19 October 2008

Discussion: A role for atmospheric CO2 in preindustrial climate forcing

Some climate spinners are no doubt having a field day with the van Hoof et al. paper Steve Bloom pointed us to. It does, after all, say something critical of the IPCC. But if you read the paper itself, you will see that spinners shouldn't be happy; a conclusion I get in reading the paper is that climate is less sensitive to solar and volcanic variations, and CO2 is more variable than previously thought.

Let us take a look at the content. I hope you already have, as I encouraged when Steve first mentioned it. In doing it myself, I'm largely reading it as a nonspecialist. While I have studied more things outside physical oceanography than typical for a physical oceanographer (on the scale a things, by the way, a fairly broadly educated bunch), the biology of plant stomata is not on that list. On the other hand, a good knowledge of how science works gets you pretty far, and you don't need to be a scientist for that.

The central experimental idea is that the density of stomata in plant leaves goes up when there is less CO2 in the atmosphere, and down when there is more. (Stoma being 'mouth' and stomata being a bunch of mouths -- the leaves breathe air in through these mouths.) The Oak (genus Quercus) has been used to infer older CO2 levels before, and these authors do so again. It's more to the novel side that they're trying to infer much shorter term variations than has typically been done before. But even that (their citations 20-24) is not entirely new by now. If you can find Oak leaves (say in swamps) and date when they're from, you then have a way of reconstructing past CO2 levels in the atmosphere that is entirely independent of ice cores. You might also have a method which doesn't have the averaging and delay problems that ice cores have. On the other hand, you have a method which probably has other problems. (All data have problems. That isn't the question; whether the problems affect your conclusions is the question.) The prime novelty in this paper is now to apply the method to the last 1000 years and consider what it may tell us about the climate system.

A quick question is how reliable the method might be. In results and discussion, and figure 1, we get an idea. On the counting stomata side, we're looking at something whose standard deviation ranges up to almost 18 ppmv (parts per million by volume -- the usual unit for CO2 concentrations; the v is often left off), with an average over the whole set of about 6 ppmv. Since the authors are looking at their largest signal being 34 ppmv, the 18 ppmv standard deviation is not small, though the 6 ppmv should be good enough. So we'll set a reminder to ourselves to see if the conclusion depends sensitively on this '34'. Turns out that a conclusion relies on 34 being different from 12; so the 6 ppmv standard deviation is definitely good enough.

In figure 1b, we're shown the regression scatter plot between stomatal index and CO2 values. Though it isn't mentioned this way, eyeball suggests that the CO2 levels inferred have less scatter at low CO2 levels (higher stomatal index levels). My being a nonprofessional, though, means that there might be some obvious, to a professional, bit of biology which says that I'm over-reading the graph. Taking the figure as given shows why there's typically a substantial standard deviation in the inferred CO2 levels -- the stomatal index doesn't have a tight correspondence to CO2.

So we have a bit of a concern about how faithful a record the oak leaves give. The authors address this by looking at how the oak leaf record they construct compares to the ice core record from Antarctica, after processing it through a filter which has a similar averaging and delay behavior. The result (figure 1d, red vs. blue curves) shows pretty good agreement. Though there's some wide error bars to use (the gray band), the two curves from wildly different sources agree pretty well in the few decades and up time scale. In saying 'few decades and up', what it means is that every single bump on the curves don't line up exactly. But if you look at averages over 30-50 years, they do compare pretty closely. Obviously an area for research is to attack exactly why there are any differences at all.

Now let's turn to the significance of the work if we take the reconstruction as given. As nonprofessionals, we can't go much farther now with whether we should do so, but we've got some pointers to ourselves about what to look at further if we were of a mind to pursue it. The ice cores, partly due to their time-averaging, only show a variation in the period discussed (1000-1500) of at most 12 ppmv. That, versus the 34 ppmv difference the authors find from their different recorder -- one we might expect to have much finer time resolution. If free variations of CO2 are much larger than previously thought, then more of the climate would be CO2-driven than previously thought. This continues an idea (as far as I know) William Ruddiman started (see citation 13). Tripling the CO2 contribution in the pre-industrial period also means that prior estimates of climate sensitivity to volcanoes and solar variations would have been overestimates.

This is why folks who are going to leap on 'IPCC was wrong' parts of the paper really shouldn't be happy. The conclusion is that climate is less sensitive to solar and volcanic, and that the natural carbon cycle is more prone to variation. Independently of any of this, however, we know that the recent 100 ppm rise was due to human activity. (See Jan Schloerer's CO2 rise FAQ if you wonder about this.) Whatever caused the 34 ppmv variation observed in this study hasn't yet been going on.

Now let's see if we know anything outside the paper that is relevant. The parts inside look reasonable. The prime thing which struck me is the 34 ppmv variation, occurring in only 120 or so years (1200 to 1320 or so by eye). That's a lot of CO2 to be released in a short period by natural means. The authors mention that it's contemporary with a warming of the North Atlantic, which is the right sign of change -- warmer water holds less CO2. But they don't present a quantitative argument about the water being enough warmer over a large enough part of the world to have released that much CO2. Time and space are limited in a PNAS paper so this isn't the issue it would be in a different source. But it's something to pursue. Prompted by this, though, I arrive at a different biological question. Or geological; or both. That is, the places where Oak leaves get buried in such a way that you can dig them up 1000 years later to analyze stomata have to be pretty special. Could such a special locale exert a local effect, say that when it's warmer stuff decomposes faster -- releasing more CO2 but only mattering to local trees? If so, then some portion of the 34 vs. 12 ppmv signal is a local effect, and the discrepancy between leaves and ice cores is reduced. Then again, this may not be a factor. Not knowing the biology leaves me in that position of needing to find more informed sources.

To that end, I'm sending this blog note and address to the corresponding author and inviting him to respond either on the blog or by email. Hank Roberts (you've seen him here a few times) brought up the idea elsewhere, that it might be a good thing for people to let authors know when their science is being discussed in a blog. Thanks for the idea Hank.

You should also see a new icon next to this note. It's from Research Blogging, and the idea is to tag blog notes about scientific papers. Then readers who'd like to see research-oriented blogging can go to the main site and have a summary feed of such postings.


Anonymous said...

I wonder what potential implications might be for ice core results, particuarly if glacial-interglacial variations in CO2 are effected significantly.

Robert Grumbine said...

I don't see anything drastic for the ice cores CO2 data themselves. We already knew that they were time-averaged. This note suggests that the time averaging is fairly accurate as compared to the leaves. What's new is to add -- with data -- that there are significant changes at the shorter periods that the ice cores average out. So now we have to consider CO2 as well as solar and volcanic effects for shorter term (decades) pre-industrial climate changes.

A few things should happen now. One is, the Antarctic ice cores are known to have particularly long averaging periods. There are other ice cores, though. Greenland has better time resolution. So now there's an additional reason to go back and analyze their CO2 records in the most recent few thousand years. Another is that some people will sit down to answer that question of how 34 ppmv could move in to or out of the atmosphere in only 120 years. This should lead us to other things to go look for (ocean temperatures warming and cooling significantly in short periods, maybe forestation/de-forestation or other land changes, ...) And another is that more people will go analyze their own Oak, and other, leaves from more parts of the world to see if they, too, show such large and rapid variations.

These multiple responses are a sign of a good science paper. In addition to saying something interesting, it raises some questions that will take more work to answer, including from other areas of science. (Ok, this question-raising is actually part of what makes a paper interesting.)

Anonymous said...

There are reasons to doubt the accuracy of the results of the stomatal studies. I commented on earlier work of some of the same authors here:


Robert Grumbine said...

Tom van Hoof's response to my blog note:

Dear Dr Grumbine

Thanks for letting me know about the blog. I think you've hit the nail with respect to summarizing my paper. I agree that the uncertainty of the proxy is a weak spot in order to give exact estimates. Biological proxies have in general a wider uncertainty level than physical proxies due to the fact that almost no processes in organisms are exactly linear or act on a one to one basis. Next to our smoothing model exercise with
the ice core we also compare to another on stomata based record from the USA to support our conclusions.

Answer to your question about the local effects: The environments in which these oaks grew are not very special, the are normal river swamps. To cope with within canopy CO2 variations we only use leaf morphotypes which come from the top-crown of the tree (sun-leaves). The way trees respond to short term CO2 alterations (within season) by opening and closing of stomata not alterations of the density/index of stomata. The average growth season CO2 concentrations are expressed in the leaves next year's stomata indices. So what I try to say is that the stomatal density/ index is an expression of the previous year's average CO2 concentration. Short term fluctuations (within a year) are coped with by opening and closing of the stomata. On average the stomata based CO2
reconstructions tend to have higher average CO2 (~300 ppmv) than ice core studies. This could be an effect of location... But we do not know yet. This is the reason we only look at rates of change rather then absolute values.

I fully agree with you that we do have comments on the IPCC BUT this paper does not deny climate change or global warming !!! My personal opinion is that this black and white view on the climate discussion which has been going on for a couple of years now is extremely harmful
to the scientific process. In science it almost never occurs that one theory is fully right, but more often new findings lead to new questions which lead to new views. This is what helped mankind forward. The blocking of the scientific debate by putting scientists in a corner; "does he support global warming or not" in my opinion is very wrong. In this paper we do have some comments on the way IPCC uses the outcomes of various studies, and in our view has a very selective way of quotation. On the other hand the results of our paper do also show that you cannot say that one other forcing factor is responsible for climate change during the past 1000 years. The most logical way is that climate is
driven by the sum of all forcing components so solar, CO2 and volcanic combined and acts with complex feedback mechanisms.

I hope I sort of answered your questions...

Best Wishes
Tom van Hoof

Robert Grumbine said...

Speaking as myself again ...

Dr. van Hoof invited me to fix grammar and typos as he had written his reply quickly. (We've already had two back and forths even though he couldn't have received my email until late Sunday/early Monday his time.) Any errors remaining, or introduced, are mine.

I was surprised that you can tell the difference between top of tree leaves and ones from lower down. But that's one of the reasons you want to talk to an expert!

But as to form of his answer, I wasn't surprised, though many nonscientists might have been. There was no I'm the expert, how dare you ask questions, or the like.

And you'll notice that he mentioned himself a point of concern that I had not raised -- that the CO2 inferred from stomata is biased high compared to other observations. It is mentioned in the paper too. I didn't mention it myself because, as he mentions, he works with the changes in CO2 rather than the absolute values. Still, we want to understand why that offset exists.

Anonymous said...

My main question is: why would this compare at all with ice cores? I thought it was well known that CO2 can vary rapidly and frequently, especially if there are tons of humans around.

The method in itself is interesting, and gives indeed (perhaps) a better high-frequency signal than ice cores, but we might want many of such reconstructions from a lot of places on earth, in order to get a planet-wide variable that then becomes comparable to ice cores.


Robert Grumbine said...

Any new method, for anything, is always (ok, I'm sure someone will find an exception to these absolutes) compared to the old method. In this case, the comparison tells us that the longer term average of the leaf indices is behaving like the ice cores do. Since we believe the ice cores are doing a good job on the longer term averages, this gives us some confidence that the leaf index is doing reasonably well at least on the long term average. If it's doing reasonably well there, where we can test it, then we have better hope that it is doing well in the higher frequencies, where we can't.

We still, as you mention, and van Hoof did, want to test against more leaf records. he only could test against one. We'll want many more, from much more of the world and from many different species.

The thing about good science is often not that it closes a question, but that it opens one. There are reasons, see tamino's note and the discussions there, to question the leaf records. But the correspondence between leaf and ice core suggests to us that the leaves are recording something important. It certainly warrants more people going out to collect and analyze more leaves. This wasn't so obvious 5 years ago.

William M. Connolley said...

I'm still a bit concerned by the stomata - icecore differences. You suggest that fig 1d - blue vs red - shows good agreement. To my eye, it shows red (the stomata proxy) wih noticably higher variation - looks to be about 2x. I don't see them discussing this issue.

Anonymous said...

The main point of discussion between the ice-core community and stomata workers is the amplitude of decadal CO2 fluctuations. Ice derived data leaves only room for CO2 fluctuations of up to ~12 ppmv in some and not all ice cores. Stomata data reveal higher amplitude CO2 shifts, not only in the past millennium but during the whole past 10.000 yrs. This specific study is significant because this is the first time that we have 3 CO2 datasets (2 stomata, 1 ice core) that all have the same timing of a CO2 change during the past 1000 yrs. The stomata data match the ice core data when firn densification based smoothing in the ice core is taken into account. This suggests the ice data underestimates the variability of the actual atmospheric CO2 signal due to local diagentic effects. This is the evidence put forward to claim that Co2 was more viariable naturally, and hence should have played a role in climate forcing as well. Note that a lott of climate model output which are used to support low air-temperature variability during the past millennium (hockey stick debate) are constrained by the low level CO2 dynamics reconstructed by the ice data. Proof of a higher dynamic CO2 regime therefore removes the constrained on past millenium air-temperature variability.

Hank Roberts said...

A question for anyone who might know -- would higher and lower CO2 measurements in a forest, compared to a polar icecap, suggest an actual flow of CO2 going on, out of or into the soil in which the trees are growing? I'd expect higher/lower measurements close to a large source/sink, I think.

> the difference between top of tree
> leaves and ones from lower down.

On this I vaguely recall (and will look for cites) that when the first "canopy crane" research started, this turned up evidence that maximum size old growth forests were in fact very actively respiring and growing, but that was only happening up in the treetops; the researchers looking at the trees lower down, from the ground or low platforms, just hadn't gotten up high enough to detect it.

And on the practical issue of more measurements of more leaves -- I have a fair number of black oaks (N. Cal) and I could start collecting leaves, even from the treetops, given instructions on how to do it. It's a little 200-year restoration hobby project I stumbled onto in my copious spare time, and I hope to hand it on. I'd be very happy to get a baseline (whatever that means for these purposes) to tempt someone later to continue collecting a time series.