Intensification and deepening of the Arabian Sea oxygen minimum zone in response to increase in ...

28 Pages·2017·10.23 MB·English

Biogeosciences,15,159–186,2018
https://doi.org/10.5194/bg-15-159-2018
©Author(s)2018.Thisworkisdistributedunder
theCreativeCommonsAttribution3.0License.
Intensification and deepening of the Arabian Sea oxygen minimum
zone in response to increase in Indian monsoon wind intensity
ZouhairLachkar1,MarinaLévy2,andShaferSmith1,3
1TheCenterforPrototypeClimateModeling,NewYorkUniversityinAbuDhabi,AbuDhabi,UAE
2SorbonneUniversité(UPMC,Paris6/CNRS/IRD/MNHN),LOCEAN-IPSL,Paris,France
3CourantInstituteofMathematicalSciences,NewYorkUniversity,NewYork,NY,USA
Correspondence:ZouhairLachkar([email protected])
Received:18April2017–Discussionstarted:2May2017
Revised:16November2017–Accepted:21November2017–Published:10January2018
Abstract. The decline in oxygen supply to the ocean as- of oxic waters in the lower epipelagic zone (100–200m) of
sociated with global warming is expected to expand oxy- thewesternandcentralArabianSea,leadingtointermittent
gen minimum zones (OMZs). This global trend can be at- expansions of marine habitats and a more frequent alterna-
tenuated or amplified by regional processes. In the Ara- tionofhypoxicandoxicconditionsthere.Theincreasedpro-
bian Sea, the world’s thickest OMZ is highly vulnerable to ductivity and deepening of the OMZ also lead to a strong
changesintheIndianmonsoonwind.Evidencefrompaleo- intensification of denitrification at depth, resulting in a sub-
recordsandfutureclimateprojectionsindicatesstrongvari- stantialamplificationoffixednitrogendepletionintheAra-
ations of the Indian monsoon wind intensity over climatic bian Sea. We conclude that changes in the Indian monsoon
timescales. Yet, the response of the OMZ to these wind canaffect,onlongertimescales,thelarge-scalebiogeochem-
changes remains poorly understood and its amplitude and ical cycles of nitrogen and carbon, with a positive feedback
timescale unexplored. Here, we investigate the impacts of onclimatechangeinthecaseofstrongerwinds.Additional
perturbations in Indian monsoon wind intensity (from −50 potentialchangesinlarge-scaleoceanventilationandstrati-
to +50%) on the size and intensity of the Arabian Sea ficationmayaffectthesensitivityoftheArabianSeaOMZto
OMZ, and examine the biogeochemical and ecological im- monsoonintensification.
plicationsofthesechanges.Tothisend,weconductedase-
ries of eddy-resolving simulations of the Arabian Sea us-
ingtheRegionalOceanModelingSystem(ROMS)coupled
to a nitrogen-based nutrient–phytoplankton–zooplankton– 1 Introduction
detritus (NPZD) ecosystem model that includes a represen-
tation of the O cycle. We show that the Arabian Sea pro- Thecombinationofstrongorganicmatterdecompositionand
2
ductivity increases and its OMZ expands and deepens in poorventilationexplainsthepresenceoflargeoxygenmini-
response to monsoon wind intensification. These responses mumzones(OMZs)intheintermediateoceanoftheeastern
are dominated by the perturbation of the summer monsoon tropical Pacific and Atlantic oceans as well as in the north-
wind,whereasthechangesinthewintermonsoonwindplay ern Indian Ocean. At low oxygen concentrations, hypoxia-
a secondary role. While the productivity responds quickly sensitivemarinespeciesaresubjecttovaryingenvironmental
and nearly linearly to wind increase (i.e., on a timescale of stressesthatcanaffecttheirgrowthandreproductivesuccess
years), the OMZ response is much slower (i.e., a timescale andultimatelycausetheirdeath(Vaquer-SunyerandDuarte,
ofdecades).OuranalysisrevealsthattheOMZexpansionat 2008).Nearcompleteoxygendepletion,suboxicconditions
depthisdrivenbyincreasedoxygenbiologicalconsumption, favoranaerobicremineralizationoforganicmatterviadeni-
whereasitssurfaceweakeningisinducedbyincreasedven- trification: a process through which nitrate is used as an al-
tilation.Theenhancedventilationfavorsepisodicintrusions ternateoxidant.Thisnotonlydepletestheoceanicinventory
of bioavailable nitrogen that is essential for phytoplankton
PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion.
160 Z.Lachkaretal.:MonsoonandArabianSeaoxygenminimumzones
growth but also releases N O, a major greenhouse gas that coastal upwelling that increases with latitude. Furthermore,
2
also contributes to stratospheric ozone depletion (Codispoti these authors show that the intensification of upwelling in
etal.,2001).Thus,theOMZsnotonlyshapemarineecosys- thewesternArabianSeaisevenhigherthanwhatispredicted
temhabitatsbutalsoimpactandregulateclimate. fortheeasternboundaryupwellingsystems.Thisstudyalso
Dissolved O is predicted to decline in the future in re- findsanintensificationofland–seathermalandpressuregra-
2
sponsetoupperoceanwarmingandincreasedstratification, dients that is consistent with the Bakun (1990) hypothesis.
which may result in the expansion of OMZs (Bopp et al., Otherstudiesrevealthatsummermonsoonwindintensifica-
2002;Keelingetal.,2010).Thesechangeswillhavedelete- tion has already been observed recently (Goes et al., 2005;
riousimpactsonmarinehabitatsandmayleadtodisruption Wangetal.,2013).Goesetal.(2005)reportedanincreasein
ofkeybiogeochemicalcycles(Keelingetal.,2010;Gruber, upwelling-favorable winds off Somalia over the period be-
2011; Doney et al., 2012). Observational evidence suggests tween1997and2005andhaveshownanassociatedincrease
that the oceanic oxygen inventory has already been declin- in productivity over the same period. Wang et al. (2013)
ing over the recent decades (Helm et al., 2011; Schmidtko found a global intensification of the Northern Hemisphere
et al., 2017) and that OMZs have expanded in several loca- summer monsoon since the late 1970s and attributed these
tions (Stramma et al., 2008; Stramma et al., 2012). Global trendstotheinteractionofthemegaElNiño–SouthernOscil-
modelprojectionsshowconsistentlydecliningoxygeninven- lationandtheAtlanticMultidecadalOscillation,inaddition
toriesthatarecommensuratewiththewarminganomalybut tohemisphericalasymmetricglobalwarming.
disagreeonthefutureevolutionofOMZs(e.g.,Coccoetal., While there is still no consensus on the magnitude and
2013).Thismayresultinpartfromthemisrepresentationof the drivers of the recent and future Indian monsoon wind
thedynamicsofOMZsinglobalmodels(Boppetal.,2013; changes, evidence from paleo-climate records overwhelm-
Coccoetal.,2013;Cabréetal.,2015;Longetal.,2016)and ingly suggests a strong link between Northern Hemisphere
thesensitivityofOMZstoregionalclimateperturbationsthat temperatures and the Indian monsoon wind intensity on
arepoorlyrepresentedinglobalsimulations. timescalesrangingfromdecadestothousandsofyears(e.g.,
An example of such perturbations is regional wind Altabet et al., 2002; Schulz et al., 1998; Gupta et al., 2003;
changes (e.g., Deutsch et al., 2014). Located near major Ivanochko et al., 2005). Generally, enhanced Indian sum-
coastal upwelling systems, the OMZs are indeed highly mer monsoon intensity was recorded during northern high-
vulnerable to changes in alongshore winds. In most east- latitude warm periods (e.g., Dansgaard–Oeschger events),
ern boundary upwelling systems as well as in the west- whilereducedsummermonsoonintensitywasfoundtocor-
ern Arabian Sea, upwelling-favorable wind intensification respondtocoldperiods(e.g.,Heinrichevents;Schulzetal.,
has been shown to occur under warming scenarios (Wang 1998). Ivanochko et al. (2005) has established a relation-
etal.,2015;deCastroetal.,2016).Bakun(1990)haslinked ship between millennial-scale oscillations in summer mon-
this to increased land–sea thermal gradient under warmer soon intensity and the position of the Intertropical Conver-
climates, strengthening the land–sea pressure gradient and gence Zone (ITCZ), where the ITCZ moves northward dur-
hence alongshore winds. Yet, previous observational and ing warm periods (interstadials) and southward during cold
model-based studies show that the amplitude of such up- periods (stadials). This relationship between the northern
wellingintensificationstronglyvariesfromonesystemtoan- high-latitude climate and the Indian monsoon seen from a
other (McGregor et al., 2007; Narayan et al., 2010; García- multitude of proxies during the last glacial period was also
ReyesandLargier,2010;Gutiérrezetal.,2011;Bartonetal., found to be valid during the Holocene (Gupta et al., 2003).
2013;Sydemanetal.,2014;Varelaetal.,2015). ThisincludesthemostrecentclimatechangesfromtheMe-
In the Arabian Sea, the summer monsoon southwesterly dieval Warm Period and the Little Ice Age. The analysis of
winds drive strong upwelling off the coasts of Oman and proxypaleo-recordsalsosuggestsastrongcouplingbetween
Somalia, giving rise to one of most productive coastal up- northernclimateexcursionsandchangesinproductivityand
wellingecosystemsintheworld(RytherandMenzel,1965). denitrificationintheArabianSeaatdifferenttimescales(e.g.,
Thishighproductivityinconjunctionwithasluggishcircu- Altabet et al., 1995, 1999, 2002; Gupta et al., 2003; Singh
lationsustainstheworld’sthickestOMZ,responsibleforup et al., 2011). For instance, Altabet et al. (1999) found that
to 40% of global pelagic denitrification despite occupying denitrification was greatest during interglacial periods and
lessthan2%oftheWorldOceanarea(Bangeetal.,2005). was probably not active during most glacial phases. Alta-
FutureclimateprojectionssuggestthatIndiansummermon- bet et al. (2002) found a strong correspondence between
soon may intensify in the future under a warmer climate changes in the productivity and denitrification of the Ara-
(e.g., Wang et al., 2013; Sandeep and Ajayamohan, 2015; bianSeaandcentury-scaleDansgaard–Oeschgereventsdur-
Praveen et al., 2016; deCastro et al., 2016). Praveen et al. ing the last glacial period. These studies show that denitri-
(2016) found that the future upwelling intensification will fication increases during warm periods concurrent with the
affect essentially the coast of Oman. Using an ensemble of summer monsoon and productivity intensification, and de-
global and regional model simulations for the 21st century, creases during cold phases. Other studies have highlighted
deCastro et al. (2016) show a strengthening of the Somali thatthefluctuationsindenitrificationandtheintensityofthe
Biogeosciences,15,159–186,2018 www.biogeosciences.net/15/159/2018/
Z.Lachkaretal.:MonsoonandArabianSeaoxygenminimumzones 161
OMZcanalsobedrivenbychangesinwintermonsoonwind ization (KPP) scheme (Large et al., 1994). Numerical
intensity(Reichartetal.,2004;KlöckerandHenrich,2006). diffusivity allows the dissipation of small-scale noise as
Despitethesepreviousstudies,theamplitudeoftheOMZ no additional background lateral diffusivity is used. The
sensitivity to potential wind changes remains largely uncer- ecological–biogeochemical model is a nitrogen-based
tain. Indeed, whether in the context of past climate fluctu- nutrient–phytoplankton–zooplankton–detritus (NPZD)
ationsorunderfutureclimatechange,theresponseofOMZ model(Gruberetal.,2006;Lachkaretal.,2016).Itisbased
toupwelling-favorablewindintensificationisdifficulttopre- on a system of ordinary differential equations describing
−
dict as such a perturbation may increase both oxygen sup- the time evolution of the following tracers: nitrate (NO ),
3
+
ply through enhanced ventilation and oxygen demand via ammonium (NH ), one class of phytoplankton, one class
4
increased biological productivity, thus leading to an uncer- of zooplankton, two classes of detritus and a dynamic
tainneteffect.IntheArabian Sea,thepictureismadeeven chlorophyll-to-carbon ratio. The two pools of detritus
morecomplicatedbytheseasonalreversalofwindsandthe represent, respectively, large organic matter particles that
potentialimportanceofchangesinwintermonsoonmixing. sink fast (10md−1) and small particles that sink at slower
Additionally, the large denitrification fluxes in the Arabian speed (1md−1). The small particles can coagulate with
Sea combined with the potential feedback of denitrification phytoplankton to form fast-sinking large detritus. The
on biological productivity, and hence on oxygen consump- sinking of the particles is represented explicitly in the
tion,furtheraddtotheintricacyoftheproblem.Finally,the model, thus allowing all tracers to be advected laterally
questionoftheOMZresponsetimescalesisessentialbutre- including in the euphotic zone. When sinking particles
mainsunanswered.Here,weaddressthesequestionsandex- reach the seafloor, they are remineralized back into am-
plore the mechanisms by which the Arabian Sea ecosystem monium at a much slower rate (0.003d−1) than in the
responds to monsoon wind changes using a regional eddy- water column (0.03d−1 for small detritus and 0.01d−1 for
resolvingmodel.Weexaminehowidealizedchangesinsum- large detritus). Finally, the model is coupled to a module
mer and winter monsoon wind intensity affect the produc- that describes the biological sources and sinks of oxygen
tivity and the volumes of hypoxic and suboxic water in the following Lachkar et al. (2016). Furthermore, at very low
ArabianSeaandexplorethebiogeochemicalandecological oxygenconcentrations(O <4mmolm−3),nitrificationand
2
implicationsofthesechanges.Weshowthattheproductivity aerobic remineralization are shut down and denitrification
increasesonatimescaleofyearswhiletheOMZexpandsand issettoconsumenitrateinsteadofoxygen.Finally,benthic
deepens on a timescale of decades. This response is essen- denitrification is parameterized following Middelburg et al.
tially driven by summer monsoon wind intensification and (1996) (see further details of the model in Lachkar et al.,
resultsfromanenhancedbiologicalconsumptionofoxygen 2016).
opposed by increased ventilation near the surface. The en-
hanced upper ocean ventilation leads to intermittent expan- 2.2 Experimentaldesign
sions of habitats in the epipelagic zone, while the OMZ in-
tensificationatdepthincreasesdenitrification,thusamplify- Themodeldomainextendsinlatitudefrom5◦Sto30◦Nand
ingthedepletionofbioavailablenitrogenintheArabianSea. inlongitudefrom34to78◦E,thusencompassingthewhole
WeconcludethatchangesintheIndianmonsooncanaffect Arabian Sea OMZ as well as the key dynamical features of
thelarge-scalenitrogenmarinebudgetondecadaltocenten- the circulation of the region such as the Somali and Omani
nial timescales, with a positive feedback on warming in the upwellingsystemsandtheGreatWhirl.Themodelhorizon-
caseofstrongerwinds. talgridhasaresolutionof1/12◦ withanaveragegridspac-
ing that is about 5 to 20 times smaller than the local first
baroclinicRossbydeformationradius(Cheltonetal.,1998).
2 Methods Thispermitsanexplicitrepresentationofalargefractionof
the region eddies. The vertical grid consists of 32σ layers
2.1 Models with an improved resolution near the surface. The seafloor
topographyisderivedfromtheETOPO2datasetsuppliedby
We use the Regional Ocean Modeling System theNationalGeophysicalDataCenter(SmithandSandwell,
(ROMS)_AGRIF (documented at http://www.croco-ocean. 1997).
org/) configured for the Arabian Sea region. ROMS solves Themodelisforcedwithamonthlyclimatology.Theab-
the primitive equations and has free surface and general- sence of interannual variability in the atmospheric forcing
ized terrain-following vertical coordinates (Shchepetkin enables us to quantify the role of internal variability asso-
and McWilliams, 2005). Advection is represented using ciated with mesoscale eddies. The temperature, salinity and
a rotated-split third-order upstream biased operator de- currents are initialized and laterally forced using the Sim-
signed to limit dispersive errors and preserve low diffusion pleOceanDataAssimilation(SODA)oceanreanalysis.Oxy-
(Marchesiello et al., 2009). The subgrid vertical mixing gen and nitrate initial and boundary conditions are derived
is represented using the non-local K-profile parameter- fromGarciaetal.(2010a,b)dataset.Theatmosphericbound-
www.biogeosciences.net/15/159/2018/ Biogeosciences,15,159–186,2018
162 Z.Lachkaretal.:MonsoonandArabianSeaoxygenminimumzones
ary conditions for heat and freshwater fluxes are based on 2016).Themodelalsocapturesthemainpatternsoftheob-
the Comprehensive Ocean–Atmosphere Data Set (COADS; served surface circulation both in summer and winter sea-
da Silva et al., 1994). We applied surface restoring to sons (Appendix A, Fig. A1). In particular, the model repro-
COADS-observed surface temperatures and salinities us- ducesthenortheastmonsooncurrentandtheSouthEquato-
ing kinematic flux corrections described in Barnier et al. rial Countercurrent in winter and the energetic Great Whirl
(1995). The model is forced with wind stress data from andSouthernGyreinsummer(SchottandMcCreary,2001).
theQuikSCAT-derivedScatterometerClimatologyofOcean Furthermore, the strength and the seasonal reversal of the
Winds(SCOW;RisienandChelton,2008). Somali Current are also correctly reproduced in the model.
The model is first spun up for 12 years and then is run Finally, the model reproduces the observed seasonal SSH
for an additional 50 years using nine different wind stress anomaliesinbothwinterandsummerseasons(AppendixA,
scenarios. In the control run, the wind stress is left unper- Fig. A2). In particular, both the spatial distribution and the
turbed.Inafirstsetoffourperturbedmonsoonsimulations, magnitudeofcoastalSSHgradientsareaccuratelycaptured.
thewindstresswasuniformlyincreasedordecreasedby20 The model captures the main patterns of the observed
and 50%, respectively. This amounts to wind speed pertur- SST from the Advanced Very High Resolution Radiometer
bationsofaround10and20%,respectively.Twoadditional (AVHRR) satellite data in winter and summer seasons (Ap-
runswereconductedwheretheperturbationsofthesummer pendix A, Fig. A3). In particular, the model reproduces the
and winter monsoon winds are set to be antagonistic, i.e., a east–west and north–south temperature gradients across the
50% increase (respectively, decrease) of the summer mon- model domain, as well as the strong surface gradients be-
soon wind stress that is concomitant with a 50% decrease tweentheArabianSeaanditsmarginalseas,i.e.,theArabian
(respectively, increase) in the winter monsoon wind stress. (Persian)Gulf(hereaftertermed“theGulf”)andtheRedSea.
Finally,twosimulationswithgradualincrease(respectively, Furthermore, the model also captures the cold SST tongue
decrease)ofwindstressatarate+1%yr−1areperformed. thatcharacterizessummerupwellingoffthecoastsofOman
Although these runs explore different wind perturbation andSomalia.
scenarios,theyarehighlyidealizedbynatureandarenotin- However, an examination of the north–south vertical dis-
tended to mimic past conditions from paleo-climatic recon- tributionoftemperaturerevealsthatthemodel(i)tendstoun-
structionsorrealisticfuturetrajectoriesbutratheraimatex- derestimatethesubsurfacewatertemperatureinthenorthern
ploringthesensitivityoftheArabianSeaOMZtomonsoon Arabian Sea particularly in winter and (ii) slightly overesti-
windintensitychangesandimprovingourunderstandingof mates the mixed layer depth in both seasons (Appendix A,
the key mechanisms that control the OMZ response and its Fig. A4). Similar mismatch with observations characterizes
timescales.Formodelevaluation,weusethelast10yearsof thesimulatedsalinityfields(Figs.A3andA4).Althoughthe
thecontrolrun. modelgenerallyreproducestheobservedsurfacedistribution
ofsalinity,itunderestimatesthedeeppenetrationofsaltysur-
2.3 Modelevaluation face waters in the northern region of the Arabian Sea. This
high-salinitywaterischaracteristicoftheGulfwaterandits
We use available satellite and in situ observations to assess underestimation in our model indicates that the model may
the model ability to reproduce observed physical and bio- underestimatethesubductionoftheGulfwaterintotheAra-
geochemical properties in the Arabian Sea domain. To this bianSea.Thismightbeduetomodellimitationssuchasthe
end,weevaluatethemodelintermsofsurfacecurrentsand lackofadequateverticalandhorizontalresolutionsaswellas
eddykineticenergy(EKE),seasurfaceheight(SSH)anoma- due to potential inconsistencies in atmospheric forcing data
lies,seasurfacetemperature(SST)andsurfacechlorophylla overtheGulfregion.
concentrations. Furthermore, we compare modeled primary The model successfully simulates the high chlorophyll
production and export fluxes with estimates based on avail- concentrations in the northern and western Arabian Sea as-
ablefielddataandsatelliteobservations.Finally,weexamine sociated with the winter and summer blooms, respectively
thedistributionsoftemperature,salinity,oxygenandnitrate (Fig. 2). The model also reproduces the high chlorophyll
intheupperoceaninthemodelandinLocarninietal.(2013), concentrations that are observed off the Indian west coast
Zwengetal.(2013)andGarciaetal.(2014a,b). in both seasons. However, the model substantially overes-
The model simulates successfully the surface EKE as it timates the observed chlorophyll concentrations off the So-
reproducesquiteaccuratelythespatialdistributionoftheob- mali coast south of 4◦N in both seasons. This is likely due
servedsurfaceeddyfield(Fig.1).Inparticular,theobserved toournitrogen-basedNPZDmodelthatdoesnotaccountfor
east–west gradient in EKE is captured by the model. How- potential limitation of productivity by silicic acid and iron
ever, the model tends to slightly underestimate the magni- inthisregionasdocumentedbyKonéetal.(2009).Besides
tudeoftheeddyactivityintheeasternandnortheasternAra- this modeldeficiency, the large-scale distributionof chloro-
bian Sea. This might be due to the resolution of the model phyll is fairly well represented in most of the Arabian Sea
that does not permit yet the representation of the full eddy despiteatendencytounderestimatechlorophyllintheopen
spectruminthisregion(Cheltonetal.,1998;Lachkaretal., ocean.Thismighthavetodowiththemodel’slackofsmall
Biogeosciences,15,159–186,2018 www.biogeosciences.net/15/159/2018/
Z.Lachkaretal.:MonsoonandArabianSeaoxygenminimumzones 163
phytoplankton functional groups that are more adapted to
oligotrophicconditions.Overall,thefidelityofthemodeled
chlorophyll is, however, in line with that simulated by state
of the art models of the Indian Ocean and the Arabian Sea
(e.g.,Resplandyetal.,2012).
Wefurthercomparesimulatedprimaryproductionandex-
portfluxeswithestimatesbasedonavailablefielddatafrom
theUSJointGlobalOceanFluxStudy(JGOFS)ArabianSea
Process Study (Lee et al., 1998). These consist in (i) in situ
measuredprimaryproduction,(ii)exportfluxesat100mes-
timatedusing234Thremovalratesand(iii)exportfluxeses-
timatedfromsedimenttrapdataat500mabovetheseafloor.
Fluxesweremeasuredessentiallyduringtheyear1995atfive
sites (M1, M2, M3, M4 and M5) along a transect extend-
ingfromthecoastofOmantothecentralArabianSea(Lee
etal.,1998).Becauseofthelimitednumberofinsituobser-
vationsofproductivity(onlyfivemeasurementsateachsite),
we also use satellite-based productivity estimates obtained
fromtwosensors(SeaWiFSandMODIS)andbasedontwo
different algorithms: the Vertically Generalized Production
Model (VGPM; Behrenfeld and Falkowski, 1997) and the
Carbon-based Production Model (CbPM; Westberry et al.,
2008).Themodelcorrectlysimulatestheobserveddecrease
in productivity associated with increasing distance to coast
(Appendix A, Fig. A5). Yet, it substantially underestimates
theinsitubasedestimatesatallsites.Someofthismismatch
may be due to the fact that the in situ estimates are derived
from asmall numberof independentmeasurements that are
allcomingfromoneindividualyear(1995).Giventheimpor-
tanceofbothmesoscaleandinterannualvariability,thesees-
timatesmaythereforenotberepresentativeofthelong-term
climatologicalconditionssimulatedbythemodel.Indeed,a
betteragreementisobtainedbetweenthemodeledproductiv-
ity and estimates based on satellite observations that have a
moreextensivetemporalcoverage(Fig.A5).Wefurthercon-
trastedthesimulatedexportfluxesat100mtoestimatesfrom
Leeetal.(1998)atthefivestations(Fig.A6a).Ourmodeled
export fluxes generally overestimate the 234Th-based esti-
matesbutremaincomparabletotheminmagnitude.Further-
more, the model reproduces quite accurately the observed
offshoregradientinexport.Itisworthhighlighting,however,
thatsimilarlytotheinsitumeasuredproductivity,theseex-
port fluxes are based on four independent measurements at
each site only, all from the same year (1995). Finally, com-
paring the modeled export fluxes in the deep ocean (500m
above the seafloor) to sediment trap data at the same sites
yieldsagoodagreementbetweenthemodelandtheobserva-
tions in all stations (Fig. A6b). It is worth noting that these
deepexportfluxestimatescanbeconsideredasmorerobust
than those at 100m, as they are based on a larger number
Figure1.Surfaceeddykineticenergy.Surfaceeddykineticenergy
ofindependentmeasurements(20–40measurementsateach
(in cm2s−2) as simulated in the model (a) and from data based
site).
onAVISOsatellitealtimeter(b)andsurfacedrifterclimatologyof
The simulated surface distribution of nitrate is consistent
LumpkinandJohnson(2013)(c).
withobservationsfromtheWorldOceanAtlasdatasetinboth
seasons (Fig. A7). In particular, the high surface concen-
www.biogeosciences.net/15/159/2018/ Biogeosciences,15,159–186,2018
164 Z.Lachkaretal.:MonsoonandArabianSeaoxygenminimumzones
Figure2.Surfacedistributionofchlorophylla.Surfacechlorophyll-a concentrations(inmgm−3)assimulatedinROMS(a,c)andfrom
SeaWiFS(b,d)duringwinter(a,b)andsummer(c,d)months.TheSeaWiFSclimatologyiscomputedovertheperiodfrom1997to2009.
trations associated with the summer upwelling in the west- ogy.Nospatialinterpolationismade,andthemodelandob-
ern Arabian Sea and the winter convection in the northern servationsarecomparedattheobservationpoints.Theresults
Arabian Sea are captured by the model. Finally, the simu- of this evaluation are summarized graphically using Taylor
lated oxygen distributions at depth compare generally well diagrams (Taylor, 2001) quantifying the similarity between
withobservations(Fig.3).Inparticular,thelocation,thesize theobservationsandmodelintermsoftheircorrelation,the
and the depth of the Arabian Sea OMZ (defined here by amplitudeoftheirstandarddeviationsandtheircenteredroot
O <60mmolm−3) are correctly reproduced. However, the meansquare(rms)difference(Fig.4).Ourstatisticalanalysis
2
model overestimates the intensity of the OMZ in the north- showsthatthesimulatedandobservedSSHanomalieshave
ern Arabian Sea and off the Indian west coast. This might similar standard deviations and correlate strongly with cor-
be driven by the model underestimated eddy activity that relationcoefficientsof0.77and0.83inwinterandsummer,
may lead to underestimated OMZ ventilation as shown by respectively(Fig.4a).Similarly,thequantitativecomparison
Lachkaretal.(2016).Thismightalsopartlyresultfromthe ofthesimulatedandobservedSSTclimatologiesshowssim-
misrepresentationandlikelyunderestimationoftheinjection ilarstandarddeviationsandsmallrmserrorsaswellasvery
oftheGulfwatersintothenorthernArabianSeaatinterme- high correlations (r >0.92) in both seasons, confirming the
diatedepths. visualcomparisonconclusions(Fig.4a).Likewise,thesimu-
A more quantitative evaluation of model skill is per- latedandobservedchlorophylldistributionsshowrelatively
formed using in situ observations from the World Ocean lowrmserrorsinbothseasonswithdecentcorrelationsrang-
Database(2013)thatwebinnedintoa0.5◦monthlyclimatol- ingfrom0.69insummerto0.74inwinter(Fig.4a).Despite
Biogeosciences,15,159–186,2018 www.biogeosciences.net/15/159/2018/
Z.Lachkaretal.:MonsoonandArabianSeaoxygenminimumzones 165
Figure3.Horizontalandverticaldistributionsofoxygen.Distributionofannual-meanoxygen(inmmolm−3;a,b)at200mand(c,d)in
◦
theupper1000malong62 EassimulatedinROMS(a,c)andfromtheWorldOceanAtlas(2009)dataset.
theunderestimationofthepenetrationofthewarmandsalty our results will be addressed in detail in the discussion sec-
surfacewatersinthenorthernArabianSeaatdepth,thelarge- tion.
scale distributions of temperature and salinity in the upper
ocean are captured by the model as indicated by low rms
errors and very high correlations with in situ observations 3 Results
(r >0.91)forbothtemperatureandsalinity(Fig.4b).Addi-
tionally, the model shows an overall high skill in reproduc- To explore the sensitivity of the Arabian Sea ecosystem to
ingtheobservedlarge-scaledistributionsofbothnitrateand changesintheintensityofmonsoonwinds,weconsidervar-
oxygen, with correlations with the World Ocean Database iousscenariosofidealizedwindperturbations.Afirstsetof
rangingintheupper200mfrom0.88fornitrateinsummer scenariosconsistsinincreasing(respectively,decreasing)the
monthstoaround0.93foroxygeninbothseasons(Fig.4b). wind stress over the whole domain and throughout the year
In summary, despite some local biases, the model gen- by20and50%,respectively.Inasecondsetofexperiments,
erally shows reasonable skill in reproducing the large-scale weincreasethewindstressby50%insummer(respectively,
features of the circulation in the Arabian Sea as well as the winter)anddecreaseitby50%inwinter(respectively,sum-
seasonal dynamics of phytoplankton blooms in the region. mer). This is to account for perturbation scenarios where
More importantly, it reproduces fairly well the location and summerandwintermonsoonwindsevolveinoppositedirec-
structureoftheArabianSeaoxygenminimumzone.Thedis- tions as suggested by multiple paleo-records (e.g., Klöcker
cussion of the potential impact of the model limitations on andHenrich,2006)showingintensificationofsummermon-
soon to be concomitant with the weakening of winter mon-
www.biogeosciences.net/15/159/2018/ Biogeosciences,15,159–186,2018
166 Z.Lachkaretal.:MonsoonandArabianSeaoxygenminimumzones
Taylor diagrams
(a) Sea surface (b) Upper ocean
Winter (JFM)
Summer (JJAS)
Figure 4. Taylor diagram displaying statistical comparison of modeled and observed fields. Taylor (2001) diagram of simulated (a) sea
surfacetemperature(datapointslabeled“1”),seasurfaceheightanomalies(datapointslabeled“2”)andsurfacechlorophylla(datapoints
labeled “3”) in winter (blue) and summer (red) months and (b) temperature (data points labeled “1”), salinity (data points labeled “2”),
oxygen(datapointslabeled“3”)andnitrate(datapointslabeled“4”)intheupper1000mandinthetop200minwinter(blue)andsummer
(red)months.ThereferencepointoftheTaylordiagramcorrespondsto(a)SeaWiFSobservationsforchlorophyll,AVHRRdataforsurface
temperatureandAVISOclimatologyforseasurfaceheightanomaliesand(b)WorldOceanDatabase(2013)forallfourvariables.Theradius
(distance to the origin point) represents the modeled standard deviation relative to the standard deviation of the observations. The angle
betweenthemodelpointandthex axisindicatesthecorrelationcoefficientbetweenthemodelandtheobservations.Finally,thedistance
fromthereferencepointtoagivenmodeledfieldrepresentsthatfield’scenteredrmswithrespecttoobservations.
soon, and vice versa. This will also allow us to disentangle duction(explainsmorethan40%oftheannuallevelswhile
theimpactsofthechangesineachmonsoonseasonandde- NEM productivity contributes by less than 33%). Second,
terminetheirrespectivecontributionstotheoverallresponse. summer productivity is more sensitive to wind changes as
Thewindperturbationsareappliedinstantlyandmaintained itisdirectlydrivenbywind-inducedupwelling.Incontrast,
over 50 years for each scenario (abrupt perturbation). The NEMproductivityisdrivenbywintertimecoolingandcon-
impactofperturbationsentailinggradualchangesinthewind vection.Hence,NEMwindintensificationenhancesvertical
intensitywillbeaddressedlaterinSect.4. mixing and surface nutrient concentrations, but also deep-
ens the mixed layer, thus potentially increasing light limi-
3.1 ResponseofNPP tation. Indeed, the winter turbulent mixed layer deepens in
the northern Arabian Sea by up to 20–25m and penetrates
The magnitude of the Arabian Sea net primary production below the euphotic zone (1% light depth) at 65–70m. This
(NPP) response is proportional to the amplitude of the per- increases the average exposure of phytoplankton to light-
turbation(Fig.5).Forinstance,NPPincreases(respectively, limited conditions, thus potentially limiting the net growth
decreases) by around 20 and 45 to 50% in response to an rate (i.e., gross photosynthetic rate minus loss terms due to
increase(respectively,decrease)inthewindstressof20and mortality, grazing, sinking and respiration) over the water
50%,respectively.However,whenthesummerwindintensi- column and hence reducing the potential biomass and pro-
fication(respectively, weakening)is concomitantwith ade- ductivity(Franks,2015).Thismayexplainthemorelimited
crease(respectively,increase)inthewintermonsoonwinds, increase in winter productivity (+38 increase in response
the annual-mean NPP still shows an increase (respectively, to 50% increase in wind stress) in comparison to summer
decrease) but of a weaker amplitude. Therefore, we con- productivity (+52 increase in response to 50% increase in
clude that the NPP response is dominated by the summer wind stress). Finally, the response of NPP to changes in In-
southwestmonsoon(SWM)perturbationandthatthewinter dianmonsoonwindintensitydevelopsoverarelativelyshort
northeastmonsoon(NEM)windchangesplayaminorrole. timescale(Figs.6andA8).Indeed,theNPPreachesaquasi-
TwofactorsexplainthestrongcontroloftheSWMperturba- steady state with limited internal variability within 3 to 5
tion over the NPP annual-mean response. First, the biolog- yearsfromthestartoftheperturbation.
ical production during the SWM dominates the annual pro-
Biogeosciences,15,159–186,2018 www.biogeosciences.net/15/159/2018/
Z.Lachkaretal.:MonsoonandArabianSeaoxygenminimumzones 167
100 100 NPP
NPP
e to control (%) 24680000 DHSSSuWWyebpn+-oo,i, xxt NNriicciEfE ivv+c-ooalluutimmonee e to control (%) 6800 DHSuyebpnooixxtriiccif ivvcooalluutimmonee
Change relativ ---6420000 0.5 x τ 0.8 x τ Win1d x s τtress 1.2 x τ 1.5 x τ Change relativ 2400
-80
0
-100 0 5 10 15 20 25 30 35 40 45 50
Time (year)
Figure 5. Biogeochemical response to monsoon wind intensity -20
changes. Relative changes in response to monsoon wind inten-
Figure6.Responsetomonsoonwindincreaseasafunctionoftime.
sityperturbationsinnetprimaryproduction(green),denitrification
Response to wind stress increase (+50%) of NPP, denitrification
(blue), suboxic volume (red) and hypoxic volume (orange). Open
andsuboxicandhypoxicvolumesasafunctionoftime.
circles(respectively,squares)indicatetheresultsfromthesimula-
tionwheresummermonsoonwindstressisincreased(respectively,
decreased) by 50% and winter monsoon wind stress is decreased
response (on a timescale of several decades) characterized
(respectively,increased)by50%.
byastronginternalvariability(Fig.6).However,theampli-
tude of denitrification response is generally larger than that
ofsuboxia(Fig.5).Forinstance,denitrificationincreasesby
3.2 ResponseofOMZanddenitrification
morethan85%whensummermonsoonintensifiesby50and
wintermonsoonweakensby50%,whereasthesuboxicvol-
Under monsoon wind intensification (respectively, weak-
umeincreasesbyonlyaround60%underthesameperturba-
ening), the OMZ (defined here as hypoxic water with
tion.Thisisconsistentwithdenitrificationfluxesdepending
O <60mmolm−3)expands(respectively,contracts;Fig.5).
2 both on the volume of suboxia as well as on the amplitude
The OMZ response is much slower and of a weaker ampli-
ofproductionfluxes.AsbothNPPandsuboxiaincreaseun-
tude than the NPP response (Figs. 6 and A8). For instance,
der monsoon intensification, the cumulative effect of their
it takes several decades for the hypoxic volume to reach a
increasesleadstoalargerdenitrificationresponse.
quasi-steady state. Furthermore, the OMZ expands (respec-
Under increased monsoon winds, hypoxia increases at
tively, contracts) by a mere 20 in response to a 50% wind
depthbutisreducedintheupperocean(Fig.7a).Similarly,
stressincrease(respectively,decrease).SimilarlytoNPP,the
thesuboxicvolumeexpandsatdepthandcontractsnearthe
OMZ response is dominated by the summer perturbation
surface. Indeed, the peak of low O located in the lower
2
(Fig.5).Thiscanbepartiallyexplainedbythelargersummer
epipelagic zone (150–200m) as seen in Fig. 3 weakens un-
productivity and its larger sensitivity to wind changes lead-
derenhancedmonsoonwinds,whereasthelowerOMZpeak
ingtostrongerperturbationoftheO demand.Additionally,
2 (<400m)expands.Thisresultsintheintensificationofden-
thedeepening(byupto25m)ofthewintertimemixedlayer
itrification at depth and its attenuation in the upper ocean
that results from NEM intensification enhances the ventila-
(Fig.7b).Thesechangesinvolvevaryingtimescalesandcon-
tionofthenorthernandnortheasternArabianSea,thuscom-
trasting levels of internal variability (Figs. 8 and A8). The
pensating the mild increase in O consumption that results
2 compression of the OMZ in the epipelagic zone (0–200m)
from enhanced winter productivity. Finally, the response of
is relatively quick (i.e., a timescale of years) and is associ-
the suboxic volume (defined here as O <4mmolm−3) is
2 atedwithastronginternalvariability.Ontheotherhand,the
similarly slow (30 to 50 years) but of much larger ampli-
OMZexpansionismuchslower(i.e.,atimescaleofdecades)
tude (Figs. 5 and 6). Indeed, the suboxic volume expands
intheintermediate(i.e.,mesopelagiczone200–1000m)and
by around 50% in response to a 50% increase in the wind
thedeep(i.e.,bathypelagiczone>1000m)oceansandisac-
stress. The amplitude of this response is even larger (more
companiedbyamuchweakerinternalvariability.
than 60%) when the summer monsoon intensification oc-
curs concurrently with a weakening of the winter monsoon
winds (Fig. 5). However, the response of the suboxic vol- 4 Discussion
umescomeswithasubstantiallystrongerinternalvariability
(Fig.6). 4.1 DriversoftheOMZchangeanditstimescale
Asdenitrificationdevelopsonlyundersuboxicconditions,
its response to monsoon changes is modulated by the re- InordertoelucidatethefactorsdrivingtheOMZchanges,we
sponse of the suboxic volume. Similarly, it shows a slow performanoxygenbudgetanalysisintheOMZvolume.We
www.biogeosciences.net/15/159/2018/ Biogeosciences,15,159–186,2018
168 Z.Lachkaretal.:MonsoonandArabianSeaoxygenminimumzones
(a) ((bb))
00.5.5xx (cid:1) τ
CCoonntrtoroll
11.5.5xx (cid:1) τ
m) 0C.o5nxt rτol m)m)
Depth ( 1.5x τ 0.5x τ Depth (Depth (
Control
1.5x τ
Area (106 km2) DDeennititrrifiifcicaattioionn ( (GGmmool lm m-1-1))
Figure 7. Changes in the Arabian Sea OMZ and denitrification as a function of depth. (a) Total area (in 106km2) occupied by hypoxic
(solidlines)andsuboxic(dashedlines)watersasafunctionofdepthunderdifferentmonsoonwindintensities.(b)ArabianSeaintegrated
annual-meandenitrification(in109molm−1)asafunctionofdepth.Notethepresenceofaweaksecondarymaximumindomain-integrated
denitrificationbetween500and800minallsimulations.Inthecontrolsimulation,thiscanbeexplainedbythefactthatatthisdepthrangethe
areaoccupiedby(potentiallydenitrifying)suboxicwaterislargest,andweakdenitrificationisstillpresentalthoughatanorderofmagnitude
weakerrate(perunitofvolumebasis)thanat200m.
firsttakeintoaccountthewholewatercolumnandthencarry 50
out the analysis separately in three layers: the epipelagic
40
zone(0–200m),themesopelagiczone(200–1000m)andthe
bathypelagic zone (>1000m). For simplicity, we consider
30
only the two most extreme scenarios where the wind stress %) Bathypelagic (>1000 m)
is increased or decreased by 50%. For each simulation, we ol ( 20
subtractthecontrolrunfromtheperturbationrunandfocus ntr
o
c 10
ontheanomalieswithrespecttothenon-perturbedcase.We o
e t Mesopelagic (200–1000 m)
quantify the annual oxygen accumulation within the OMZ ativ 0
with respect to the control run and determine the contribu- el 0 5 10 15 20 25 30 35 40 45 50
tionsoftransportandbiologytothis(Figs.9andA9). ge r -10 Time (year)
n
a
This analysis shows that when considering the whole Ch
-20
OMZ, the oxygen accumulation is generally negative (cor- Epipelagic (0–200 m)
responding to a net loss of O2) over the first 2 decades -30
whenmonsoonwindsareincreased(Fig.9a).However,this
-40
term becomes gradually small and oscillates around zero
during the last decade of the simulation as the OMZ tends
-50
towards a different steady state. This explains the fast in-
Figure8.OMZresponseindifferentverticallayersasafunctionof
creaseintheOMZvolumeoverthefirstfewdecadesandits
time.Changesinhypoxicvolume(in%)underwindstressincrease
quasi-stabilizationbytheendofthesimulation.Thisoxygen
(+50%) in the epipelagic zone (0–200m), the mesopelagic zone
lossisdrivenbyanimbalancebetweenincreasedO supply
2 (200–1000m)andthebathypelagiczone(>1000m).
fromtransportandenhancedO consumptionduetobiology
2
(Fig.9a).Overthefirst20to30yearsunderincreasedwinds,
the enhanced supply of oxygen through ventilation under-
acterizes the OMZ oxygen content (and hence its size and
compensates the additional sink of oxygen due to biology.
intensity)isassociatedwiththeventilationcontribution.
Thisisbecausethebiologicalresponseismuchquickerthan
The oxygen budget in the upper (0–200m) OMZ reveals
theventilationresponse.Indeed,ittakesaround30yearsfor
thattheO accumulationtermispredominantlypositiveover
the transport contribution to reach a quasi-equilibrium and 2
the first few years of the simulation and is driven by a sup-
only around 3 years for the biological response to reach its
plyofoxygenthroughtransportthatexceedsbiologicalcon-
steadystate.Thisdifferenceintheresponsetimescaleofbi-
sumption (Fig. 9b). The oxygen source and sink reach a
ologyandcirculationisresponsiblefortheslowexpansionof
quasi-balance within a couple of years. This explains the
OMZ.Additionally,mostoftheinternalvariabilitythatchar-
rapidresponseoftheOMZintheupperoceaninourmodel.
The strong internal variability that characterizes the time
Biogeosciences,15,159–186,2018 www.biogeosciences.net/15/159/2018/