Albert Gabric, Nicholas Murray, Lewi Stone and Manfred Kohl
The model predictions suggest that the concentration of DMS in marine surface waters is indeed governed by bacterial metabolism. Environmental factors that affect the bacterial compartment are thus likely to have a relatively large influence on dissolved DMS concentrations. The ecological succession is particularly sensitive to the ratio of phytoplankton to bacterial nutrient uptake rates as well the interaction between herbivore food chain and the microbial loop. Importantly for the design of field studies, the model predicts that peak DMS concentrations are achieved during the decline of the phytoplankton bloom with a typical time lag between peak DMS and peak phytoplankton biomass of 1 to 2 days. Significantly, the model predicts a relatively high DMS concentration persisting after the phytoplankton bloom due to excretion from large protozoa and zooplankton, which may be an additional explanation for the lack of correlation between DMS and chlorophyll a field measurements. Comparison of the model predictions has been made with tank algal bloom experiments.
Dimethylsulphide (DMS; CH3SCH3 ) is an important sulfur-containing trace gas produced by some classes of marine phytoplankton [ Keller et al., 1989; Turner et al., 1988; Iverson et al., 1989; Belviso et al., 1990 ]. It is present in oceanic surface waters at concentrations sufficient to sustain a considerable net flux to the atmosphere which is currently estimated to be 0.5 ± 0.3 Tmol Syr-1 compared to the global natural (marine + terrestrial + volcanic) flux estimate of sulfur to the atmosphere of 0.78 Tmol S yr-1 [ Bates et al., 1992 ].
Charlson et al. [1987] ; have suggested that a major source of cloud condensation nuclei (CCN) over the oceans is the DMS produced by planktonic algae in seawater. DMS is oxidized in the atmosphere to form non-sea-salt sulfate ( nss SO42- ) and methane sulfonate (MSA) aerosols. Because the formation of clouds is sensitive to CCN density, it has been postulated that biological regulation of the climate is possible by affecting albedo and thus the Earth's radiation budget. This would be due to the effect of temperature and solar radiation on phytoplankton growth; however, it is still unclear whether the change in albedo will cause a positive or negative feedback on climate.
Although the theory of a DMS-climate link has been somewhat controversial [Schwartz, 1988; Bates et al., 1987a; Monaghan and Holdsworth, 1990], support for the hypothesis that a change in DMS flux can affect global temperature has come from analysis of Vostok ice core sulfate data [Legrand et al., 1988]. Examination of the nss SO42- profile along the 160-kyr Antarctic ice core showed that a 20-46% increase in nss SO42- content in the Antarctic atmosphere during the full glacial (compared with interglacial) conditions was consistent with higher DMS emissions from marine biota productivity. Further evidence of the connection between atmospheric DMS and aerosol sulfur species has been presented by Ayers et al. [1991] who demonstrated a strong coherence between sea level DMS concentration, MSA, and nss SO42- at Cape Grim, Tasmania.
In some industrialized regions such as the Baltic Sea [Leck and Rodhe, 1991] it has been found that anthropogenic sulfur emissions dominate the biogenic flux. Regional studies in the North Sea [Turner et al., 1988] and the east coast of the U.M.A. [Iverson et al., 1989] have been undertaken in order to estimate regional DMS fluxes and to attempt to correlate these with biological parameters such as phytoplankton biomass (chlorophyll (chl) a ), however good correlations were found only for coccolithophores and dinoflagellates. Leck et al. [1990], in a study of the Baltic Sea, found a significant correlation of DMS concentration with zooplankton biomass, supporting the observation of Dacey and Wakeham [1986] that grazing may be a major factor in the liberation of DMS from phytoplankton cells. Nguyen et al. [1988] suggested that algal cell senescence may also promote DMS liberation and reported that the ratio of DMS concentrations during lysis to that during the growth phase was 7 for a diatom species and 26 for a dinoflagellate species.
At the present time the complex mechanisms responsible for DMS cycling are not well understood, and the ecological ramifications are even less so. Here we attempt to synthesise some of the diverse information reported in the literature and place it within the broader context of a (model) aquatic ecosystem. The model is designed to demonstrate the interdependence of the DMS cycle and the major ecological processes that occur over a range of scales in time and space within the framework of the recent conceptual picture of marine food webs [Ducklow, 1983; Azam et al., 1983; Cole et al., 1988], where microorganisms are recognized as significant components (for a general review see Fenchel [1988]).
Aquatic ecosystems were originally conceived as a linear trophic food chain in which major nutrients move vertically from primary producers to zooplankton to fish. However, it is now realized that bacteria and protozoa can control large nutrient fluxes through a myriad of pathways. The microbial loop has the potential to divert a large proportion (over 50%) of primary production and nutrients out of the classical food chain. Further, because microorganisms have very high rates of turnover (~d-1) they have the ability to rapidly recycle and remineralize nutrients at the base of the food web.
The role of bacteria in the anaerobic removal of DMS has been known for some time [Zeyer et al., 1987; Zinder and Brock, 1978], and its aerobic catabolism has been reviewed by Taylor and Kiene [1989]. However Taylor and Gilchrist [1991] and Visscher et al. [1992] have reported that dimethylsulfonium propionate (DMSP), the precursor compound to DMS, can also be aerobically biodegraded by demethylation to 3-methiolpropionate (MMPA) and 3-mercaptopropionate (MPA). Kiene and Service [1991] have suggested that DMS is not the major transformation product of DMSP and reported estuarine experiments where less than 30% of the DMSP was converted to DMS, presumably owing to the alternate demethylation pathway.
Recently, a number of studies have reported that there is a definite link between the microbial food web and the oceanic cycling of DMS. Kiene and Bates [1990] have shown that DMS turnover in seawater is of the same order as microbial activity (~ 1/day) and 3-430 times faster than turnover by atmospheric ventilation. In a similar vein, the results of Belviso et al. [1990] suggest that ciliates are largely responsible for DMS production. The production of DMS is apparently closely linked to the dynamics of the marine pelagic food web, and it is clear that the microbial loop plays an important role in the DMS cycle.
Existing mass balance estimates of DMS in the oceans and atmosphere are often based upon assumptions of a steady state and in this respect can be misleading. Marine environments are noted for their variable and nonequilibrium conditions. Algal blooms, nutrient upwelling, fluvial nutrient injections, and a whole range of transient and seasonal events are typical, and often vital, occurrences in aquatic ecosystems. This variability naturally manifests in the spatio-temporal distribution of DMS [Turner et al., 1988]. As Kiene and Bates [1990, p704] argue, & quot; Plankton populations that interact to affect DMS concentrations.... are not likely to be in a steady state, and therefore predicting effects of climate change will involve complex models that consider ecological interactions as well as physical and chemical factors." For this reason our model is time-dependent and applies for non-equilibrium conditions and provides an initial attempt at quantifying the oceanic components of the overall DMS-climate hypothesis.
Model Structure and Assumptions
In order to analyze the dynamics of DMS production in the upper ocean, we have chosen a flow network approach and adapted the ecological structure of the nitrogen- based plankton community model of Moloney et al. [1986] to include the production, bacterial consumption, chemical transformation, and ventilation of DMS to the atmosphere. The ecological structure has been kept as simple as possible consistent with the overall aim of the modelling exercise, which is to concentrate on factors affecting DMS production and cycling.
The model consists of eight compartments, three of which are abiotic: dissolved inorganic N, DMSP and DMS (see Figure 1). The biotic compartments comprise the primary producers (generic phytoplankton) and four consumer groups (bacteria, zooflagellates, nonphotosynthetic protozoa and micro and mesozooplankton). For simplicity, no higher trophic levels are considered although we have modified the Moloney et al. [1986] model by including zooplankton export through grazing predators such as fish.
The various fluxes between compartments are described in Table 1. The cycling of nitrogen and sulfur (as DMSP or DMS) occurs in a mixed layer whose bottom boundary coincides with a thermocline at 20-m depth. Nitrogen (as nitrate) is injected into the euphotic zone at the start of the simulation, mimicking the rapid input of nutrients that occurs during wind mixing or coastal upwelling. Phytoplankton growth is assumed not to be limited by light or temperature in the 20-m layer, and diurnal variability is not explicitly modelled. It is clear that vertical attenuation over the depth during very intense blooms may invalidate this assumption. Loss of particulate organic matter occurs by sedimentation below the thermocline. Nitrogen (as ammonium) is regenerated via the microbial loop; however we have not distinguished between nitrate and ammonium in the nutrient uptake parameterization. The model assumes spatial homogeneity, although the inclusion of vertical variability is feasible.
In order to incorporate the cycling of DMS via the microbial loop, the model uses dual elemental currencies of nitrogen and sulfur. This necessitates the use of conversion factors to transform from nitrogen to sulfur currencies or vice versa. The approximate molecular weights of DMS and DMSP are 62.1 and 134.2, respectively. Thus the elemental ratios for DMS and DMSP are
C:S (DMS) = 0.75 mg C mg S-1
C:S (DMSP) = 1.87 mg C mg S-1
while for equal weights of DMS and DMSP the ratio of S is given by
S(mg DMS) : S(mg DMSP) = 2.16
To estimate the growth of bacteria due to DMS uptake, we first calculate the proportion of carbon in DMS and then assume the ratio of carbon to nitrogen uptake is the same as the elemental composition of the biotic compartment, i.e., C:N equals 5.0 for bacteria and 6.0 for phytoplankton [Strickland, 1960; Nagata, 1986].
The flow of N or S between the various compartments is described by a system of coupled ordinary differential equations (1)-(8) given in the appendix. Where both N and S fluxes occur in the same equation, a conversion factor has been applied for dimensional consistency, for example in equations (2) and (8). Since the magnitude of nitrogen fluxes associated with the sulfur cycle is relatively small, we neglect the fluxes of DMSP and DMS ( namely, F17, F18, F47, F57) in equations (1), (4), and (5). The sole coupling between the N and S cycles is via bacterial uptake of S in (2). Equations (7) and (8) are in S(DMSP) and S(DMS) currencies, respectively, and the flux conversion factor in (8) is given by b = 2.16.
The Sulfur Cycle
Dissolved DMS and its biogenic precursor DMSP are both phytoplankton metabolites although the actual mechanism of release is unclear. As was noted above, DMSP can be either aerobically biodegraded to MMPA and MPA (flux F72) or alternatively transformed to DMS (flux F78) by enzymatic cleavage both within the algal cell (at a slow rate) and in the water column [Dacey and Wakeham, 1986; Wakeham et al., 1987]. Here we follow Kiene and Service [1991] and take the ratio for these two transformation pathways, F72:F78 to be approximately 2.
We assume DMSP to be present in the excreta of zooplankton and large protozoa, both of which consume DMSP-containing phytoplankton, while both DMSP and DMS are released to the water by phytoplankton. The microbial loop plays a crucial role in the model providing the link between the N and S cycles, since we assume that bacteria can consume (and grow on) dissolved DMSP and DMS as well as nitrogen.
Loss mechanisms for dissolved DMS include photochemical oxidation in the euphotic zone [Brimblecombe and Shooter, 1986] and ventilation to the atmosphere. DMS flux to the atmosphere is a function of the concentration gradient across the air-sea interface and the wind speed at the water surface. Given the large difference between measured oceanic and atmospheric DMS concentrations, the atmospheric concentration is usually neglected in determining the flux [Bates et al., 1987b]. Particulate DMS/DMSP in the phytoplankton and zooplankton compartments can also be lost from the system by sinking of fecal material or dead cells below the thermocline or, in the case of zooplankton, through grazing and export by pelagic predators.
The DMS photo-oxidation rate constant (2.07 day-1) quoted by Brimblecombe and Shooter [1986] was estimated in the absence of detailed information on the wavelength dependence of the process. It is thus not clear over what depth photo-oxidation occurs. We use a depth-averaged estimate of the photo-oxidation rate by noting that the average light intensity over the 20-m mixed layer, assuming a vertical extinction coefficient ke = 0.042 m-1 [Parsons et al., 1984] appropriate for clear oceanic waters, is 0.67 times the surface value, that is 1.4 day-1. Because of the diurnal change in solar radiation, Brimblecombe and Shooter [1986] suggest that photo-oxidation occurs for about one third of the time thus the depth-averaged value has been divided by 3.
Keller et al. [1989] report cellular concentration of DMSP for 12 classes of phytoplankton and note that there is a strong correlation with taxonomic position, the highest levels being found in Dinophyceae and Prymnesiophyceae. The DMSP content per cell is shown for a selection of species in Table 2. The measured cell volumes are only approximate because osmotic volume was not estimated by Keller et al. [1989]. The carbon content per cell volume has been calculated from regression equations for phytoplankton [Strathmann, 1967], given by
Diatoms
log C = 0.758 logV - 0.422
Other species
log C = 0.866 logV - 0.460
where C is given in picograms C/cell and V is the cell volume in cubic micrometers.
The C:N ratio for phytoplankton is used to derive nitrogen content and hence the S(DMSP):N ratio g for each species. It is interesting to compare these ratios with the most commonly quoted stoichiometry for plankton of 106 C:263 H:110 O:16 N:1 P:0.7 S by atoms [Stumm and Morgan, 1981] which corresponds to an elemental ratio by weight of 10 N : 1 S, namely, g = 0.1.
Phytoplankton Kinetics
The functional form of fluxes follows the prescription given by Moloney et al. [1986]. Michaelis-Menten kinetics are assumed for N uptake by phytoplankton and heterotrophic bacteria; thus the rate of nutrient uptake n is given by
n = Vmax N/(Ks + N)
where Vmax is maximum uptake of nutrient N, and Ks is the half-saturation concentration. Following the suggestion of Moloney et al. [1986], a two-parameter asymptotic formulation,
Fmn = ki Xn [1 - exp(kj Xm )]
has been used to avoid computational difficulties in the integration. The correspondence between the Michaelis-Menten parameters, V max and Ks, and the parameters of the asymptotic formulation is, Vmax = ki and Ks = ln(2) / kj.
Simpler single parameter expressions have also been used where possible. A Lotka-Volterra expression is employed to model consumption of prey compartment n by predator compartment m, namely, Fmn = ki Xn Xm. The release of N or S from compartment n to compartment m is modelled by the expression Fnm = ki Xn, that is, by assuming first-order kinetics. Dacey and Wakeham [1986] found that starved zooplankton contained less than 4% of the DMSP in animals with a full gut, suggesting that DMSP does not accumulate in zooplankton body tissue but is mostly excreted. We have thus used the phytoplankton value of the S(DMSP):N cell ratio ( g ) also for computing protozoa (F47) and zooplankton (F57) DMSP excretion fluxes.
The parameter values assumed for the mixed layer model are given in Table 3 , and a range has been shown where data are available. Uncertainty exists in the biological parameters associated with phytoplankton and bacteria nutrient uptake (k23, k24, k25, k26) which vary according to cell size and physiological status as well as ambient conditions such as temperature and nutrient concentration [Moloney and Field, 1989; Parsons et al., 1984]. The values used here are based on those chosen by Moloney et al. [1986] and Moloney and Field [1989], allowing where appropriate for the different mixed layer depth used by these authors, and may be considered relevant to coastal phytoplankton in temperate waters.
The model has been used to predict resulting dissolved DMS concentrations
during a phytoplankton bloom in two distinct cases. The first
case uses parameter values appropriate to the conditions pertaining
in a shallow mixed layer on the continental shelf while the second
case attempts to reproduce the conditions in the tank experiments
carried out by
Case 1 : Mixed Layer
Initial values for the eight state variables were X1(0)=3, X2(0)=2, X3(0)=0.1, X4(0)=0.2, X5(0)=2, X(6)=500, X(7)=0, and X(8)=0. The biotic compartments have been given arbitrarily low initial values, while the initial dissolved inorganic nitrogen (DIN) level corresponds to a nutrient concentration of 25 mg N m-3 (equivalent to 1.8 mM NO3 ). Initial DMSP and DMS pools were both set to zero.
Because of the wide range in magnitude of the rate parameters governing the various processes (see Table 3), the system of ordinary differential equations (1) - (8) in the appendix was numerically integrated using a stiff solver (D02EBF) available in the Numerical Algorithms Group (NAG) library. This algorithm uses a variable time step to ensure numerical stability throughout the integration.
The ecological succession over a 40-day period is shown in Figure 2. Phytoplankton take up the injected DIN and reach a minor peak at about 8 days. Bacteria peak at 9 days, followed by zooflagellates (which are bacterivores) at 12 days. There is a second, higher phytoplankton peak at 14 days. Nutrient regeneration via the microbial loop is evident between days 16 and 22, with bacteria reaching a maximum at day 33 having no competition for the remaining DIN once the phytoplankton have declined as a result of predator grazing. This perhaps surprising bacterial peak has been observed in recent studies on the Red Sea, where Weisse [1989] found bacterial production to be at least as high as particulate primary production. Both Scavia [1988] and Strayer [1988] have suggested that summed secondary production (including bacterial production) may exceed primary production during certain periods in the open oligotrophic ocean.
The relationship between dissolved DMS and phytoplankton is shown in Figure 3. Peaks in DMS follow those in phytoplankton by about 1.5 days. Significantly, the model correctly predicts that peak DMS concentrations occur while phytoplankton concentrations are declining. The decline is due to predation, as nutrients are still relatively high. Although Nguyen et al. [1988] have suggested that a high DMS release rate is likely to coincide with the senescence phase, leading presumably to cell breakdown, here the influence of ecological succession during the bloom period is also demonstrated to be a contributing factor.
Importantly, the model predicts that the dissolved DMS concentration (and hence DMS flux across the sea-air interface) does not fall to zero at the end of the phytoplankton bloom at 25 days but persists as a result of inputs from higher trophic levels, namely, the large protozoa (which peak at 18 days) and zooplankton (which peak at about 22 days). Such an influence of food web dynamics, notwithstanding the taxonomic variability of algal cell DMSP content, may explain the apparent lack of correlation between chlorophyll a and dissolved DMSP/DMS reported in some field measurements [e.g., Andreae and Barnard, 1984; Iverson et al., 1989]. The time lag between peak chlorophyll and peak DMS suggests that a lagged correlation of field data may be more appropriate.
While use of the model to make accurate quantitative comparisons with field measurements of DMS concentrations awaits a more complete field database that is currently available, it is encouraging to note that the model predictions for peak DMS (see Figures 5 a through 5e ), which are in the range 0.01 - 0.1 mg S(DMS)m-3 for our 20-m-deep mixed layer, generally compare well with the measured values given in Table 4. It is also interesting to note that while cell DMSP content can vary by 2 orders of magnitude between species (Table 2), except for the Antarctic data of McTaggart and Burton [1992], the measured DMS concentrations display a much tighter range (Table 4). This may reflect the taxonomic heterogeneity of most natural phytoplankton communities or alternately may suggest that a separate mechanism is tending to level dissolved DMS concentrations.
The reliability of predictions of multiparameter ecological models such as the system presented here can be properly gauged only given a knowledge of the sensitivity of the results to changes in the rate parameters, some of which are only estimates because experimental data were not always available [Moloney et al., 1986; Pace et al., 1984]. Sensitivity analysis can also indicate which parameters should be known to great accuracy and which can permit some degree of uncertainty. We have systematically conducted a empirical sensitivity analysis [Platt et al., 1981] varying each of the model parameters by plus or minus 50% the from "reference" value quoted in Table 3. The metric we have used to gauge sensitivity is the absolute difference in the time-integrated DMS concentration, X8av , calculated with upper and lower estimates of each parameter.
Figure 4 shows the values of this sensitivity metric plotted in ascending order; thus the parameters appearing at the right of the abscissa are the ones that are most determinant of time-integrated DMS concentration. The time-integrated DMS prediction is most sensitive to the parameters k23, g, k3, k14, k17, k11, k8, k28, and k9, all of which have a metric value > 10. As would be expected, the algal cell DMSP concentration reflected by g is an important parameter. In the context of nitrogen cycling, Moloney et al. [1986] noted that the protozoa play a pivotal role in linking the microbial loop and herbivore food chain. The most important parameters in this context are k3 and k17, which are involved in the link between the phytoplankton-protozoa and zooflagellate-protozoa fluxes. The parameters k11, k 14, k8 and k9 are all associated directly with the bacterial compartment or regeneration of dissolved nitrogen which is a microbially mediated process. Clearly the bacterial compartment is central to DMS production. The Vmax for nitrogen uptake by phytoplankton k23 is also an important parameter. DMS is also clearly sensitive to k28, which describes the rate of DMS consumption by bacteria.
Of the abiotic processes, the DMS photochemical oxidation rate k29 is the most important parameter (metric @ 6). It is worthwhile noting that DMS is relatively insensitive (metric<2) to changes in k30 the rate of ventilation to the atmosphere, which confirms the hypothesis of Kiene and Bates [1990] regarding the relative importance of microbial turnover versus flux to the atmosphere.
While the time-integrated concentration is probably the most appropriate sensitivity measure, extra insight can be gained by examining how the temporal evolution of DMS is influenced by varying the most sensitive parameters. When one parameter has been varied, all the others are kept at their reference values. The DMS time evolution is shown in Figure 5a where the model has been run with three different initial DIN concentrations, X6(0). Changing the initial DIN level has a dramatic effect on the dynamics of the phytoplankton bloom since the initial nitrogen uptake rates F61 and F62 depend on X6 (see equations (25) and (26) in the appendix). The timing as well as the magnitude of the DMS peak are affected, with maximum DMS achieved later as the initial nitrogen pool is decreased. The proportional change in peak DMS concentration (and hence flux to the atmosphere) between maximum and minimum nutrient loading is approximately a factor of 3, equal to the proportional change in initial DIN. This result supports the observations of Turner et al. [1988] that during bloom conditions (high available nutrients) the flux to the atmosphere can be 2-5 times that during normal nutrient loading.
Figure 5b demonstrates the influence of varying the maximum phytoplankton N uptake rate, k23, on the timing and magnitude of peak DMS concentration. For high N uptake rates, peak DMS is achieved about 10 days earlier than for the reference case, and the peak concentration is also higher. The situation is very different under the low N uptake scenario where, bearing in the mind that bacteria compete with phytoplankton for DIN in our model, primary productivity is kept quite low and thus little DMSP or DMS is produced.
Clearly, the ratio of phytoplankton to bacterial N uptake plays a key ecological role and consequently is also very important for DMS production. Moloney and Field [1989] have presented allometric equations that relate cell mass to nutrient uptake rate. Their allometric equation for maximal nutrient uptake rate Vmax pertaining to phytoplankton and bacteria of cell mass M is
Vmax = 3.6 M - 0.25
Assuming a typical ratio of bacteria to phytoplankton cell mass of 1:1000, then the ratio of Vmax (phytoplankton) to V max (bacteria) is approximately 0.17. To test the effect of a higher value of Vmax for bacteria the model was run with k25 increased to 6.6 day-1 and the results are shown in Figure 5c. Bacteria now outcompete phytoplankton for DIN, and primary productivity is depressed with the level of DMS also much reduced.
The model sensitivity to a change in phytoplankton species is shown in Figure 5d, where the DMSP cell content g has been varied. There is no shift in the timing of the peak DMS concentrations, which varies linearly in proportion to the change in g. It should be noted that we have assumed here that the other phytoplankton parameters, such as nutrient uptake rate and zooplankton grazing rate, are constant, which may not necessarily be so for different species. In the case of a monospecific bloom of dinoflagellates, which as a class have much higher DMS cell content than diatoms, one might expect a higher dissolved DMS concentration. However, as was noted by Legendre [1990], because these species may be unpalatable to zooplankton, grazing may be reduced and, as a consequence, the release of DMSP/DMS impeded [Dacey and Wakeham, 1986]. In the absence of species-specific data for the rate parameters it would thus be unwise to try to predict the net effect of a species shift.
Figure 5e shows the effect on DMS concentration of a change in the microbial consumption rate k28. There is a very slight retardation of the timing of the peak at low consumption rates. The DMS concentration decreases (nonlinearly) with increasing bacterial uptake as more DMS is cycled via the microbial loop back to the particulate phase. The microbial metabolic rate and, by inference, the population level, are thus very important controls on dissolved DMS concentration and its flux to the atmosphere. Numbers of heterotrophic bacteria in seawater can range from 1 x 105 mL-1 in oligotrophic oceanic waters to 5 x 106 mL-1 in eutrophic coastal waters [Fenchel, 1988]. Higher bacterial population levels in eutrophic waters may act as a counter to the increased primary productivity that should normally lead to higher DMS levels but which are in fact observed to be similar to other regions [Andreae, 1985].
Case 2: Tank Experiments
One of the few experimental determinations of DMS production during a monitored bloom was carried out by Nguyen et al. [1988] in 2-m3 tanks covered by a neutral fiberglass plate. The tanks were filled with seawater from Villefranche Bay which had been filtered through a 150 mm screen. The filter prevented mature zooplankton (copepods) from entering the tanks, although juveniles could have been present. However, grazing is likely to have been very much reduced compared with the natural ocean environment. The filter also prevented larger diatoms from entering the tank, thus biasing the phytoplankton population to smaller-celled species. Three experiments were reported, two consisting of diatom blooms and the other an unidentified flagellate species. In the experiments the tank seawater was enriched with nitrate, phosphate, and silicate and from the results presented it was evident that the phytoplankton phosphate uptake rate was slightly higher than that for nitrate or silicate. Clearly, our model does not purport to reproduce the complex nutrient dynamics in these tank experiments, as we have assumed nitrate to be the only limiting nutrient.
In the first diatom experiment the phytoplankton biomass reached a maximum after 83 hours and DMS peaked about 14 hours later. Measurements were stopped when algal biomass started to decline at 97 hours, although the dissolved DMS concentration was still significant at this stage. The timescale for the experiment was thus approximately 4 days.
In order to apply the model to the first experiment reported by Nguyen et al. [1988] the depth dependent parameters were rescaled to the tank depth of 1.2 m. There was no loss of matter from the tanks, so the sedimentation parameters k7 and k22 and zooplankton export k32 were all set to zero. The light field will not be much attenuated over 1.2 m, so that the photo-oxidation rate k29 has been set to the surface value given by Brimblecombe and Shooter [1986] adjusted for diurnal variability, of 0.7 day-1. Ventilation of DMS depends on wind speed which will be negligible in the covered tanks so k30 has been set to zero. Initial values for the phytoplankton, DIN, and DMS compartments were measured in the experiment, and the other variables have been set to arbitrarily small values.
Nguyen et al. [1988] measured total chlorophyll
a as an indicator of phytoplankton biomass. Since there
is large uncertainty in C:chl a and N:chl a ratios
for phytoplankton [
Model predictions for normalized DMS are shown in Figure 6a with experimental results in Figure 6b. The predicted peak biomass is about 30 (compared with the observed peak of 40), and maximum DMS is about 5 (compared to 6.3). Bearing in mind the limitation of assuming a single limiting nutrient, the model has reproduced the peak phytoplankton and DMS ratios reasonably well. The timescale of the theoretical predictions is slightly compressed compared to the experiment, probably as a result of a longer phytoplankton growth cycle in the tanks due to multiple nutrient (N, P, Si) availability.
The model presented here attempts to place oceanic DMS production and its flux to the atmosphere into a proper ecological context. Ecological models can be used to gain insight into the various mechanisms that might determine DMS in the ocean as well as to predict the absolute DMS concentration for a particular site. Conversely, future studies on the production and consumption of DMS in the ocean may lead to useful insights into the dynamics of the microbial loop.
The model is meant to be a heuristic aid rather than a strict predictive tool. The many processes affecting DMSP/DMS concentration and removal can be seen in better perspective by the model's ability to examine the totality of the system, notwithstanding the uncertainty in the absolute value of some of the ecological rate parameters.
One of the key conclusions that may be drawn from the model simulations is that ecological succession in the pelagic food web will be a primary determinant of the temporal variation in DMS. This is because higher trophic levels such as protozoa and zooplankton will continue to release DMSP in their excreta for some time after the peak in phytoplankton has passed. This has clear significance for the design of sampling programs and the interpretation of field data and suggests that attempts at simple correlations of DMS concentration/flux with phytoplankton biomass or primary productivity will not be successful unless allowance for a temporal lag is included.
The observation by Kiene and Bates [1990] of the central role of the microbial loop in controlling dissolved DMS concentration is supported by our model simulations. Ventilation to the atmosphere seems to be a minor influence on the dissolved DMS concentration. This implies that the DMS flux may be significantly altered by any changes that might occur to the dynamics of the marine food web. Thus factors affecting bacterial growth and biomass could be indirect determinants of DMS level in the ocean. Although we have not explicitly modelled the effects of temperature here, it is well known that bacterial metabolism is greatly suppressed at low temperatures [Pomeroy and Wiebe, 1988], so one might expect DMS levels at high latitudes to be relatively high. Support for this hypothesis is provided by Bates et al. [1987 b] who report mean DMS concentrations in the Bering Sea and Gulf of Alaska of 4.7 and 5.0 nmol L-1, respectively, compared with 3.1 nmol L-1 in tropical Pacific waters. Similarly the extremely high DMS concentrations measured by McTaggart and Burton (1992) in Antarctic waters may be partly a consequence of suppressed microbial activity.
Sieracki and Sieburth [1986] have found that sunlight can reduce the growth rate of natural marine bacteria, with photoinhibition operating even at 16% of surface light intensity, effectively over the whole of the euphotic zone. Photochemical removal of DMS will also increase under high light intensities, but this process is thought to be confined to the upper meters of the water column [Brimblecombe and Shooter, 1986]. The net effect of decreased consumption over most of the euphotic zone and increased photo-oxidation in the upper few meters may well be that mean DMS concentration could be increased in strong light conditions. This hypothesis may explain the apparent paradox that a linear relationship exists between solar radiation and DMS flux [Bates et al., 1987a], yet correlation of DMS with phytoplankton biomass or productivity has generally been poor [Andreae, 1986; Andreae and Barnard, 1984].
An interesting insight into the complex role played by plankton ecology in DMS production is the importance of the ratio of phytoplankton to bacterial nutrient uptake. Moloney and Field [1989] have shown that cell mass can be related to nutrient uptake rate, with generally higher uptake rates for smaller organisms, such as heterotrophic bacteria. Thus small-celled Prymnesiophyceae (see Table 2) competes better for dissolved nitrogen with the bacteria than would the larger diatoms, and the resulting higher phytoplankton peak will produce higher dissolved DMS concentrations. While DMSP cellular concentration is certainly high in Prymnesiophyceae, because of their small cell size the mass of DMSP per cell is low. The model suggests that algal-bacterial competition for DIN may be an alternate explanation for the higher amounts of DMS produced by this class of phytoplankton.
BACK on your client or go back to the dms home page.
Please feel free to contact us Nicholas Murray for more information, or use our feedback form.