Worst graph ever

ResearchBlogging.orgA couple of scientists at the University of Montana say they have detected a small but non-negligible decline in global terretrial “net primary production.” NPP is basically a way of measuring plant growth — how much carbon they’re removing from their surroundings and turning into biomass. To my mind, there are two noteworthy aspects to their research, which just appeared in Science. Both led to me to the phrase that is the title for this post, although each use carries distinct meanings.

First, “Drought-Induced Reduction in Global Terrestrial Net Primary Production from 2000 Through 2009” begins by pointing out it used be thought that as carbon dioxide levels in the atmosphere rise, plants were expected to grow more. That idea has long since been undermined by evidence that there are other growth-limiting factors, although it still sees the light of day in pseudoskeptical arguments against doing something about anthropogenic climate change in the form of “CO2 is plant food.”

Indeed, write the authors, Maosheng Zhao and Steven W. Running, NPP “increased from 1982 through 1999, in part due to eased climatic constraints on plant growth.” But not anymore.

The past decade (2000 to 2009) has been the warmest since instrumental measurements began, which could imply continued increases in NPP; however, our estimates suggest a reduction in the global NPP of 0.55 petagrams of carbon.

That works out to a decline about about 1%, which doesn’t sound like much. But the idea that NPP isn’t rising anymore, despite an annual growth rate in CO2 emissions of 2 or 3%, is troubling. If these findings, which are based on data from satellite, are verified by other researchers, we can add falling NPP — with all the disastrous consequences for agricultural output and wildlife habitat — to the list of things to worry about if warming trends continue. This could prove to be an important piece of research. News of a worst kind.

The second thing that caught my attention is the first graph in the paper. Here it is. Please look closely at it before drawing any conclusions:


On first glance, it looks like NPP and CO2 levels are closely correlated. Which they are, which is why Zhao and Running plotted them in this fashion. It’s the sort of thing that good scientists do: choose a visualization technique that best expresses what one is trying to communicate. The troublesome part is the decision to invert the CO2 scale. This makes sense if the primary message to be conveyed is absolute correlation between plant growth rates and CO2 levels. Of course, the correlation is negative — the more CO2 rises, the more NPP falls. The graph’s legends states this explicitly and the right-hard y-axis is inverted. But given how easy it is to misinterpret even data that’s presented straight-ahead, you can pretty much guarantee that this graph, either in whole or cleverly edited or cropped, will find its way into presentations and blogs as evidence that NPP and CO2 are positively correlated and so is plant food after all.

So 9 out of 10 for good intentions, but minus several million for poor style.

Perhaps it is unfair to ask every scientist to take into account how their work will be distorted by those with political agendas when crafting their papers. Surely they should be encouraged to always use the most precise and scientifically appropriate methods, those that maximize the signal-to-noise ratio, to describe their work? There’s merit to that argument. But I can’t help thinking about Randy Olson’s book, Don’t Be SUCH a Scientist. On a matter as important as climate change, it wouldn’t hurt to think a bit beyond the lab when writing up one’s research.

Zhao, M., & Running, S. (2010). Drought-Induced Reduction in Global Terrestrial Net Primary Production from 2000 Through 2009 Science, 329 (5994), 940-943 DOI: 10.1126/science.1192666

24 Replies to “Worst graph ever”

  1. Yikes! Dreadful graph. No, I don’t think it’s too much to ask any scientist who’s involved in research that touches on global warming to think seriously about how their findings will be presented to the public. Both because they can predict that someone will mis-represent it if that’s advantageous, and because they should be wanting to communicate as clearly as possible to the general public on this issue.

  2. I like the graph. Maybe it’s my background in archeology: We invert stuff all the time.

    An XY plot, however, is what is needed here to address the correlation, along with an appropriate regression model.

  3. Stuart Saniford has some other concerns. I think his point is a good one. I agree that it’s important to get the paper out there – as you say, even if NPP is flatlining in response to increased CO2, that’s concerning enough that we should be following closely. But aren’t these findings somewhat premature – or shouldn’t the tentative nature of the findings have been made more explicit?

  4. I must say, the trees I’m seeing this summer (driving 8,000 km across Canada and back) look unusually unhealthy with bare twigs, leaves growing mostly near the tips of branches, and not nearly dense enough. In addition, I’m noticing a lot of dead trees and dead branches on sick-looking trees. I’ve been noticing for about twenty years that trees are looking more and more translucent throughout the summer, without dense leave growth, but the symptoms are accelerating. I don’t know if it’s the unusual heat, a new form of acid rain, or what. It’s not confined to any one species. Trees in sheltered and damper areas and some trees growing on people’s lawns (more fertilized and watered?) are a little more likely to look better–not up to standard, but better. I fear that in three to five years some trees won’t leaf out at all.

    Any thoughts?

  5. Saniford indeed makes some good points. Most interesting to me is the comment from Running that he and Zhao, and Science, it would seem, believe the issue is of such importance that including uncertainty ranges in their estimates of NPP trends was not necessary. My point is, if you think it’s so darn important as a public policy issue, why not pay attention to how the information is presented to the public?

  6. I don’t see anything wrong with the graph. It conveys the information it’s intended to convey clearly. If graphic representations are going to be criticized over how they may possibly be misinterpreted, the same criticism could be leveled against any communication.

    I believe that marine NPP is also falling due to AGW. Warmer, less dense surface water is preventing the upwelling of colder, more nutrient rich water, leading to a decline in phytoplanktonic NPP. These trends in both terrestrial & marine systems are potentially very troubling.

  7. The word ‘Drought’ in the paper title is important.

    Plant net productivity should drop during periods of extreme hydric stress. And it does. 2002 was a record year for high temperatures and drought in many places – including the US and Canada, part of on-going record dry period that began as early as 1997 and continued through mid-2006, with a break in 2000 and 2004, years which were still warm, but wet, with a concurrent drop in the number of fires and increase in productivity due to release of biomass accumulated and soil organic carbon nutrients into soils from fire activity.

    Record warmth years also drives energy use for cooling, and because the majority of energy is petroleum fuel supplied, that also contributes to jump in anthropogenic CO2-inputs.

    Atmospheric CO2 went UP due to an jump in the number of very large arboreal fires that occurred during this period of hydric stress – particularly in the period 2000-2004. Savannah burning is also correlated to water availability, vapor pressure, and nutrient availability (discussed in detail in the somewhat suspect 2006 paper in interannual fire variability 2000-20006; there has been quite a bit of technical discussion of MODIS data filters and comparison of ASTR and MODIS fire detection datsets at about the time this reference paper was published).

    Net primary productivity gains in the previous decade has been ascribed to decrease in cloud cover, primarily in the tropical rain forest, with an increase in UV/solar radiation. However, there has been some discussion in the global climate change literature on the possibility of aerosol reflectance inducing artifact in satellite detection of cloud cover.

    This is at the periphery of my expertise; my interest lies in environmental effects of fires and other aerosol-forming processes.

  8. Isn’t there a dimensional problem here? As in your discussion, I’d expect primary productivity to be related with the amount of atmospheric CO2. But why should I expect it to be correlated with growth in atmospheric CO2, as the graph has it?

    Put differently, you restate the graph’s point as “the more CO2 rises, the more NPP falls”, but according to the labels the graph seems to be saying “the faster CO2 rises, the more NPP falls.”

    I’m missing something which is likely glaringly obvious to those in the know.

  9. My main concern with the graph is that there are no error bars. Is it possible to measure NPP this precisely? Secondly, I totally agree with Douglas McClean, what could be the cause of this correlation between CO2 rise and NPP?

  10. It’s not a very good graph…it tells the story and obscures it at the same time. Surely, the uninverted graph combined with a footnote giving a correlation coefficient would be as useful?

  11. Surely this graph must show that in years with low NPP less carbon is taken up by the biosphere and thus more stays in the atmosphere. As Douglas points out an effect of CO2 on NPP should depend on CO2-concentration, not growth rate.

  12. While I agree with Thomas on the probable direction of the causal effect, the graphing technique itself is a perfectly kosher way of illustrating anticorrelation between two time series, and the legend clearly labels one scale as being inverted.

    I have a couple of beefs with the layout (lack of horizontal grid lines and symbols that would be hard to distinguish in a black-and-white printout), but by and large those are minor nitpicks.

    – Jake

  13. I agree with Thomas and Jake. To the OP: I don’t think this graph means what you think it means.

    Of course, the correlation is negative — the more CO2 rises, the more NPP falls.

    It is probably more accurate to say that the more NPP falls, the more CO2 rises. This is because the NPP removes CO2 from the atmosphere, so if there is less NPP by plants, more CO2 sticks around in the atmosphere. Thus, the graph *does* show that CO2 is plant food. Sorry.

    (It doesn’t disprove global warming, though, and I agree that this graph might be misused by climate skeptics.)

  14. It would have been just as clear and far more intuitive to leave the data as is, with no inversion, and let the relationship be revealed by the mirror imaging of the lines.

  15. I agree with Curt F that the graph might be used by climate skeptics – as an example of scientists who are so convinced of the bad effects of CO2 that they cannot even interpret their own data properly. Yes obviously the graph shows causality running from NPP to CO2. If you wanted to provide evidence the other way you wouldn’t remove the trends. You would want people to see that the increased CO2 levels in 2009 were associated with lower NPP. But what if the data didn’t show that? Then you might remove the trend and just compare the anomalies, giving your graph a scary anti-skeptic interpretation. Do that and you have something you can publish.

    I am not an AGW skeptic. But I am rapidly coming to the view that climate science is about as respectable and apolitical as is nutrition science.

  16. I think that the graph is drawn properly. There is no defense against liars, and no point in attempting to obstruct those who deliberately abuse the truth. If you write “NPP is negatively correlated with CO2”, what’s to stop them from reporting that “NPP is correlated with CO2”? The world is a complicated place and if you phrase your statements in a way that cannot possibly be twisted, you end up sounding like a politician. Compare the amount of trust that the public has in politicians with the trust they invest in scientists, and you can see why it’s best to just tell the truth as clearly as possible.

  17. Perhaps it is unfair to ask every scientist to take into account how their work will be distorted by those with political agendas when crafting their papers.

    No, that’s a job for tech-writers. /shameless_plug

  18. In my opinion, James is reading the graph wrong and his power of suggestion has confused the readers. 🙂 Read the graph from left to right like most Americans are accustomed too. That is, if NPP rises, you’d expect CO2 rates to decline because plants are absorbing less of it … and that is what the graph shows. The inverse is also true. If NPP declines, you’d expect CO2 rates to accelerate … and they do. James read the right axis first and misunderstood what it means (IMO). The right axis doesn’t represent absolute CO2 growth, but growth related to NPP, which we’d expect to be inverted.

  19. This is the sort of article that makes me want to puke. Have another look at the figure. No error bars. Suppose they had started in 2001 instead of 2000: the whole signal goes away. Look at hemispheres in suppl material. If they had only the northern hemis they would say growing. (that fig is misleading further in that the scales are different.) One could go on and on.

  20. Raging Bee is right. We need professional writers, preferably technical writers, to translate “science language” into plain talk. One of the great pleasures of reading Einstein’s Relativity is that the writing is so clear.

  21. It would be interesting to see what their findings will be now. Also how the use of gas engine vehicles affect this planet and what people can do to stop it. Limousine services in Chicago are still using 4.6 engines, even the new 2013 MKT Lincoln town car.

Comments are closed.