Abstract. We present in situ electrode measurements of sediment resistivity, pore water oxygen, ... and calcium carbonate in pelagic sediments of the western.
GLOBAL BIOGEOCHEMICAL
CYCLES, VOL. 10, NO. 3, PAGES 527-541, SEPTEMBER 1996
Calcite dissolutionin sedimentsof the Ontong-Java Plateau: In situ measurementsof pore water Oz and pH Burke Hales Lamont-Doherty EarthObservatory of ColumbiaUniversity,Palisades, New York
Steve Emerson Schoolof Oceanography, Universityof Washington, Seattle
Abstract. We presentin situelectrode measurements of sediment resistivity, porewateroxygen, andporewaterpH fromthreestations between2300and3000m depthontheOntong-Java Plateauin thewesternequatorial Pacific.Oneof thesestations is alsothesiteof a concurrent benthicchamberincubation experiment [Jahnkeet al., 1994]. Theporewateroxygendataanda steadystatediffusionandreactionmodelconstrain thedepth-dependent rateof oxicrespiration in
thesediments andimplya diffusive fluxofoxygen tothesediments of 10-21gmolcm-2yr-1. Giventheserespiration rates,theporewaterpH datacannotbeexplained withoutcalcite dissolution drivenby metabolically produced CO2. Thedissolution necessary to explainthe
observations, quantified bya statistical approach, is3.5-6gmolcm-:yr-1,whichcorresponds toat least20-40% of the calciterain to thesesediments.Over 65% of the total dissolutionis drivenby
metabolic CO:. Oxygenfluxesandnetcalcitedissolution constrained by theelectrode dataare compatible withthebenthicchamber measurements of Jahnkeetal. [ 1994]. Thedissolution flux, whilea significant partof theearlydiagenesis of calcitein thesesediments, is lessthanwouldbe predicted by earliermodelsof dissolution, andJahnkeetal. [1994]probably couldnotdistinguish it from zerowith thebenthicchambertechnique.The dissolution ratesfoundin thisstudyare lowerthanpreviousestimates because therespiration reactionis concentrated nearthesedimentwaterinterface,andthe calcitedissolution rateconstants arevery small. The statisticalevaluation of theporewaterpH dataandmodelconstrain thecalcitedissolution rateconstant to 0.005-0.16%
d'1,followingthegeneraltrendof lowervaluesdetermined byin situtechniques ratherthanby laboratorymethods.
Introduction
We describea studyof the early diagenesisof organiccarbon and calcium carbonate in pelagic sedimentsof the western equatorialPacific. Our goals were (1) to determinethe total respirationand dissolutionfluxes within the sedimentsand comparethem to the ratesof supplyto, and burial within, the sediments;(2) to separatethe effects on calcium carbonate dissolution due to bottom water undersaturation from those due to
metabolicaddition of CO2 within the sediments;and (3) to determinethe empiricalformsof the kineticsof thesereactions within the sediments.
Understanding the mechanismof calcitedissolutionis critical becausethe balancebetweenthe influx of alkalinity from rivers andthelossby oceanicburialof calciumcarbonatecandetermine
theCO2partialpressure (Pco2) of thesurface wa•ers andhenceof
metabolically producedCO2 can drive atmospheric CO2 differencesbetweenglacialandinterglacialperiodsif the ratioof the supplyratesof organiccarbonand calciteto the sediments changedover thosetimescales.It is clear that calciumcarbonate accumulationin the sedimentsis linked to glacial cycles [e.g., Farrell and Prell, 1989]; however, separatingthe effects of dissolution(eitherdue to bottomwatersaturationstatechangesor increased supply of organic matter to the sediments)from changesin productivityis difficult without quantificationof the kinetics of dissolution.
These are not new questions,but there is a fair amountof disagreementbetween historical attempts to answer them. Emersonand Bender[ 1981] first quantifiedthe amountof calcite that could dissolvein sedimentsin responseto metabolically producedCO2. Subsequent in situpH electrodemeasurements in pore watersof sediments[Archer et al., 1989;Hales et al., 1994;
theatmosphere on glacialtimescales [ArcherandMaier-Reimer, Cai et al., 1995] both above and below the saturation horizon 1994; Emersonand Archer, 1992; Broecker, 1989;Boyle, 1988; could not be explained without dissolution in responseto Broeckerand Peng, 1987]. Archer and Maier-Reimer [1994] metabolicallyproducedCO2. Studiesof pore water chemistry, argued that dissolution in pelagic sediments driven by measuredin pore waterscollectedwith in situ samplers,also implicateddissolutiondriven by metabolicCO2 [Sayles,1980, 1985; Saylesand Curry, 1988;Martin and Sayles,1996]. Bulk Copyfight1996by the AmericanGeophysical Union.
sediment calcite content and sediment dissolution indices based
on foraminiferal assemblages,however, often do not show significantchangesabove the depth of the saturationhorizon
PaperNumber96GB01522. 0886-6236/96/96GB-01522512.00 527
528
HALES AND EMERSON: CALCITE DISSOLUTION IN SEDIMENTS
Pacific Ocean
5øN
......
Ontong-JavaPlateau
In Situ Electrode Measurements (this study) Benthic Flux Measurements (Jahnke et al. 1994)
•k ElectrodeandBenthicFluxMeasurements
•
•500 40
00-2øS
• • I
155øE
160 ø
' 165 ø
Figure 1. Map of theOntong-Java Plateaustudyarea.Depthcontours in meters.
[Berger et al., 1982]. Jahnke et al. [1994], basedon direct measurementsof calcium and alkalinity fluxes at the seafloor, questioned earlierestimatesof metabolicdissolution.They found significant alkalinity and calcium fluxes out of sedimentsat depthsbelow the saturationhorizonbut nonefrom depthsnearor abovethe saturationhorizon. Fluxes predictedwith respiration and dissolution rate models used previously [Emerson and Bender, 1981; Archer et al., 1989; Archer and Maier-Reimer, 1994] were significantly greater than the benthic chamber measurements, while modelsincludingdissolutiondrivenonly by bottom water undersaturationwere in agreement with these observations.
The rateof calciumcarbonate dissolution in seawater Recis given by the empiricalexpression:
Rd, c=kd, c[CaC03(s) ](1-•c)nc
whereKsp.c is thethermodynamic solubility of calcite[UNESCO, 1987]. The rate constantlq.c and the reactionordernc are determinedempirically. Field studiesof calcitedissolutionhave generallyassumedthat the exponent,nc, is 4.5 [Archeret al., 1989; Hales et al., 1994; Jahnke et al., 1994;Berelsonet al.,
1994;Caietal., 1995;MartinandSayles, 1996],followingKeir [1980],although otherlaboratory studies haveshown thatnc may be as low as 2 [Walter and Morse, 1985]. There is less
agreementabout the value of kd.c. Keir [1980]performed laboratory measurementsof the rate of dissolutionof several
differentkindsof naturalandsynthetic calcite,findingthatthe best value for bulk sedimentwas at least 1000% d'l, which
appears tobein accord withthe•4Cages of themixed layers of deeppelagiccarbonatesediments[Keir and Michel,'1993].
(1)
Determinations of kd. cbased onfielddatagenerally haveyielded much lower values. Measurementsof in situ benthic fluxes
[Keir, 1980; Morse, 1978; Walter and Morse, 1985]. The saturationstate of the surroundingwater with respectto calcite
[Jahnkeet al., 1994;Berelsonet al., 1994],in situmicroelectrode porewaterpH profiles[Archeret al., 1989;Cai et al., 1995;
FIc is definedby
Haleset al., 1994],or carbonate chemistry of porewaters collectedwith in situsamplers[Martinand Sayles,1996], interpreted with modelsof the porewaterreactions, give estimates of k•cfrom0.2to 100% d'•.
ac--[Ca2+][CO32-] (2) Ksp, C
HALES AND EMERSON: CAI•ITE
DISSOLUTION IN SEDIMENTS
529
Table 1. Ontong-JavaPlateauStationDescriptions BottomWater Properties
Station
Depth,
m
Oxygen,
gmol kg-1
ALK,
geqkg-1
2A
2322
128
2404
2B 3
2335 2966
128 140
2404 2411
DIC,
•
2345 2345 2358
91_+ 13
gmol kg-1
C,bw'
%
*
91_+ 13 75_+ 11
*Theuncertainty in •C,bw isduetotheuncertainty incalculating carbonate ionconcentration from the alkalinity(ALK) and dissolvedinorganiccarbon(DIC) measurements and the uncertainty in the 1-atmcalcitesolubilityproduct.
Setting Three stationswere occupiedon the Ontong-JavaPlateau,a topographic highin thewesternequatorial Pacific(Figure1), on a cruiseby the R/V Moana Wave in Juneof 1991. Two of the stationsat --2300 m are spatiallyseparatedby less than 1 km, verticallyby lessthan15 m, andareconsidered duplicates. The plateauhashighcalcite(>70%by dry mass)from 1500m to nearly5000 m, andis in a uniformlylow productivityportion of the world's oceans[Berger et al., 1987]. This location is ideally suited to addressthe stated goals because(1) the sedimentsare well characterizedhere, including estimatesof dissolution from sedimentary indicators[e.g.Bergeret al., 1982] and rates of sedimentaccumulationand benthicmixing [Berger
and Killingley, 1982], and(2) Jahnkeet al. [1994] basedtheir conclusionsregardingmetabolicdissolutionon benthicflux chamber results from two sites here. One of these, the station
referredto by Jahnkeet al. as"NS",is thesameasourstation3. This studyrepresents the first successful attemptto quantify the rate of calcite dissolution with both benthic chamber and in
situ electrode measurementsat the same site. The pertinent bottom water characteristics of the stations are summarized in
The vertical spacingbetween measurementsis known very well; the locationof the interfacerelativeto eachoxygenandpH electrodewas determinedfrom their respectiveprofiles' Each profile includeda seriesof measurements in the overlyingwater that showed little change as the depth increased. The first measurement that deviatedsignificantlyfrom thesewas assumed to be below the sediment-water interface, and the location of the
interfacewas determinedby interpolation(estimatedaccuracyis about 20% of the vertical interval or 0.5-1.0 mm). The interface
for the resistivityelectrodewas determinedby interpolationas the depthwherethe resistivitywas 10% higherthanit wasin the
bottom water[Archer,1990].Theinterface determined by measuringthe lengthdifferencebetweenthe resistivityelectrod e and each of the others and assuminga fiat seafloor was not systematically differentfrom the abovemethod.
Electrode driftwasassessed bymakingtwomeasurements of each electrode in the overlying water after the profile was completed. If thesevaluesdiffered from thosein bottom water prior to entering the sediment by more than 10% (15% for resistivity) of the total changerecordedby the electrode,that profile was discarded. Two of the four oxygen electrodes performed acceptably at each station. Two pH electrodes
Table 1. The rangein estimatedbottomwater saturationstate with respectto calciteis due to uncertainties in boththe bottom water carbonate ion and the estimated solubility product for
performed acceptably at station 2A,butonlyoneperforme d
calcite.
relationbetweenthe signalin bottomwater(concentratio n
Methods
acceptablyat stations2B and3. Oxygenelectrodeswere calibratedin situby assuminga linear
determinedby Winklet titration)and the opencircuitreading
(zero oxygenconcentration).We have shownthat the assumption of an open circuit reading at zero oxygen is valid with In Situ Electrode Measurements measurements in oxygendepletedsedimentsin PugetSound(B. The microelectrode measurements were madein situby an Hales, unpublished data, 1990). The pH electrodes were calibratedon the ship prior to deploymentwith pH 7 andp H 8 updatedversion of the profiler used by Hales et al. [1994], deployed on a free-vehicle lander. This a substantial buffers,madeup to the ionic strengthof seawaterwith KC1. The improvement overthebox-coredeployment methodfollowedby differencebetweenthesesolutionsand seawatermay meanthere Hales et al. [1994] andArcher et al. [1989], becauseof decreased are differentliquidjunctionpotentialsfor the electrodes between wire time and the decreasedpotentialfor introducingartifacts the buffersand seawater.This is not a significanterror,however, associated with takinga core. On eachdeployment, the profiler since we are interestedin the pH differencesbetweenbottom water and pore water (ApH), not the absolute value of the was configuredwith one resistivity,four oxygen,and fourpH electrodes.All electrodes usedwereof thesametypeasusedby seawaterpH. The slope thus determinedwas correctedfor the Hales et al. [1994]. Electrodemeasurements were made at 0.25- temperaturedifferencebetweenthe laboratoryand the bottom 0.5-cm vertical intervals from a few centimeters above the water using the Nernst equation. Error bars on each electrode sediment-water interface to as much as 9 cm below. Each measurementbelow the interfaceincludethe uncertaintydue to measurement is theaverageof thelastfive readingsoutof a suite both short-termvariability (i.e., the standarddeviationof the five readingsof eachelectrodeat eachdepth)andthe uncertaintydue of 15 readingsmadeonceevery10 s.
530
HALES AND EMERSON: CALCITE DISSOLUTION IN SEDIMENTS
to the differences between the measurements in bottom water
all cases were bottom water concentrations
at the interface and
beforeand after profiling(i.e., "drift"). Most of the uncertainty zero-derivativesat the deepestpart of the model. Both models derives from the latter. were solvedby discretizingthe differentialequationsdescribing the systemandinvertingtheresultingmatricesusinga tridiagonal Bulk Sediment Measurements on Retrieved Cores approach[Presset al., 1989]. The oxygenmodel was linear and independentof the pH model, so it could be solveddirectly. Sedimentsampleswere taken from subcoresof Soutarbox coresat two locations. One set of sampleswastakenat 2-3-mm Numerical solutionswere verified by comparisonto analytical resolutionoverthe upper10 cm of the sediment,andanotherwas solutions. The pH modelrequiredsimultaneoussolutionof the takenat =1 cm resolutionover the upper30 cm of the sediment. Bothsetsof sampleswereanalyzedfor particulate organiccarbon andtotalcarbonby a Carlo-ErbaCHN analyzer.Organiccarbon wasmeasuredon samplesthathadbeentreatedwith HC1vapor followedby directadditionsof ultrapureHC1to eachsampleto removeall solid carbonates[Hedgesand Stem, 1984]. Calcium carbonate content was calculated assuming the difference betweenacid-treatedand untreatedsampleswas entirelydue to calciumcarbonate.Porositywascalculatedfrom measurement of thedry massof thefine-resolution samplesandthesaltcontentof the samples. Bottom Water Properties
Temperature and salinity were taken from the nearest Geochemical Ocean Sections Study (GEOSECS) station, at corresponding depths. Bottomwater oxygenwasdeterminedon the ship by Winkler titration on bottle samples. Bottomwater alkalinity (ALK) and dissolvedinorganiccarbon(DIC) were determinedby Gran titrationwith ultrapureHC1in a closedcell monitoredby a RosspH electrode. The titrationwascalibrated with ALK andDIC standards (A. Dickson,ScrippsInstitutionof Oceanography,personalcommunication,1991). Precisionof replicatealkalinity and DIC measurements was lessthan0.25%. There was an artifact associated with the volume of the titration
cell, which appearedto be systematicallyincreasingin volume overthe courseof thecruise. As a result,samplestitratedlaterin the cruiseyieldedestimates of ALK andDIC thatwereup to 1% higher than samplesfrom the same locations titrated earlier. However, the most importantparameterfor this study was the
distributions of ALK, DIC, anddissolved calcium, Ca2*. Species
included in ALK and/orDIC wereCO2'(CO2(aq) q-H2CO•), HCO•',CO•2',andB(OH)4'.Thespecies HCO•'andB(OH)4' were eliminatedalgebraicallyfrom the equations,usingthe acid-base equilibriumrelationships betweenthemandthe otherspeciesand the dissociationconstantsfor carbonicand boric acidssuggested by UNESCO [1987]. This left a combinationof threecoupled,
nonlinear differential equations withCO2',CO32', andCa2*asthe dependentvariables. Becauseof the nonlinearity,the tridiagonal inversionof the matricesdefinedby thesediscretizedequations had to be executediteratively,usingNewton's method[Presset al., 1989]. The pore water pH data are presentedas ApH, the differencebetweenpore waterpH and bottom water pH. For comparisonto the data,modeledpore waterApH wascalculated
fromtheresulting modelsolution for CO•2'andCO2'fromthe relationship
where the subscriptsBW and PW distinguishconcentrations in bottom water and pore water, respectively. This result is relatively insensitiveto the choice of equilibrium constantsfor carbonicand boric acids,as long as a consistentbuffer systemis employed. Diffusivity in the pore waters is approximatedby
Di.•,w = Di/F [Bemer,1980;McDuffandEllis, 1979]whereD•is the diffusioncoefficientfor speciesi in "clear"waterandF is the sedimentformationfactor,determinedby measurements with the resistivityelectrode. Diffusion coefficientswere takenfrom Li
carbonate ionconcentration ([CO32']).Theaccuracy withwhich andGregory[1974](forCa2+,HCO•'andCO32'); Broecker and [CO32'] couldbedetermined wasrelatively insensitive tothiscell Peng[1974](forCO2');andWiseandHoughton [1966](for02). volume error, since it affectedALK and DIC similarly. There
Borateion (B(OH)4') was assumedto have the samediffusion coefficientas HCO3'. The modelsincludeda diffusivesublayer titrationsperformedearlierandlater in the cruise. Calculationof that was 0.5 mm thick; diffusivitiesin this layer were assumedto [CO•2']isverysensitive toerrors in ALKandDIC;theanlaytical equal the clear water diffusivities;respirationand dissolution precisionof 0.25% in both measurementsyields uncertaintyin rateswere zero in the diffusivesublayer.
wasno significant difference between [CO•2']determined from
the calculated [CO•2-] of about+6%. Uncertainties in the dissociation constants of carbonic and boric acids [UNESCO,
1987]havelesseffect,sincecalculation of [CO•2']fromALKand DIC dependson the ratios of the constantsand not the absolute
values.Evengiventheuncertainty in calculation of [CO32'], these stationsare more undersaturatedwith respectto calcite than similardepthsat the nearestGEOSECSstation,whichis nearly 2000 km east of the Ontong-JavaPlateau. This follows the generaltrend of a westward-shoaling of the calcite saturation horizonevidentin theGEOSECSequatorialdata. Pore Water
Models
The pore water 02 andp H data were interpretedusingonedimensional, steady state mathematical models simulating diffusionand reactionsin the sediment.Boundaryconditionsin
Results
Bulk sedimentpropertiesfor the top 10 cm of the sediments are presentedin Figure 2. Calcite contentsare fairy constant with depthat 85-92%. Thereis slightlyhighercalciteat station2 than
at station
3, but the difference
is similar
to the
reproducibility of each measurement(based on duplicate measurements of the same sample). Organic carbon is consistently low (0.1-0.4%) at bothstations,with station3 being slightly enrichedover station2. The organiccarboncontentat bothstationsdecreases weaklyoverthetop 10 cm,butthechange is small relative to the measurementerror, and the decreaseis not
by any meanscontinuous.For thisreasonthe organiccarbondata were not usedto constrainthe ratesof respirationwithin, or the flux of metabolizable carbon to, the sediment.
HALES AND EMERSON: CALCITE DISSOLUTION IN SEDIMENTS
g CaC03/100g
g OrganicCarbon/100g 0.1 I
80
0.5
0.3 l•
ß
I
531
85
90
95
B) O
I
$I
•
ll•
I
/
W
I
I
ß
4
Depth (cm)
Sta. 2
5
ll
Sta. 3 $
ll
I
/
!
II
/
',
I
....
I
....
I
....
I
,
Figure 2. Solidcarboncontentof Ontong-Java Plateausediment,as percentages of the total solids: (a) organic carbonand (b) calcium carbonate.
thepH profilesandthereturntoward,andevenbeyond(at station 2A), their bottom water values at depths >2 cm below the interface. This could not happenwithout dissolutionof some carbonatemineralat depthsof a few centimeters in the sediment. These stations are strongly undersaturatedwith respect to aragonitein thebottomwaters,andit is unlikelythata significant 1 amountof thismineralcouldpersistto thesedepths[e.g.Hales et • = F •' (4) al., 1994]. Indeed,Berger et al., [1982] observedno aragonitein [Andrewsand Bennett,1981];porositymeasuredon subsamples the sedimentsthey collectedfrom the Ontong-JavaPlateau at thesedepths. The only conclusionthat can be drawn is that from box coresdecreasedto constantvaluesof about70% by 3-4 cm below the interface (data not shown), implying that the calciteis activelydissolvingwithinthe sediments. exponentb is about2. Oxygen (Figure 3b) decreases smoothly throughoutthe measurementregion, showingcurvatureeven at Discussion the deepestpenetrationof the electrodes. The pH data is We interpretedthe pore water databy simulatingthe vertical presentedas ApH, the differencebetweenpore water pH and distributions of oxygen and pH with steady state, onebottomwaterpH (Figure3c). Negativevaluesof ApH meanthat the pore water is more acidic than the bottom water; positive dimensional diffusion and reaction models to evaluate the pertinent reaction rates in the sediment. The approachis valuesindicatethe opposite.All electrodesat all stationsshowa essentiallythe same as used by Hales et al. [1994], and the shift towardlower pH below the interfaceto a minimum, and a governing equationsand numerical solutionsare describedin gradualreturn towards(in one caseeven beyond)bottomwater values. There are severalinterestingqualitativefeaturesof these detailby Hales [ 1995]. data. The first is the significantdifferencebetweenthe two sets of oxygenprofilesat Stas.2A and2B, implyingdifferencesin the The Pore Water O2Model
The microelectrode dataare shownin Figure3. The formation factor (Figure 3a), defined as the ratio of resistivity in the sedimentsto that in the overlyingwater,reachesa fairly constant value of about 2 by 3-4 cm below the interface at all stations. Porosityq• is relatedto formationfactorF by therelationship
oxygenflux into the sedimentsbetweenthesetwo stations. The pH profilesare alsodifferentbetweenthe two stations,with thepH minimumat station2B beingroughlytwice as low as at station 2A.
Pore wateroxygensimulationsare sensitiveto the depthdependent rateof oxygenconsumption Ro2(Z ) assumed to follow the empiricalform:
Since the saturation state of calcite must be similar at
the two sites,the moreacidicporewatersat stationB mustbe due to the greaterrespirationratesin the sediments, asimpliedby the oxygengradients. Also remarkableis the existenceof minima in
Ro2(Z ): (1-q•)(rl e-r2z +r3e-r4z );
(5)
whereq•is the porosity(calculatedfrom the formationfactordata
532
HALF3 AND EMERSON:CAI_E:ITEDISSOLUTIONIN SEDIMENTS
Station 2A-- 2322m Formation Factor Oxygen(gmol/kg) 1 1.5 2 30 50 70 90 110130 -1 b.O, . ß [ .... • , ._ P ......... '......... '......... '......... '.......t•"-1
ApH -0.04
-0.02
0
"'''1 ......... I....... • .......
c)
I
I
•'
ß
I
=
I
d•
3
t
Depth (cm) 4
-
I
oi
ß
I
I I
I I i
I
;.I .........
I .........
I .........
I .........
I ........ .I ..... ,•
Station 2B-- 2335m Formation Factor 1 1.5 2 -1 '•' ' ' ' .... [''
Oxygen(gmol/kg) 30
50
70
90
ApH
110130
-0.04 .... [ ........
-0.02
0
[ ..........
• .... :
,
ß
I
I
I
2
-
I I
I
I
I
Depth 3:, (cm)4:'
I I I
I
.
-[
I
I
.'[
I
I
5
I
'[
6
I
I
I
I
I
I
I
I
-
I
I
I
I
I
7
I
-
-
-
I
[
,
I I I '
I
...I
.........
I .........
I ........
'
Figure 3. In situelectrodedatafromthe Ontong-Java Plateau:(a) formationfactor,definedastheratioof the resistivitymeasured in the sediment to thatmeasured in theoverlyingwater,(b) porewater oxygen,and(c) porewater pH, expressed asApH,thedifference between porewater pH andbottomwaterpH. In figures3a-3c, depthis expressed positivedownward,with the sediment-water interfaceat zero (solidhorizontalline). The vertical dashed linesrepresent thebottom watervalueof eachproperty. Different shaped symbols in thesameplot represent measurements madeby separate electrodes.Profilesrepresented by the diamonds will be referredto as
profile1,thoserepresented bythecircles willbereferred toasprofile2. In allplots,thelarger,opensymbols 1 cm .abovethe interfaceare measurements madeaftercompleting the sediment measurements, represented by the smallersolidsymbolof thesameshape.In somecases, thesepointsarepartiallyobscured by themeasurements prior to penetrationof the sediment.All measurements belowthe sedimentwaterinterfacehaveerrorbars;those
notvisiblearesmaller thanthesymbols themselves. Sincethemeasurements abovetheinterface areall replicate measurements of bottomwater,theirscatter isindicative of theirreproducibility anderrorbarsarenotgiven.
HALF3 AND EMERSON: CALCrrE DISSOLlrrION
IN SEDIMF_.aN•
533
Station 3-- 2966m 1
Formation 1.5
Factor 2
-1•'• '''....''' 0 oA)
ApH
Oxygen(gmol/kg)
-0.04
30 50 70 90 1101)0
u......... I......... I......... I......... I........ '1"•
-0.02
0
"'"1 I........ • ...... : C)......... ß
_
1._• .!
ß
I
'
ß .!
D•
,.
ß
ß
ptl•'-' ;.
g
,
-
I
ß
!
ß
!
.
. !
6•
ß
i
I
ß
i
I
.
7"-!
i
L
-!
.
8"
,!
I
I
i
I
L
i
I
I
I
I
I.
I
I
I
I
I
i
9" ;I ......... I......... I ......... I......... I ......... I.... h•
: ....
I .........
I .........
I ........
.
.
.
i
Figure 3. (continued)
and (4)) and rl, r2, r3, and r4 are adjustable parameters.This expression of Ro2is similarto thatusedby Hammond et al., [ 1996]andis slightlymoredetailedthanthatusedby Haleset al.,
explicitly include the dependenceon the solid fraction of the model sediments, of organic (1- {). carbon Although degradation conceptually [Berner, similar1980], to a "multi-G" our goal
[ 1994],whousedat mostthreeadjustable parameters, anddid not
was only to matchthe oxygenprofile as accuratelyas possible. Determinationof the optimum combinationof four adjustable parameters manually is difficult, so the oxygen model was executedby an optimizationroutine(Powell'smethod,Presset al., 1989] that adjustedrl, r2, r3, andr4 to minimizethedeviation betweenthe data andthe modeloutput. Figure4 illustratesthe needfor inclusionof all four adjustable rate parameters(two exponentialterms)in orderfor the modelto fit the oxygendatathroughoutthe profile;resultsfor othercases
Oxygen (gmol/kg) 60
80
100
120
140
are summarized in Table 2, where there are some trends worth
noting. In all simulations, most (>70%) of the oxygen consumption is dueto the first exponentialtermin (5), whichhas a scaledepth(1! r2) of 1.7-3.5 mm. This meansthat mostof the CO2 wasgeneratedvery nearthe interface.As expected,oxygen fluxes calculatedfrom the model outputat station2B are greater than those at station 2A. This difference is greater than the differencebetweenindividualprofilesat eachstation,but is only
Depth (cm) Figure 4. Exampleoxygenmodeloutputfor station3, compared to oxygenprofile 2. Model outputwith respirationratesthat are constantwith depth (heavydashedline); a single,exponentially decayingfunctionof depth(heavysolidline); or a combination of a constantand a singleexponential(light dashedline) do not provideadequateagreementwith the data over the entiredepth range,while that with a two-exponentialformulationdoes(light solid line). The inflection near the interfacein the profile with respirationdescribedby a singleexponential(heavysolidline) is due to the fact that the gradientin formationfactoris significant
relativeto thereactiontermin thatdepthinterval,Notethateach exponentialterm containstwo empirical parameters(see, e.g. (5)).
534
HALES AND EMERSON: CAI_L'TrE DISSOLUTION IN SEDIMF_2q'I•
Table 2. OxygenModel Results OxygenConsumption Rate Parameters*
r •,
r2,
rs,
r4,
Integrated(Net) Consumption,
gmol02
gmol02
(cmssed.solids) 4
(cmssed.solids) 4
gmol02
Station
Profile
yr4
cm4
yr-•
cm4
cm-2yr4
2A
1 2
111 192
2.85 4.25
8.50 6.57
0.352 0.277
9.7 9.6
2B
3
1
262
2.87
6.86
0.318
18.7
2
305
3.47
2.69
0.132
16.8
1 2
281 854
4.18 6.17
3.09 3.45
0.200 0.228
14.0 21.1
*The rate parametersr•, r 2, r s, and r4from (5) are determinedby an optimizationroutineas the combinationthat minimizesthe deviationbetweenthe model simulationand the porewateroxygendata [seeHales, 1995]. The integrated consumption is calculatedfromthisbest-fitsimulationandis oppositeandequalto the flux at the sediment-water interface.
slightlygreaterthanthe differencebetweenthe two estimates at station3, implyingthatthe variabilityin benthicoxygenflux is as largeover spatialscalesof a few centimeters as it is overa few thousand meters. The two estimates of oxygen flux to the
carbon Rco2. Uncertainties in the magnitude and depth dependences of these reactions must be addressed before attemptingto ,quantifythe calcitedissolutionimplied by thepH
sediment at station3, 14 and21 gmolcm'2yf•, arein good agreementwith the flux predictedby the benthic chamber
Assumingthat the order of the dissolutionreactionnc is known to equal 4.5, the importantparameterscontrollingthe
measurements, 18+8gmolcm'2yr4.
calcitedissolution ratearethe dissolution rateconstant ka,c and
Consideringthe detailedsensitivityanalysisof thepH model that follows, some discussionof the uncertaintiesassociatedwith
the ri parametersof (5) is justified. There are, of course, uncertainties in theseparameters for a givenoxygenprofilethat are not shownin Table 2. Becauseof the shallowscaledepthof the first exponentialterm in (5), (even when the increasein the term (l-q>)is considered,the scaledepthof this term is of the orderof 2.5-5 mm), the parametersr• and r2 are the leasttightly
data.
the saturationstateof the bottomwater with respectto calcite flc,bw.The rate constanthas a wide rangeof reportedvalues.
Keir [1980] determined that1•,c wasover 1000%d'• with laboratory measurementsof the dissolution rate of natural sediments,while Jahnkeet al. [1994] reporteda rangeof 0.05-
0.5% d'•, basedon in situmeasurements of seafloorcalciumand
alkalinity fluxes.Thisis a relative difference of over104,leaving essentially no predeterminedlimits on this parameter. We
constrained, and we will limit the rest of this discussionto those
examinedrate constants in the range10'3-104% d'•, which
terms.
includesall previousestimates.The bottomwatersaturationstate
Constraint
of these
values
is difficult
to address
unequivocally,becausetheyare sodependent on oneanother.If, for example,r• is held constant,adjustments of only a few percent in r2 generatemodel profiles that do not approximatethe data acceptably;the sameis true if r2 is held constantand r• varied.
depends onboththethermodynamic solubility product Ksp,c and the bottom water carbonate and calcium ion concentrations. uncertainties
The
in the measured bottom water DIC and ALK lead to
However, if both are allowed to vary in such a way that the implied 02 flux to the sedimentremainswithin a few percentof the optimum,r• andr2 valuesas muchas 20% differentfrom the optima in Table 2 yield acceptable approximationsof the
uncertaintiesin the calculatedcarbonateion concentration of up to 6%. The UNESCO [1987] solubility product is about 8% higherthan the lower limit of estimatesbasedon Mucci's [1983] 1-atmsolubilityandsimilarlylower thanthe upperlimit of those basedon Plath et al.'s [1980] 1-atm solubility measurements.
electrode data. These uncertainties,however, are smaller than the
This meansthatourestimateof flc,bw hasuncertainties of about
differences implied by inter-electrode variability. In the followingsection,uncertainties in Ro2will be limitedto only the different setsof ris implied by the different electrodes.In that way, Ro2is treatedas a singleinput parameterto thepH model, ratherthanfour, with limits that are definedby the optimumfits to thetwo observed02 profilesat eachstation.
+14%, which we used to define the practical limits. This assumesthat Millero's [1983] pressuredependence is correct;at thesedepths,the pressuredependenceof Ingle [1975] predicts only abouta percenthighersolubility. MetabolicCO2 productionis linearly proportionalto the rate of oxygenconsumption:
The Pore Water pH Model Model simulationsof pore water pH are sensitiveto two
reactions, thedissolution rateof calcite(Ra, c, asexpressed by (1)) and the rate of CO• productionduring metabolismof organic
Rco2= flRo2, where B is a stoichiometric
constant.
(6) At each station there are
two estimatesof Ro2,determinedby the two oxygenprofilesat each station(seeTable 2). As statedpreviously,the differences
HALES AND EMERSON: CALCITE DISSOLUTION IN SEDIMENTS
ApH -0.1
-0.05
ApH 0
-0.1
-0.05
535
ApH 0
-0.1
-0.05
2
Depth (cm) 3
Figure 5. Model pH curvesat (a) station2A, (b) station2B, and (c) station3, for threescenarios:no dissolution
(heavysolidlines),nometabolic CO2production (heavydashed lines),andthebestfit to thedata(lightsolidline).
betweenthesetwo estimates of Ro2aregreaterthanuncertainties in eachestimate,and the practicallimits on this parameterare definedby the two estimates.The constant13can be between 0.55 (alkaneequivalentorganiccarbon;seeHaleset al. [1994]or 0.77 (Redfield organiccarbon). A recentwater columnstudy [Andersonand Sarmiento,1994] estimatedthis parameterto be 0.69, anda benthicflux chamberexperiment[Hammondet al., 1996] gave an estimateof 0.67, but the reportederrorbarswere large enough that these estimatesdid little to discriminate betweenthe upper and lower limits statedabove. We usedthe aboverange(0.55 to 0.77) asthe practicallimitson 13.
contrast to all the observations. Furthermore, the no dissolution
simulations greatly overpredict the acidification of the pore waters. It is clear that neither case excluding the effects of metabolic dissolutioncan reproducethe observations,and that dissolutiondriven by metabolic CO2 productionmust play a significantrole in determiningthe porewater chemistryat these stations.
To addressthe apparentdisagreement betweenthe pH data and the no metabolic dissolution scenarios, we determined the
constraints posedby thepH dataon the dissolutionflux predicted by the model,giventhe practicallimits on the four modelinputs The primary goal of the pH modelingexercisewas to assess discussedabove. Figure 6 demonstratesthe sensitivityof the the importance of calcite dissolution driven by metabolic modelsimulationsto theseparameters.Sensitivityto 13(Figure processes relativeto thatdrivenby bottomwaterundersaturation. 6a) over its entire practical range is similar to a factor of 2 As a first step, we addressedtwo simplecasesthat includedno adjustment in the rateconstant(Figure6b). The mostsensitivity, metabolic dissolution of calcite.
These are (1) the "no
dissolution" case, where the respiration reaction produces carbonicacid,but no calciumcarbonatedissolves(effectivelythe
however, is to the bottom water saturation state. A similar shift
in modelpH resultsif this valueis alteredonly abouta percent
fromits optimumvalue. This sensitivity to f•c.bw meansthat,in sameas assuming thatkd.c is zero),and(2) the "no CO2"case, somecases,thepH datathemselvesprovidea betterconstrainton
of alkalinity where the respirationreactiondoesnot producecarbonicacid this parameterthanthe bottomwatermeasurements while consumingoxygen(effectivelythe sameasassumingthat13 and DIC. The first step in constrainingthe model simulationswas is zero), and calcitedissolvesonly in responseto bottomwater undersaturation.The rangein pH for thesetwo casesis presented choosinga statisticthat quantifiedthe deviationbetweenmodel in Figure5, alongwitha scenario including metabolic dissolution simulations and the data. We chose the "average relative that providesthe best fit to the pH data in the range of input absolutedeviation"(ARAD), definedby: parameters investigated (described below). The two no (7) dissolutioncurvesrepresentthe rangeof possiblesolutionsgiven N i=l O'i the two different estimatesof the oxygen consumptionrate at eachstationand the practicalrangeof 13;the two no CO2curves representthe rangeof solutionsfor the saturationstateof the whereN is thenumberof datapointsin a givenprofile,mi is the modeloutputat depthi, di is the valueof the dataat thatpoint, bottomwatersgiven in Table 1 and dissolutionrate constantsof 0.05%and0.5%d4 (therangesuggested byJahnke etal. [1994]). and oi is the analytical standarddeviation of the data at that It is worthnotingthat no combinationof dissolutionparameters point, as describedin the methodssection. The ARAD is a linear in the no CO2 casecan generatenegativevaluesof ApH, in version of thefamiliarZ2whichisthenaveraged overthenumber
ARAD__I•Imi-dil
536
I-•ES ANDEMERSON: CALCITEDISSOLUTION IN SEDIMENTS ApH -0.04 '"'1
ApH
-0.02
.........
-0.04
I .........
'''l
mmmmmmmmm
ApH
-0.02
.........
0
-0.04
ß ß
B = 0.77
(cm)
• I
0
m.
1
Depth
-0.02
I .........
.
k =0.02
I
ß
!
i
•
i $
i
$
i
•
i
i I
$
i
%
!*,
,.
.,, t
'.
ß
I
'k = 0.01
' B-0.S: .
!' ',
mm'
•'
I
I
m
m-
'
'
-
•",
m
m-
_
'
e"-
m
m
ß
•
m
m
.
,.
'...I ........
m
m
le !
I
I !
i I-
I
I .
I
I
s
ß
ß
ß
I ,•..l...ml.'
..., .......
l.,
i
...I .....
m
.
.
'
m I I m i m, m m
i
m mI , m
Figure6. Demonstration of thesensitivity of thepH modeloutputat station3 to (a) thedepth-dependent rateof CO2generation, (b) dissolution rateconstant, and(c) bottomwatersaturation. In all cases, thelightsolidlineis the
model output withkd, ½= 0.014% d'x,•½,bw = 76%,B= 0.64,andtheoxygen consumption ratedetermined byoxygen profile1. The othercurvesin figures6a-6carethe sameasthiscurve,with oneof theseparameters altered.In Figure6a,theheavysolidlineis theresultif Bis 0.77'theheavydashed linehasB= 0.55. Theheavydottedlineis
output if theoxygen consumption ratedetermined byoxygen profile2 isused.Figure6B shows theresultif kd,½ =
0.01% d'•(heavy solid line)or0.02% d'•(heavy dashed line).Figure 6Cshows theoutput for•½,bw = 75%(heavy dashed line)and•½,•w= 77%(heavysolidline).
of datapointsin the profile. It is a more'robust'statisticthan
ApH -0.02
-0.01
0 I
I
stepwasto searchtherangeof inputparameters untilwe founda
I
minimumvalueof theARAD. Because everydataprofilehas different analyticalerror bars and different degreesof "noisiness", therewasno universally acceptable valueof this statisticfor all profiles,and we visuallyevaluatedthe model outputto assess the "goodness" of fit. Thisis preciselywhat Presset al. [1989]referto as"chi-byeye",butthepresence of variability in theobservations thatwecannot expectthemodelto
I I
fit
%%
, I I
%
'acce' I
3
Depth (crn) 4
higher-order statistics liketheZ2,because thedependence onIm idllis lesssensitive to outlying pointsthanhigher-order termslike (mi_di)2 [Presset al., 1989]. (Thisstatistic is theoneminimized in theoxygenratedetermination described above.)The second
I I
reproduce withintheanalyticaluncertainties of themeasurements
I
(e.g.,thevariability in thepH dataat station 2A below4 cm)left
I
us with no other alternative.
I
I I
The minimum ARAD, or "best fit" case was evaluated for
eachpH profiletodecide whether ornotthefit wasgoodenough
I I
I I
I
Figure 7. Examplep H model profilesdefiningthe "region of agreement"with the data. The solid line is the model solution with the lowest ARAD,
or best fit.
The two dashed lines are
modelsimulationsthat fit the dataacceptably;simulationsthat fit worsethan thesedashedlines have greaterARAD statisticsand are deemedunacceptablereproductions of the data. Data shown here are p H profile 2 from station 2A; this procedurewas repeatedfor pH profile 1 at this stationand the profile at station 3.
15M
3M
?5
•C,bw
(%)
?5
•C,bw
80
Of•C,bw
pre_determined lower 85
3M
85
from oxygen profile 1 ••Ro2 from oxygen profile 2
90
15M
90
log(ka,cl(%d-•) )
-1
-2
-3
-1
log(ka,cl(%d-•) ) Limits on
B) 7o
?5 ?5 :ainedupper
80 of
(%) 80
85
90
Agreement with pit :2 Agreement with pH 1
Region ofagreement; B=0.55 ofagreement; B .•Re gion B= =0.64 0.T
-3
90
-1
-2
0
-3
-2
-1
log(k,•,cl(%d43) this
1øg(ka'c/(%d4)) the dissolution rate parameters, demonstratecl for station Figure 8. The procedure for constraining Contour plot of the bRAD statistic for profile 2, with oxygen of the log of the dissolution rate constant and the bottom water ß es inFigure 7.Solutions was also followed for station 3.(a) of the bRAD and corresponds to procedure the maximum value hedashed lin urs represent consumption rate 1,[5 0.64, as afunction allowable •dding ines; soluU 3Mrepresen . saturation state. The contour labeled ....
.
ß solutions ma,,--
k and•c,• ....
thedashed 1 .....
'OhS outsloe m •-
ß
t tooactb•,•-
.
will mours labeled _ _•adW,F•res 8b-Sd
commnauuns o[_. a,c data •etter_than •_ RAD't•at•s,c•,., .... :••im•re •. potct• w contour fitthe ev•u
