Is this the measurement error?

The Mann hockey stick produced a trend that suggested that the MWP and LIA did not happen. But the MWP and LIA are not the product of statistical proxies. They are the result of a very well documented historical record that goes back to the Viking colonisation of Greenland and Vinland in Canada. So why did Mann and the others continue in the face of the evidence and why did the IPCC accept their findings?

Right, enough of displaying my lack of erudition. The BBC were at it again today. (5/5/07) According to their so-called environmental correspondent, the IPCC have established that the temperature is now rising so fast …. you know the rest. Today one of our leading chains of garden centres announced that they were going to stop selling outdoor gas space heaters. The story on Radio 4 was of course accompanied by an expert from some green outfit who said we just had to find out how dangerous to the planet these heaters were etc etc. BTW, the will stop selling them in 2008 when their sticks run out so the end of the world is put off for another year.

More from the BBC today (6/4/07). With the impending publication of the IPCC report (appears to be the second of three) I noticed some more equivocal statements such as the climate maybe warming and it might be man’s thought. But then we had a leading scientist from WWF (experts on CC) whose name sounded like Hansen (not the one from NASA) who opened by saying that the impact of CC was evident now. First example, Gore’s polar bears that are dieing out. At this point I turned off the radio for fear of causing an accident in the car.

I keep adding to this post as my internet connection is on the blink.

Happy Easter

Paul M

PS Insurance industry Windstorm forecasters have continued to warn the industry that an over-active Atlantic hurricane season is on the way.(5/4/07). Can’t wait.

15. Posted Apr 7, 2007 at 4:37 AM | Permalink

Steve, Re your 9, there is relevant info on my USSR pages;
“USSR High Magnitude Climate Warming Anomalies 1901-1996”
http://www.warwickhughes.com/climate/ussr1.htm
Take link to “Lake Baikal Region”
http://www.warwickhughes.com/climate/baikal.htm
Apart from the obvious UHI affected Irkutsk and Ulan Ude, I could see little long term warming.
There may be more stations used in Jones & Moberg 2003 but station ID’s are secret.

16. Posted Apr 7, 2007 at 6:03 AM | Permalink

Re: 14

I’ll restrict my response to stuff I know or reason about.

9. How is the tree rings’ sensitivity to temperature calibrated? Since I presume that the subject trees are in areas where they are no other direct temperature measurements, do they calibrate tree ring growth against some more modern temperature record and then extrapolate backwards?

The basics of the method seems to go like this: Temperature changes cause some of the changes in tree ring widths. That is, we have a relationship

Delta(RW) = f(Delta(T)| other variables)

Thus, the claim goes, there must also be a relationship:

Delta(T) = g( Delta(RW) ) (not causal, just correlational)

Just based on gut feeling, this where I first raise an eyebrow, but let’s continue.

Further, this relationship is assumed to be linear or transformable to linear, and a regression of Delta(T) on Delta(RW) is run in the period where there are both T and RW measurements. Here is another sticking point for me: The regressions do not use Delta(T) (AFAIK; maybe I misread the whole thing), but deviations from some base period mean (some use 1950-80, some 1972-1996). Then, temperature is extrapolated backward in time using this “inverse regression”.

This sort of calibration may be useful in a laboratory setting where the levels or rates of change of all those pesky “other variables” can be controlled, but I am not sure what it tells us here.

11. And what is the noise’ red or otherwise, that is referred to frequently. Is this the measurement error?

No, that is not measurement error. It would be great for all researchers in all fields if measurement error could be white noise, but in most instances, it is not.

White noise is simply random pertubations whose distribution does not ary across time or between observational units with constant expected value and constant variance (normally assumed to be 0 and 1, respectively).

Red noise is usually associated with an AR(1) process whereby the time series of a variable follows a relationship of the form

x_t = x_(t-1) + white noise

Incidentally, there is an interesting paper on this as relates to detecting climate regime shifts: http://www-pord.ucsd.edu/~rdavis/publications/Red_Noise.pdf.

Hope this helps.

Sinan

17. Posted Apr 7, 2007 at 8:23 AM | Permalink

We do, however, believe that our results show that the existence of changes deemed significant by the composite analysis is not evidence for anything more than Gaussian red noise with stationary statistics.

I had this kind of thought in this post..
http://www.climateaudit.org/?p=1285#comment-99188

Wilson 07 refers to Rodionov 2004. Chap 2., Student’s t-test. That had something to do with sample and population means of normal population, if I remember correctly. That’s fine, we are testing i.i.d Gaussian against i..d. Gaussian. I’d be interested to see what happens if the true underlying process is AR1, for example. Many many regime shifts?

18. jae
Posted Apr 7, 2007 at 8:41 AM | Permalink

8:

As cited by numerous others here and elsewhere, until someone can establish a causal relationship between tree-ring widths and the many sources of decadal-scale variation, tree-ring widths cannot be used as proxies for any climate variables, including local temperature.

I think it is very well established that tree rings are useful for drought studies. But I think more and more that the moisture factor generally greatly overwhelms all other factors, and therefore, any temperature signal is way too weak to find in nearly all trees. And, even if there is a temperature signal, you have that upside down quadratic relationship to deal with. And I don’t think that is a small matter, even at supposed “temperature limited sites.”

19. jae
Posted Apr 7, 2007 at 9:13 AM | Permalink

AVERAGE June temperatures in Leadville, CO, which is above 10,000 feet elevation is 10 C. This is about the elevation of supposed temperature limited trees in CO. The average June daily maximum is 24 C, which is at or over the threshold for declining growth.

20. bender
Posted Apr 7, 2007 at 9:58 AM | Permalink

The average June daily maximum is 24 C, which is at or over the threshold for declining growth.

If the “threshold for declining growth” is defined in terms of an AVERAGE (day & night) (e.g. 10C), then it is not correct to compare it to a mean daily maximum (e.g. 24C). What proof do you have that *daily* growth declines when the *daily* maximum exceed 24C? None, I suspect. Your conflation of daily and seasonal time scales appears problematic. A seasonal average max temp of 24C does not have the sample implications as a daily max temp of 24C.

This is what the dendros are complaining about when it comes to “misinformation” and the CA time-sink. It takes gigajoules of energy to continuously correct the misguided analyses by inexpert commenters.

But I think more and more that the moisture factor generally greatly overwhelms all other factors

What one *thinks* is largely immaterial. That you think it “more and more” is equally unworthy of publication. It’s what the data have to say that matters. And when there is a lack of data, which is the case here, the focus must be on quantifying the uncertainty. You keep repeating that temperature reconstruction is not possible, and this is just wrong. It may be impossible to reconstruct temperature with absolute certainty. But nobody’s claiming that. Let’s grow up a little.

Fighting AGW alarmisim with such vague skepticism does no good. What did Steve M say about trying to eliminate the weakest link from your arguments? You always go one step too far, jae. Temperature reconstruction with tree rings is possible. The question is how accurate these reconstructions are.

21. Ken Fritsch
Posted Apr 7, 2007 at 11:33 AM | Permalink

Re: #20

Bender, you are giving all sides in this issue hell of late (not that it is undeserved or inappropriate) but the Gators reign supreme in football and basketball. If that hasn’t improved your dispostion what do we have to look forward to in the future.

I am an impatient man so getting directly to the point is something I appreciate. And please do not hesitate to let me know when I am getting out of hand, even though I am a nice old man who is simply here to learn and thinks that the current Bears’ QB might make me forget Sid Luckman.

22. Steve McIntyre
Posted Apr 7, 2007 at 12:07 PM | Permalink

#20,21. I want to endorse bender’s position here and repeat a point I made in a recent comment. When you post a comment, remember that visiting readers aren’t “fair”. Many third party readers are looking for reasons to dismiss this site. So they will point to some over-the-top comment by one of more zealous readers and say that this discredits CA – regardless of whether it’s a view that I’ve ever espoused. They wouldn’t do that at realclimate – nobody would say that Gavin is discredited or realclimate is conveying misinformation because Lynn Vincentnathan said something silly. But they’re looking to do that here. This site is being judged not just by what I say, but what everyone else says.

While this annoys me, no one ever told us that life is fair. So, folks, let’s understand the situation. People can raise their game in simple ways (and I’m going to keep at people about this and appreciate bender’s comments):

(1) don’t be angry; my own experience is irony is more effective anyway

(2) avoid adjectives – stick to descriptions and facts

(3) don’t spout off. For example, on water vapor issues, I think that there are interesting issues, lots to discuss, but I’m not going to make categorical statements based on what I know and I don’t think that any other readers here should either.

(4) avoid going a bridge too far.

(5) remember that people judge you by your worst point not your best point, So before you push the send button, re-read your post and delete the worst point and the extra adjectives.

23. bender
Posted Apr 7, 2007 at 12:49 PM | Permalink

Re #21
1. I know for a fact that the dendros are visting CA, so it is time to step the level of argumentation up a notch. There are some unanswered questions still that they can help with.
2. The dendros have claimed CA is a tool for propagating misinformation on climate science. I think it is a tool that is subject to the occasional abuse, but I do not for one second believe that is Steve M’s intent.
3. When the discussion focuses on climate physics, I read. When it shifts to dendroclimatology, I write. It is a coincidence that Wilson & Pisaric were discussed when the Gators were busy cleaning Big 10 clock. (Next football season’s going to be a lot of fun.)

24. Ken Fritsch
Posted Apr 7, 2007 at 3:10 PM | Permalink

Re: #23

I agree with yours and Steve M assessments of the discussion with the dendroclimatologists and keeping it on target by avoiding less than substantiated claims as that invariably will allow the discussion to get side tracked — and that does test my patience. I have attempted to listen to and learn from the dendro conversation, although I have not been able to overcome my urge to comment when I see what I think are weak excuses for avoiding some of the hard questions ‘€” and that again tests my patience. On the other hand, like you and Steve M have pointed out, we sometimes have only ourselves to blame.

Re: #22

(5) remember that people judge you by your worst point not your best point, So before you push the send button, re-read your post and delete the worst point and the extra adjectives.

I xxxx xxxxxxxx xxxxxxxxx xxx xx xxxxxxx xxxxxxxxxxxxxxx xxx …

25. Andrey Levin
Posted Apr 8, 2007 at 12:51 AM | Permalink

Re:12

Bender:

Direct and precise CO2 measurements are going on for half a century. CO2 fertilization effect is known for more than a century. Temperature proxies in question are ‘€” what? ‘€” 25 years old? How “it is something that they have only recently begun to examine in the detail”?

The reason why question of CO2 fertilization continues to pop-up on this site is very simple. It is not “second order effect”, as you claim. It is first order effect. There are hundreds of articles on the subject, pointing out that trees are the most responsive to carbon fertilization from all plants, and increase from 290 ppm CO2 to 380 ppm in last century is responsible for 10, 20, 30, for some tree species even 40% and more increase of biomass growth. See for example compilations of Idso with tons of references here:

http://www.co2science.org/scripts/CO2ScienceB2C/subject/f/foresttemp.jsp

http://www.co2science.org/scripts/CO2ScienceB2C/subject/f/forests.jsp

“The next major development in the continuing saga was the finding of Graybill and Idso (1993) that
very long ring-width chronologies (some stretching back nearly 1800 years) of high-altitude long-lived bristlecone, foxtail and limber pine trees in Arizona, California, Colorado and Nevada all developed an unprecedented upward growth trend somewhere in the 1850s that continued as far towards the present as the records extended. In this case, too, like the ones that preceded it, comparisons of the chronologies with temperature and precipitation records ruled out the possibility that either of these climatic variables played a significant role in enhancing the trees’ growth rates, strongly implicating the historical rise in the air’s CO2 content as the factor responsible for their ever-increasing productivity over the prior century and a half.” :

http://www.co2science.org/scripts/CO2ScienceB2C/articles/V8/N16/EDIT.jsp

You wrote:

“In terms of statistical approaches to reconstructing temperature, how the heck are you going to
disentangle the independent effects of two inter-correlated upward-trending variables (CO2 & Temp)? You
aren’t, that’s how.”

Sorry, CO2 and temperature are inter-correlated and upward-trending only in the movies. It is only CO2 which is rising remarkably uniformly. Local temperatures, which only matter to tree growth, are moving on both yearly and on multidecadal levels up and down. The most important thing is that they moving down for some periods. Combined with nearly linear increase in CO2 levels for about 80 years, it is not a problem by using proper statistical tools to disengage temperature and CO2 induced tree ring growth.

Modern stomata proxies of historical CO2 levels are going centuries and even thousands years back, with remarkable resolution and precision. Simple re-run of already existed tree-ring temperature series together with these stomata proxies could yield very interesting results.

Yet it is not done.

26. Paul Maynard
Posted Apr 8, 2007 at 2:07 AM | Permalink

Dear Bender #20

I sometimes think that Steve M should be rechristened St S for his forbearance and willingness to be open to all argument and for his desire to let real science and the scientific method show through. On this measurement then Mann, the IPCC and the hosters of RC are in league with the ….whoops.

Returning to your post “The question is how accurate these reconstructions are.” exactement! You’re implication must be that they have some accuracy but then your assertions are immediately contradicted. I’d like to know how accurate and I’d like to know what type of accuracy in the HS eliminates the MWP and the LIA.

Although the HS appears to have disappeared from AR4, the output of the dendros is still taking prime position. If they are the basis upon which the IPCC is saying that the rate of warming of increased and the consequent actions by loony governments like the one we have in the UK, then they must be challenged and their accuracy’ must be demonstrated.

As an idea to begin with why cannot the IPCC post on one giant website all of the data that supports all of the papers that it has reviewed. I’d also like to know what changes it has made in the reports and what work has been rejected.

Best wishes

Paul

27. Paul M
Posted Apr 8, 2007 at 5:50 AM | Permalink

SUNDAY TELEGRAPH

Excellent article by Bob Carter in today’s ST. You should be able to pick it up on the website tomorrow.

Regards

Paul

28. James Erlandson
Posted Apr 8, 2007 at 6:24 AM | Permalink

Re 27: Bob Carter in Sunday Telegraph. Available here.

Posted Apr 9, 2007 at 10:27 AM | Permalink

RE: #19 – That site appears to be quite similar to Eastern Sierran and White Mountain ones. Average temp is similar to what you find at the so called “latitudinal tree line” and yet, the summer highs tend to be higher. I’ve experienced temps in the 90s at 6 thou, 80s at 10, repeatedly, through the years, in these sorts of places.

30. beng
Posted Apr 9, 2007 at 10:59 AM | Permalink

RE 12: Bender says:

And that’s why a second-order effect like C is not included in models designed to reconstruct first-order inputs

Andrey Levin in #25 is pretty convincing about CO2 fertilization being a 1st-order effect, assuming that the present experimental results are fairly reliable (I think they are). Do you disagree, bender? I understand the potential mine-field of trying to shoe-horn FACE-like experimental results even onto same-species ring-width series. But…

31. bender
Posted Apr 9, 2007 at 11:17 AM | Permalink

beng,
How can I argue this point? After all, *I’M* the one suggesting C (and N, and interactions, and nonlinearities) be included in the modeling! Levin is choosing to take exception to my choice my words, rather than address with my overrall argument, with which he apparently agrees. Why any one would do that is for him to explain, not me. First-order, second-order, to some degree it’s semantics. My point is that statistical approaches are not as powerful as process-based simulation approaches when it comes to multivariate nonlinear interactions. It’s the data, and the models by which the data are interpreted, that will determine the magnitude of these effects, not human rhetoric – a rhetoric that is not at all convincing to me.

32. jae
Posted Apr 9, 2007 at 12:02 PM | Permalink

20: Chill a little, bender. If the dendros consider my speculation as misinformation, then that is a flaw in their logic.

What I was trying to say is that growth is retarded on days when temperatures are above a certain level:

The 12 zones of the map indicate the average number of days each year that a given region experiences “heat days”-temperatures over 86 degrees (30 degrees Celsius). That is the point at which plants begin suffering physiological damage from heat. The zones range from Zone 1 (less than one heat day) to Zone 12 (more than 210 heat days).

If the average June maximum temperature is 24 degrees, then I will bet that temperature on many days exceeds 30 degrees.

33. bender
Posted Apr 9, 2007 at 1:19 PM | Permalink

Chilling.

I agree with your point on 30 as a common daily max when the average daily max is 24. My point is that if this happens on very few days in the growing season, then don’t think that you are on the downside of the *seasonal* parabolic response curve. Remember: it’s the seasonal-scale response curve, not the daily-scale response curve, that matters when you are doing annual-resolution climate reconstructions. i.e. You have to have many, many hot, dry days in the season before you get out to the downside of the parabola on the seasonal response curve. You probably know this, but your comments seemed to suggest otherwise.

34. MarkR
Posted Apr 9, 2007 at 1:20 PM | Permalink

#25 & 21 Bender and Levin

Modern stomata proxies of historical CO2 levels are going centuries and even thousands years back, with remarkable resolution and precision. Simple re-run of already existed tree-ring temperature series together with these stomata proxies could yield very interesting results.

Yet it is not done.

The reason is that to do what Levin suggests would involve formally accepting that temperature changes are independent of CO2 level, something which would be anathema to the warmers.

It doesn’t matter if you use statistical analysis or process based simulation, the CO2 level is independent of temperature.

In fact the only CO2 data they want to use is Ice Core and Mauna Loa, they ignore about 200 years worth of chemically analysed CO2 measurement data that is available.

Why?

Because it shows that temperature changes are INDEPENDENT of CO2, CO2 rise doesn’t cause rise in temperature. End of current Global Warming theory.

35. Willis Eschenbach
Posted Apr 9, 2007 at 2:42 PM | Permalink

Bender and jae, a couple of points about the quote:

The 12 zones of the map indicate the average number of days each year that a given region experiences “heat days”-temperatures over 86 degrees (30 degrees Celsius). That is the point at which plants begin suffering physiological damage from heat. The zones range from Zone 1 (less than one heat day) to Zone 12 (more than 210 heat days).

Saying that plants begin suffering physiological damage from heat at 30°C is a huge oversimplification. There are several different complicating factors:

1. Moisture, moisture, moisture. The number assumes adequate moisture. While in some sites this is common (e.g. Pacific Maritime regimes of the Western US), in other regimes it is not (e.g. Qilianshan Junipers). From jae’s site:

Most important, the AHS Plant Heat-Zone ratings assume that adequate water is supplied to the roots of the plant at all times. The accuracy of the zone coding can be substantially distorted by a lack of water, even for a brief period in the life of the plant.

It is also very site-specific. In particular, the assumption that elevational treelines are “temperature limited” is belied by the fact that elevational treelines are often dry, on sloped, fast draining sites, and with soils with poor water retention. I grew up in forested mountains, and in the summer I’d often see treeline trees losing turgor, or with browning needles, despite temperatures far below 30°C, simply because it was dry up at the treeline.

2. Species. Some trees do well at high temps, others do not.

3. Timing of heat and water. Certain periods of a plant’s life (budding, flowering, etc.) are more critical than others w.r.t. temperature and moisture.

4. Individual plant variations. The site jae cited says “Plants vary in their ability to withstand heat, not only from species to species but even among individual plants of the same species!”

Finally, the underlying assumption that the elevational/latitudinal treeline is “temperature-limited” is an oversimplification. The treeline is the line beyond which the trees cannot survive the combined stressors. These include temperature, moisture, wind, soil nutrients, soil moisture holding capability, partial pressure of CO2, snow cover, slope, exposure, sunlight hours, and rain shadow. Even the response to temperature is not simple, as trees can be stunted by too little heat in the summer as well as too much cold in the winter, or by late or early frosts or hot spells at critical times.

I do not see any way to disentangle these various aspects to get an annual average temperature from tree rings. UC, I think it was, suggested a Kalman filter, and I’m looking into that, but given the generally low cross-correlation between local monthly temperatures, I don’t see how to get to an annual result even if I can get a monthly result.

w.

36. jae
Posted Apr 9, 2007 at 3:40 PM | Permalink

Finally, the underlying assumption that the elevational/latitudinal treeline is “temperature-limited” is an oversimplification.

I totally agree with this. As Veiser has pointed out, most vegetation in the extra tropics is moisture limited. Heat AND low moisture is especially limiting.

37. Andrey Levin
Posted Apr 9, 2007 at 9:44 PM | Permalink

Re #31

Bender:

My apology if I misunderstood you.

As for influence of antropogenic nitrogen emissions (plus non-linear interaction of N with CO2), it is probably better to choose remote from big cities tree locations and leave N aside altogether. There are already too many variables, no need to add another one.

For increasing CO2 ‘€” we can not leave it aside, it is global and ever-present. My impression from numerous articles of Idso is that response for carbon fertilization is basically linear. It simplifies things a lot.

38. Posted Apr 10, 2007 at 3:37 AM | Permalink

#35

UC, I think it was, suggested a Kalman filter, and I’m looking into that, but given the generally low cross-correlation between local monthly temperatures, I don’t see how to get to an annual result even if I can get a monthly result.

I’m not sure if Kalman filter clears this mess but I think that filtering theory is a right direction to go. We can start from simple things, take a discrete stochastic dynamical system

$x_{k+1}=\varphi (x_k,t_{k+1},t_k)+w_{k+1}$

where $w$ is white Gaussian sequence. We want to estimate $x_k$ using noisy observations

$y_k=h(x_k,t_k)+v_k$

where $v$ is again white Gaussian. Here we have a simple model for temperature reconstruction: scalar $x_k$ is temperature at year $k$, and vector $y_k$ contains proxy readings (and instrumental temperatures and so on). Next step is to find functions $\varphi$ and $h$, and variances of $w$ and $v$. If those functions are linear, it is quite straightforward to derive optimal solution, estimate of x along with error covariance matrix. We’ll obtain estimates of past (optimal smoother), current temperature (optimal filter), and future values (optimal predictor).

Like I said before, ( see http://www.climateaudit.org/?p=1342#comment-102410 ), finding the function $\varphi$ is very problematic, and clearly$h$ is not linear. But if the other choice is variance adjustment, I’d go with this.

And with this framework we can talk about climate issues so that even rocket scientists will understand:

Runaway effect – System x is not stable, system x is not controllable

We can’t reconstruct past temperatures with those observations – System x is not observable

We can reconstruct past temperatures, but CIs will go to infinity at some point – P is not bounded

Nonclimatic influences on tree-ring record – Variance and autocorrelation of $v$

Nonlinear response to temperature – Linearity of $h$

..and the list continues. It cannot be something completely different.

39. MC
Posted May 1, 2007 at 7:12 AM | Permalink

A tree ring does not tell you how cold winter was that year, or when it ‘started’ in relation to that tree.
It can only give you how much it grew that year during a certain temperature spell in which it could grow (normally spring and summer)

How is this a good record of average temperature for the year then? Summers could get warmer and be offset by a colder winter.
Tree rings don’t tell you that or at least I haven’t seen a function to tell me that. At what temperature does a particular tree
stop growing and what other environmental conditions are involved? Deciduous and ever-greens being different.

Why are we using them if the basic growth relationships are not well understood. Also why are we teaching kids in tne (English)
curriculum that tree rings are a good indictor of average yearly temperature if at the same time they are being taught that trees
don’t grow in winter?

40. bender
Posted May 1, 2007 at 7:19 AM | Permalink

Re #39
This point has been discussed many times at CA, and the dendros are perfectly aware that trees are at best a summer temperature proxy. Although conditions during winter – wind & UV exposure, snowpack, soil thaw-freeze – can affect summer growth rates. Read the blog. Use the search tool. Tree responses may not be unambiguous, but they are annually resolved.

41. Steve McIntyre
Posted May 1, 2007 at 7:27 AM | Permalink

#40. bender, twq, our Chinese dendro, argued that trees could be a proxy for preceding winter temperature, a view espoused in Gou et al 2007, and,as you recall, got quite cross at any suggestions to the contrary. I agree that the literature is pretty consistent on summer temperatures.

I agree with bender that the annual resolution of tree rings is a feature which should not be dismissed as readily as some readers do. When one sees the dating problems with sediments, it becomes easy to see the attraction of extracting info from tree rings although this does not justify many dendro recipes as we’ve discussed at length.

42. MarkW
Posted May 1, 2007 at 9:39 AM | Permalink

I would say that while temperatures definitely affect the rate at which trees grow, so do half a dozen other things.
Trying to extract the portion of the data that is caused by temperature, to the exclusion of everything else is a difficult trick.
One that the dendros have not convinced me, that they have mastered.