In a recent post here at RealClimate, Simon Lewis wrote regarding a 2010 paper by Samanta et al. on the effect of single-year drought conditions on the Amazon. Samanta et al. claimed to have contradicted a 2007 paper by Scott Saleska et al., and to have thereby overturned some IPCC conclusions.
Lewis showed why Samanta’s paper did not contradict the IPCC, even if it may have correctly identified an error in Saleska et al. Now Saleska has written to say that, actually, Samanta et al.’s results do not identify any error in their work: the results agree completely. With our apologies for the journalistic whiplash, Simon Lewis and I are convinced he’s right. The more general point though, is that the the balance of evidence shows that the Amazon is sensitive to drought, and the IPCC’s statements about it remain valid.
Here is Saleska’s commentary in full
——-
Guest Commentary by Scott Saleska, University of Arizona
The title of the Lewis post (“Up is Down, Brown is Green”) is perhaps even more true than the insightful commentary by my colleague Simon Lewis indicates! The Samanta et al paper says brown, but in fact their own data (when you dig it out of the supplement) shows green, consistent with (and indeed virtually indistinguishable from) our original findings published in Science (Saleska et al., 2007).
Samanta et al. misrepresents our work on many levels (one of which is to assert, falsely, that we did not filter out atmosphere-corrupted observations when in fact we did), and we intend, of course, to present an appropriate response in the peer reviewed literature, where the technical details of our differences may be evaluated by anyone who wishes. But for the moment we will, for the sake of argument, accept their analysis at face value and ask: even if Samanta et al. are 100% correct in their critique of our methods (which we of course dispute), what are the implications? Does the alternative to our method which Samanta et al. advocate, or the recent update in the MODIS satellite data (to version 5 from version 4), make any difference for the main conclusion of our paper? With due respect to our friends and colleagues at Boston University, the answer is no, it does not.
First: the actual relevant Samanta et al data (which comes from their Supplement, Table S3) is this:
| Table S3 (Samanta et al. 2010, supplement) | |||||
|---|---|---|---|---|---|
| Year | Rain defecit (%) | Area Green (%) | Area Brown (%) | Area unchanged(%) | Area with valid pixels (%) |
| 2000 | 0.99 | 5.19 | 6.13 | 23.75 | 35.09 |
| 2001 | 6.09 | 5.15 | 5.68 | 24.24 | 35.09 |
| 2002 | 10.5 | 5.08 | 6.05 | 23.95 | 35.09 |
| 2003 | 5.34 | 8.05 | 4.12 | 22.90 | 35.09 |
| 2004 | 4.68 | 7.56 | 6.72 | 20.80 | 35.09 |
| 2005 | 87.04 | 10.80 | 3.89 | 18.98 | 33.68 |
| 2006 | 26.46 | 4.95 | 3.86 | 26.2735.09 | |
| 2007 | 41.59 | 4.76 | 6.43 | 23.88 | 35.09 |
| 2008 | 18.95 | 3.10 | 6.57 | 25.40 | 35.09 |
Note that the green area in the drought region increases to its maximum (10.8% of the total area = 10.8/33.68 = 32% of the valid area) in 2005. In other words, the Samanta et al data contradict the Samanta et al text and title (which states that Amazon forests did not green up): not only do forests in the drought region green up, they green up alot, more than any other year since the MODIS satellite sensor was launched.
Second, how does this compare to Saleska et al. (2007), which Samanta et al claim to rebut? Here are the numbers (again, taken directly from Samanta et al, Table S3 and Saleska et al., 2007):
Fraction of valid pixels in the 2005 drought region that are “green” (> + 1 Standard deviation)
Saleska et al. (2007): 34% (p<0.000001)
Samanta et al. (2010): 32% (p<0.004)
The bottom line is that their observed 2005 result (32% greenness) is indistinguishable from ours (34%). I.e. Samanta et al effectively reproduce the results of Saleska et al.
This summary response, of course, begs some very interesting questions about tropical forest function under climatic variability and change (indeed the most interesting questions of all!): what caused the anomalously disproportionate green-up in the drought region? And, even if satellite “green up” does in fact represent an increase in photosynthesis (as we think), could this in fact be a symptom of the trees compensating for the increased stress of the drought? The bottom line “carbon balance” of a tree depends on both photosynthetic uptake and respiratory losses, and it is almost certainly the case that those losses (which were not seen by the satellite) increased under the hotter and drier conditions of the drought as well.
Thus, the most intriguing idea to me is that the short-term satellite-detected green-up, and the longer term increase in net carbon loss reported in the Phillips et al paper (discussed by Simon Lewis) are not in conflict at all. It might well be that they represent different parts of a coherent forest response to drought, in which the longer term losses are larger than the satellite-detected attempt to compensate for them by increasing photosynthesis, and in the end, increased tree mortality is the result.
In conclusion I would like to reinforce Simon’s point about Samanta et al and the IPCC. More important than whatever they say about our one short paper, Samanta et al. truly and egregiously misrepresent the implications, of both their work and ours, when they claim that a single paper on short term vegetation response somehow rebuts the IPCC’s review of the large scientific literature on how Amazonia might respond to long-term shifts in the mean climate state. It is an illogical and misguided claim on many levels, one that is already and deservedly attracting the opprobrium of many of my colleagues, talented scientists who study Amazon forests and climate (see Scientists speak: Amazon “myths” are not debunked).
In sum:
— Samanta et al data show a drought region green up that is on average indistinguishable from Saleska et al (but they call it NO green up).
— Samanta et al data almost exactly reproduce Saleska et al’s most salient bottom-line result (but they say what we did was not reproducible).
— the Samanta et al paper, based on a three-month drought response, says not one word about long-term climate change scenarios reviewed in IPCC (but they advertise their analysis as “reject[ing] claims” put forward by the IPCC).
Calculating the standard deviation for AreaGreen/AreaValid, I get 17% +/- 7%.
Therefore p~0.05 ….. how do you find p<0.004 ??
many scientists of Woods Hole Research Center signed Amazon myths are not debunked
and then you do some research…………
For those who are interested, here’s more on the whiplash effect that sometimes surrounds discussions of the greenhouse effect:
http://dotearth.blogs.nytimes.com/2008/07/29/climate-research-media-focus-whiplash/
http://dotearth.blogs.nytimes.com/2010/01/13/climate-whiplash-in-the-greenhouse/
Press offices at institutions can either feed or help modulate this tendency driven by journalists’ normal hunger for “the front-page thought.”
Just today in a chat about another arena – social science work on human responses to disasters – Denis Mileti of the U. of Colorado gave the most important guidance for assessing trends in complex science: “A single study can produce ‘findings,’ but ‘knowledge’ only comes from reading across all the studies in a particular area of inquiry.”
.. Kind of what the IPCC tries to do. ; )
[Response: Andy, thanks for cogent comments. Well put.–eric]
So a drought gets the response from trees that restricted food intake gets from people? If it isn’t too bad, the organism fights harder to live and lives longer? If rain is plentiful, the tree gets a bit lazy? Interesting.
But Samanta couldn’t believe his eyes? Understandable.
More research into the genetics of the trees is called for.
Thanks for the link to the Amazon researcher’s statement.
The only thing the IPCC did wrong was to reference the review article rather than its original citations. Ok, so who hasn’t done this at least once please raise their hand.
Why didn’t Samanta and Saleska work more closely with tropical ecologists? These two articles seem to fall into the trap of reporting a new research method but then using their example data to draw conclusions about a different research question for which it is ill-suited.
The techies and plant press luggers need to get together here.
very good article! thanks for this kind of work!
Sorry its a bit offtopipc:
Lindzen ans Choi have now an response to Trenberth et al. submitted to Journal of Geophysical Research
“On the observational determination of climate sensitivity and its implications” http://www.legnostorto.com/allegat /Lindzen_Choi_ERBE_JGR_v4.pdf
I think it would be of broder interest if you could cover this paper! Its a bit curious, although they state that most of Trenberth et al. points were correct, they raise exactly the same conclusions…
Thank you for writing this. I especially liked the summary points at the end which can be laid out as a quick response that most can easily understand. Nicely done with the take-home message. Perhaps more posts could have summary points to make it easier for journalists to pick up and print?
[Response: Thanks for that suggestion. I think we could do a better job of summarizing the essential points in our posts. I agree that Saleska did a particular good job of this.–eric]
It might be helpful for nonscientist readers if someone could list the coauthors of the various papers in the last five years that all appear to be talking about the same area and using some of the same data sets. It’s normal in the sciences but likely to confuse the heck out of readers unfamiliar with how papers get extracted from data sets and how coauthors get involved.
For example:
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.71.3129&rep=rep1&type=pdf
GEOPHYSICAL RESEARCH LETTERS, VOL. 33, L06405, doi:10.1029/2005GL025583, 2006
Amazon rainforests green-up with sunlight in dry season
Alfredo R. Huete, Kamel Didan, Yosio E. Shimabukuro,
Piyachat Ratana, Scott R. Saleska, Lucy R. Hutyra, Wenze Yang, Ramakrishna R. Nemani, and Ranga Myneni
compared to
Science 26 October 2007:
Vol. 318. no. 5850, p. 612
DOI: 10.1126/science.1146663
Amazon Forests Green-Up During 2005 Drought
Scott R. Saleska, Kamel Didan, Alfredo R. Huete, Humberto R. da Rocha
@Daniel Andrews @8 and eric’s answer.
Coulnd’t agree more, summarizing essential points would be fantastic. Some of us have day jobs and families to prevent us from reading blogs 24/7 you know :) (Same goes for you guys I realize. Anyway thanks for the fine work you all do.
> coauthors … area … data sets
See also:
http://scholar.google.com/scholar?q=samanta+nemani+myneni
Must have been a case of good quality Henna. Goes on Green comes out Brown.
Good post!. I teach AP Chemistry in High School and am putting together a section on the scientific method and the literature. this type of feedback and demonstration of the method of debate in the published literature is a great example. I want to emphasize Andy’s point about findings and knowledge. Of course not all findings are confirmed, some are discarded.
There is a big error in Saleska’s comment, computing 10.8/33 has no meaning, 10.8 is already for the entire forest !
He is assuming that 10.8% is 10.8% of the 33% of the valid pixels, but in fact it is 10.8% of the entire forest !
AreaGreen+AreaBrown+AreaUnchanged=AreaValid !
[Response: This is fine isn’t it? 10.8% of the total forest is definitely green — but it could be much higher since 67% of the forest has invalid pixels. At least 10.8% of the forest is definitely green, and 10.8/33 provides an estimate of the likely total fraction of the forest that is green.–eric]
Definitely well put, Andy. I recall a recent sceptic on Dot Earth saying something to the effect of, well, that’s just a survey of papers, not an actual paper.
It’s long past time to contact Prof. Myneni for his statement on the papers and the press release.
This is the kind of post I can reference to show people how science is done, seriously. Yet more proof that RealClimate is a treasure. It will be interesting to see the responses in comments/peer lit.
When I look at table S3 (Samanta et al. 2010, supplement) I conclude that the suggestion that these forests may be more tolerant of droughts than we previously thought isn’t wrong. Just look at the raindeficits and the green, brown and unchanged area’s. It looks to me that the 2005, 2006, 2007 and even the 2008 measured raindeficits are a bit more than a “light” reduction.
Surely more scientific work is needed to assess what is happening at the base of these forests.
Question for the writers and bloggers who read this: How do you approach the unscientific skepticism (versus useful scientific skepticism) and denialism that now surrounds every conversation available in the public discourse on climate change. Were well past the point where informed individuals accept anthropogenic interference is affecting earths climate cycles: How do you as a journalist/blogger/scientist whatever approach having meaningful conversation and debate on this subject, by which I mean, debate and conversation which we lead to action towards addressing those human caused factors which are making our climate less hospitable towards an advanced human civilization?
As disturbing as it is to think so, this is rapidly getting to the point where charges of research misconduct start to be thought. At a minimum Prof. Myneni has to explain what was done and by whom.
@Benjamin, #14 – You state that the 10.8% is for the entire forest. However, since the sum of green, brown and unchanged percentages equate to the ‘valid area’, it would appear that the calculation is correct although the statement in the text is misleading.
Where it says: “10.8% of the total area = 10.8/33.68 = 32% of the valid area) in 2005” it probably should read: 10.8% of the total valid area = 10.8/33.68 = 32% of the valid area) in 2005.
(If the pixels cannot be read, then they cannot be read as ‘green’. Percentages do need to be clarified to what they refer to, I agree.)
@Ibrahim #18 – I also looked at the following years, particularly the still high reductions in rainfall, and the impact on the forest. Most definitely more analysis linking these observations to what is happening on the ground in the Amazon would be very helpful. I’m wondering how well these techniques for analysis can be validated on the ground in dense, multi-story forests. I understand a lot of work is being done in this regard and would be interested if anyone has any links to papers on same.
Google Results … about 15,200 for samanta amazon green
and counting
Correction: last para should be ‘dense, multi-storey forests’, although undoubtedly the dense forests hold many stories as well :D
#14
Let N = total number of pixels, V = number of valid pixels, G = # green pixels. According to the table, in 2005 G/N = 10.8% and V/N = 33.68%. Therefore, G/V = 10.8/33.68 = 0.32 (the N’s cancel). Saleska’s math is correct — 32% of the valid pixels are green. And, as Eric says in his response, it is reasonable to assume that a similar ratio holds for the invalid pixels.
Eli #20, if Ian Plimer’s book and public pronouncements didn’t attract charges of misconduct, what will? Answer: something that attacks the interests of fossil fuel.
Bodo #7: your URL is broken, should be http://www.legnostorto.com/allegati/Lindzen_Choi_ERBE_JGR_v4.pdf
Regarding comment #18 on precipitation deficit. A point of clarification might help. In all cases (whether for vegetation response as recorded by EVI or for precipitation as recoreded by TRMM), the columns represent the fraction of pixels in the drought region that experience anomalies more extreme than +1 (EVI green) or -1 (Precip, or EVI brown) standard deviations from the mean. On average (with a normal distribution) we expect this to be 16% of the valid pixels (16% more than +1 SD, and 16% less than -1 SD, with 68% in the middle). So years with values more extreme than ±16% are anomalously large, and values less extreme are anomalously small. So the the years in which the precip deficit is less than 16% (ie. 2000-2004) mean there is actually *more* precip than average.
Regarding comment 1 by Benjamin on the statistics: we used a different null hypothesis than assumed by Benjamin, but my post was obviously unclear on the details of the statistical analysis. The post might be amended to contain more detail as follows:
“Saleska et al. (2007): 34% (p<0.000001) (2.2 million km2 of valid pixels in drought region)
Samanta et al. (2010): 32% (p<0.004) (0.7 million km2 of valid pixels in drought region)
Note: Unlike our analysis in Saleska et al 2007, Samanta et al provided no objective criterion for distinguishing green-up from no green-up. So we quantified the statistical significance of observed EVI anomalies, for both their data and ours, by comparison to the null hypothesis that independent forest observations are as likely to exhibit positive (=greening) as negative (=browning) EVI anomalies. P-values are calculated from the binomial probability of seeing at least the observed fraction of positive anomalies by random chance, given the number of independent observations (seeing green vs brown is analogous to thinking of each observation as a coin flip with equal probability of getting heads vs tails).
What is the number of independent observations? The scale at which different patches of forest “greenness” can be treated as independent presumably corresponds to the scale of individual hydrological catchments – possibly larger than a MODIS pixel (1 km x 1 km), but almost certainly smaller than a 1deg x 1deg square (110 km on a side). Conservatively assuming that independence is achieved only as patch area approaches the larger of these, we analyzed the probability of observing the reported fraction %positive/(%positive+%negative) out of the total valid area exhibiting change. (This explains why p-value for the Samanta et al results above shows less statistical power: their method filters out a majority of the forest area, compared to ours, which keeps most of it.) "
(note that this statistical approach is also described in our original Science paper supplement, see: http://www.eebweb.arizona.edu/faculty/saleska/docs/Saleska07.SOM_Drought.Greenup_Science.pdf)
I think Eric (and Sou’s comment at 21) probably already answered Benjamin’s question at #14 about how the fraction is calculated, but just to make sure it’s clear:
Samanta et al data show that 10.8% of the total forest area registers as green. That 10.8% is all in the valid area (as it must be, since by definition we can’t tell whether the area deemed “invalid” is brown or green), so this means that 10.8/33.7 = 32% of the valid area is green. What is the fraction green of the whole forest area? We don’t know, obviously, but logically it could range from 10.8% (if none of the invalid area is green) to 66.3%+10.8% = 77.1% (if all of the valid area is green). In the absence of further information, one estimate follows from assuming that the invalid area is like the valid area, in which case it would also be 32% green. In any case, the objective information we have is that, of the forest that we can observe in the 2005 drought region, almost three times more of it is green than brown, however you choose to calculate the separate percentages.
Scott Saleska says:
21 March 2010 at 5:27 PM
“A point of clarification might help. In all cases (whether for vegetation response as recorded by EVI or for precipitation as recoreded by TRMM), the columns represent the fraction of pixels in the drought region that experience anomalies more extreme than +1 (EVI green) or -1 (Precip, or EVI brown) standard deviations from the mean.”
________________________________________________________________________
Could you please clarify what period/area “mean” refers to? Is it the mean for an individual image, an individual year, or all images from all years? Something else?
Can someone explain why 67% of the pixels are invalid? If an instrument is out or giving invaild readings on 2/3 of its input leads, why is anyone using it to do any kind of reseach?
Similar to the work by Huerte et al (2006 GRL 33: L06405), our earlier work on seasonal leaf phenotypes of tropical rainforest trees showed that greening occurs in several species during and in advance of an ensuing dry season (Graham et al. 2003. PNAS 100:572.) We have suggested that this greening response is the result of programming for two leaf types – a dry season and a wet season phenotype, each provisioned for photosynthesis under sunny and cloudy conditions, respectively. Our studies have shown that cloud cover limits the productivity of tropical rainforest during the rainy season, and it seems reasonable that carbon fixed during the early to mid dry season represents a substantial portion, possibly the majority, of annual carbon uptake. Optimal allocation of resources should result in the production of leaves with higher photosynthetic capacity for use during the dry season, and this is what we have found in trees and understory shrubs. Perhaps I am coming in on this discussion a little late, but it seems plausible that the greening you have observed is a community-level manifestation of programming for the dry-season leaf phenotype. This programming could result from either natural selection or developmental canalization, or both. This response would have limited meaning in the context of resistance to drought. Simply put, the greening may be unrelated to drought resistance, but is instead a response to maximize photosynthesis under clear skies.
It’s great that we are starting to see some real science debate on this site. A paper is publicised that raises some questions, then authors of the paper they were responding to rebut the paper. It’s all very well to insist that all of this should go through peer review but if the other side (a) doesn’t care if something could pass peer review or (b) that a flawed paper can get through the filter, you are in a losing position if you insist on playing the old way.
What I would very much like to see is a break away from the old model of publishing where a paper only sees the light of day after review and editorial corrections, at which point more flaws could be exposed but the fact that the paper is deemed published gives it a gravitas that is hard to shake especially if you insist that “peer reviewed” is a minimum standard.
In an electronic publishing era where versions don’t have to be frozen, as long as you have a way of citing the version you used as an authority, it should be possible for a paper to go through phases of draft for comment, formal review, revision and new revisions if fixable errors are discovered. Strangely enough this is not a new idea: network standards for example follow much this path.
Thin Blue Line…the best of the series so far for basic atmosphere info.
Scott,
I know that the period is far to short to really make any judgements, but I found the behavior around 2005 as interesting as 2005 itself. In particular, there are questions (in my mind) about when you might see more impact from a drought, in the same your or the subsequent year or two. There’s also a hint that the ongoing rain deficits, building on 2005, are causing problems down the road, as the green area drops below 2000-2003 levels, while the area brown increases. I suppose measurements aren’t available before 2000? Sorry, I don’t have easy access to your paper, so if this info is already there, forgive me for asking.
I’d also offer an alternative answer to the “green up” that would be interesting to consider. Not being a biologist, take this with a grain of salt and all that, but… when resources are depleted, I can think of three mitigation strategies; migration, hibernation, and competition.
Migration obviously means leave the area. On the surface, plants can’t do this. In actuality, however, as a species they can by producing more seeds that spread. I swear that the evil pine tree (pine cone city) and nasty oak trees (the litterbugs of the tree world, dropping volumes of leaves, branches and acorns) in my yard respond exactly this way after dry years. I’m definitely wasting more time cleaning up the mess after hot, dry summers.
My thought on migration would be that, to support the plant’s efforts to produce more seeds, the tree also greens up (more photosynthesis). The idea is basically “make kids before I die, so that if things get better, the children can carry on, and possibly farther from this spot, which doesn’t look so great right now.”
The second strategy, hibernation, would involve browning (less greening) in an effort to conserve resources and wait it out. This is what people seem to expect to happen.
The third strategy, competition, would involve trying to get as much of the diminishing resource for oneself as one can, right away, thus depriving competitors, driving them “out of business,” and so having less competition next year, which will be good whether the drought ends or not, but is best if the drought continues… less competition means “more for me.”
Of course, the problem with strategies one and three is that if the drought continues, the surviving trees are in trouble, having depleted a resource that does not replenish as needed. If this is the case, you’d see extreme browning and less greening in subsequent years, when the flaws in the strategies are exposed.
Thoughts?
Another thought…
Has anyone performed a similar study, using the same technology, on other regions (rain forests, temperate forests, grasslands) to compare responses to varying precipitation conditions? It would be interesting to see and compare how a variety of environments react in terms of “greening/browning” to a variety of conditions (consistent rainfall, single drought, etc.).
This may help, I don’t know. (A lot of us readers aren’t Science subscribers so we can’t see the main paper, though we can see the supporting material PDF at the main link).
But as there are a lot of other presentations out there, like this one, perhaps the same information can be found in publicly available form?
http://modis.gsfc.nasa.gov/sci_team/meetings/201001/presentations/land/didan.pdf
One of those slides shows:
Data filtering & Methodology
• Pixel reliability is based on a decision tree that uses the following information (Didan & Huete, 2005, White paper ‐ MODIS C5 planning)
– Pixel QA
– The VI values
– Viewing geometry
• Generates data reliability classes
– Ideal (No issues)
– Good data
– Marginal data
– Cloudy
– Snow/Ice
– No Data
——-
I do repeat, there are far more papers and presentations online pulling information out of what I think is this same set of information than we’ve talked about, with overlapping authors — it’s not ‘camp A’ vs. ‘camp B’ and, like much of the climatology talked about at RC over the years, people contending on one issue may be coauthors on other issues.
Hank, #36,
Excellent find, thank you. Now I need a 34 hour day instead of just 33, to fit this in as well. Way to go. :)
Using a single year’s remote sensing data to draw conclusions about the forest ecology of a massive area with very different forest types and large regional and annual differences in rainfall is not science.
[Response: That’s not what they did and wouldn’t disqualify it as science if they had. Read the papers–Jim]
The authors of these studies have about as much a chance of drawing a correct conclusion as they would hitting a single cowbird with a stone flung at a flock of a million grackles.
[Response: Never trust anyone that says this or that study, published in a scientific journal, ‘is not science.’ I have colleagues that do this too; it’s a petty put-down, and it’s point-scoring, but its not an argument.–eric]
I think there is a major factor here that is being ignored: do these different authors even use the same definition of “the Amazon” i.e. regional area?
A guy at Discovery News has been writing some good commentary on climate issues:
Regional Rainfall in a Warming World
Analysis by John D. Cox
Fri Mar 5, 2010 05:15 PM ET
Slowly but surely, a picture of climate change at the regional scale — where it really matters — is beginning to take shape.
Drought is indeed linked to precipitation changes, so you would think, rather than relying on a single source of data, the authors would have done the usual thing, looked at multiple types of data, not just satellite models. Since they didn’t do this essential ground-truthing, which would involve a lot of fieldwork all across the Amazon in fairly difficult conditions, this work is mostly schlock, on either side. Most people know you have to do satellite analysis of photosynthetic production very carefully to get anything like a reliable answer – “color green” alone doesn’t do it.
Furthermore, the short time period makes it even less likely that the study is applicable to climate change – RC authors know that year to year variability in temperatures is no indication of climate trends; neither is year-to-year variability in carbon fixation and respiration.
[Response: Pretty strong words there. First of all, the fact that these two papers use remote sensing methods to evaluate forest conditions is not in itself worthy of criticism. Satellite image analysis is an increasingly vital information source for monitoring the planet, including its vegetation. Other folks, such as Oliver Phillips and Simon Lewis (and others) are working on the ground- and biometrically-based methods of change detection, and still others are immersed in the flux tower and micro-meteorological work. Like the parable of the blind man and the elephant, each group is looking at a part of the whole picture, and reporting what they find with their approaches. Synthesizing a common understanding from the various, and sometimes contradictory data, that emerges, then follows, in the normal process of science. As for year to year variation, there is no one inherent temporal scale deserving all the attention at the expense of others–all are worthy of study.–Jim]
In other news, the British government has appointed a former Shell executive to head their “climate email science panel” – and this is another one of those who promote the fraudulent carbon capture schemes so heavily:
http://news.bbc.co.uk/2/hi/science/nature/8579929.stm
CCS boosters are appointed to investigate climate scientists for fraud over stolen emails related to tree ring data sets…
Will the also appoint a chemical engineering group to investigate CCS boosters for fraudulent scientific claims, which is a whole lot easier to demonstrate? Somehow, I doubt it.
After all, CCS is the #1 greenwashing propaganda line in the fossil fuel lobby’s arsenal – so why does RC boost it as well?
[Response: Do we? I certainly don’t! Most of us at RC are pretty skeptical about these sorts of ‘solutions’.–eric]
[Response: Baloney–I’ve seen no one here proposing that, and I’ve explicitly commented that I’m against it in relation to other C mgt strategies.–Jim]
Still hoping for an explanation on why 67% of the pixels are invalid.
[Response: Because the EVI data had to fall within areas defined as qualifying as both drought affected and undisturbed forest. Perhaps there were cloud and cloud shadow effects as well.–Jim]
Another question would be of the remaining 33%, unless the pixels are very carefully selected, how can the vaild pixels be a representation of the whole area? Perhaps it may be a good sample for parts of the Amazon, but are you telling us random chance gave a representive sample for the entire Amazon? Somone with a statistics background may wish to leap in at this point but I don’t see how the sample can be valid without knowing how the invaild/valid pixels are distributed, or that the data means anything at all. Soory to be a wet blanket but will be happy to educated.
[Response: The probability of sampling one third of your analysis area with imagery and getting a strong bias wrt to vegetation response to env. drivers is very low.–Jim]
Hmmmm, backtracking from the link I posted above, I find this pair of papers has been on an agenda. I wonder if there’s also a poster/PDF for Myneni in the same file somewhere? (Google searches happen on this kind of thing)
http://modis.gsfc.nasa.gov/sci_team/meetings/201001/landAgenda.pdf
So there’s been some discussion among the MODIS people. These two are on the agenda side by side.
Ranga Myneni, BU
‐ Amazon Forests Green‐Up During 2005 Drought ‐ Fact or Fiction?
Scott Saleska, AU
‐ Amazon Forests During 2005 Drought: MODIS collection 5 confirms Green‐up fact
Googling finds the Myneni presentation too:
http://modis.gsfc.nasa.gov/sci_team/meetings/201001/presentations/land/myneni.pdf
——————-
One might speculate …. but that would be wrong.
[Never trust anyone that says this or that study, published in a scientific journal, ‘is not science.’ I have colleagues that do this too; it’s a petty put-down…not an argument.-eric]
Agreed – just say their timescale isn’t applicable to their conclusions, so it’s a bit schlocky.
[Response: To who’s? Second time you’ve said this now. Saleska et al were studying the effect of severe drought in one year, using imagery and precip. station data. Their time scale is what it is.–Jim]
However, the study may still be useful if their datasets are of high quality, right? Often, science is really more about getting the data and/or material samples. The ice core archives produced by Lonnie Thompson’s expeditions to high altitudes & the many Vostok and Greenland cores – well, they have to write that up, submit grants for more research funding, but that’s almost incidental to the data collection effort. Notice here that data collection alone – without analysis – is also “not science”, but it’s a critical component. The same goes for satellite data. One good rule is never trust conclusions about complex systems like the Amazon based on a single type of data.
[Response: Nobody is, as I discussed in your previous comment. And the satellite data is extremely high quality data.–Jim]
This is also very true of any kind of paleoclimate reconstruction. The underlying problem? Well, one likely explanation why there’s only one dataset is that the others didn’t agree with the author’s bias/agenda/previously published work, etc.
[Response: Or maybe that they can’t launch their own Terra/Aqua/Landsat with ETM or MODIS??–Jim]
Don’t fall into that trap – apparently, the entire “climategate inquiry” (run by Shell) rests on the tree ring dataset, which was truncated early as it deviated from the warming trend later on. It would have been far better to include all the data – then you could speculate as to why. Was there a CO2 fertilization effect? What tree species were you looking at? etc. If you brush the dirt under the rug, someone will come across it sooner or later.
The underlying issue, climate change forced by changes in atmospheric composition -> radiative forcing -> convective feedbacks -> cryosphere/ocean/precipitaion -> biosphere/agriculture feedbacks, is completely unaffected by the political wrangling and propaganda warfare, however.
I accidentally left out the precipitation projections – take a look – the lowland Amazon has the largest projected drop in rainfall, but the higland area indicates increased precipitation.
Here’s the reference for that figure:
Zheng et al. (2010) Indian Ocean Dipole Response to Global Warming: Analysis of Ocean–Atmospheric Feedbacks in a Coupled Model
ABSTRACT
Oh, heck, why not, I’ll speculate.
I think we’re seeing tip-of-the-iceberg still and that there are a whole lot of papers and presentations, public and private.
I think the real issue being discussed is the satellite imagery, how well it can be used, and whether it can eventually replace fieldwork on the ground.
So I think we’re seeing a little bit of what nonscientists often don’t see at all — typical scientific hard argument with much back-and-forth, and we’re seeing more of it than usual because, well, the world has changed and people can look stuff up.
And I’ll further speculate that someone lacking sufficient perspective on how back-and-forth happens over time as scientists work, possibly a new grad student, was (ahem) framing this in terms of “war” or “team sport” rather than being aware that over time what we have is usually “shifting alliances of fiercely individualistic scientists.”
I’d bet some such confused soul mistook the most recent paper for a heart’s desire final nail in the IPCC’s coffin, got in a position to draft the press release, blew that badly including the misquote that’s been silently disappeared from BU’s press release, and _maybe_ also hyped the original to Hannity and maybe others we don’t yet know about (tho’ the gmail address reeks of misdirection to me).
Whew. Okay, that’s all wild speculation on my part, no cite for any of it.
Anyone got a mapping tool that can track how this stuff spreads over time by detecting links? It’d be fascinating:
http://www.google.com/search?q=samanta+amazon+green
I have tried to make a summary of what I have been reading in the threads “Up is down, brown is green” (including the guest commentary from S. Lewis) and “Saleska responds”:
Recently, three articles have been published on the effect of short-term extreme drought (in 2005) on the Amazon forests. The conclusions from these articles appear to be rather contradictory, to say the least. Phillips et al (Science 2007) claimed a massive tree mortality that would temporary change the forest from a CO2 sink (2 billion tonnes absorbed yearly) to a CO2 source (3 billion tonnes released). Saleska et al. (Science 2007), however, found a large-scale green-up and concluded: “Coupled climate-carbon cycle models suggest that Amazon forests are vulnerable to both long- and short-term droughts, but satellite observations showed a large-scale photosynthetic green-up in intact evergreen forests of the Amazon in response to a short, intense drought in 2005. These findings suggest that Amazon forests, although threatened by human-caused deforestation and fire and possibly by more severe long-term droughts, may be more resilient to climate changes than ecosystem models assume.” And recently Samanta et al. (GRL, 2010) disagreed with Saleska’s conclusion because they found “no evidence of large-scale greening of intact Amazon forests during the 2005 drought” and they added that the changes that they found “are also not unique – approximately similar changes are observed in non-drought years as well.” In other words, nothing much had been changing.
To make things even more complicated, Saleska now comments that in his view the results from Samanta are NOT different from his own earlier conclusions on the increased greenness in 2005. But regardless of whether there was an increase in greenness or no change at all, it seems to me that in both cases this is very difficult to reconcile with the claim made by Phillips et al. about massive tree mortality and strongly decreased CO2 consumption
It is therefore hard enough to bring together the conclusions of these three publications. But Simon Lewis wants us even to believe with respect to all three articles that “Overall the conclusions in the IPCC 2007 Fourth Assessment Report are strengthened”. Apart from the fact that the IPCC report was dealing with possible effects of LONG-TERM droughts, how can an increased greenness support a projection of strong deterioration ??
[Response: Couple of points. First, you are looking at the normal process of earth/ecological science–different research groups approaching complex, large scale phenomena with different data, tools, approaches and specific questions. Sometimes the results are disparate and thus require either more evaluation or new explanations. Read Stephen Mulkey’s post #31 here, and also the work of Nepstad and others that forms the basis of the IPCC AR4 statements, and also the recent statement made by Nepstad and others regarding this whole topic, the link to which now escapes me.–Jim]
Orwell would probably have used the following Newspeak line: Brown or Green, the IPCC is always right!
Ok, my comment was trash, I admit that. Definitely not the first time I’ve gobbed up a blog post and probably not the last.
Here is my criticism of the work. I read the Saleska paper when it was first published, so it’s been a while:
The paper proposes an effect based on a single occurrence of drought. This is problematic because vegetation response to rainfall differs dependent upon actual soil moisture conditions which are not always cleanly predicted by rainfall amount.
[Response: Saleska et al. discuss this, with reference to a site where the seasonal course of the soil water potential at depth was monitored in relation to the seasonal precipitation.–Jim]
They are further complicated by the effect of temperature, prior moisture conditions, and prior seasonal growth conditions (overall plant condition going into the drought) on the observed plant response. In other words, there is a lot of natural variability to take into account and drawing conclusions with such a short data set is may provide an answer totally at odds with reality.
[Response: The only conclusions they are really drawing is that there is an apparently enhanced chlorophyll (“green-ness”) signal associated with a drought event, which is counter-intuitive.–Jim]
I see a lot of criticism from climate scientists to those who would make future predictions based on short temperature, rainfall, etc. records due to natural variability. The same standard should have been applied to this paper.
The authors propose some mechanisms for the effect based on the site’s characteristics. These do not include an ecological mechanism, yet the conclusions suppose to report an ecological response of the forest to drought.
What would I have done differently?
I’d have included more data or at least more supporting ground information that related the remotely sensed data to observed plant growth responses so that a cause and effect was elucidated (ie. what caused the green-up observed – vine growth, more leaves, more chlorophyll, what?). I’d have included more background information regarding the targeted forest community types and what was known about climate constraints on plant growth, survival and reproduction.
[Response: People need to understand that these are big, complex topics and no one group or person can do more than a part of what’s required for a coherent picture to emerge. This is hard work–Jim]
I very much appreciate this guest post which provides a viewpoint from an ecologist, but I don’t think his conclusions or criticisms of the papers are being addressed in the comments or replies to comments.
#43 Hank
Maybe it an overzealous grad student was involved – for instance, that appears to explain the Marengo quote which was a surprise to everyone, including Marengo.
But none of the authors have done anything to correct misimpressions. In particular, senior author Myneni apparently told the National Post’s Terence Corcoran that the IPCC Amazon statement was “alarmist”, and hasn’t bothered to correct any of Corcoran’s other misinterpretations as far as I know.
So your explanation may be valid in part, but appears incomplete.
Oops, I mean Hank #43.
And here’s the National Post link:
http://network.nationalpost.com/NP/blogs/fullcomment/archive/2010/03/12/terence-corcoran-remember-amazongate.aspx
I feel like i am missing something with respect to the claims that the new paper by Samanta et al contradicts IPCC claims about the vulnerability of the Amazon to drought. In fact, it would seem to me that their results are *more* supportive of the IPCC report statement that the Amazon may be highly sensitive to drought.
Saleska et al (207) were the ones whose results hinted that the forest could be more resilient to extreme (if short-lived) drying than many ecosystem models suggested. IF anyone’s results were to be used to wag fingers at the IPCC, it seems like it should have been theirs. If Samanta et al’s results show that the forest did not green up during the 2005 drought, then their results tip the balance back toward the Amazon being very sensitive to rainfall and, as the IPCC report indicated, prone to conversion to savanna in a warmer, drier tropics. I really don’t understand why they (or their press department) think that their results undermine the IPCC statements.
The other question I have (perhaps RC could solicit a post from the Samanta, et al authors?) is, if they believe the Amazon didn’t green up during the 2005 drought, do they have similar uncertainties about the seasonal patterns of enhanced greenness that were described in the even earlier paper (Huete, A.R., Didan, K., Shimabukuro, Y.E., Ratana, P., Saleska, S. R., Hutyra, L.R., Yang, W., Nemani, R.R., and Myneni, R.(2006) Amazon rainforests green-up with sunlight in dry season. Geophysical Research Letters, 33, L06405, doi:10.1029/2005GL025583)?
Some of the authors who are now disagreeing about the impact of the 2005 drought were co-authors on the seasonal cycle study, which seems to be based on the same/similar data. How does the current disagreement affect the interpretations of that paper?
[Response: I had the same initial response as you. It’s all pretty confusing frankly.–Jim]
From the Saleska et al. Science abstract:
“These findings suggest that Amazon forests, although threatened by human-caused deforestation and fire and possibly by more severe long-term droughts, may be more resilient to climate changes than ecosystem models assume.”
This statement directly conflicts with Dr. Lewis’ findings and those of the scientists who recently signed a statement disagreeing with reporting that the IPCC was flawed (i.e. Amazongate). Doesn’t it? Or am I missing something (totally possible – not a rhetorical question)?
The study uses green up as an indicator of health. I believe that it may supply data that backs up the hypothesis that the forest is sensitive to variations in rainfall. But I don’t believe the observed sensitivity indicates the forest is or isn’t resilient against changes in species composition (forest health) caused by climate change.
[Response: It’s not attempting to do that (certainly not addressing compositional changes, which at any rate do not equate to “forest health”). It’s simply describing green-ness with respect to drought–throwing some observations out there if you will.–Jim]
For example, the rainfall deficits continue past 2005 and the amount of greening steadily drops in response. What is going on here? I can’t tell without a lot more data, but it doesn’t seem like a simple relationship.
I have experience conducting vegetation studies in seasonal, deciduous swamps (subtropical). These areas respond very differently to changes in rainfall, often in complex ways. For example, if we have a dry winter and the swamp dries down early, leaf flush occurs earlier as leaf growth is positively affected by the lack of soil anoxia. If we get enough rain to keep the soil moist during the summer and fall, then the swamp will stay green (I have no idea what the leaf color signature MODIS would see, but I assume at least green for longer). But is this good for the swamp forest? If moist but less inundation persisted, the forest would obviously change to become dominated by a drier plant community as these exist just slightly uplope of the swamp. A very fine gradient from swamp to dry forest is observed and is driven by average inundation period (soil conditions change little in permeability and moisture retention and are probably not a factor).
[Response:
Your example is exactly why the ground-based demography work of Phillips and others, in relation to climatic monitoring, is critical. Neither of the remote sensing papers are attempting to address these longer term questions you discuss.–Jim]
Also rainfall wouldn’t be the best climactic factor to use to correlate forest change with, rather duration and intensity of soil anoxia and moisture deficit are.
[Response: I already mentioned that Saleska et al provide evidence of the relationship between gauge data and soil water (and satellite data)–in their supplemental naterial, which should be available.–Jim]
In drier climates just to the south, such swamp forests don’t exist. In these areas similar landforms (large depressions with impermeable soils that are inundated by rainfall) contain marshes with shrubs instead of forest as the extreme swings between long inundation during wet periods and multi-year dry periods (which don’t occur to the north in the swamp forests) preclude tree growth.
At least where I work I can make a prediction that less annual rainfall or longer drought periods will equal less swamp forest. I would question anyone using remotely sensed greeness data to say otherwise regardless of whether it was publised in Science.
[Response: Again, they’re not addressing long term type conversion.–Jim]
Ecology is the dominant journal in my field. Tropical Ecology is another. Science – not so much.
Comment from Jim (#44)”.. you are looking at the normal process of earth/ecological science–different research groups approaching complex, large scale phenomena with different data, tools, approaches and specific questions. Sometimes the results are disparate and thus require either more evaluation or new explanations.” I agree completely, but if the results are completely contradictory then it is not easily understood that they are assumed to fit in one pattern, and together would support the IPCC projection. Green IS green, and is not the same as dead trees.
[Response: I agree–that’s why the public statement linked to in the post discusses the primacy of ground-based demographic census data in relation to the effects of drought on tree mortality.–Jim]