SEARCH

SEARCH BY CITATION

Keywords:

  • dimethyl sulfide;
  • climate change;
  • ocean ecosystem;
  • CLAW/GAIA

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Our Modeling and Results
  5. 3. Analysis and Discussion
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[1] Dimethyl sulfide (DMS) is one of the major precursors for aerosols and cloud condensation nuclei in the marine boundary layer over much of the remote ocean. Here we report on coupled climate simulations with a state-of-the-art global ocean biogeochemical model for DMS distribution and fluxes using present-day and future atmospheric CO2 concentrations. We find changes in zonal averaged DMS flux to the atmosphere of over 150% in the Southern Ocean. This is due to concurrent sea ice changes and ocean ecosystem composition shifts caused by changes in temperature, mixing, nutrient, and light regimes. The largest changes occur in a region already sensitive to climate change, so any resultant local CLAW/Gaia feedback of DMS on clouds, and thus radiative forcing, will be particularly important. A comparison of these results to prior studies shows that increasing model complexity is associated with reduced DMS emissions at the equator and increased emissions at high latitudes.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Our Modeling and Results
  5. 3. Analysis and Discussion
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[2] The Southern Ocean is a setting for strong teleconnections among the systems of global climate change. Primary production, carbon drawdown and convective return of nutrients are planetary in scale in this region [Longhurst, 1998]. Large areas of oceanic surface waters are severely iron limited such that atmospheric dust inputs and human intervention may both be capable of modulating major element geocycling [Gabric et al., 2010; Wingenter et al., 2004]. The mid-latitude westerlies force the Antarctic Circumpolar Current (ACC) and generate Ekman upwelling which may drive Antarctic ice shelf loss [Alley et al., 2008], and perhaps also play into the global meridional overturning [Toggweiler and Russell, 2008]. Sulfur cycle climate feedback linkages, such as the CLAW hypothesis [Charlson et al., 1987; Erickson et al., 1990; Gabric et al., 2001] are also largest in the Southern Ocean because sea-air transfer of dimethyl sulfide (DMS) tends to rise with biological productivity, and the source of sulfate aerosols in the atmosphere over the Southern Ocean is dominated by oceanic DMS emissions (as opposed to anthropogenic sulfur emissions).

[3] Models for the distribution of marine DMS have lately been increasing in number and complexity such that a regional portrait of their evolving climate response is constructible. We present here initial results of the consequences of climate change from the most sophisticated ocean sulfur cycle model yet reported, as well as the apparent connection between the sophistication of the model and the predicted response of the sulfur cycle to climate change. The various models, grouped into three generations based on their complexity and formulation, are described in the auxiliary material. The changes in their DMS emissions in response to climate change are somewhat varied, and are summarized in Table 1.

Table 1. Annual Average Increase in DMS Flux Going From a Present-Day Climate to a Future Climate, in 10° Latitude Bands, for the Models Described in the Auxiliary Materiala
GenerationReferenceDegrees South LatitudeInterpretation
80-7070-6060-5050-4040-3030-2020-1010-Eq
  • a

    The percentage changes are taken directly from text interpretations in the original work wherever possible, and rounded to the nearest 5%.

  • b

    ML stands for ‘mixed layer’.

  • c

    No zonal integrations presented so that samples were taken along meridians central to the Pacific, Atlantic and Indian basins.

  • d

    Their run duration was fifty years, which therefore had less greenhouse gas buildup than the other models (see Table S1).

  • e

    Zonal average flux perturbations reported most directly in the text.

  • f

    Parenthetical values indicate interpolation.

  • g

    Checks performed against zonal concentration integrations.

1stGabric et al. [2001, 2003] +30 +5    Ice cover domiates
2ndGabric et al. [2004] +50+105+30+10+5+5+5MLb changes dominate
2ndVallina et al. [2007]−50+5+500+5+5MLb changes dominate, notesc,d
3rdBopp et al. [2003] 0(+10)+30+10(0)(−10)−15See text, notese,f,g
3rdKloster et al. [2007]>+30+10−2000−10−10−10See text, notese,g
4thThis study+170+70−15+50−10−10−10See text.

[4] Unfortunately, it is hard to determine the best model using observations because experiments using seawater from different ecosystems with elevated CO2 and/or temperature show varied responses too [Lee et al., 2009; Kim et al., 2010; Hopkins et al., 2010; Vogt et al., 2008; Wingenter et al., 2007]. Most of the experiments showed increases in DMS concentration, but a major complicating factor in understanding those results is that the studies each strained out different sizes of the mesozooplankton.

2. Our Modeling and Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Our Modeling and Results
  5. 3. Analysis and Discussion
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[5] We report here on the first marine sulfur simulations performed in the Community Climate System Model (CCSM) [Collins et al., 2006]. We used the most recent version of CCSM available at the time, which was an unreleased version between v3.5 and v4.0. CCSM contains the Parallel Ocean Program (POP) [Smith and Gent, 2004] as its ocean general circulation model and utilizes a carbon/nitrogen/silicon/iron ocean ecosystem model with biological resolution exceeding any of the other models discussed in this paper (diatoms, coccolithophores, diazotrophs and collective picoplankton are all distinguished [Moore et al., 2004]). We then attached the DMS mechanism developed and validated by Elliott [2009] for global flux studies, which includes the high sulfur producer Phaeocystis, and is described in the auxiliary material. Though still highly parameterized, this DMS model constitutes a steady advance in detail, even relative to the latest studies, and we consider it to be a fourth generation model.

[6] The version of CCSM we used had not been fully tuned and exhibited a large Arctic cold bias that resulted in unrealistically extensive and persistent sea ice. However, comparisons of the climate in the Southern Hemisphere with results from standard CCSM3 simulations (including sea ice extent, SST, zonal wind stress, Drake Passage transport, and a number of other quantities) were sufficiently good that we feel justified in presenting our results for the Southern Hemisphere in this paper. Furthermore, even though we expect the high northern sulfur chemistry to be just as compelling scientifically as that of the Southern Ocean, the DMS effect on cloud brightness is often overshadowed in the northern hemisphere by anthropogenic sulfate [Schlesinger, 1997].

[7] We spun up the ocean model without biogeochemistry for one hundred years from rest using climatological states for the atmosphere, sea ice, and land components. The final ocean state was then adopted as the initial condition for a 30 year fully coupled simulation where atmospheric carbon dioxide was set at the late twentieth century value 355 ppm. Major element geocycling in the ocean was also initiated at this point, with initial conditions derived from a variety of data sources and idealized distributions [Moore et al., 2004]. The end state of this three decade coupled spinup was then taken as the initial state for two time-slice simulations. In one simulation, atmospheric carbon dioxide remained at 355 ppm. In the second simulation the CO2 concentration was set at 970 ppm in order to simulate a possible climate for the end of the 21st century (based on the IPCC SRES scenario A1FI given by Intergovernmental Panel on Climate Change [2001]). This is at the high end of IPCC SRES estimates, but is entirely plausible given recent estimates of actual emissions [Raupach et al., 2007; Friedlingstein et al., 2010]. If CO2 concentration growth is different than this scenario, then our simulations represent the date at which 970 ppm occurs. Both calculations were carried forward sixty years with constant CO2 forcing such that the physics and geochemistry of the upper ocean attained an approximate steady state. Using steady-state rather than trend simulations was computationally expedient, avoided ocean spin-up issues for trend runs, and makes the comparison between the two runs much cleaner. This does mean we ignored any lag in system response to the CO2 forcing, but the primary effect should be just a delay in timing and shouldn't affect our main conclusions about DMS sensitivity to CO2 forcing. To compare the changes between the two runs we averaged all quantities over the final ten years of the two runs, and we verified that the differences we note in this study exceeded local standard deviations.

[8] We validated the results of our contemporary simulation using the climatology of Kettle and Andreae [2000] and the strategies outlined by Elliott [2009] and Le Clainche et al., [2010] for uncoupled simulations. Our agreement with the data was comparable to those two works. The surface ocean DMS concentration fields are shown in Figure S1 (auxiliary material) for the two carbon dioxide levels. There are a couple of features worth commenting on here. We see rings of high DMS near Antarctica due to the inclusion of a Phaeocystis parameterization. This class of organism generates several times the typical DMSP level (dimethyl sulfoniopropionate, a major DMS precursor) and favors cold water habitat [Matrai and Vernet, 1997]. We see a DMS minimum zone between 40°S and 60°S that is due in part to the dominance of diatoms, which have very low sulfur content. This circumpolar minimum is less prominent in the DMS climatology of Kettle and Andreae [2000], but is present in the work of Kloster et al. [2007].

[9] Figure 1 (top) shows the difference in DMS flux between the two simulations at the ocean-atmosphere interface. The annular pattern is due almost entirely to poleward shifts in the phytoplankton community structure, which are important in many marine ecological contexts [e.g., Hallegraeff, 2010]. As the ocean warms, the diatoms (which dominate south of 40°S) migrate toward the south, allowing smaller phytoplankton with greater sulfur content to contribute more to the biomass, resulting in elevated DMS flux in the band between 30°S and 50°S. The alternating bands between 55°S and Antarctica are due the southward shift of Phaeocystis, which becomes dominant under ice retreat due to its cold water preference, but loses habitat toward the north as the ocean warms. Integrated over the entire Southern Hemisphere, the total amount of DMS transferred to the atmosphere in the 970 ppm run dropped by 3.5% compared to the 355 ppm case.

image

Figure 1. (top) Absolute difference in DMS flux (nanomoles/m2/s) between the final decades of the 60 year CCSM simulations. Positive values indicate that the 970 ppm simulation transfers more DMS to the atmosphere than the 355 ppm case. (bottom) Percentage difference of DMS flux for the 970 ppm simulation relative to the 355 ppm simulation. Note: values above +150% have been clipped so that the small and moderate magnitude features can be seen.

Download figure to PowerPoint

[10] Formal intercomparisons between DMS models are only beginning to appear for even contemporary DMS distributions [Le Clainche et al., 2010], and there have been no attempts to extend them into the future. Hence, we opt for a qualitative approach in Table 1, focusing on significant climate change over time scales spanning many decades. Comparison of our results to the third generation models shows that our runs amplify their dynamic ecology results. However, distinctions may be drawn with respect to the first and second generation statistical approaches, which tended to be regional and uni- or else bivariate. Their flux changes were consistently positive and in some cases very low in magnitude. Given increasing model domain, resolution, and ecosystem complexity moving downward through the table, it appears that sourcing may be reduced in the central gyre by order ten percent while increases of tens to hundreds of percent are possible at higher latitudes.

3. Analysis and Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Our Modeling and Results
  5. 3. Analysis and Discussion
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[11] Indirect aerosol effects are thought to be critical to climate evolution, and DMS is one of the major precursors for aerosols and cloud condensation nuclei in the marine boundary layer over much of the remote ocean. In this regard, it is clear from our model that a meridional redistribution of DMS flux in the warming world may occur across the entire Southern Hemisphere, with potentially significant effects on high latitude clouds. In global estimates involving constant upward or downward DMS flux changes, average planetary surface temperatures separate by three or more degrees Celsius [Charlson et al., 1987; Gunson et al., 2006]. Since the Southern Ocean is more cloudy than the global average, and an absorptive surface (the ocean) lies below the clouds, the regional importance of DMS emissions is likely to be even greater.

[12] DMS-albedo coupling has been investigated in models ranging from the conceptual to full atmospheric chemistry-climate simulations [Charlson et al., 1987; Gunson et al., 2006]. Typically, emissions have been raised or lowered by the same proportion at all locations. However, there are several ways in which regional contrasts may be of greater importance, and seasonal swings in brightness will be superimposed [Gabric et al., 2001; Bopp et al., 2003; Vallina et al., 2007; this study]. Since our emissions of DMS follow shifts in the distribution of the cold-loving Phaeocystis, if the resulting sulfate is able to cool the local ocean, then it is possible that the production of DMSP is an ecosystem adaptation for habitat maintenance. This has echoes of the CLAW/Gaia hypothesis [Charlson et al., 1987].

[13] The Southern Ocean is a place where the hydrosphere, cryosphere, atmosphere and marine biosphere interact in a myriad of complex ways. A partial list of coincident features includes: our DMS changes; annular ecosystems and massive circumpolar currents [Longhurst, 1998]; the atmospheric Southern Annular Mode (SAM) which is in turn constrained/guided by the Drake passage (60° south [Toggweiler and Russell, 2008]); the implied geostrophic westerly winds; ocean acidification; seasonal migration of the ice edge; whale/krill fisheries driven foodweb structure; and Ekman pumping along the Antarctic coast - which may well be implicated in ice sheet destabilization [Alley et al., 2008]. Because of the interconnections of all these features, the conclusion of a shift in DMS production causing south polar cooling may be premature. The only real way to evaluate the full effects will be in Earth systems models (ESMs) that couple in atmospheric chemistry and the aerosol indirect effects on clouds.

[14] We must also consider whether any results will be robust to upcoming improvements in the ecodynamics models, which currently constitute the weakest link in such ESM simulations. In a preview of other exercises we have conducted, the DMS shift is usually reinforced with even more sophisticated models. For example, in this paper the low DMSP cyanobacteria are only significant in warmer areas, with their maximum contribution fixed at one half of local biomass based on measurements [Elliott, 2009]. But, parameter settings in complete and non-sulfurous ecology schemes [Gregg et al., 2003] suggest that future stratification will favor the prokaryotes more dramatically. Thus, gyre concentrations may be overpredicted.

[15] Several authors have noted that flagellates can broaden their influence in a warmer world along with Phaeocystis [Gabric et al., 2003]. Both classes of organism are notably sulfur rich. The ice algae are also intense producers [Levasseur et al., 1994], and it might be expected that as their habitat retreats some degree of compensation should ensue. We are now configuring ice biogeochemical dynamics codes to simulate this prospect. In early runs the effects appear to be locally critical but two dimensional in nature – they are most important along a thin band of surface water tracking the ice margin in the springtime. The latter arguments deal exclusively with the phytoplankton and their cellular make up.

[16] Literature reviews have recently shown that microbial ecology will also be critical to the comprehension of sulfur during the next century [e.g., Stefels et al., 2007]. It is possible that bacterial trace element demand, demethylation yield, ultraviolet or temperature sensitivity, and taxonomy must all be accounted with fidelity. Most of these processes are now lacking for all the models we have described. Unfortunately, the number of metabolic and chemical channels which must be simulated in ocean ecosystem models is already computationally expensive, the requisite parameter sets remain highly uncertain, and both problems grow rapidly with increasing model sophistication. Future models will therefore benefit from experimental data that can constrain the critical parameters for these processes and the behavior of the overall system to climate changes. Examples of valuable, measurable quantities include the reduced sulfur and precursor compounds (particularly concentrations of DMSPp, DMSPd, DMS, and DMS flux to the atmosphere), producing/consuming organism densities, and sulfur processing rates using isotope injection. To test responses to climate change, it would be valuable to sample areas naturally high in ocean acidity such as upwelling regions [Hauri et al., 2009] as a proxy for future changes to phytoplankton physiology and ecosystem community structure.

4. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Our Modeling and Results
  5. 3. Analysis and Discussion
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[17] We have reported here on a state-of-the-art global ocean biogeochemical simulation of DMS distribution and fluxes for atmospheric CO2 concentrations of 355 ppm and 970 ppm, corresponding to present-day and a possible 2100 scenario respectively. We find changes in DMS flux to the atmosphere of 50% or more over large regions of the Southern Ocean due to concurrent sea ice changes and shifts in ocean ecosystem composition. A comparison of these results to prior studies shows that increasing model complexity is associated with reduced DMS emissions at the equator and increased emissions at high latitudes.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Our Modeling and Results
  5. 3. Analysis and Discussion
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[18] We thank two anonymous reviewers for their suggestions. National laboratory authors were supported by the U.S. DOE OBER SciDAC project. Wingenter was supported by the NMIMT Geophysical Research Center. We used the Oak Ridge Leadership Facility at the Oak Ridge National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under contract DE-AC05-00OR22725. Part of this study was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under contract DE-AC52-07NA27344.

[19] The Editor would like to thank two anonymous reviewers for their assistance in evaluating this paper.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Our Modeling and Results
  5. 3. Analysis and Discussion
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Our Modeling and Results
  5. 3. Analysis and Discussion
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

Auxiliary material for this article contains a text file, a table, and a figure.

Auxiliary material files may require downloading to a local drive depending on platform, browser, configuration, and size. To open auxiliary materials in a browser, click on the label. To download, Right-click and select “Save Target As…” (PC) or CTRL-click and select “Download Link to Disk” (Mac).

Additional file information is provided in the readme.txt.

FilenameFormatSizeDescription
grl27981-sup-0001-readme.txtplain text document1Kreadme.txt
grl27981-sup-0002-txts01.pdfPDF document808KText S1. Descriptions of the models discussed in this work.
grl27981-sup-0003-ts01.pdfPDF document15KTable S1. Summary of the key features of the published models discussed in this paper.
grl27981-sup-0004-fs01.pdfPDF document790KFigure S1. Ocean DMS concentrations simulated by our model for atmospheric CO2 concentrations of 355 ppm and 970 ppm.
grl27981-sup-0005-t01.txtplain text document1KTab-delimited Table 1.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.