New production due to N₂-fixation in the ocean can result in a net draw down of atmospheric CO₂ and a net export of carbon from surface waters because the nitrogen and carbon that fuel this export ultimately come from the atmosphere.  Direct rate measurements suggest that the input of new nitrogen from N₂-fixation can be comparable to the nitrate-based upwelling flux in oligotrophic tropical waters, and recent geochemical and modeling studies indicate that the rates are even higher.  If carbon export associated with N₂-fixation is indeed significant, then we must understand the global controls on N₂-fixation, and find ways of representing it in models which are currently being developed in the U.S. JGOFS Synthesis and Modeling Project (SMP).

Recognizing this fact, U.S.JGOFS (with support from both NSF & NASA) recently sponsored a 3.5-day workshop on N₂-fixation at the Wrigley Laboratory on Catalina Island, CA .  The overarching goal of this meeting was to bring together modelers interested in incorporating oceanic N₂-fixation into their efforts with researchers familiar with the ecology and physiology of oceanic N₂-fixation.  Some 30 scientists (see list below) attended what turned out to be a tremendously successful, exciting and enjoyable meeting. Although we had defined a set of very specific goals at the outset, what ultimately emerged was far greater, and potentially much more significant than anticipated.  In the following paragraphs we summarize the results from the workshop and make recommendations for future field and modeling studies.

The Emerging Picture

Presentations and discussions at the workshop underscored the increasing evidence that the flux of new nitrogen (and carbon) due to N₂-fixation is much greater than previously thought.  Recent geochemical (N*) studies indicate that N₂-fixation rates in the North Atlantic ocean are at least twice as high as rates derived from direct measurements (Gruber, Michaels).  These high values are consistent with rates required to account for the observed summertime depletion of DIC at the Bermuda Atlantic Time-Series (BATS) station (which happens in the absence of measurable NO₃) (Michaels, Hood), and they are further supported by natural isotope abundance (d¹⁵N) which have revealed anomalously low values in particulate matter in surface waters of the tropical and sub-tropical North Atlantic (Montoya).  The combined evidence suggests annual rates on the order of 20 - 40 Tg N/y, which equates to a potential net carbon export of 170 - 340 Tg C/y (assuming an export C:N ratio of 10) for the North Atlantic alone.  If it is assumed that about 25% of the total global N₂-fixation occurs in the North Atlantic, then this gives a global net carbon export of about 0.7 - 1.4 Pg C/y.

If a global new production of about 16 Pg C/y is assumed then carbon export due to N₂-fixation constitutes a significant fraction (4-8%) of the total.  It is important to remember, however, that NO₃-based new production does not generate a net carbon export. Rather, most of the carbon that is exported is balanced by an influx of DIC along with the new NO₃.  It is therefore likely that net carbon export due to N₂-fixation constitutes a much larger fraction of the total, biologically-mediated net C export, i.e., much greater than 4-8%.

Direct N₂-fixation rate estimates have been based principally on the observed distribution and biomass of the diazotrophic (i.e. N₂-fixing) cyanobacterium, Trichodesmium.  It is clear that the rates associated with this organism have been substantially underestimated in past extrapolations (Capone, Carpenter).  However, there is increasing evidence that the "missing N₂-fixation" is, at least in part, associated with diazotrophic species other than Trichodesmium.  Genetic studies (Zehr) and epifluorescence and light microscopy (Carpenter) have recently revealed a large and diverse microbial diazotrophic community which appears to be composed of a variety of autotrophic, heterotrophic and symbiotic forms.  These organisms are wide spread in both tropical Atlantic and Pacific waters and they are derived from many different phylogenetic groups, including the Eubacteria (particularly cyanobacteria) and Archaea.

Contradictions and Gaps in Our Knowledge

If N₂-fixation rates in the ocean are as high as the geochemical estimates suggest, it raises major questions.  How, for example, can these diazotrophs grow in the absence of measurable phosphate?  Numerous P-addition assays have been performed with Trichodesmium and, as yet, have not consistently revealed significant P-limitation.  A variety of hypotheses have been proposed for delivering P from depth to the near-surface ocean where Trichodesmium grows, including vertical migration to the phosphocline ("phosphorus mining") and potential utilization of dissolved organic sources (Letelier).  For some species of Trichodesmium, given their reported collapse pressure for their gas vesicles, P mining does not seem an option (Villareal).  As yet, none of these hypotheses have been conclusively tested.

Our inability to resolve the phosphorous question is compounded by the lack of data on the uptake and stoichiometric ratios of C:N:P:Fe for Trichodesmium and other diazotrophic organisms.  For example, it is not possible to determine whether or not phosphorous mining is a viable mechanism for obtaining P without knowing how much Trichodesmium can store.  Moreover, the N₂-fixation rates predicted by both geochemical and numerical models are critically dependent upon the value of these ratios.

Geochemical (N*) evidence indicates that N₂-fixation is lower in the Pacific Ocean than in the Atlantic, and this has been attributed to differences in the patterns and quantities of Fe deposition from the atmosphere.  This conclusion follows logically from the high Fe requirement of the nitrogenase enzyme and the general global correspondence between regions of high Fe deposition and high N₂-fixation.  However, the evidence for Fe-limitation of N₂-fixation is largely circumstantial. Results from most Fe addition experiments that have been carried out to date are inconclusive.  Further, recent analysis of Fe content in natural populations of Trichodesmium reveal neither a significantly higher Fe quota, nor a geographical pattern in cell content consistent with depositional patterns (Sanudo-Wilhelmy & Kustka).

High rates of DON release have been reported for Trichodesmium and modeling studies suggest that measurable DON anomalies (i.e., greater than 0.3 µM above background) should develop in oligotrophic regions where N₂-fixation rates are high (Hood).  Recent observations have revealed such anomalies (0.5µM above background) in stratified regions of the southwestern Equatorial Atlantic and southwestern Pacific where Trichodesmium has been observed (Hansell).  However, measurable DON anomalies do not appear to develop at BATS even though the geochemically derived rates and models predict that they should.

Although it is clear that there are many pelagic species other than Trichodesmium that are capable of fixing nitrogen in the marine environment, with the exception of a few more conspicuous symbiotic forms (i.e., Richelia/Hemialus), there is no direct evidence that these putative N₂-fixers substantially contribute to the input of new nitrogen to the ocean.

Recommendations for field studies:

To answer these questions requires that we undertake a series of key measurements, field experiments and modeling activities. First and foremost, we must find ways to quantify the biomass and N₂-fixation rates associated with non-Trichodesmium diazotrophs.  The sensitivity of ¹⁵N uptake rate measurements has improved considerably in the last decade and it is now feasible to measure N₂-fixation rates of the entire diazotrophic community (including Trichodesmium) using large volume (e.g. 1-10 liter) incubations.  Large volumes will be required if low rates are due to relatively small heterotrophic or autotrophic organisms in the picoplankton, free Trichodesmium filaments or small symbiotic associations.

It is entirely possible that the high rates indicated by geochemical studies and models are, in fact, largely due to Trichodesmium (as we now assume) and that the traditional methods simply underestimate their biomass and/or the rates.  Perhaps the most significant problem with the direct rate estimates is that shipboard measurements do not provide sufficient temporal and spatial coverage of the oceans to adequately characterize the distribution of Trichodesmium.  One possible solution, that is now being pursued, is to take advantage of the unique optical characteristics of Trichodesmium and use satellites to help us better characterize their distributions (Subramaniam), which can then be used to derive new global rate estimates (Hood, Capone).

The physiology of P-uptake and utilization is another area where a simple series of targeted experiments could vastly improve our understanding.  In particular, it is crucial to determine Trichodesmium's capacity for P-storage because this capacity largely determines whether or not P mining is possible. To our knowledge only two experiments have been carried out which measured P uptake and storage directly in Trichodesmium and they gave dramatically different results.  Phosphate uptake and storage capacity can be measured using radioisotopic (³²P or ³³P uptake) and time-course P depletion assays.  Ideally, such experiments would be complemented by measurements of C:N:P:Fe uptake and stoichiometry.

Proving that N₂-fixation is Fe-limited in some ocean regions will be a more difficult challenge, particularly from steel research vessels!  To do this it will be necessary to use trace metal clean techniques in Fe-addition/limitation assays, and the concentration and speciation of Fe must be monitored in both the treatments and in controls.  Only then can Fe-limitation be conclusively shown. Although expensive and time-consuming, these methods can and should be integrated into ongoing field studies.  As in the Iron-Ex experiments, Fe-limitation assays could be augmented using "pump and probe" or PAM fluorometric methods which provide an index of potential Fe-limitation.

Natural isotope abundance studies (d¹⁵N) also hold great promise for improving our understanding of the importance of N₂-fixation, and showing how new nitrogen derived from it moves through the food web.  Methods for measuring the d¹⁵N of DON and NO₃ are currently under development and will soon provide valuable new information on the contribution of N₂-fixation to two major nitrogen pools in the ocean.  These kinds of measurements will be extremely useful for validating ecosystem models with N₂-fixation.

Recommendations for modeling studies:

Although our understanding of the role of diazotrophic organisms in the marine environment is still very limited there are many key questions that can be addressed with models at this time.Table 1 lists a hierarchy of potential model formulations.  The "0 Order Models" include diagnostic models where the net (N₂-fixation - denitrification) rate is calculated by restoring to the observed N* distribution, formulations where the rates are simply specified based upon observed (or N* inferred) rates, and models which directly calculate N₂-fixation rate based upon the distribution of Fe flux to the sea surface.  An obvious disadvantage of the first two of these formulations is that they cannot be used to predict the future, and the Fe-flux model would require additional constraints to prevent N2-fixation at high latitudes/low temperatures.

Table 1.
 

Model	Parameterization (Jnfix)	Assumptions	Advantages	Disadvantages
0 Order Models:
diagnostic	1/t (N* - N* (obs))	reasonable physics	simple ->net rates	no prognostic capabilities
specified rates	d(x,y,z)	distribution of rates can be characterized	simple	no prognostic capabilities
Fe-flux	r(N:Fe) Fsfc(Fe)	Fe is only limiting factor	simple	no I or P limitation
1st Order Models:
Statistical	a+b1T+b2I+b3[Fe]+?	strong correlations exist	permits multiple controls	requires data base
Niche Model	d(probability of occurrence)		ecological basis	p(occurrence) is not a rate
Quasi-Mech. limitation	aI[Fe] if N<Ncrit           and T>Tcrit	Tricho is I and Fe limited	simple,quasi-mechanistic	no P limitation
Ecosystem Models:
NPZDT	aGTT - STT	GT = d (I)	Resolves Trichodesmium biomass	large number of parameters
Phys./Allocation		optimization of growth rates	physiological basis	data lacking

Although the combination of direct N₂-fixation rate measurements and N* distributions are beginning to reveal coherent global patterns which suggest a relationship to the supply of atmospheric dust, diagnostic models can provide direct 3-D estimates of the global distribution of the net (N₂-fixation - denitrification) rate, and Fe-flux models could help to determine if Fe from atmospheric sources is, indeed, controlling N₂-fixation.  Ocean biogeochemical models with specified N₂-fixation and associated carbon export flux could also be used to determine if N₂-fixation driven carbon export has a significant impact on global carbon export and uptake of atmospheric CO₂.

At the next level of complexity "1^st Order Models" attempt to dynamically calculate N₂-fixation rate in response to GCM predicted variations in environmental parameters such as temperature, irradiance and Fe concentrations.  These models include empirical/statistical formulations (e.g., derived using multiple regression techniques) that predict N₂-fixation rate directly, as well as "Niche"-based models which predict the probability of occurrence of Trichodesmium. The advantage of these kinds of models over the simpler "0 Order" formulations is that they are prognostic, and can include multiple controlling factors.  The disadvantage is that they must be formulated using empirical relationships derived from the present day ocean, and that they will work only if we can find strong and robust correlations between GCM-predictable environmental factors and N₂-fixation and/or Trichodesmium biomass.

An alternative type of formulation would be a "quasi-mechanistic" model where N₂-fixation rate is scaled, for example, to irradiance, and Fe concentration.  This type of formulation might also include "switches" on DIN concentration and temperature, i.e., N₂-fixation only occurs where DIN concentrations are below some critical level, and temperatures are above, say, 20^o C.  We refer to these models as quasi-mechanistic because the rate is determined by linear dependence on factors which are known to control N₂-fixation, but such models do not explicitly represent the biomass of the diazotrophs.  A potential problem with this approach is that it is not possible to directly include phosphorous as a linear controlling factor because P concentrations are often zero in regions where relatively high N₂-fixation rates are observed.

At the highest level of complexity we have, for example, "Ecosystem models" (Hood) and "Physiological/Resource Allocation models" (Armstrong).  The former includes formulations that attempt to explicitly model the biomass of diazotrophs and rates of N₂-fixation in an ecosystem/biogeochemical context, whereas the latter operate at the organism level and include models which are designed to maximize growth through optimal allocation of a limiting nutrient, such as Fe.  The advantage of these models is that they allow one to explicitly include observed functional relationships with controlling factors such as light, and they can be used to assess the potential impact of N₂-fixation on the entire pelagic ecosystem.  Although both require parameters and functional relationships which have not yet been adequately characterized, the development of these models should be pursued because they help define where critical information is lacking, and they can be used to guide the development of simpler models for inclusion in GCMs.

Although we are still lacking critical information that we need to construct models there is much that can and should be done at this time toward our goal of incorporating N₂-fixation into models that are currently being developed as part of the U.S.JGOFS SMP.   All of the formulations that have been outlined here are within our current computational means, and the experimental work we have suggested should substantially improve our ability to model N₂-fixation in the near future.

Click here for humorous quotes from the meeting.

Participant list:
 

Capone, Doug (Chair)	WIES/ USC
Hood, Raleigh (Co-chair)	HPL/ UMCES
Armstrong, Rob	Princeton
Berelson, Will	USC
Burns, Jay	WIES/ USC
Carpenter, Ed	MSRC/SUNY
Clayton, Tanya	Old Dominion Univ.
Deutsch, Curtis	Princeton
Gruber, Niki	Princeton
Gunderson, Troy	USC
Hansell, Dennis	BBSR
Haug, Gerald	USC
Kleypas, Joanie	NCAR
Krauk, Jamie	USC
Kuska, Adam	SUNY/MSRC
Laws, Ed	U. Hawaii
Letelier, Ricardo	U. Oregon
Lipschultz, Fred	BBSR
Mahowald, Natalie	UCSB
Michaels, Tony	WIES/USC
Montoya, Joe	Georgia Tech
Mulholland, Margie	SUNY/MSRC
Sanudo-Wilhelmy, Sergio	SUNY/MSRC
Siefert, Ron	CBL/UMCES
Stott, Lowell	USC
Subramaniam, Ajit	WIES/USC
Villareal, Tracy	Univ Texas
Zehr, Jon	UCSC