Noble, Abigail E., Carl H. Lamborg, Dan C. Ohnemus, Phoebe J. Lam, Tyler J. Goepfert, Chris I

22 Pages·2012·8.57 MB·English

Limnol.Oceanogr.,57(4),2012,989–1010
E2012,bytheAssociationfortheSciencesofLimnologyandOceanography,Inc.
doi:10.4319/lo.2012.57.4.0989
Basin-scale inputs of cobalt, iron, and manganese from the Benguela-Angola front to
the South Atlantic Ocean
Abigail E. Noble,a,b,1 Carl H. Lamborg,a Dan C. Ohnemus,a,b Phoebe J. Lam,a Tyler J. Goepfert,a,b
Chris I. Measures,c Caitlin H. Frame,a,b Karen L. Casciotti,a Giacomo R. DiTullio,d Joe Jennings,e
and Mak A. Saitoa,*
aDepartmentof MarineChemistry andGeochemistry, WoodsHoleOceanographicInstitution, WoodsHole, Massachusetts
bMassachusettsInstituteofTechnology-WoodsHoleOceanographicInstitutionJointPrograminChemicalOceanography,Department
of MarineChemistryand Geochemistry, WoodsHoleOceanographic Institution,WoodsHole, Massachusetts
cDepartmentof Oceanography,University ofHawaii, Honolulu,Hawaii
dDepartmentof Biology,College of Charleston,Charleston, South Carolina
eCollege ofOceanic &Atmospheric Sciences,Oregon StateUniversity, Corvallis, Oregon
Abstract
We present full-depth zonal sections of total dissolved cobalt, iron, manganese, and labile cobalt from the
South Atlantic Ocean. A basin-scale plume from the African coast appeared to be a major source of dissolved
metals to this region, with high cobalt concentrations in the oxygen minimum zone of the Angola Dome and
extending2500kmintothesubtropicalgyre.Metalconcentrationswereelevatedalongthecoastalshelf,likelydue
to reductive dissolution and resuspension of particulate matter. Linear relationships between cobalt, N O, and
2
O ,aswellaslowsurfacealuminumsupportedacoastalratherthanatmosphericcobaltsource.Lateraladvection
2
coupledwithupwelling,biologicaluptake,andremineralizationdeliveredthesemetalstothebasin,asevidentin
twozonaltransectswithdistinctphysicalprocessesthatexhibiteddifferentmetaldistributions.Scavengingrates
withinthecoastalplumedifferedforthethreemetals;ironwasremovedfastest,manganeseremovalwas2.5times
slower, and cobalt scavenging could not be discerned from water mass mixing. Because scavenging, biological
utilization,andexportconstantlydepletetheoceanicinventoriesofthesethreehybrid-typemetals,pointsources
of the scale observed here likely serve as vital drivers of their oceanic cycles. Manganese concentrations were
elevated in surface waters across the basin, likely due to coupled redox processes acting to concentrate the
dissolved species there. These observations of basin-scale hybrid metal plumes combined with the recent
projections of expanding oxygen minimum zones suggest a potential mechanism for effects on ocean primary
productionandnitrogen fixationvia increasesin trace metalsource inputs.
Cobalt, iron, and manganese play essential roles as either as inorganic cobalt (Saito et al. 2005) or as the
micronutrients in ocean ecosystems. These metals are metallocenter of vitamin B in high-latitude environments
12
involved in photosynthesis, nitrogen fixation, and many (Bertrandetal.2007;Panzecaetal.2008).Cobaltcanalsobe
other cellular processes(Morelet al.2003), butcanexist in incorporated into the active site of carbonic anhydrase in
such low concentrations in seawater that they may limit or diatoms(Moreletal.2003),andisrequiredbytheabundant
co-limitthegrowthofprimaryproducersintheopenocean marine cyanobacteria Prochlorococcus and Synechococcus
(Moore and Braucher 2008; Saito et al. 2008). Iron is (SundaandHuntsman1995;Saitoetal.2002).
predictedtolimit40%ofoceanproductivityandmuchofthe Recent decades of field sampling and careful, low-level
marine nitrogen fixation capacity (Moore et al. 2009). analyses have identified sources and sinks to the water
PhotosystemsIandII(PSIandPSII)areironintensive,and column trace metal inventory that include aeolian depo-
iron is required in many enzymatic pathways of the sition, hydrothermal venting, riverine discharge, an-
microbialnitrogen loop, includingnitrogen fixation (Morel thropogenic-derived input, reductive dissolution, sediment
et al. 2003; Saito et al. 2011). Manganese, while typically resuspension, biological uptake and remineralization, and
abundant in the upper ocean of tropical and subtropical scavenging(BrulandandLohan2003).Withtheseobserva-
waters,isusedtooxidizewatertoO inPSII,todetoxifycells tions, dissolved metal profiles have been categorized
2
from harmful O{ radicals via its incorporation into into conservative,nutrient-like, scavenged,andhybrid-type
2
superoxide dismutase, and it is an electron acceptor for (BrulandandLohan2003).Iron,cobalt,andmanganesefall
manganese-oxidizingbacteria(Teboetal.1984;Moffettand into the hybrid-type category since their distributions are
Ho1996;Moreletal.2003).Cobalthasbeendemonstrated controlled by a combination of phytoplankton uptake,
to co-limit the growth of some phytoplankton in culture which creates nutrient-like distributions in the photic zone,
studies (Saito and Goepfert 2008), and some field studies and scavenging, which creates scavenged-like distribu-
tions in intermediate and deep waters. These three metals
aredistinctfromothertraceelementmicronutrients,such
*Corresponding author:[email protected] as zinc and nickel, because they are redox active, and
1Present address: Department of Earth, Atmospheric and Plane- havesmalloceanic inventories and whole-oceanresidence
tary Sciences, Massachusetts Institute of Technology, Cambridge, times on the order of 40–150 yr due to scavenging
Massachusetts (Brulandetal.1994;SaitoandMoffett2002).Scavenging
989
990 Noble et al.
plays an important role in preventing accumulation of respectively. Shallow and deep casts alternated at a 2.5-
hybrid-type metals with thermohaline circulation, which degree longitudinal spacing. This was followed by a short,
is seen in the accumulation of macronutrients like phos- north-to-south coastal transect (Sta. 20–23) and a short,
phate and nitrate, and nutrient-type trace metals like east-to-west southern transect at 25uS (Sta. 23–27, Figs. 1,
cadmium and zinc between ocean basins. The reduced 2). Acid-cleaned, Teflon-coated X-Niskin sampling bottles
forms of these hybrid-type metals can be introduced to (OceanTest Equipment) of 8-liter capacity were deployed
oxygenated seawater by reductive dissolution within on an epoxy-coated rosette (Sea-BirdElectronics) attached
anoxic pore waters (Heggie and Lewis 1984), exchange to a nonmetallic line and were programmed to trip at
withparticles(Lametal.2006;LohanandBruland2008), predetermined depths and record sampling depths.
organic matter decomposition, and photoreduction All bottles used to store seawater prior to analysis were
(Sunda and Huntsman 1988), while scavenging removes soaked overnight in the acidic detergent Citranox, rinsed
them from the water column. The biotic oxidation of thoroughly with Milli-Q water (Millipore), filled with 10%
manganeseandcobaltbymanganese-oxidizingbacteriais HCltosoakfor10d,rinsedthoroughlywithMilli-Qwater
thought to be an important removal mechanism for these adjusted to pH 2, and double-bagged. Low-density
two metals (Tebo et al. 1984; Moffett and Ho 1996). polyethylene bottles were used for trace metal sample
Complexation by organic ligands is thought to protect storageandhigh-densitypolyethylenebottleswereusedfor
ironandcobaltfromscavengingtosomedegree(Johnson nutrient sample storage. Immediately following collection,
et al. 1997; Saito and Moffett 2001), and recent work has the X-Niskins were pressurized with 99.999% N , and
2
suggested that manganese may also be complexed by seawaterwasfilteredthroughTeflontubinganda142-mm,
some marine siderophores (Parker et al. 2007), although 0.4-mm polycarbonate plastic sandwich filter (Geotech
theinfluenceofcomplexationonmanganesescavengingis Environmental Equipment) into sample storage bottles in
still largely unknown. a positive pressure class-100 clean environment. All tubing
A primary goal of this study was to investigate how the andfilterswereacid-washedpriortouse.Samplesintended
relativedistributionsofcobalt,iron,andmanganesechange fordissolvedironandmanganeseanalyseswereacidifiedto
acrossastrongbiogeochemicalgradientfoundintheoxygen pH 1.7 with high-purity HCl (Seastar) within 6 months of
minimum zone (OMZ) of the South Atlantic Ocean. This sampling, and stored acidified at room temperature for at
wasaccomplishedbycollection andanalysisofafull-depth least 8 months prior to analysis. Samples intended for all
zonal cross section of these dissolved metals during the cobalt analyses were not acidified and were kept at 4uC in
Cobalt, Iron, and Microorganisms from the Upwelling to darkness until analysis.
theGyre(CoFeMUG)cruiseinNovember2007.TheSouth
Atlantic Ocean is understudied with respectto trace metals Cobalt analyses—Concentrations of total dissolved and
(Moore et al. 2009), and understandingthebiogeochemical labile cobalt were determined using a previously described
cycling of metals near the Angola Dome and Benguela cathodic stripping voltammetry (CSV) method (Saito and
Upwellingisofeconomicinteresttoduetothepresenceofa Moffett 2001; Saito et al. 2004) both during the cruise and
major fishery. An OMZ extends from the coast, generated upon returning to the laboratory. Measurements were
byacombinationofhighproductivityandpoorventilation madeusingtheEco-ChemiemAutolab-IIIsystemsconnect-
(Lass and Mohrholz 2008; Mohrholz et al. 2008), and this ed to Metrohm 663 VA Stands equipped with hanging
redox gradient likely affects hybrid-type metal distribu- mercury drop electrodes and Teflon sampling cups.
tions. The South Atlantic has also been presumed to not Standard additions were carried out with Metrohm 765
be subject to iron limitation due to its low macronutrient Dosimats using a programmed dosing procedure (Noble
abundancesandaeolianinput;however,thelowerlimitof et al. 2008).
iron concentrations during the low dust season had not For total dissolved cobalt analyses, samples were
previously been investigated in detail. The cruise track ultraviolet (UV) irradiated for 1 h prior to analysis using
was chosen to cross the oligotrophic gyre and penetrate aMetrohm705UVdigestertodegradetheorganicligands
theBenguelaUpwellingduringaperiodofhighupwelling that bind cobalt and allow binding by the added electro-
and low dust deposition. We present evidence of active cobalt ligand, dimethylglyoxime. Samples were
significant sources of metals to the South Atlantic Basin analyzed in 8.5-mL aliquots with the addition of 30 mL
from the African coast and discuss their biological recrystalized dimethylglyoxime (DMG; 0.1 mol L21 in
interactions and relative removal rates. methanol), 1.5 mL purified sodium nitrite (1.5 mol L21 in
Milli-Q water), and 50 mL purified N-(2-hydroxyethyl)pi-
Methods perazine-N-(3-propanesulfonic acid) (EPPS) buffer
(0.5 mol L21 in Milli-Q water) (see Saito and Moffett
Samplingtechniques—TheCoFeMUGcruiseaboardthe 2001 for cleanup protocols). Analysis began with a 180-s
R/V Knorr (16 November–13 December 2007; Chief purge with 99.999% N . Each sample was conditioned at
2
Scientist, Saito) employed a trace metal sampling program 20.6Vfor90satastirrateof2500revolutionsperminute
for three transects in the South Atlantic Ocean. The main (rpm) followed by a 10-s equilibration step and a linear
transect crossed the basin from west to east, beginning at sweep from 20.6 V to –1.4 V at a rate of 10 V s21. Cobalt
11uS,330uE,andendingat14.75uS,12.2uE(Figs. 1,2).This concentrations were determined by the standard additions
transectconsistedof19stationsthatweresampledto800m technique,withinitialconcentrationsmeasuredintriplicate
or to near bottom in profiles of 11 or 22 depths, followed by four 25 pmol L21 cobalt additions.
Basin-scale inputs of Co, Fe, and Mn 991
While the precision of this method using a single set of non-OMZ samples can be successfully preserved by this
reagents is excellent (e.g., mean and standard deviation of storagetechnique.Infuturestudies,careshouldbetakento
triplicate analyses of a surface-water sample 5 32 6 minimize the number of reagent batches used to avoid
0.7pmolL21),carefulanalysisofreagentblanksisacrucial systematic errors, to carefully constrain and minimize the
component of the cobalt method, and presented a new cobalt reagent blank, and to continually monitor for small
challenge for application to the analysis of an ocean but important systematic error caused by contamination.
section. The analytical blank was determined by analyzing Forlabilecobaltanalyses,8.5mLoffilteredsamplewas
seawater that had been UV irradiated for 1 h, equilibrated added to clean Teflon vials preconditioned with a small
overnightwithpreparedChelex-100resinbeads(Bio-Rad), aliquot of sample water. Thirty microliters of DMG was
and UV irradiated a second time to degrade any leached added to each vial and allowed to equilibrate overnight in
synthetic ligands. Blanks for each batch of reagent (nitrite, the dark (Saito et al. 2004). Analyses were then performed
DMG, EPPS) were subtracted from the initial sample as described for total concentrations with the addition
concentration. During the cruise, two batches of reagents of the remaining two reagents and use of the standard
were contaminated as observed by both an abrupt and addition technique.Previously, we determinedthat natural
systematicincreaseinsampleconcentrationsatSta.13and cobalt ligands in seawater have conditional stability
15 relative to adjacent profiles and by confirmation by constants of . 1016.8 (Saito et al. 2005), suggesting that
reanalysisofthereagentblankwhileatsea.Uponreturning thecobaltisboundverytightlytothecobaltligands.Thus,
to the lab, the systematic offsets for these samples were we define labile cobalt as the fraction of total dissolved
againconfirmedusingtwoindependenttechniques:reanal- cobalt that is either bound to weak organic and inorganic
ysisinthelabbyCSV,andanalysisbyinductivelycoupled ligands in seawater or present as free Co(II), and is then
plasma–mass spectrometry (ICP-MS). This in-lab CSV exchangeable with the complexing agent (DMG) used for
reanalysis of seven profiles (114 samples), performed with analysis(Saitoetal.2004,2005).Thedifferencebetweenthe
one set of reagents with a blank of 6 6 1.7 pmol L21 (n 5 totaldissolvedcobaltandthelabilecobaltcanthenbeused
19), quantified the contamination of the reagents as an as an estimation of the strong cobalt ligand concentration.
average offset of 28 and 47 pmol L21 relative to the pre- Three sets of reagents were used for these analyses, with
contaminated blanks (both 6 pmol L21), and was blanksof461.9pmolL21(n54),560.7pmolL21(n5
consistent with the at-sea reanalysis of the blank (where 3),and761.7pmolL21(n54).
theoffsetwasdeterminedbytheaveragedifferencebetween At the time of this expedition, low-level total seawater
the at-sea and in-lab analyses for a given reagent batch). cobalt intercalibration data were not yet available. In
However, an additional artifact was uncovered by this recentyears,ourlabhasparticipatedintheGEOTRACES
comparison of at-sea to in-lab analyses: data points in the Intercalibration effort (www.geotraces.org), which, upon
100–500-m OMZ range (five data points per profile) were our urging, examined and documented the importance of
found to have a signal loss of cobalt during reanalysis, UV oxidation for total cobalt analysis, and our results are
potentially due to oxidation or precipitation in the sample presented in Table 1 (Bruland 2010).
bottles with time and were excluded from the offset
calculation(NobleandSaitoinprep.).Theaveragedblank Iron and manganese analyses—Total dissolved iron and
concentrations for the seven reagent batches used in this manganese were measured using ICP-MS, as described by
data set were 4 6 1.9 pmol L21 (n 5 4), 6 6 1.7 pmol L21 Saito and Schneider (2006). Briefly, 13.0-mL aliquots of
(n519),761.7pmolL21(n54),761.7pmolL21(n5 acidifiedseawaterwereweighedintocleanedpolypropylene
4), 8 6 0.9 pmol L21 (n 5 3), 28 6 1.4 pmol L21 (n 5 4), centrifuge tubes and 57Fe was added for isotope dilution
and 49 6 6.6 pmol L21 (n 5 2). Twelve stations were analysis and equilibrated overnight. Concentrated ammo-
analyzed 15 months after sampling (4, 6, 8, 10, 12, 14, 16, nium hydroxide (Seastar) was added to induce Mg(OH)
2
23,24,25,26,27).Ofthosestations,10,12,14,and16have and trace metal coprecipitation for 3 min, followed by
been excluded from the data set presented here because centrifugation for 3 min at 3000 rpm (1460 3 g) using an
samples appeared to have experienced some loss of cobalt EppendorfCentrifuge5810R.Thesupernatantwasdecant-
withtimewithintheOMZdescribedabove(,100mmolO ed,thesamplewascentrifugedanddecantedagain,andthe
2
kg21), which was apparent by comparison to adjacent pellet was dissolved in 5% nitric acid (Seastar) containing
profiles that were measured at sea. In contrast, Sta. 4, 6, 1 mg L21 indium. The 5% nitric acid resuspension solution
and 8 (found outside the OMZ) were oceanographically was used to estimate the 56Fe, 57Fe, and 55Mn reagent and
consistent with adjacent profiles that were analyzed at sea, instrument blank, and signal suppression due to matrix
and have been included, as well as Sta. 23–27. With the effects was accounted for by using a ratio of the indium in
exception of two data points from Sta. 23 and four data theblanktotheindiumintheresuspendedsamplesolution,
points from Sta. 24,the in-lab analyses from these stations as described in Saito and Schneider (2006). The Sampling
were from waters with oxygen concentrations above and Analysis of Fe (SAFe) seawater intercalibration
100mmolkg21.Additionally,goodreproducibilitybetween standards were analyzed at the beginning of each analysis
shipboard and shore-based determinations was observed daywithironconcentrationsof0.9660.095nmolL21for
previously for samples from the well-oxygenated Ross Sea the D2 seawater standard and 0.125 6 0.046 nmol L21 for
. 17 months after sampling. In this case the shore-based the S1 seawater standard (n 5 10), which are within the
recovery was 97% 612% relative tothe atsea values (n 5 reported ranges of 0.91 6 0.17 nmol L21 and 0.097 6
9,A.E.NobleandM.A.Saitounpubl.).Thissuggeststhat 0.043 nmol L21 (Johnson et al. 2007).
992 Noble et al.
Basin-scale inputs of Co, Fe, and Mn 993
Fig. 2. Full-depthoceansectionsofmacronutrients,andoxygenfortheCoFeMUGexpedition.ThereisanOMZintheeasternend
of the basindueto highproductivityand subsequentorganicmatter decomposition andthe limited watermass ventilation.
Sampling and analysis of N O—Nitrous oxide (N O) stoppers (MicroLiter Analytics) and aluminum crimps.
2 2
sampleswerecollectedfromthetracemetalrosettebytwice Poisoned samples were stored for at least 2 months before
overfilling glass 165-mL serum bottles (Wheaton) using analysis; however, previous research has demonstrated the
Tygon tubing. They were poisoned with 100 mL of integrity of this storage method for up to 6 months
saturated HgCl solution and then sealed with butyl (McIlvin and Casciotti 2011). Isotopic analyses of N O
2 2
r
Fig. 1. Cruisetrackwithstationlocationsandfull-depthoceansectionsoftotaldissolvedcobalt,labilecobalt,totaldissolvediron,and
total dissolved manganese for the CoFeMUG expedition across the Southern Atlantic Ocean. A large plume of cobalt, with
corresponding but smaller plumes of iron and manganese, originates at the Southwest African coast and extends to the center of the
basin.Coastalsourcesare likelyduetoacombination ofreductivedissolution andresuspension ofparticulatematterinthesediments
along the shelf. A plume of iron and manganese also extends westward from the Mid-Atlantic Ridge, with no corresponding elevated
concentrationsforcobalt(M.A.SaitoandA.E.Nobleunpubl.).Notethatthemaintransect(E–W),themeridionaltransectalongthe
Namibiancoast(N–S),andthesouthernzonaltransect(W–E)areconnectedtoformacontinuouswraparoundsectionforeachanalyte,
bycruise track distancefrom Sta. 1,numbered sequentially. Theasterisk marksthe endof the maintransect.
994 Noble et al.
Table 1. SAFe (S1 and D2) and GEOTRACES (GS and metal rosette Sea-Bird SEACAT SBE-19 sensor. Titration
GD) standard seawater cobalt analysis results for UV-digested water samples were taken immediately after rosette
samples. All valuesare inpmol L21.nd,no dataavailable. recovery by overfilling volume-calibrated flasks by one
flask volume and then sealing with ground-glass stoppers.
Current
Standard Winkler reactions with iodate standards were
Sample Dec08 Aug10 Oct10 consensus
performed and used to calibrate the Sea-Bird sensor and
S1 761.4(n55) 462.9(n54) nd 5.462.2 SEACAT sensor data. In cases where data were available
D2 4965.8(n55) 4562.9(n52) nd 45.463.8 frombothsensors,theSBE-43sensordatawereusedrather
GS 3161.7(n54) 2561.2(n54) 3162.9(n57) 3567
thantheSEACATsensordataduetoitscloserrelationship
GD 7065.3(n54) 5965.2(n56) 7065.2(n59) 6568
tothe calibration titrations (r2 5 0.9995 and 0.995 for Sta.
1 and 13, respectively).
were made using a Finnigan DeltaPLUS XP isotope ratio
mass spectrometer (McIlvin and Casciotti 2010). Particle trap collection and analysis—Sinking particulate
matter was collected using a surface-tethered, drogued-at-
Nutrient and oxygen analyses—Nutrient samples were depth system and acid-cleaned polycarbonate particle
filtered and frozen in acid-washed 60-mL high-density interceptor tubes, similar to that used during JGOFS and
polyethylenebottlesuntilanalysis.Thefrozensampleswere at Hawaii Ocean Time Series and Bermuda Atlantic Time
thawedinawarmwaterbathandstoredinthedarkfor20– Series (Steinberg et al. 2001). Collection tubes were held in
24 h prior to analyses. This protocol has been found to polyvinylchlorideframesatdepthsof60,150,and500mat
increase the recovery efficiency of silicic acid in frozen Sta.13,andonlytheshallowertwodepthsatSta.19.Sixto
samples, and has no observed adverse effects on the other eighttubesweredeployedateachdepth,andheld250mLof
nutrients. Immediately before analysis, aliquots of the a double-strength seawater brine (made by freezing filtered
samplesweretransferredto15-mLpolypropylenecupsand uncontaminated local seawater)at thebottom of each tube
an Alpkem autosampler. Technicon AutoAnalyzer IITM below a layer of filtered seawater that was buffered with
components were used to measure phosphate and ammo- borate to pH 8.2 and that doubled the alkalinity. No
nium; and Alpkem rapid flow analyzer (RFA) 300TM additionalpreservativewasused,andwehavefoundthatfor
components were used for silicic acid, nitrate + nitrite, and short-duration deployments, brine without preservative
nitrite. All five of the macronutrients were analyzed gives comparable fluxes of carbon, nitrogen, and thorium
simultaneously. The nutrient methods were essentially as did brines preserved with formaldehyde (Lamborg et al.
those employed by the Oregon State University lab during 2008).Thesystemwasdeployedfor3datSta.13and4dat
the World Ocean Circulation Experiment and Southern Sta. 19. Two capped tubes, loaded with filtered water and
Ocean Joint Global Ocean Flux Study (JGOFS) cruises. brine,wereonboardthearrayandusedasaprocessblanks,
The phosphate method was a modification of the molyb- and subjected to all the same processing protocols as the
denum blue procedure of Bernhardt and Wilhelms (1967), samples.Afterrecovery,thetubeswereallowedtostandfor
in which phosphate was determined as reduced phospho- . 1 h and brine and particle mixture screened (350 mm) to
molybdic acid employing hydrazine as the reductant. The remove zooplankton swimmers. The remaining particles
nitrate + nitrite analysis used the basic method of were collected onto acid-cleaned and pre-weighed 25-mm-
Armstrong et al. (1967). Sulfanilamide and N-(1-napthyl) diameter, 1-mm-pore-size polycarbonate Nuclepore filters.
ethylenediamine dihydrochloride react with nitrite to form Following the filtering, the membranes were rinsed with a
a colored diazo compound. For the nitrate + nitrite small amount of pH 8.2, borate-buffered Milli-Q water.
analysis, nitrate is first reduced to nitrite using an open Prior to analysis, samples were desiccated and weighed
tubular cadmium reductor and imidazole buffer as cleanly in a laminar flow bench. Minor and trace elements
described by Patton (1983).Nitrite analysiswas performed weredeterminedbyICP-MSasdescribedinLamborgetal.
on a separatechannel, omittingthe cadmiumreductor and (2008).
the buffer. The determination of silicic acid was based on
that of Armstrong et al. (1967) as adapted by Atlas et al. Suspendedparticlecollectionandanalysis—Size-fraction-
(1971). Addition of an acidic molybdate reagent forms ated particles were collected via in situ pumps (McLane
silicomolybdic acid, which is then reduced by stannous Research Laboratories, Inc.) on McLane-style filter hold-
chloride. An indophenol blue ammonium method was ers. The . 51-mm size fraction was collected on acid-
modified from Alpkem RFA methodology, which refer- cleaned polyester prefilters (Sefar), followed in-line by
ences Methods for Chemical Analysis of Water and collectionofthe1–51-mmsizefractiononcombusted,acid-
Wastes, March 1984, EPA-600/4-79-020, ‘‘Nitrogen Am- cleanedquartzmicrofiberfilters(WhatmanQMA).Process
monia,’’ Method 350.1 (Colorimetric, Automated Phe- blank filters were a complete filter set that was loaded in a
nate). A detailed description of the continuous segmented plastic housing, attached to a pump, and deployed. These
flowproceduresusedcanbefoundinGordonetal.(1994). ‘‘dippedblank’’filterswereexposedtoseawaterbutdidnot
Oxygen sensor data were calibrated by five Winkler have water actively pumped through them, and were
titrations performed at Sta. 1, 2, 13, 22, and 24. Oxygen otherwise processed identically. QMA subsamples were
data were obtained from the conductivity, temperature, digested in Teflon bombs (Savillex) in 50% v v21 nitric
depthrosetteSea-Birdsensor(SBE-43)ateverystation (0– acid (Seastar Baseline) for 4 h at 135uC, centrifuged to
800m)andateverydeepstation(fulldepth)fromthetrace remove filter slurry, and diluted to 5% v v21 HNO for
3
Basin-scale inputs of Co, Fe, and Mn 995
analysisbyICP-MS(ThermoElement2).Indiumwasused Dissolvedmetaldistributionsandregionaloceanography—
as an internal drift monitor, and concentrations were The western and central region of the main transect
determined using multielement external standard curves. sampled the oligotrophic subtropical gyre and was
The recovery of manganese, cobalt, and iron from the characterized by surface drawdown of total cobalt to
MESS-3 sediment standard using this digestion procedure 6 pmol L21, labile cobalt to below detection, and iron to
was 97%, 94%, and 84%, respectively, of certified values. 0.05 nmol L21, while surface manganese concentrations
All reported concentrations have had the process blanks remained consistently elevated across the entire main
removed.Samplehandingandanalyseswereallconducted transect with values ranging from 1.93 to 4.04 nmol L21
in class-100 clean environments following trace metal (Figs. 1, 3). Along the coast of Namibia, elevated
procedures. concentrations of all three metals were found that
extended far into the South Atlantic basin along the main
Data analysis and repository—Linear relationships transect.Forexample,intheupper60mofSta.18and19,
between cobalt and phosphate, oxygen, N O, and d15N concentrations reached as high as 200 pmol L21 Co,
2
were determined by applying a two-way linear regression 4.04 nmol L21 Mn, and 3.03 nmol L21 Fe with a second,
least squares fit in MATLAB (MathWorks Inc.) using the spatiallypronounced,westwardpropagatingfeatureforall
script lsqfitma.m. The cobalt plume size calculations were three metals observed around 400 m (163 pmol L21 Co,
determined by creating a grid of the South Atlantic basin, 3.1 nmol L21 Fe, 1.1 nmol L21 Mn; Figs. 1, 3, 4). Of the
where each sampling point represented the corner of an three metals, the subsurface cobalt extended farthest into
enclosed area (either trapezoidal or triangular in shape). the basin, for a distance of approximately 2500 km from
The area of each box was determined by depth and the Namibian coast. This cobalt plume comprised 37% of
distance between data points, and the total cobalt in each the upper 1000 m of the zonal section area, but
box was determined by calculating trapezoid areas to disproportionately contributed 53% of the total cobalt
interpolate between sampling points across the basin. The therein (Fig. 4; see Methods for calculations). The
plume was defined by the 100 mmol kg21 O contour potential contributions to this cobalt plume are discussed
2
wheretheelevatedcobaltwasfound.Thecobaltwithinthe in later sections.
plume was divided by the sum of the cobalt within the WithintheAngolaGyre,theedgeoftheAngolaDomeis
upper 1000 m to estimate relative inventories. The trace identifiedbyahightemperaturesignatureatSta.17and18
metaldatapresentedinthispaper(429totaldissolvediron between 0- and 50-m depth (Lass and Mohrholz 2008;
values, 429 total dissolved manganese values, 346 total Fig. 5g). Cobalt, iron, and manganese concentrations at
cobaltvalues,and290labilecobaltvalues,withfewerthan 400 m were 2.2, 3.8, and 3.8 times higher in the Angola
7 outliers for each species analyzed) have been deposited Gyre (Sta. 15–19), respectively, than at comparable depths
into the Biological and Chemical Oceanography Data inthecenteroftheoligotrophicsubtropicalgyre(Sta.3–5),
Management Office database under the CoFeMUG and 2.6, 2.5, and 1.9 times higher than at comparable
program (http://bcodmo.org/). Ocean sections were creat- depths off the coast in the southern transect (Sta. 25–27).
ed in Ocean Data View with the exception of these few The elevated metals generally co-occurred with the oxygen
outliers (Schlitzer 2011). depletion that occurs as a result of organic matter
remineralization and poor ventilation in the shadow zone
Results near the African coast (Figs. 2, 4, 5a; Lass and Mohrholz
2008; Mohrholz et al. 2008). South Atlantic Central Water
The CoFeMUG expedition consisted of three transects: (100–500 m) carries this low oxygen signature south and
The main zonal transect (Sta. 1–19; main transect eventuallywestwardupon convergence withEasternSouth
hereafter), a short meridional transect along the coast Atlantic Central Water (Mohrholz et al. 2008). Additional
within the Benguela Upwelling (Sta. 19–23), and a short oxygen consumption is driven by the Benguela Upwelling
transectsouthofthemaintransectextendingfromtheshelf to the south, which also injects nutrients into surface
backintothegyre(Sta.23–27;southerntransecthereafter). waters.
The cruise proceeded from west to east across the basin, In contrast to the high abundances observed offshore in
from regions of low productivity in the oligotrophic the main transect through the Angola Gyre (Fig. 4), the
subtropicalgyretohigherproductivity intheAngolaGyre cobalt, iron, and manganese concentrations were lower in
and finally to the highest productivity regions associated the offshore southern zonal transect (Fig. 1). Metal
with coastal upwelling of macronutrients (Figs. 1, 2). The concentrations were high along the coast at Sta. 23,
high-productivitycoastalregionswerefurtherdistinguished reaching203pmolL21Co,7.9nmolL21Fe,and2.8nmol
bywatermassesoftheAngolaDome(thesouthernedgeof L21 Mn, and dissolved oxygen was low (as low as
which is observed in Sta. 17 and 18), the Benguela 11 mmol kg21), but this signal did not extend westward at
Upwelling region that covers much of the Namibian and this latitude. This likely reflects the differences in physical
SouthAfricanAtlanticcoast(WeeksandShillington1994), processes where the main transect encounters surface and
and the dynamic Angola-Benguela Frontal Zone (ABFZ) subsurface westward advection resulting from the conver-
that lies between them (roughly between 14.5uS to 18.5uS gence of the southward-flowing Angola Current and the
for the northern and southern extents varying by season northward-flowing Benguela Coastal Current (at the
and climatology; Kostianoy and Lutjeharms 1999; Lass ABFZ) and from the northwestward-flowing Benguela
and Mohrholz 2008). Ocean Current that is driven by the strong trade winds
996 Noble et al.
Fig. 3. Representative vertical profiles from the CoFeMUG data set, showing the
subsurface maxima forcobalt, iron, andmanganese with proximityto the coast.
(Peterson and Stramma 1991; Mohrholz et al. 2008). In 1:4260,pCo:dCo<1:180,pMn:dMn<1:14withinthe
contrast, the southern zonal transect crossed the north- gyre,p5particulateandd5dissolved;Fig. 6).Particulate
ward-flowingBenguelaCoastalCurrentwhereitlikely still ironconcentrationsincreasedatSta.19tovaluesthatwere
follows the coastline north, and water mass movement 7 times higher than the dissolved counterpart (Fig. 6g).
westward is more restricted (Peterson and Stramma 1991). While particulate cobalt concentrations also increased at
A discussion of the hydrothermal contributions of iron the bottom sampled depths (Fig. 6f), the dissolved coun-
and manganese plumes found above the Mid-Atlantic terpart was , 20 times higher than the particulate
Ridge (M. A. Saito unpubl.) will be presented elsewhere. concentrations, and a similar trend was observed for
Lastly, strong correlations between cobalt and phosphate manganese (dMn , 10 times pMn) (Fig. 6h).
are observed in surface waters, and a discussion of the At the open ocean station, particulate profiles exhibited
ecological stoichiometry of cobalt in this region relative to surfacemaximaforcobaltandphosphorusbutnotforiron
the North Atlantic (A. E. Noble unpubl.) will also be and manganese (Sta. 13, Fig. 6a–d). Instead, particulate
presented elsewhere. iron and manganese showed steady increases in concentra-
tion from the surface to 600 m (Fig. 6c,d), and the
Particulate distributions—Particulate phosphorus, co- particulate concentrations were respectively 140- and 6-
balt, iron, and manganese concentrations are presented fold less than the respective coastal particulate concentra-
for a coastal station (Sta. 19) and an open-ocean station tions (P. Lam and D. Ohnemus unpubl.; Fig. 6c,d,g,h).
(Sta. 13). For all elements, the particulate concentrations
observed at the coastal station were higher than were Particle fluxes—Sediment trap data are reported at 60-,
observed at the open-ocean station. 150-, and 500-m depth foran open-ocean station (Sta. 13),
At the coastal station, dissolved and particulate cobalt, and at 60- and 150-m depth for a coastal station (Sta. 19).
iron, and manganese showed parallel trends with depth Attheopen-oceanstation,thetrapdatashowedadecrease
(Sta. 19, Fig. 6f–h), and in all cases but that of iron, the in flux with depth between the 60-m and 150-m traps
relative concentrations of the particulate material were (Fig. 7b–e). The integrated input from remineralization
orders of magnitude lower than that of the dissolved was calculated by taking the difference between the fluxes
counterpart (pP:dPO3{ < 1:780, pCo:dCo < 1:25, anddividingitbythedepthbetweenthetwotraps.Bythis
pMn:dMn < 1:6 al4ong the coast and pP:dPO3{ < estimate, remineralized metal fluxes of 0.013 pmol Co
4
Basin-scale inputs of Co, Fe, and Mn 997
Fig. 4. Expanded upper water column sections for the northern transect. Metal
concentrations decrease with distance from the coast, and the coincidence of the OMZ with
iron, manganese, and particularly cobalt, is evident. For reference, the Mid-Atlantic Ridge is
centered over15uW.
L21d21,8.5pmolFeL21d21,0.11pmolMnL21d21,and Discussion
11.2 nmol P L21 d21 were delivered to the water column
between these sediment trap depths. The overarching scientific objective of this research
At the coastal station, the 60-m flux was indistinguish- cruisewastoexaminethedistributionsofcobalt,iron,and
able from the open-ocean flux within the error of the manganeseintheOMZoftheSouthAtlanticandBenguela
measurement. The sediment trap data did not detect UpwellingRegion,motivatedbypreviousobservationsof
remineralization for iron, manganese, and phosphorus elevated cobalt in the OMZs of both the Costa Rica
(Sta. 19, Fig. 7b,d,e); however, the total mass flux at Dome and the Peru Upwelling Region (Saito et al. 2004,
150 m was larger than that at 60 m, with no noticeable 2005). The metal distributions within three broad geo-
change in the iron, manganese, and phosphorus fluxes graphic distinctions are presented in this paper: the low
(Fig. 7a,b,d,e), which may be due to particulate cycling surface concentrations in the oligotrophic gyre, the large
processes other than remineralization. plumesobservedwithinthesubsurfaceOMZoftheSouth
998 Noble et al.
Fig. 5. Hydrographicdistributionsinthenorthern,coastal,andsoutherntransectsnearthe
Southwest African coast for the (a–c) dissolved oxygen (mmol kg21) in the upper 800 m, (d–f)
salinity in the upper 300 m, and (g–i) potential temperature (uC). Oxygen concentrations are
muchhigherinthesoutherntransectthaninthenorthernandcoastaltransects,andlow-oxygen
watersalongthecoastdonotappeartobeadvectedwestwardinthesouthastheyareinthenorth.
Atlantic,andthelocalized,elevated concentrations along its uniformity across the region is perhaps surprising.
the coast. The salient, large basin-scale features apparent Aeolian deposition isoften invoked as a primary source of
in the ocean section are presented first, followed by the manganese to the surface ocean (Statham et al. 1998).
coastal observations and discussion of potential supply Given the low dust deposition predicted for this region
mechanisms. (Fig. 8a), these observations suggest that there is likely an
additional mechanism working to maintain the uniform
Upper Ocean distributions of dissolved cobalt, iron, surfacemaxima,suchasredoxcyclingprocessesthatactto
and manganese—The oligotrophic waters of the western retain dissolved manganese at the surface. Manganese-
andcentralregionsofthemaintransectwerecharacterized oxidizing bacteria are photoinhibited in the mixed layer,
bysurfacedrawdownoftotalcobalt,labilecobalt,andiron but below that, they are known to oxidize manganese,
due to biological utilization (Figs. 1, 3). These low surface creatingsmallneutrallybuoyantmanganeseoxideparticles
cobalt and iron concentrations (6 pmol L21 and below that may be re-entrained into the mixed layer and
detection for total and labile cobalt, respectively, photoreduced (Sunda and Huntsman 1988; Tebo et al.
0.05 nmol L21 iron) were within a range that can be 2004). The combination of aeolian input from above and
limiting to some phytoplankton growth (Sunda and entrainmentandrecyclingofmanganeseoxidesfrombelow
Huntsman1995), althoughafewstations mayreflectsome could act to create and maintain the elevated dissolved
aeolian deposition that could relieve micronutrient stress manganese surface-water inventory we observe across the
(Sta. 9, 10 m; 0.34 nmol L21 Fe, 6 pmol L21 labile Co). gyre, similar to redox processes that can concentrate
These low surface concentrations are consistent with dissolved and particulate metals at distinct strata in
modeling studies by Mahowald et al. (2005) that predict sedimentary environments.
low dust input to the South Atlantic (Fig. 8a), relative to Elevated concentrations of cobalt, iron, and manganese
large inputs like the Sahara Desert to the north (Vink and were found along the coast of Namibia and extended into
Measures 2001). the South Atlantic basin in the main transect (Fig. 4).
In contrast to cobalt and iron, manganese remained Subsurface cobalt concentration at around , 400 m
consistently elevated in surface waters across the entire extended farthest into the basin, for a distance of ,
main transect, with values ranging from 1.93 to 2500 km from the Namibian coast. As described in the
4.04 nmol L21. While this surface maximum is typical of Results section above, this cobalt plume spatially made
manganese profiles from tropical and subtropical regions, up 37% of the upper 1000 m of the zonal section, but