http://dotearth.blogs.nytimes.com/2011/04/26/on-plankton-warming-and-whiplash/

On Plankton, Warming and Whiplash

Last summer, a paper published in Nature, “Global phytoplankton decline over the past century,” caught the attention of climate campaigners and some media outlets because it concluded that warming of the seas over the last century was linked with a big and near-global decline in oceanic phytoplankton.

If that finding stood the test of time, it would indeed be momentous; the vast clouds of tiny photosynthesizing organisms in the seas are an important part of the carbon cycle and underpin the marine food chain. There had already been some work, on short time scales, pointing to a blunting of plankton productivity in warmer periods.

Now, three “brief communications,” essentially rebuttal papers, have been published in Nature pushing back strongly against the paper’s core conclusion. Links to the summaries are below. I’ve queried the authors of those papers and the original analysis and will post an update when that discussion begins.

Climate naysayers have latched onto the rebuttals just as breathlessly as climate hawks latched onto the initial, if tentative, findings.

What’s up? Science.

I’ve used a variety of analogies to describe the ugly process by which knowledge evolves, and devolves — a process including lots of tussles and negative learning (often the more you study, the less you know firmly).

A new finding or theory is like a lamb shank lowered into a piranha pool. Competing teams nibble away until only the sturdy bone is left. A rugby scrum forms over the ball and strangely lumbers up or down a field even as heaps of burly players fight epically for the advantage. It’s a journey, with the state of knowledge held in dynamic tension, either crystallizing or dissolving depending on new sources of data or new ways of thinking. What’s your favorite description of the scientific process?

The phytoplankton analysis, by Daniel G. Boyce, Marlon R. Lewis and Boris Worm at Dalhousie University, drew on a variety of types of data, from thousands of water transparency readings taken since the end of the 19^th century by lowering a white disk into the sea to direct and satellite measurements of chlorophyll. The conclusions were powerful and disturbing:

The long-term global declines observed here are, however, unequivocal. These results provide a larger context for recently observed declines in remotely sensed [chlorophyll] and are consistent with the hypothesis that increasing ocean warming is contributing to a restructuring of marine ecosystems, with implications for biogeochemical cycling, fishery yields and ocean circulation. Such consequences provide incentive for an enhanced in situ and space-borne observational basis to reduce uncertainties in future projections.

I held off writing about the finding last year because, after I sent it out to a batch of marine scientists for reactions, they expressed deep reservations about several aspects of the research. I didn’t want readers here to suffer from whiplash journalism (or science).

Here’s one of the cautionary notes from last August, from Paul Falkowksi at Rutgers University (it was a response to the group):

Dear All

Inspection of the data reveal (not too surprisingly) large gaps is several areas of the oceans and I seriously doubt their conclusion that phytoplankton biomass declined by 40% over the past century.  Were that so, we almost certainly wouldn’t be seeing the deoxygenation of large areas of the open ocean today.  Moreover (and I am surprised that Marlon Lewis didn’t bring this out) the loss of chl (were it real) would decrease the rate of warming in the upper ocean.   Further, I didn’t find the same trend in my analysis of the long term trends in chlorophyll from the North Pacific (my paper with Cara Wilson, Nature 358, 741 – 743 (27 August 1992).  Moreover Venrick et al reported a large increase in chl in the central N. Pacific gyre starting around 1978 – the causes of this increase ([i]f true) have never been resolved (Science, 238, Oct. 2, 1987, pp. 70-72)

So, while the Boyce et al paper is certainly provocative, I would wait another several years to see what the long-term trends in chlorophyll are from satellite retrievals of ocean color.  Here is a link to another paper with a completely opposite view.

I shared a batch of such reactions with the authors of the Nature paper and Boyce graciously sent a detailed reply. I should have posted on the exchange at the time, but better late than never. If you read Falkowski and Boyce, along with the summaries of the new challenges to the plankton paper below, you can see how the pushmi-pullyu dynamic of science works.

The first followup analysis — Is there a decline in marine phytoplankton? — comes from 14 marine scientists at institutions in England, the Netherlands, the United States, Canada and Bermuda. They describe a variety of long-term data sets on phytoplankton that were not included in the original analysis and show plankton trends in the opposite direction:

Eight decades of data on phytoplankton biomass collected in the North Atlantic by the Continuous Plankton Recorder (CPR) survey², however, show an increase in an index of chlorophyll (Phytoplankton Colour Index) in both the Northeast and Northwest Atlantic basins^{3, 4, 5, 6, 7} (Fig. 1), and other long-term time series, including the Hawaii Ocean Time-series (HOT)⁸, the Bermuda Atlantic Time Series (BATS)⁸ and the California Cooperative Oceanic Fisheries Investigations (CalCOFI)⁹ also indicate increased phytoplankton biomass over the last 20–50 years. These findings, which were not discussed by Boyce et al.¹, are not in accordance with their conclusions and illustrate the importance of using consistent observations when estimating long-term trends.

The second paper — “Does blending of chlorophyll data bias temporal trend?” — is by David L. Mackas of the Canadian Institute of Ocean Sciences and concludes that “some or much” of the trend identified by Boyce and his co-authors could be the result of problems arising from drawing on two disparate types of data. “Although real changes may have occurred, their proper quantification requires removal of the bias component,” Mackay concludes.

The third brief paper — “A measured look at ocean chlorophyll trends,” by Ryan R. Rykaczewskiand John P. Dunne – echoes Mackas’s analysis and concludes rather bluntly: “Our results indicate that much, if not all, of the century-long decline reported by Boyce et al.¹ is attributable to this temporal sampling bias and not to a global decrease in phytoplankton biomass.”

The eagerness to find the “Pearl Harbor moment” or line of evidence that jogs people to act on the long-term risk of human-driven climate change, combined with the “tyranny of the front-page thought,” will long cause the kind of reaction that the initial plankton paper engendered — and that past papers on frog extinctions, Atlantic Ocean currents, ice-sheet behavior, hurricane dynamics and other facets of the climate puzzle have done.

It’s useful to step back and take the long view on such work, if only to avoid neck pain.

This practice can also help sustain faith that the strength of science, including climate science, lies in such struggles over new ideas and information.

Postscript: Ryan R. Rykaczewski, an author of one of the new critiques, and Bryan A. Franz, who co-authored a Nature commentary accompanying the Boyce et al. paper when it was published, have contributed fresh thoughts on the trio of new studies:

Franz:

The Boyce et al. analysis and subsequent communications highlight the challenges to interpretation of historical proxies for phytoplankton biomass. The Siegel and Franz letter simply emphasizes the new challenges associated with measuring changes in phytoplankton biomass in the satellite remote sensing paradigm, where sampling frequency is much improved, but limited time-series duration and uncertainties associated with instrument calibration and data processing algorithms complicate the interpretation.

Rykaczewski:

My thoughts on this issue are mostly consistent with those expressed by Falkowski and Antoine below. Boyce et al. should be complimented for their effort to address a pressing issue and for their use of a historical data set, but the main finding of their paper is not robust.

Mackas and Dunne and I noted the significant bias between the Secchi disk method and in situ fluorometry which, when combined with a shift from one sampling technique to the other during the 1900s, induced a spurious centennial scale trend in the data. This emphasizes the importance of consistent measurement techniques (or a well understood calibration between similar techniques) when investigating long-term trends. A precious few observational programs have made sampling consistency a priority during the last several decades; these programs were noted by McQuatters-Gollop et al. in their comment. So together, this trio of ‘brief communications’ articles acts to underscore the value of long-term, consistent observations of the marine environment.

Conducting regular observations is not as simple as it sounds. Standard (or perhaps ‘boring’) observational programs are not considered to be as gratifying as short-term, innovative projects, yet these observational programs require continuous funding even during financial crises. The value of these historical records increases over time, but it takes prudence to justify such records in a budget year after year. Additionally, observational programs are faced with the challenge of sustaining historical time series by sampling with seemingly obsolete instruments in order to maintain comparability over the decades while also utilizing the newest technologies. The fact that biological records tend to be representative rather than definitive also burdens ecosystem studies. Measuring physical properties such as temperature or salinity is relatively straightforward in comparison to biological properties such as biomass or productivity. Although biologists have been studying ecosystems for hundreds of years, methodologies have changed frequently and remain difficult to standardize globally.

Concerning your question about the confidence I have in Boyce et al.’s original conclusions: I have no confidence that the analysis demonstrated a global, long-term decline in chlorophyll over the centennial scale. Marine biogeochemists and biological oceanographers have a good understanding of the processes which control phytoplankton biomass, such that a decrease in global phytoplankton biomass of the magnitude described by Boyce et al. cannot be explained by other physical, chemical, and biological changes which have been observed. Secondly, I expect that the changes in instrumentation, inaccuracy of a single Secchi-depth-to-chlorophyll conversion, and variation in sampling effort prevent identification of trends over the last century. Relatively few regions of the ocean were sampled prior to the 1960s, and Boyce et al.’s extrapolation of apparent trends in these limited regions to draw conclusions about a global trend was particularly disturbing.

Using a relatively primitive (and inaccurate) instrument to estimate long-term chlorophyll changes would require frequent sampling over the period of interest. I am aware of one paper from the CalCOFI program which did utilize regular Secchi disk measurements to carefully discern a slight, long-term shoaling of Secchi depth since the 1940s (Aksnes and Ohman, 2009), and these authors were cautious not to attribute this changes to any single factor such as a long-term increase in phytoplankton biomass.

I do have confidence in Boyce et al.’s conclusion that large-scale climate variability (e.g., El Niño Southern Oscillation, North Atlantic Oscillation, and others) has influenced chlorophyll anomalies since the 1980s. This was noted in two paragraphs in the latter half of the original paper. Over this shorter time period, sampling has been comprehensive enough to accurately estimate chlorophyll variability related to climate phenomena. This was not the case in earlier decades. Boyce et al. are incorrect in suggesting that the ability of their dataset to represent this seasonal to interannual variability adds confidence to their main result. The ability of a sampling method to accurately measure seasonal variability does not indicate that the method is valid for estimating trends over centennial time scales.

I asked Rykaczewski whether the issues with the original paper reflect problems with the process of peer review and he added this note:

I prefer not to speculate on why the original paper was published by Nature, but I think it does hint at issues with the editing and peer-review process. On the other hand, Nature was willing to publish these additional comments and deserves credit for recognizing the level of skepticism within the oceanographic community concerning the original, highly publicized conclusions. Our present conversation gives me some hope that the peer-review process is not completely defective.

John P. Dunne, Rykaczewski’s co-author on the plankton critique, added this thought:

The only point I might underscore beyond Ryan’s comprehensive summary is in relation to your question of what kind of information we would need to have in order to corroborate a 40% decline in phytoplankton chlorophyll over the last century. On that issue, the potential complexities regarding phytoplankton community composition, ecological interactions, and anthropogenic impacts (such as changing iron supply, pollution, fishing) make definitive statements extremely uncertain (and I would guess one of the sources of humility in the reviewers that would allow them to acknowledge the possibility that the Boyce result was possible). However, physical drivers such as light supply, nutrient supply, and stratification are known to have an extremely important role on determining phytoplankton chlorophyll – dominating it on seasonal to inter-annual time scales. If one frames the question differently, and asks whether there is any known physical climate mechanism that might have caused such a precipitous global decline over the last century, the answer would be definitively ‘No’. The climate signal in our current generation of earth system models doesn’t really kick in for another few decades, as was discussed in a recent comparison of satellite and model-based chlorophyll:

Henson, S A., Jorge L Sarmiento, John P Dunne, L Bopp, I Lima, S C Doney, Jasmin John, and C Beaulieu, February 2010: Detection of anthropogenic climate change in satellite records of ocean chlorophyll and productivity. Biogeosciences, 7(2), 621-640.