Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and...

29
* Corresponding author. Tel.: #32-2650-5988; fax: #32-2650-5993. E-mail address: lancelot@ulb.ac.be (C. Lancelot). Deep-Sea Research II 48 (2001) 2745}2773 Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the Southern Ocean E. Hannon, P.W. Boyd, M. Silvoso, C. Lancelot * Ecologie des Syste % mes Aquatiques, Universite & Libre de Bruxelles, CP221, Bd du Triomphe B-1050, Belgium Department of Chemistry, National Institute of Water and Atmosphere, Centre for Chemical and Physical Oceanography, University of Otago, Dunedin, New Zealand Abstract The impact of a mesoscale in situ iron-enrichment experiment (SOIREE) on the planktonic ecosystem and biological pump in the Australasian-Paci"c sector of the Southern Ocean was investigated through model simulations over a period of 60-d following an initial iron infusion. For this purpose we used a revised version of the biogeochemical SWAMCO model (Lancelot et al., 2000), which describes the cycling of C, N, P, Si, Fe through aggregated chemical and biological components of the planktonic ecosystem in the high nitrate low chlorophyll (HNLC) waters of the Southern Ocean. Model runs were conducted for both the iron-fertilized waters and the surrounding HNLC waters, using in situ meteorological forcing. Validation was performed by comparing model predictions with observations recorded during the 13-d site occupation of SOIREE. Considerable agreement was found for the magnitude and temporal trends in most chemical and biological variables (the microbial food web excepted). Comparison of simulations run for 13- and 60-d showed that the e!ects of iron fertilization on the biota were incomplete over the 13-d monitoring of the SOIREE bloom. The model results indicate that after the vessel departed the SOIREE site there were further iron-mediated increases in properties such as phytoplankton biomass, production, export production, and uptake of atmospheric CO , which peaked 20}30 days after the initial iron infusion. Based on model simulations, the increase in net carbon production at the scale of the fertilized patch (assuming an area of 150 km) was estimated to 9725 t C by day 60. Much of this production accumulated in the upper ocean, so that the predicted downward export of particulate organic carbon (POC) only represented 22% of the accumulated C in the upper ocean. Further model runs that implemented improved parameterization of diatom sedi- mentation (i.e. including iron-mediated diatom sinking rate, diatom chain-forming and aggregation) sugges- ted that the downward POC #ux predicted by the standard run might have been underestimated by a factor 0967-0645/01/$ - see front matter 2001 Published by Elsevier Science Ltd. PII: S 0 9 6 7 - 0 6 4 5 ( 0 1 ) 0 0 0 1 6 - 9

Transcript of Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and...

Page 1: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

* Corresponding author. Tel.: #32-2650-5988; fax: #32-2650-5993.E-mail address: [email protected] (C. Lancelot).

Deep-Sea Research II 48 (2001) 2745}2773

Modeling the bloom evolution and carbon #ows duringSOIREE: Implications for future in situ iron-enrichments

in the Southern Ocean

E. Hannon�, P.W. Boyd�, M. Silvoso�, C. Lancelot��*

�Ecologie des Syste%mes Aquatiques, Universite& Libre de Bruxelles, CP221, Bd du Triomphe B-1050, Belgium�Department of Chemistry, National Institute of Water and Atmosphere, Centre for Chemical and Physical Oceanography,

University of Otago, Dunedin, New Zealand

Abstract

The impact of a mesoscale in situ iron-enrichment experiment (SOIREE) on the planktonic ecosystem andbiological pump in the Australasian-Paci"c sector of the Southern Ocean was investigated through modelsimulations over a period of 60-d following an initial iron infusion. For this purpose we used a revised versionof the biogeochemical SWAMCO model (Lancelot et al., 2000), which describes the cycling of C, N, P, Si, Fethrough aggregated chemical and biological components of the planktonic ecosystem in the high nitrate lowchlorophyll (HNLC) waters of the Southern Ocean. Model runs were conducted for both the iron-fertilizedwaters and the surrounding HNLC waters, using in situ meteorological forcing. Validation was performed bycomparing model predictions with observations recorded during the 13-d site occupation of SOIREE.Considerable agreement was found for the magnitude and temporal trends in most chemical and biologicalvariables (the microbial food web excepted). Comparison of simulations run for 13- and 60-d showed that thee!ects of iron fertilization on the biota were incomplete over the 13-d monitoring of the SOIREE bloom. Themodel results indicate that after the vessel departed the SOIREE site there were further iron-mediatedincreases in properties such as phytoplankton biomass, production, export production, and uptake ofatmospheric CO

�, which peaked 20}30 days after the initial iron infusion. Based on model simulations, the

increase in net carbon production at the scale of the fertilized patch (assuming an area of 150 km�) wasestimated to 9725 t C by day 60. Much of this production accumulated in the upper ocean, so that thepredicted downward export of particulate organic carbon (POC) only represented 22% of the accumulatedC in the upper ocean. Further model runs that implemented improved parameterization of diatom sedi-mentation (i.e. including iron-mediated diatom sinking rate, diatom chain-forming and aggregation) sugges-ted that the downward POC #ux predicted by the standard run might have been underestimated by a factor

0967-0645/01/$ - see front matter � 2001 Published by Elsevier Science Ltd.PII: S 0 9 6 7 - 0 6 4 5 ( 0 1 ) 0 0 0 1 6 - 9

Page 2: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

of up to 3. Finally, a sensitivity analysis of the biological response to iron-enrichment at locales with di!erentinitial oceanographic conditions (such as mixed-layer depth) or using di!erent iron fertilization strategies(single vs. pulsed additions) was conducted. The outcomes of this analysis o!er insights in the design andlocation of future in situ iron-enrichments. � 2001 Published by Elsevier Science Ltd.

1. Introduction

The limitation of primary production by iron supply in the Southern Ocean was "rst hy-pothesized by Gran (1931). Laboratory and shipboard iron-enrichment studies indicate thatphytoplankton growth in the open high nitrate low chlorophyll (HNLC) waters of the SouthernOcean is severely limited by iron availability (e.g. de Baar et al., 1990, 1995; Buma et al., 1991;Martin et al., 1990). There also have been reports of other factors controlling phytoplanktongrowth in these waters, such as light limitation (Mitchell and Holm-Hansen, 1991), iron-lightco-limitation (Sunda and Huntsman, 1997; Lancelot et al., 2000), and/or limitation byboth iron and silicic acid supply (Boyd et al., 1999; Sedwick et al., 1999). In order to resolve thisissue, there was a need to conduct an in situ iron-enrichment experiment in these waters (Frost,1996).

The Southern Ocean Iron RElease Experiment (SOIREE) is the "rst in situ mesoscale ironfertilization experiment conducted in the Southern Ocean (Boyd et al., 2001a). It took place in theAustralasian-Paci"c polar waters (613S, 1403E), over 13 d (10}22 February 1999) and consisted offour iron infusions (Boyd and Law, 2001a). The SOIREE study revealed a signi"cant increase inalgal biomass and production within the fertilized patch (Gall et al., 2001a, b) and a shift inplanktonic community structure towards large diatoms (Gall et al., 2001a). Elevated productionresulted in a subsequent substantial drawdown of macronutrients (Frew et al., 2001) and inorganiccarbon (Bakker et al., 2001). However, no signi"cant impact of iron fertilization on the magnitudeof export production was evident over the 13-d experiment (Charette and Buesseler, in press;Nodder and Waite, 2001). Thus, the timescale of SOIREE monitoring was insu$cient to assesswhether a signi"cant increase in export production*as put forward by the iron hypothesis ofMartin (1990)*was the fate of the SOIREE bloom. Indeed, the timescale of decoupling of algalblooms and the eventual export of algal carbon varies between oceanographic provinces (Buesseler,1998). Furthermore, the iron-accumulated algal biomass might be much larger than estimated byday 13 of SOIREE, as indicated by SeaWiFS composite image by 21/22 March 1999 (day 42) ina 1100-km� ribbon-shaped feature with estimated chlorophyll a of up to 3 mg m�� attributed tothe SOIREE bloom (Abraham et al., 2000).

In this study, we investigate the impact of SOIREE on the planktonic ecosystem and thebiological pump, through model simulations over a 60-d period following an initial iron infusion.This duration (4.5 times longer than the SOIREE site occupation) was chosen in order to warrantassessment of the impact of the iron fertilization in these waters on timescales that cover bloomformation and decline. For this purpose, we have used the SWAMCO model (Lancelot et al., 2000),revised to take into account the speciation of the carbonate system and the air}sea exchange ofCO

�(Hannon et al., submitted). This model, driven by in situ hydro-meteorological conditions and

dissolved iron inputs (from both the atmosphere and intermediate waters), explicitly takes into

2746 E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773

Page 3: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

account the biological processes that control the formation and remineralization of carbonbiomass in surface polar waters, as well as those driving export production (POC) and air}seaexchange of CO

�. It describes the cycling of C, N, P, Si and Fe through 19 chemical and 10

biological compartments of the planktonic ecosystem. As far as we know, it is the "rst publishedbiogeochemical model adapted to polar waters that considers explicitly iron limitation, in additionto macronutrients.

In addition to model simulations run for both the iron-fertilized waters (now referred to as &thepatch') and the surrounding waters, we have explored how the outcome of SOIREE might havebeen di!erent if a site with di!erent physical oceanographic properties had been chosen, or ifa di!erent fertilization strategy had been used. These sensitivity analyses provide useful informa-tion for choosing the site, season, and design of future experiments. Finally, the importance ofdi!erent sedimentation mechanisms in determining the magnitude of the downward #ux ofphytoplankton was explored by testing di!erent parameterizations of diatom sedimentation andaggregation. Speci"cally, a possible link between iron depletion and diatoms sinking rate (seeMuggli et al., 1996) was considered and the importance of cell chain formation and particleaggregation was investigated using the model of Kriest and Evans (1999).

2. Methods

2.1. Model description

General concepts, basic assumptions and parameterization of the ecological SWAMCO modelare detailed by Lancelot et al. (2000). In summary, this model integrates the mechanisms control-ling the upper mixed-layer dynamics, biological productivity, and the structure of the planktonicecosystem in the Southern Ocean. In particular, it takes into account the key role of iron in drivingthe structure and functioning of the planktonic system, and related biogeochemical cycles. Statevariables include major macronutrients (ammonium, nitrate, silicic acid and phosphate) anddissolved Fe, permitting many nutrient limitations to be simulated, including any temporal shiftbetween limiting nutrients (such as iron and silicic acid). Two phytoplankton groups (nano#agel-lates and large diatoms) are considered in SWAMCO and are distinguished on the basis of theirphysiology (iron-uptake, sinking rate) and mode of grazer control. Nano#agellates are ingested bymicrozooplankton for which grazing, production and excretion rates are explicitly described in themodel. In contrast, the mesozooplankton grazing constitutes a closure term of the model and isdescribed by a "rst-order loss function of diatom and microzooplankton biomass. Limitation ofphytoplankton growth by irradiance and nutrients is expressed by a double Michaelis}Mentenfunction, accounting for the availability of both intracellular carbon monomers (produced byphotosynthesis and reserve catabolism) and limiting inorganic nutrient. The latter is determined bythe Liebig law of the minimum and corresponds to the nutrient that displays the lowest ambientconcentration compared to the half-saturation constant for phytoplankton uptake. The microbialfood web is represented by dissolved and particulate organic matter (both have two classes ofbiodegradability), bacteria, nanophyto#agellates, bacterivorous nano#agellates and microzooplan-kton, with the latter feeding on both auto- and heterotrophic nano#agellates according to theirrelative abundance (no food preference).

E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773 2747

Page 4: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

The SWAMCO model was primarily developed for application to the Atlantic sector of theSouthern Ocean (Lancelot et al., 2000) and was validated using data from JGOFS cruise ANTX/6(RV Polarstern, austral spring 1992; Smetacek et al., 1997).

Its parameterization (see Tables 2}4 in Lancelot et al., 2000), is derived from our presentunderstanding of relevant biological processes in the Southern Ocean and other HNLC areas, andwas thus assumed valid for application to the SOIREE site/study. However, some minor amend-ments were made to the parameterization as described by Lancelot et al. (2000) in order to considersome process-level results obtained during SOIREE, and contemporary published information onphytoplankton iron-uptake physiology. Firstly, the phytoplankton light adaptation coe$cient wasincreased by 50%, based on "ndings during SOIREE, suggesting cells were adapted to low ambientirradiances (M. Gall, unpubl. data). This most probably re#ects the photo-acclimation of cells tothe lower mean irradiances recorded in the Australasian sector compared with the Atlantic sector.The half-saturation constant for iron assimilation by autotrophic nano#agellates was set to 0.3 nM(i.e. 10-fold higher than used in Lancelot et al., 2000). This value was derived from shipboardiron-enrichments (Boyd et al., submitted) in the Australasian sector, and recent laboratory studiesof algal iron uptake (M. Maldonado, pers. comm.).

In the present version of SWAMCO, nano#agellate and microzooplankton feeding is describedby a Holling-III sigmoid function parameterized according to a synthesis of grazing experimentsreported by Becquevort (1999) for the Atlantic and Indian sector of the Southern Ocean. Thisfunction was preferred to the Michaelis}Menten formulation with a grazing threshold used byLancelot et al. (2000) because it was shown to provide robust stabilization of low phytoplanktonbiomass limit in the HNLC Southern Ocean. During SOIREE, grazing losses of diatom andmicrozooplankton stocks were small as neither krill nor salps were reported (Zeldis, 2001).A constant daily removal of 0.7% of standing stock (phytoplankton) was derived from shipboardgrazing experiments run by Zeldis et al. (2001). In the model, the sedimentation of diatoms isexpressed by a constant sinking rate, "xed to 1.0 m d�� based on the mean sinking rate of largecells ('22 �m) inside the patch (Waite and Nodder, 2001). Finally, we have employed theupgraded version of SWAMCO (Hannon et al., submitted), that includes carbonate systemspeciation and air}sea exchange of CO

�, which permits simulation of the response of the biological

pump to iron fertilization. The speciation of the carbonate system is computed on the basis of thesurface-water temperature, salinity, dissolved inorganic carbon (DIC) content, and alkalinity. DICand alkalinity are, respectively, corrected for biological uptake of carbon and nitrate. The air}seaexchange of CO

�is then estimated from air and sea partial pressure of CO

�(pCO

�) and wind

velocity using the Wanninkhof (1992)'s relationship.

2.2. Model runs

In order to apply the SWAMCO model to SOIREE, two simulations were conducted: one forthe iron-fertilized patch, and the other for the surrounding HNLC waters (not a!ected byiron-enrichment). These two simulations are now referred to as IN and OUT, respectively. Thecoupled model was implemented in a 1D-2 layer hydrodynamical framework composed ofa well-mixed upper layer and a strati"ed deeper layer down to the maximum depth of the euphoticzone (110 m) (Lancelot et al., 1993). The time resolution and vertical discretization were, respective-ly, "xed to 1/2 h and 1 m. Numerical integration was run according to the Runge}Kutta fourth

2748 E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773

Page 5: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

order procedure. Identical initial conditions and hydro-meteorological forcing were applied toboth simulations. Thus, the only di!erence between the IN and OUT simulations is that of the ironreleases (a!ecting simulation IN) and the subsequent losses from the patch as it increases in size (seeAbraham et al., 2000).

Initial conditions were derived from observations in the vicinity of the SOIREE site prior to theiron release. Diatoms and autotrophic nano#agellates stocks were derived from data of mixed-layer integrated carbon biomass estimated from microscopy analysis and size-fractionation (seemethods in Gall et al., 2001a). Since the resolution of the model distinguishes only two algal groups(diatoms and autotrophic nano#agellates), we grouped the size classes de"ned in SOIREE asfollows: cells '20 �m were assumed to be diatoms; smaller size classes (0.2}2 �m, 2}5 �m,5}20 �m) were summed and assumed to be autotrophic nano#agellates. This assumption isvalidated by microscopy observations (Hall and Sa", 2001) indicating that nano#agellates domin-ated the 2}20 �m fraction. On the other hand, all diatoms were larger than 20 �m and dominatedthis algal size class (see Gall et al., 2001a). The pico-phytoplankton (all eukaryotes, Hall and Sa",2001) group is implicitly considered within the state variable `nanophyto#agellatesa, assuming thatthe iron-uptake physiology of these two classes is relatively similar, due to their small size (Sundaand Huntsman, 1997).

The initial biomass of microzooplankton and heterotrophic nano#agellates (grazers of nanof-lagellates and bacteria, respectively), was derived from data on protozooplankton biomass classi-"ed into ciliates and heterotrophic #agellates (Hall and Sa", 2001). The observed taxonomicpartitioning could not be exactly incorporated into the model, due to the simpli"ed trophicresolution of the microbial food web. Indeed, picophytoplankton, a signi"cant component of thephytoplankton community during SOIREE, is not a state variable of the model (it is implicitlyincluded in the `autotrophic nano#agellatesa state variable). The `autotrophic nano#agellatesa areonly grazed by the `microzooplanktona in the model. Yet picophytoplankton is considered to bea prey for nano#agellates. Thus, initializing microzooplankton biomass to the observed stock ofciliates would not permit a su$cient top-down control of the aggregated variable `autotrophicnano#agellatesa. Therefore, the partitioning of the protozoan biomass between the microzooplan-kton and heterotrophic nano#agellates was adjusted to yield an optimal prediction of autotrophicnano#agellates biomass over the 13-d SOIREE. So, the initial biomass of microzooplankton andheterotrophic nano#agellates, respectively, was "xed to 35% and 65% of the protozooplanktonbiomass prior to the iron-enrichment (Hall and Sa", 2001).

The initial concentration of nitrate, orthophosphate, silicic acid and total dissolved iron insurface waters was set to 25, 1.5, 10 �mol l�� and 0.08 nmol l��, respectively, based on pre-releaseunderway data (Boyd and Law, 2001a). The initial surface water content of dissolved inorganiccarbon was set in the model at 2137 �mol kg�� (Bakker et al., 2001).

Physical constraints of the model (mixed-layer depth, sea-surface temperature, solar radiation,atmospheric pressure, sea-surface salinity, wind velocity) were extracted from CTD casts duringSOIREE and the vessel's underway data-acquisition system (including meteorology) (E. Abraham,unpubl. data), for the 10}22 February 1999 site occupation of SOIREE. For the 60-d extended run,the atmospheric forcing was set*from days 14 to 60-using archived meteorological data for 61S140E (NCEP/NCAR project; Kalnay et al., 1996). In the absence of data, sea-surface temperatureand salinity were assumed unchanged from days 14 to 60. The mixed-layer depth was "xed to 65m based on a ship-of-opportunity XBT record 50-d (mid-March 1999) after the onset of SOIREE

E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773 2749

Page 6: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Fig. 1. Decrease in the peak SF�

concentration during SOIREE: "eld observations (�) and adjusted exponentialdecay (*).

(N}S transect along 1403E meridian), which measured a 65 m mixed layer at this time (Abrahamet al., 2000).

The release of dissolved iron inside the patch was considered in the model by increasing theambient dissolved iron concentrations in the mixed layer to the mean level recorded underwayseveral hours after each of the four iron infusions (see Bowie et al., 2001). These in situ ironmeasurements were preferred to theoretical values (calculated from the total amount of ironreleased and the estimated volume of the patch) for constraining the model, due to uncertaintiesregarding the patch volume, the dynamics of dispersal, and the chemistry of the iron in sea water(i.e. possible rapid conversion to particulate form).

In addition to the iron supplied from four infusions, two natural sources were considered:atmospheric deposition of continental dust and input from intermediate waters due to upwellingand vertical di!usivity. Aeolian deposition was "xed to 1 mg Fe m�� yr�� (Duce and Tindale,1991), vertical di!usivity to 0.24 cm� s��, and upwelling velocity to 10�� cm s�� (de Baar et al.,1995). Sub-surface ((100 m) dissolved iron concentrations were assumed to be 0.3 nM, based onvertical pro"les obtained during SOIREE (Frew et al., 2001). These natural sources of iron wereconsidered for both the IN and OUT simulations.

The dispersal of iron, nutrients and biological components was parameterized using the SF�

signal at the center of the patch (Abraham et al., 2000). This dispersal resulted mainly from thehorizontal di!usivity and the strain rate of the patch (Abraham et al., 2000). Two mechanisms ofdispersal were distinguished during SOIREE. Both di!usivity and the strain were extending thearea of the patch during the "rst 3 days of SOIREE, then after day 3 strain became the dominantprocess (Abraham et al., 2000). The dispersal rate was therefore adjusted independently for eachperiod, assuming that for both periods there was an exponential decay of the conservative tracerSF

�(Fig. 1). This resulted in a dispersal rate of 0.013 h�� during days 0}3 of SOIREE, and

0.0013 h�� for days 4}13. This latter value was used to extrapolate beyond day 13 for the extended60-d simulation.

2750 E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773

Page 7: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

These dispersal rates were applied to each state variable of the model, throughout the simulation,by considering the di!erence in their concentrations between the patch center (IN) and thesurrounding waters (OUT), following the equation:

�d>

�dt �

��������

"(>i��

!>i�

)r��������

(1)

with >�

being the state variable i, and r��������

the dispersal rate.The dispersal process therefore acts as either a loss term (e.g. iron and phytoplankton cells) or

a source term (e.g. major nutrients, in particular silicic acid) for elements that may sustain thebloom at the patch center.

3. Results

3.1. Phytoplankton

Figs. 2 and 3 show IN and OUT 60-d SWAMCO predictions of chlorophyll a (Fig. 2) andprimary production (Fig. 3) associated to diatoms ('20 �m) and autotrophic nano#agellates((20 �m) and integrated over the mixed-layer depth. As a general trend, predictions compare wellwith available observations of IN and OUT chlorophyll a and IN primary production measuredduring the 13-d SOIREE site occupation. In particular, the timing and the magnitude of iron-mediated increase in both chlorophyll a (Fig. 2) and primary production (Fig. 3) associated with thediatom fraction are well reproduced by the model. Predicted diatoms, which remain at very lowlevels outside the patch (Chl a'20 �m&0.05 �g l��), become the dominant algal group insidethe patch, as observed during SOIREE (Gall et al., 2001a). Supporting this, SWAMCO predictionsof biogenic silica production rates inside the patch (Fig. 4) are within the range of rates measuredduring SOIREE (Gall et al., 2001b). Also the predicted timing of the increase in uptake rates isconsistent with observations, con"rming a robust simulation of diatom growth rates. As shown bythe 60-d simulation, the dominance of diatoms inside the patch becomes greater after the end ofSOIREE (i.e. day 13). A maximum diatom biomass of 3.4 �g Chl a l�� is predicted around day 23(Fig. 2), beyond which diatom biomass is shown to decrease exponentially up to chlorophylla stocks close to initial value.

On the other hand, observations of autotrophic nano#agellates stocks indicate a peak inchlorophyll a around day 10 (Fig. 2; Gall et al., 2001a), i.e. prior to the maximum in the simulation(around day 17; Fig. 2). Also the maximum for nanoplankton predicted by the model (1.2 �g Chla l��; Fig. 2) is signi"cantly lower than that predicted for diatoms but slightly higher than observed(0.9 �g Chl a l��; Fig. 2). Examination of simulated trends in protozooplankton biomass (Fig. 6)suggests that this discrepancy might be due to an underestimation of the herbivory response to ironenrichment by the model.

SWAMCO predictions of chlorophyll a stocks after day 13 may be compared with those fromSeaWiFS ocean color (Abraham et al., 2000). The composite image 42 days after the onset ofSOIREE (mid-March 1999) displays a 230 km� region with chlorophyll a'1 �g l��, and peakconcentrations of '2 �g Chl a l�� (based on algorithm). The model predicts around day 42

E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773 2751

Page 8: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Fig. 2. Time series of chlorophyll a integrated over the mixed layer, outside (OUT) and inside (IN) the fertilized patch, forthe large ('20 �m) and small ((20 �m) algae: modeling (*) vs. "eld observations (�). The dashed line indicates the endof the SOIREE site occupation.

a mean chlorophyll a of 1.33 �g l��, which corresponds to the declining phase of the iron-stimulated bloom.

3.2. Dissolved iron and silicic acid concentrations

The model predicts a rapid decrease in dissolved iron from 4 to 1.25 nM within the 4 days thatfollow the initial iron release (Fig. 5a). This trend results from the high dispersion rate appliedduring days 0}3 of SOIREE, reducing iron concentrations, as observed in situ (Bowie et al., 2001).The predicted decrease of iron concentration after each of the three subsequent iron infusions ismuch slower, due to the 10-fold decrease in the dispersion rate after day 3. Moreover, during bothdays 0}3, and later, the predicted decline in iron levels in the patch is less abrupt than observed

2752 E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773

Page 9: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Fig. 3. Time series of net primary production integrated over the mixed layer, inside the fertilized patch, for the large('20 �m) and small ((20 �m) algae: modeling (*) vs. "eld observations (�). The dashed line indicates the end of theSOIREE site occupation.

(Fig. 5a). Such observations imply that the rapid loss of dissolved Fe inside the patch was due toa rapid oxidation of the added Fe(II) to Fe(III) and the subsequent formation of Fe(III) oxyhydrox-ide species, a large fraction of which were probably bound to colloidal and particulate ('0.2 �m)phases. These chemical processes are not yet taken into account by the current SWAMCO model.

At the center of the patch, the model predicts that dissolved iron reaches limiting concentrationsfor diatom growth (i.e. 10% of the Fe half-saturation constant of the diatoms) after 25 days(Fig. 5a). At that time, when the decline of the diatom biomass starts (Fig. 2), silicic acid alsoapproaches limiting concentrations for diatom growth (1.7 mmol m�� at day 23). This potentiallimitation by silicic acid is then progressively relaxed following increases in ambient silicic acidconcentration (Fig. 5b) due to mixing with sub-surface waters and entrainment of surroundingHNLC waters (see Abraham et al., 2000). In contrast dissolved iron remains at a low concentration(&0.1 nM), comparable to those in the surrounding waters (&0.08 nM), thus preventing thediatom bloom to further develop.

E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773 2753

Page 10: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Fig. 4. Time series of silicic acid uptake integrated over the mixed layer, inside the fertilized patch: modeling (*) vs. "eldobservations (�).

3.3. The microbial food web

Since the model variables &microzooplankton' and &heterotrophic nano#agellates' do not corres-pond stricto sensu to their &real world' equivalent, due to the trophic resolution of the model (seedetails in the above section &Model runs'), they cannot be validated directly. However, the sum ofpredicted micro- and nanozooplankton biomass (Fig. 6) can be directly compared to protozoop-lankton biomass recorded during SOIREE (Hall and Sa", 2001). Protozoan biomass predicted bythe model agrees well with observations outside the patch, but does not match the 2-fold increase inbiomass observed by 10 inside the patch (Fig. 6). Although the model predicts such an increase, itcorresponds to the timing of the simulated autotrophic nano#agellates (i.e. a peak in microzoop-lankton is seen on day 26 in Fig. 6), and thus it lags the peak observed at day 10 during SOIREEand is about 20% higher. Also, the model predicts a few days time lag between the maximumbiomass of protozooplankton and autotrophic nano#agellates (Fig. 6). Such a delay was notobserved during days 10}13 of SOIREE, when protozooplankton (Fig. 6) and autotrophicnano#agellates (Fig. 2) were blooming together. This suggests that the current parameterization forthe micrograzers fails to reproduce the timing of micrograzer control of autotrophic nano#agel-lates and needs further investigation.

Predicted bacterial biomass remains constant over 13 d (Fig. 6), in agreement with observations(Hall and Sa", 2001). However, the model predicts a signi"cant increase in bacterial productionand biomass inside the patch by day 22, suggesting that bacteria might take bene"t of the ironfertilization as well. Indeed, such an increase is not simulated for the surrounding waters (Fig. 6).

3.4. Downward export (POC) production

All measurements of POC export at 100 m [#oating sediment traps (Nodder and Waite, 2001);���Th (Charette and Buesseler, 2000) and ���C (Trull and Armand, 2001) signature of trapping

2754 E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773

Page 11: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Fig. 5. Time series of dissolved iron (a) and silicic acid (b) concentrations (mixed-layer average) inside the fertilized patch,predicted by the model. Under-way observations of dissolved iron inside the patch during SOIREE (13 days) aredisplayed on plot (a). The dashed line indicates the end of the SOIREE site occupation.

particles] go to prove low export of POC during the SOIREE diatom bloom. Accordingly,SWAMCO does not predict increase of POC export at 110 m inside the patch (Fig. 7). Yet a slightincrease in the export of POC is well simulated at the base of the mixed layer (65 m) from the patchon day 13 (Fig. 7). In the 60-d simulation, the highest POC export (65 m depth) of33.7 mmol C m�� d�� occurs around day 37, i.e. about 15 days after the bloom maximum (Fig. 2).Interestingly, the predicted timing of the peak in export #ux is sensitive to the depth at which this#ux is recorded, due to the time required by particles settling at a rate of 1 m d�� to reach greaterdepths. As shown in Fig. 7, the maximum predicted export #ux at 65 m on day 37 is not recorded at110 m (the depth at which sediment traps were deployed; Nodder and Waite, 2001) within the 60-dsimulation period. Rather model simulations suggest an exponential increase of the downward #uxof POC at 110 m after day 42 (Fig. 7), i.e. &20 days after the predicted diatom bloom (Fig. 2).

E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773 2755

Page 12: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Fig. 6 . Time series of protozooplankton, autotrophic nano#agellates and bacteria biomass (mixed-layer average),outside (OUT) and inside (IN) the fertilized patch, predicted by the model. For comparison, observed protozooplanktonbiomass is plotted as well. The dashed line indicates the end of the SOIREE site occupation.

3.5. Carbonate chemistry and uptake of atmospheric CO�

In response to iron fertilization, the model predicts a signi"cant decrease in dissolved inorganiccarbon (DIC) concentration (!37 �mol kg��) and pCO

�(!79 �atm) (Fig. 8). Minimum values

are predicted by day 36 (much later than the end of SOIREE), followed by increases of DIC andpCO

�until day 60, due to decreases in phytoplankton stocks and primary production, and

dominance of heterotrophic processes after day 30 (Figs. 2, 3 and 6). The predicted DIC content ofsurface waters outside the patch shows little variation (i.e. 2136}2139 �mol kg��, Fig. 8). A smallincrease in pCO

�(4.5 �atm) is predicted for waters outside the patch between days 0}13 due to the

observed warming of surface waters (Bakker et al., 2001). In the model, this trend of increasingpCO

�stops after day 13, since surface temperature is kept unchanged from days 13}60. Predicted

DIC and pCO�

changes during days 0}13 both inside and outside the patch agree well with the

2756 E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773

Page 13: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Fig. 7. Time series of POC export (recorded at 65 and 110 m), predicted by the model, outside (OUT) and inside (IN) thefertilized patch.

trends from the underway surveys during SOIREE. Yet the observed gradient between IN andOUT waters (i.e. the envelope of the underway data in Fig. 8) exceeds that simulated by SWAMCO(Fig. 8). This might be due to the model underestimating of primary production by large diatoms(Fig. 3). However it should be remembered that the model simulates average conditions occurringinside the patch and not the in situ variability recorded during the underway survey (Bakker et al.,2001).

As a consequence of the substantial decrease of the DIC content inside the fertilized patch (days10}35), SWAMCO predicts greater uptakes of atmospheric CO

�(Fig. 8). The maximum daily CO

�uptake occurs on day 43 inside the patch (47 mmol m�� d��). The &jagged' pro"le of CO

�exchange across the air}sea interface (Fig. 8) is due to the changing wind "eld, which together with�pCO

�(between the atmosphere and the sea surface), drives the CO

�#ux, following the Wannin-

khof (1992) relationship.

E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773 2757

Page 14: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Fig. 8 . Time series of surface water pCO�

(a), dissolved inorganic carbon (mixed-layer average) (b), and sea}air exchangeof CO

�(c), outside (OUT) and inside (IN) the patch, predicted by the model. Under-way observations of pCO

�and

dissolved inorganic carbon during SOIREE (13-d) are shown on plots (a) and (b), respectively. Variations of theunder-way signal re#ect the ship movements crossing the patch center and the surrounding waters.

2758 E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773

Page 15: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Fig. 9 . Carbon budget outside (OUT) and inside (IN) the fertilized patch, integrated over 13 and 60-d, predicted by themodel. Uptake of atmospheric CO

�, size fractionated and total primary production (d: diatoms; n: nano#agellates, t:

total phytoplankton), mineralization, export (POC) production, changes of the standing stocks of dissolved inorganiccarbon and total organic carbon. Unit: mmol C m��.

3.6. A carbon budget of the wind-mixed layer

The impact of SOIREE on the functioning of the biological pump was quanti"ed by comparingthe predicted carbon budgets inside and outside the fertilized patch. These budgets were calculatedby integrating net carbon production, POC export, and air}sea exchange of CO

�over the mixed

layer, for the 13-d of the SOIREE study and the extended 60-d simulation (Fig. 9, Table 1). Table1 also compares SWAMCO carbon production by diatoms and autotrophic nano#agellates in thepatch for 13-d, with those obtained by integrating SOIREE primary production measurements ofFig. 3. Both methods show reasonable agreement, although SWAMCO estimates of primaryproduction by diatoms and autotrophic nano#agellates, respectively, are 22% and 6% lower thanthose based on in situ measurements. Also the model predicts a primary production by autotrophicnano#agellates higher than that by diatoms over 13-d, which is contradictory with the in situestimation. This excess production results of the underestimate by SWAMCO of the micrograzercontrol of autotrophic nano#agellates.

As a general trend, the magnitude of all carbon #uxes is signi"cantly higher inside the patch thanin the surrounding waters, especially when calculated for the 60-d simulation. Net carbon produc-tion (the di!erence between primary production and mineralization) accumulated during 60-dreaches 5497 mmol C m�� inside the patch, but only 90 mmol C m�� outside. This di!erence ismainly due to a large increase in primary production inside the patch (6946 vs. 575 mmol C m��),of which 86% is due to diatoms. In contrast, during the 13-d simulation primary production inside

E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773 2759

Page 16: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Table 1Carbon budget of the mixed-layer predicted by the SWAMCO2 model, outside and inside the fertilized patch, integratedover 13- and 60-d. Estimates of primary production inside the patch from in situ measurements are indicated in bold.Units: mmol C m��

Outside Inside

13 d 60 d 13 d 60 d

Primary production (diatoms) 68.9 155.8 316.4(405) 5976.4Primary production (nano#agellates) 159.2 419.4 338.8(360) 970.1Primary production (total) 228.1 575.2 655.2 6946.5Mineralization 130.5 484.7 152.4 1449.6Net carbon production 97.6 90.4 502.8 5496.9Export (POC) production 13.3 98.8 21.3 1200.8Uptake of atmospheric CO

�64.5 232.6 72.2 1204.3

the patch is shared almost equally by diatoms and autotrophic nano#agellates (316 and339 mmol C m��, respectively).

These di!erences in terms of the contribution of algal groups to community production ontimescales of 13 or 60-d are a consequence of the #oristic shift driven by both top-down control ofautotrophic nano#agellates and iron stimulation of poorly grazed diatoms, which occurs after day13 (Figs. 2 and 6). Outside the patch, there is little di!erence in the relative contribution of algalgroups to community production; model predictions integrated over 13 or 60-d indicate thatdiatoms account for 27}30% of primary production (Fig. 9).

The predicted higher production inside the patch over 60 days results in a DIC depletion of4293 mmol C m�� and a simulated accumulation of the standing stocks (i.e. TOC) of4296 mmol C m��. Consequently, the sink of atmospheric CO

�and the downward POC export of

carbon from 65 m are 1204 and 1201 mmol C m��, respectively, over 60-d. Interestingly, these#uxes are very similar, indicating a carbon uptake #ux at the sea surface of the same order as exportproduction in the patch. When integrated over 60-d, 22% of net production in the patch is exportedto depth (deeper than 65 m); in contrast only 4% is exported over 13-d. Again, this illustrates theneed for caution when interpreting the SOIREE observations since the e!ect of iron enrichment onthe carbon pump could be small over 13-d. However, our simulations are inconclusive as to the fateof the carbon "xed inside the patch, since 78% of net production remains as standing stocks (i.e.TOC) after the 60-d simulation. Also this model is unable to predict the fate of exported carbon inmidwater or the deep ocean.

4. Discussion

4.1. Impact of SOIREE on the pelagic carbon cycling

The model results permit an assessment of the impact of SOIREE on the carbon cycling at thespatial scale (10-km length scale) of the fertilized patch. Net production, uptake of atmospheric

2760 E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773

Page 17: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Table 2Impact of the iron fertilization on the net carbon "xation, uptake of atmospheric CO

�and carbon export production, at

the regional scale (150 km�)�

13 days 60 days

Net carbon production 905 (#729) 9893 (#9725)Uptake of atmospheric CO

�130 (#4) 2164 (#1682)

Export (POC) production 38 (#14) 2161 (#1984)

�Numbers within parentheses indicate the e!ect of the iron enrichment. Units : t C.

CO�, and downward POC export were estimated using model results of Table 1, and assuming an

area of 150 km� (based on the patch size, which expanded to about 150 km� during the 13-dexperiment). Based on these assumptions, a net production of 9893 t C, with a subsequent uptake ofatmospheric carbon of 2164 t, and a downward export of 2161 t C are predicted in response to theaddition of &1.7 t of Fe by day 60 (Table 2). This corresponds to a 59-fold increase of netproduction, a 4.5-fold increase of atmospheric CO

�uptake, and a 12-fold increase of POC export,

in comparison with the outside waters simulation (Table 2). From this, only 9%, 6% and 2%,respectively, of global (60-d) net production, uptake of atmospheric CO

�, and POC export occur

during the 13-d SOIREE site occupation (Table 2). Finally, the SWAMCO prediction of905 t C accumulated in the upper waters of the patch by day 13 compares relatively well with the800 t C estimated from primary production measurements (Boyd et al., 2000).

One may consider these estimates from the SWAMCO as conservative, since the comparisonbetween predictions and observations of primary production (Fig. 3, Table 1), DIC and pCO

�(Fig. 8) suggest that the model slightly underestimates primary production. Hence, carbon produc-tion and export induced by SOIREE might be higher than indicated by our estimates. Also, theSOIREE patch size actually increased after day 13, as suggested by SeaWiFS images in late March.Indeed, 42 days after the initial infusion, these images indicate chlorophyll a'1.0 �g l�� over anarea of 230 km�. However, model predictions for the patch center cannot be extrapolated to sucha wide area, since the forcing on primary production induced by the iron release declines from thecenter of the patch to the margin due to mixing and dilution. A better estimation of the local impactof SOIREE on the carbon cycling would require a more elaborated hydrodynamical resolution(2D).

4.2. Inyuence of fertilization strategy and hydrodynamics on ecosystem response

The model simulation of the SOIREE bloom overall provides a reasonably good "t (themicrobial food web excepted) to most of the SOIREE biological observations. This result, as well asformer successful application of the model to contrasted hydrodynamical and chemical conditionsoccurring in the Atlantic sector (Lancelot et al., 2000), suggest that SWAMCO may be used toexplore how the initial conditions and hydrodynamics at a study site and/or the iron releasestrategy adopted might in#uence the outcome of such an in situ iron-enrichment experiment.Model scenarios presented below constitute a preliminary approach to this topic; they should be

E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773 2761

Page 18: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

�Fig. 10. Time series (60-d) of mixed-layer averaged chlorophyll a (a, b), dissolved iron (c, d), and silicic acid concentra-tions (d, e) predicted by the model for di!erent scenarios of iron fertilization: 1. single iron release of 2, 4 and 6 nM (a,c and e); 2. successive iron releases (b, d, f): four additions of 1 nM spread over 10 days, 13 daily additions of 0.3 nM, 20daily additions of 0.3 nM.

considered, however, with caution, in view of the fact that a good agreement between modelpredictions and observations on a limited number of study cases and over short periods does notguarantee the validity of the model for extrapolation.

4.2.1. Experimental designThe design of iron enrichments, i.e. the timing and magnitude of the iron release(s), is expected to

in#uence the iron-mediated response of the planktonic system. The extent of this impact wasestimated by running 60-d simulations for di!erent enrichment strategies. We "rst considered threescenarios for iron enrichment, consisting of one single pulse that raised dissolved Fe concentrationswithin the patch by 2, 4 or 6 nM Fe. Then we conducted three scenarios mimicking multiplesequential infusions: four successive infusions that raised concentrations in the patch each by 1 nMFe (4 nM Fe) over 10 days; 13 daily infusions to raise concentrations each time to 0.3 nM Fe (4 nMFe); and 20 daily infusions to raise concentrations each time to 0.3 nM Fe (6 nM Fe). Irrespective ofhow much iron was added in each case, the dispersal rate of the enriched waters for each scenario isas reported during SOIREE (from the SF

�signal). The single infusion of 4 nM, and the multiple

infusions (i.e. 4�1 nM Fe and 13�0.3 nM Fe) correspond to a total addition of 4 nM of iron.These di!erent designs induce quite di!erent responses (Figs. 10 and 11). As might be expected,

in the case of a single infusion there is a clear correlation between the amount of iron added and theresulting amplitude of the phytoplankton bloom (Fig. 10) as well as corresponding increases in therate of primary production, POC export, etc. (Fig. 11). The daily infusion scenario of 0.3 nM Feover 13-d delays the peak in algal biomass by 5-d, but results in a 2-fold greater chlorophyll a level(Fig. 10), primary production and POC export rates (Fig. 11), compared to the single infusion of4 nM Fe. If the daily infusion of 0.3 nM Fe is carried out for 20 days, silicic acid becomes depletedto 0.5 �M Si (Fig. 10). This result suggests that the bloom maximum of 6.4 �g Chl a l�� predictedfor this scenario constitutes a theoretical limit of the phytoplankton response to a summer ironenrichment in this sector of the Southern Ocean, due to the shift towards another limiting nutrient(Fig. 10b and f). In comparison, the SOIREE design produces a response very similar to themultiple infusions of 4�1 nM Fe or 13�0.3 nM Fe (Fig. 10), but the silicic acid minimumpredicted by the model for the SOIREE is higher (i.e. 1.7 �M Si).

From these scenarios, it can be concluded that the amplitude of the ecosystem response to ironfertilization is determined by the length of the period over which the dissolved iron is maintainedabove limiting concentrations for diatom growth. This non-limitation period depends on the totalamount of iron initially released or the successive infusions (frequency and amount). Successive Feinfusions appear to produce a greater phytoplankton response than a single infusion (whichcorresponds to a similar total addition of iron). It is likely that daily additions of iron aretechnically di$cult to conduct and interpret. Here the SWAMCO model provides a useful tool forselecting the best iron-release strategy to be conducted, under de"ned logistical requirements*toobtain an optimal response of the ecosystem.

2762 E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773

Page 19: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773 2763

Page 20: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Fig. 11. Net primary production, carbon export production, and uptake of atmospheric CO�

(integrated over 60 d)predicted by the model for di!erent scenarios of iron fertilization: SOIREE; single iron release of 2,4 and 6 nM (a, c and e);successive iron releases (b, d, f): four additions of 1 nM spread over 10 days, 13 daily additions of 0.3 nM, 20 dailyadditions of 0.3 nM.

At this stage it should be mentioned, however, that our fertilization scenarios were simulatinginfusions of bioavailable iron. As indicated by SOIREE, some rapid chemical transformation of theadded iron (speciation, transfer of dissolved iron to particulate form) could alter the availability (i.e.the concentration of dissolved Fe available to phytoplankton) in the fertilized patch. Althoughbeyond the scope of the present study, these chemical processes deserve further investigationtowards incorporation into a future version of the model.

4.2.2. The light climate and mixingIn addition to iron supply, the light climate in the upper ocean associated with both incident

irradiance and the wind-driven mixing regime of the water column (both varying regionally andseasonally) may set the upper limit of algal biomass accumulation (see Mitchell et al., 1991;Lancelot et al., 2000) during a mesoscale iron fertilization such as SOIREE. In order to assess thein#uence of light availability on the response of the phytoplankton to the SOIREE iron release, weconducted four simulations of constant mixed-layer depth (40, 65, 100, 120 m). In each case, thetotal amount of iron released is similar to this of SOIREE. Such mixed-layer depths encompass therange predicted for the Australasian-Paci"c sector of the Southern Ocean in summer (Boyd et al.,2000). As shown in Fig. 12, the shallowest mixed-layer (40 m) simulation favors the development ofhigh autotrophic nano#agellates biomass without signi"cantly modifying the diatom bloomrelative to the standard conditions (65 m). This is due to a rapid growth of autotrophic nano#agel-lates under full light and iron non-limiting conditions and the delayed-response of the micro-grazers. As a consequence of this #oristic shift the higher primary production predicted for the 40m mixed layer is not re#ected in the magnitude of POC export, which is less than that for thestandard 65 m simulation (Fig. 13). These SWAMCO scenarios suggest that the magnitude of thediatom bloom, stimulated by the SOIREE experiment, was solely controlled by the iron supply (seeBoyd et al., 2000 Fig. 5), whereas light more than iron was limiting autotrophic nano#agellate

2764 E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773

Page 21: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Fig. 12. SOIREE: sensitivity of the diatoms and autotrophic nano#agellates bloom development to the mixed-layerdepth (from 40 to 120 m), predicted by the model.

growth. At lower mean water column irradiances (simulations of 100 and 120 m), both the diatomsand autotrophic nano#agellates populations attained lower biomass than for the 40- and 65-msimulations (Fig. 12), indicating that the e!ect of the iron addition is negated by these reduced lightlevels.

Results of these model scenarios are in general agreement with the conclusions from theshipboard iron/light experiments during SOIREE, which have shown similar trends in increases inchlorophyll a for the in situ simulated &40-' and &65-m' incubations, and little increase in chlorophylla levels for the &100-m' treatment (Boyd et al., 2000). These model scenarios point to the importanceof site selection in determining the outcome of such iron enrichments. Large regions of thecircumpolar Southern Ocean, especially the Antarctic Circumpolar Current, experience frequentstorm events. The associated wind-mixing events deepen surface layer to depths sometimes inexcess of 100 m, preventing bloom development, independent of iron concentrations (Mitchell and

E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773 2765

Page 22: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Fig. 13. SOIREE: sensitivity of the net primary production, export (POC) production and uptake of atmospheric CO�

(integrated over 60 d) to the mixed-layer depth (from 40 to 120 m), predicted by the model.

Holm-Hansen, 1991; Lancelot et al., 1993). Such waters may not be suitable to conduct in situmesoscale iron enrichment. A series of unpublished SWAMCO scenarios with changing ironconcentration and wind speed (thus ambient light in the upper mixed layer) suggests that regionswith mean wind velocity below 10 m s�� would respond positively to mesoscale iron enrichment.Areas with such favorable weather conditions can be identi"ed from the reconstruction of themonthly wind climate based upon archived meteorological data.

4.3. Inyuence of sedimentation mechanisms on bloom dynamics and carbon export

One important result of 60-d simulations with respect to the impact of the iron-fertilizationexperiment on the biological pump is the low POC-export #ux relative to the large accumulation oforganic carbon in surface waters. This low e$ciency of the biological pump might be due toincorrect SWAMCO parameterization of food web export mechanisms. Iron-mediated diatomsinking rate (Muggli et al., 1996) and diatom chain-forming and aggregation (Smetacek, 1985), aretwo mechanisms that have been suggested to increase the downward POC export. In the followingsection we investigate the in#uence of these processes on diatom bloom dynamics and POC export#ux by including their empirical expression (based on recent studies) in the model.

4.3.1. Sensitivity of export production to iron-mediated changes in diatom sinking ratesSedimentation of POC is parameterized in SWAMCO by a constant sinking rate ("xed at

1 m d�� for this SOIREE study; see Waite and Nodder, 2001). However, the dominant bloomingspecies (Fragilariopsis kerguelensis) displayed reduced sinking rates after iron addition, which thenincreased after 13-d, concurrent with the onset of algal iron stress (Waite and Nodder, 2001). Inorder to evaluate the possible impact of iron supply on diatom sinking rates, and any subsequentchange in the magnitude of POC export, we ran di!erent SWAMCO scenarios in which we alteredthe relationship between diatom sinking rates and iron algal stress. For these scenarios, weconsidered a minimum sinking rate of 1 m d�� for replete iron conditions, and 1.7 and 2.3 m d��

2766 E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773

Page 23: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Fig. 14. Various sinking rate to iron relationships for testing the sensitivity of the phytoplankton bloom to thesedimentation process; see detail concerning the choice of the functions in the text.

to characterize iron-depleted diatoms. The elevated value corresponds to the highest measuredsinking rate on Fragilariopsis kerguelensis in the surrounding HNLC waters during SOIREE(Waite and Nodder, 2001). For each case, we then chose two functions relating algal sinking rate toiron concentration (Fig. 14). One is quasi-linear, the other approximates an inverse exponential,increasing the sinking rate markedly below dissolved iron levels of 0.5 nM. Considering the twovalues chosen for sinking rates of iron-stressed diatoms, we ran in total four scenarios withiron-mediated functions of the diatom sinking rate (Fig. 14).

Interestingly, simulations of changes in diatom biomass are altered little by using either function,and remain similar to predictions using a constant sinking rate of 1 m d�� (Fig. 15). Changes in thesimulated rates of primary production are more sensitive to alteration of diatom sinking rate byiron levels (Table 3). The simulation corresponding to the function �3 (Fig. 14) exhibits a reduc-tion of integrated primary production of 19% (Table 3). POC export increases for each of the fourscenarios relating diatom sinking rate to iron stress, in comparison with the standard run (Fig. 15).As a consequence, a rapid increase in particle #ux (up to 57 mmol C m�� d��) is simulated forscenarios with iron-modulated diatom sinking rate to a maximum of 2.3 m d�� (Fig. 15, scenarios�3 and 4). In this case, cumulative POC export integrated over 60 days was 1803 mmol C m��(Table 3), i.e. 50% more than the standard simulation (constant sinking rate of 1 m d��). Finally,a comparison of simulations with changing parameterization of the iron-modulated sinking rate ofdiatoms indicates a high sensitivity of POC export to the selected sinking rate for iron-stresseddiatoms (Fig. 15) but not to the mathematical function (linear vs. exponential, Fig. 14) used tocalculate the sinking rate under iron-replete and deplete conditions.

4.3.2. Sensitivity of POC export yux to diatom chain forming and aggregation processesThe settling of organic matter is related to particle size and density. Particle size is determined by

the intrinsic cell size but also the ability of individual cells to form chains and colonies or aggregatesupon collision. Video-analysis of samples from surface-tethered free-drifting sediment traps

E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773 2767

Page 24: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Fig. 15. Sensitivity of the diatom bloom development and the carbon export production to the iron-mediated sinkingrate of the diatoms. (0) standard scenario (constant sinking rate"1 m d��); (1), (2), (3) and (4): sinking rate related to ironavailability according to relationships displayed in Fig. 14.

deployed during SOIREE indicates the presence of cell aggregates, including chains of Fragilariop-sis kerguelensis up to 40 cells long (Waite and Nodder, 2001). Aggregation, by changing the sizedistribution of the particles, may thus have greatly a!ected the average sinking rate of ungrazeddiatoms during SOIREE and therefore accelerated the export production. In order to test thishypothesis we implemented in SWAMCO the simple parameterization of phytoplankton aggrega-tion recently developed by Kriest and Evans (1999) and based on the model of Jackson (1990). Insummary, this approach relies on an independent treatment of cells abundances and biomass, eachobeying its own conservation law. Average size and sinking rate of particles change as particlesaggregate or the largest particles sink out. The size distribution of the aggregates is assumed tofollow a hyperbolic law, whose exponent is a function of the average particle size. Equations and

2768 E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773

Page 25: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Table 3Sensitivity of the diatom primary production (PP) and the export (POC) production (EP), integrated over 60 days, to theiron-mediated sinking rate of the diatoms, and to the aggregation process. Units: mmol C m��

PP EP EP/PP (%)

Standard run (constant sinking rate) 5970 1202 20

Iron-mediated sinking rate�Function (1) 5276 1605 30Function (2) 5472 1549 28Function (3) 4819 1803 37Function (4) 5153 1745 34

Aggregation�

p"0.15, s"0.05 3023 1645 54p"0.15, s"0.2 2183 1469 67p"0.3, s"0.05 4884 1989 41p"0.3, s"0.2 3354 2060 61

�See Fig. 14.�p: probability of separation of the cells; s: stickiness factor.

parameters are described and discussed in Kriest and Evans (1999). Essentially, two parameters aredeterminant in this aggregation model: the stickiness factor (s) characterizing the probability thattwo colliding particles will aggregate, and the probability that two dividing cells will separate (p).As a "rst test we choose to run four scenarios of aggregation, in each of which we altered theseparameters (s"0.05 and 0.2; p"0.15 and 0.3, all chosen among literature values, e.g., Jackson,1990; Riebesell and Wolf-Gladrow, 1992; Kriest and Evans, 1999) in order to investigate howaggregation mechanisms might play a role in determining the timing and magnitude of diatombloom and POC export #ux during SOIREE. The minimum sinking rate (for a single diatom cell)was "xed to 1 m d�� (similar to the standard SOIREE run).

The incorporation of an explicit description of diatom aggregation in SWAMCO simulatesa diatom bloom of lower magnitude than the standard SOIREE run (which has no aggregation)(see Fig. 16). In the former, the subsequent bloom decline is more rapid, and the POC export #ux ishigher (Fig. 16). Smaller values of p result in increased diatom aggregation and correspondinglymore cells are exported (Fig. 16, scenarios 1 and 2). This substantial POC export constitutes a lossterm, which keeps the diatom stock at low biomass ((2 �g Ch l a l��). The decrease trend in themagnitude of the diatom bloom at low value of p is reinforced by a higher s with, however, littlee!ect on the POC export #ux (Fig. 16, scenario 2). On the contrary, a high p results in feweraggregates and hence more diatom biomass is predicted to accumulate in the mixed layer (Fig. 16,scenarios 3 and 4). Here, low or high stickiness does not modify much the height of the diatombloom, but slows down or accelerates the decline (Fig. 16, scenarios 3 and 4). Interestingly, thecombination of high p and s induces a short but massive export event occurring only a few daysafter the diatom peak (Fig. 16, scenario 4). This leads to a POC export #ux accounting for 61% of

E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773 2769

Page 26: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Fig. 16. Sensitivity of the diatom bloom development and the export (POC) production to the aggregation process(p: probability of separation of the cells; s: stickiness factor). (1) p"0.15, s"0.05; (2) p"0.15, s"0.2; (3) p"0.3,s"0.05; (4) p"0.3, s"0.2; (0) no aggregation, constant sinking rate"1.0 m d�� (standard scenario).

primary production (relative to 20% in the standard SOIREE run). Finally, it may be noticed thatnone of these scenarios predict signi"cant export production during the 13-d simulation (i.e. timeperiod of the SOIREE observations) (Fig. 16). A longer period of observation at the site would havebeen needed to detect a signal in export production.

These exploratory scenarios show the importance of the aggregation processes on the bloomdecline and the subsequent downward #ux of particles to the deep ocean. In particular, they pointto the importance of diatom cell characteristics such as cell size, chain forming and stickiness on theintensity and duration of export events. Unfortunately, little is known on these aspects of diatomphysiology, and it deserves further research.

2770 E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773

Page 27: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

5. Concluding remarks

SWAMCO is a conceptual mechanistic biogeochemical model that was developed for describingcarbon cycling in the HNLC waters of the Southern Ocean. It was "rst calibrated in the Atlanticsector to the conditions met during the JGOFS cruise ANTX-6 of RV Polarstern in Spring 1992(Lancelot et al., 2000). Its present successful application to SOIREE in the Australasian Paci"csector in February 1999 demonstrates the robustness of the model structure and parameterizationfor predicting algal blooms in these waters and assessing the ecosystem response to an in situ ironfertilization beyond the period of "eld monitoring. From the SOIREE simulation we indeedshowed that the e!ects of the iron fertilization are incomplete over the timescale of the bloommonitoring in SOIREE after the initial iron infusion (13-d). The impact of the iron fertilization interms of increased algal biomass, primary production, export production, uptake of atmosphericCO

�, depletion of major nutrients, and dissolved inorganic carbon reaches its maximum several

weeks after the initial infusion. Experimental design of future in situ iron-enrichment experimentstherefore should include tools to allow monitoring of the post-release stage over a su$cient periodwith respect to the timescale of these mechanisms. This may include remote sensing, but alsoprobably implicates repeated visits to the site of the experiment. Of particular importance is thequanti"cation of the increased carbon-export production following an iron-release experiment.Indeed, neither the results from the sediment traps deployed during SOIREE nor the SWAMCOsimulations over 60 days were able to assess the fate of the iron-stimulated primary productionwith con"dence. With respect to the model, this is due to the inadequate parameterization of themechanisms modulating the mass sedimentation of diatoms such as aggregation. Further experi-mental research on diatom characteristics and physiology is, however, needed to improve thisrepresentation.

Finally, model scenarios with changing ambient light and iron fertilization strategy suggestSWAMCO to be a useful mathematical tool for selecting the most probable area and season fora positive response to a future iron fertilization experiment and guiding the fertilization strategies.For instance, multiple iron infusions at a few days interval rather than one single addition wasa good strategy to conduct on the SOIREE site in late summer 1992. This strategy could di!erbetween areas of the circumpolar Southern Ocean. Hence SWAMCO runs prior to a planned insitu iron-enrichment experiment could be very useful for optimizing the ecosystem response.

Acknowledgements

This modeling work would not have been possible without the enthusiastic contribution of thosecolleagues who took part in the SOIREE cruise. We express our gratitude to Mark Gall, ScottNodder, Julie Hall, John Zeldis, Edward Abraham (NIWA, NZ), Tom Trull (University ofTasmania, AU), Matthew Charette, Ken Buesseler (WHOI, USA), Maite Maldonado (Maine,USA) and Cli! Law (CCMS, UK) for the availability of their unpublished manuscripts and theiruseful comments on their data. This work is part of the European Union CARUSO project fundedunder contract No. ENV4-CT97-0472 by the Environment and Climate Program of the EuropeanCommission and the Belgian research project BELCANTO funded under contract No.A4/DD/B12 in the scope of phase IV of the Belgian Research Program on Antarctic. We also thank

E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773 2771

Page 28: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Hein de Baar (NIOZ, NL), the project leader of CARUSO, for his scienti"c support. We are verygrateful to the referees for their constructive comments and for manuscript proo"ng.

References

Abraham, E.R., Law, C.S., Boyd, P.W., Lavender, S.J., Maldonado, M.T., Bowie, A.R., 2000. Importance of stirring in thedevelopment of an iron-fertilized phytoplankton bloom. Nature 407, 727}730.

Bakker, D.C.E., Watson, A.J., Law, C.S., 2001. Southern Ocean iron enrichment promotes inorganic carbon drawdown.Deep-Sea Research II 48, 2483}2507.

Becquevort, S., 1999. Importance du reH seau trophique microbien dans l'OceH an Atarctique: ro( le du protozooplankton.Ph.D. Dissertation, UniversiteH Libre de Bruxelles.

Bowie, A.R., Maldonado, M.T., Frew, R.D., Croot, P.L., Achterberg, E.P., Mantovra, R.F.C., Warsfold, P.J., Law, C.S.,Boyd, P.W., 2001. The fate of added iron during a mesoscale fertilisation experiment in the Southern Ocean. Deep-SeaResearch II 48, 2681}2701.

Boyd, P.W., Crossley, A.C., Di Tullio, G.R., Gri$ths, F.B., Hutchins, D.A., QueH guiner, B., Sedwick, P.N., Trull, T.W.,. E!ects of iron supply and irradiance on phytoplankton processes in subantarctic waters south of Australia. Journalof Geophysical Research, submitted for publication.

Boyd, P., LaRoche, J., Gall, M., Frew, R., Michael, R., McKay, L., 1999. Role of iron, light and silicate in controlling algalbiomass in subantarctic waters SE of New Zealand. Journal of Geophysical Research 104 (6), 13395}13408.

Boyd, P.W., Law, C.S., 2001a. The Southern Ocean Iron RElease Experiment (SOIREE)*introduction and overview.Deep-Sea Research II 48, 2425}2438.

Boyd, P.W., Watson, A.J., Law, C.S., Abraham, E.R., Trull, T., Murdoch, R., Bakker, D.C.E., Bowie, A.R., Buesseler,K.O., Chang, H., Charette, M., Croot, P., Downing, K., Frew, R., Gall, M., Had"eld, M., Hall, J., Harvey, M.,Jameson, G., La Roche, J., Liddicoat, M., Ling, R., Maldonado, M.T., McKay, R.M., Nodder, S., Pickmere, S.,Pridmore, R., Rintoul, S., Sa", K., Sutton, P., Strzepek, R., Tanneberger, K., Turner, S., Waite, A., Zeldis, J., 2000.A mesoscale phytoplankton bloom in the polar Southern Ocean stimulated by iron fertilisation. Nature 407, 695}702.

Buesseler, K., 1998. The de-coupling of production and particulate export in the surface ocean. Global BiogeochemicalCycles 12, 297}310.

Buma, A.G.J., de Baar, H.J.W., Nolting, R.F., Van Bennekom, A.J., 1991. Metal enrichment experiments in the WeddellSea: e!ects of Fe and Mn on various plankton communities. Limnology and Oceanography 36, 1865}1878.

Charette, M.A., Buesseler, K.O., 2000. Does iron fertilisation lead to rapid carbon export in the Southern Ocean?Geochemistry, Geophysics, Geosystems.

de Baar, H.J.W., Buma, A.G.J., Nolting, R.F., CadeH e, G.C., Jacques, G., TreH guer, P., 1990. On iron limitation of the SouthernOcean: experimental observations in the Weddell and Scotia Seas. Marine Ecology Progress Series 65, 105}122.

de Baar, H.J.W., de Jong, J.T.M., Bakker, D.C.E., LoK scher, B.M., Veth, C., Bathmann, U., 1995. Importance of iron forplankton blooms and carbon dioxide drawdown in the Southern Ocean. Nature 373, 412}415.

Duce, R.A., Tindale, N.W., 1991. Atmospheric transport of iron and its deposition in the ocean. Limnology Oceanogra-phy 36 (8), 1715}1726.

Frew, R., Bowie, A., Croot, P., Pickmere, S., 2001. Macronutrient and trace-metal geochemistry of an in situ iron-inducedSouthern Ocean bloom. Deep-Sea Research II 48, 2467}2481.

Frost, B.W., 1996. Phytoplankton blooms on iron rations. Nature 383, 475}476.Gall, M.P., Boyd, P.W., Hall, J., Sa", K.A., Chang, H., 2001a. Phytoplankton processes. Part 1: Community structure the

Southern Ocean Iron RElease Experiment (SOIREE). Deep-Sea Research II 48, 2551}2570.Gall, M.P., Strzepek, R., Maldonado, M., Boyd, P.W., 2001b. Phytoplankton processes. Part 2: Rates of primary

production and factors controlling algal growth during the Southern Ocean Iron release Experiment (SOIREE).Deep-Sea Research II 48, 2571}2590.

Gran, H.H., 1931. On the conditions for the production of plankton in the sea. Rapport du Conseil International pourl'Exploration de la Mer. 75, 37}46.

Hall, J.A., Sa", K., 2001. The impact of in situ Fe fertilisation on the microbial food web in the Southern Ocean. Deep-SeaResearch II 48, 2591}2613.

2772 E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773

Page 29: Modeling the bloom evolution and carbon ows during SOIREE ... · Modeling the bloom evolution and carbon #ows during SOIREE: Implications for future in situ iron-enrichments in the

Hannon, E., Stoll, M.H.C., de Baar, H.J.W., Veth, C., Lancelot, C. . Control of the CO�

drawdown in the Southern Oceanby iron and wind: a modeling study. Global Biogeochemical Cycles, submitted for publication.

Jackson, G.A., 1990. A model of the formation of marine algal #ocs by physical aggregation processes. Deep-SeaResearch 37, 1197}1211.

Kalnay, E., et al., 1996. The NCEP/NCAR 40-years reanalysis project. Bulletin of the American Meteorological Society77, 437}471.

Kriest, I., Evans, G.T., 1999. Representing phytoplankton aggregates in biogeochemical models. Deep-Sea Research I 46,1841}1859.

Lancelot, C., Hannon, E., Becquevort, S., Veth, C., de Baar, H.J.W., 2000. Modelling phytoplankton blooms and relatedcarbon export production in the Southern Ocean: application to the Atlantic sector in Austral spring 1992. Deep-SeaResearch I 47, 1621}1662.

Lancelot, C., Mathot, S., Veth, C., de Baar, H.J.W., 1993. Factors controlling phytoplankton ice-edge blooms in themarginal ice-zone of the north western Weddell Sea during sea ice retreat 1988: "eld observations and mathematicalmodelling. Polar Biology 13 (6), 377}387.

Martin, J.H., 1990. Glacial}interglacial CO�

change: the iron hypothesis. Paleoceanography 5, 1}13.Martin, J.H., Fitzwater, S.E., Gordon, R.M., 1990. Iron de"ciency limits phytoplankton growth in Antarctic waters.

Global Biogeochemical Cycles 4, 5}12.Mitchell, B.G., Holm-Hansen, O., 1991. Observations and modelling of the phytoplankton crop in relationship to mixing

depth. Deep-Sea Research II 38, 981}1008.Muggli, D.L., Lecourt, M., Harrison, P.J., 1996. E!ects of iron and nitrogen source on the sinking rate, physiology and

metal composition of an oceanic diatom from the subarctic Paci"c. Marine Ecology Progress Series 132, 215}227.Nodder, S.D., Waite, A.M., 2001. Is Southern Ocean organic carbon and biogenic silica export enhanced by iron-

stimulated increases in biological production? Sediment trap results from an in situ iron enrichment experiment.Deep-Sea Research II 48, 2681}2701.

Riebesell, U., Wolf-Gladrow, D., 1992. The relationship between physical aggregation of phytoplankton and particle #ux:a numerical model. Deep-Sea Research I 41 (2), 335}357.

Sedwick, P.N., DiTullio, G.R., Hutchins, D.A., Boyd, P.W., Gri$ths, F.B., Trull, T.W., Queguiner, B., 1999. Limitation ofalgal production by iron de"ciency in the Australian sector of the Subantartic Southern Ocean. Geophysics ResearchLetters 26, 2865.

Smetacek, V.S., 1985. Role of sinking in diatom life-history cycles: ecological evolutionary and geological signi"cance.Marine Biology 84, 239}251.

Smetacek, V., de Baar, H.J.W., Bathmann, U.V., Rutgers van der Loe!, M.M., Lochte, K., 1997. Ecology andbiogeochemistry of the Antarctic Circumpolar Current during Austral spring : Southern Ocean JGOFS cruiseANTX/6 of R.V. Polarstern. Deep-Sea Research II 44 (1}2), 519.

Sunda, W.G., Huntsman, S.A., 1997. Interrelated in#uence of iron, light and cell size on marine phytoplankton growth.Nature 390, 389}392.

Trull, T.W., Armand, L., 2001. Insights into Southern Ocean carbon export from the ���C of particles and dissolvedinorganic carbon during the SOIREE iron release experiment. Deep-Sea Research II 48, 2655}2680.

Waite, A.M., Nodder, S.D., 2001. The e!ect of in situ iron addition on the sinking rates and export #ux of SouthernOcean diatoms. Deep-Sea Research II 48, 2635}2654.

Wanninkhof, R., 1992. Relationship between wind speed and gas exchange over the ocean. Journal of GeophysicalResearch 97 (5), 7373}7382.

Zeldis, J., 2001. Mesozooplankton community composition, feeding, and export production during SOIREE. Deep-SeaResearch II 48, 2615}2634.

E. Hannon et al. / Deep-Sea Research II 48 (2001) 2745}2773 2773