Response of O2 and pH to ENSO in the California Current System in a high-resolution global ...

18 Pages·2017·18.06 MB·English

OceanSci.,14,69–86,2018
https://doi.org/10.5194/os-14-69-2018
©Author(s)2018.Thisworkisdistributedunder
theCreativeCommonsAttribution4.0License.
Response of O and pH to ENSO in the California Current System
2
in a high-resolution global climate model
GiulianaTuri1,MichaelAlexander2,NicoleS.Lovenduski3,AntoniettaCapotondi2,JamesScott4,CharlesStock5,
JohnDunne5,JasminJohn5,andMichaelJacox6
1NationalSnowandIceDataCenter,Boulder,CO,USA
2NOAA/ESRL,Boulder,CO,USA
3DepartmentofAtmosphericandOceanicSciencesandInstituteofArcticandAlpineResearch,
UniversityofColorado,Boulder,CO,USA
4CIRES,UniversityofColoradoatBoulder,andNOAA/ESRL,Boulder,CO,USA
5NOAA/GFDL,Princeton,NJ,USA
6UniversityofCalifornia,SantaCruz,CAandNOAA/SWFSC,Monterey,CA,USA
Correspondence:MichaelAlexander([email protected])
Received:3August2017–Discussionstarted:17August2017
Revised:12December2017–Accepted:21December2017–Published:2February2018
Abstract.Coastalupwellingsystems,suchastheCalifornia affectedbyupwelling,explainingtheconfinednatureofthe
Current System (CalCS), naturally experience a wide range pHsignalclosetothecoast.Below∼100m,wefindcondi-
of O concentrations and pH values due to the seasonality tionswithanomalouslylowO andpH,andbyextensionalso
2 2
ofupwelling.Nonetheless,changesintheElNiño–Southern anomalouslylowaragonitesaturation,duringLaNiña.This
Oscillation (ENSO) have been shown to measurably affect result is consistent with findings from previous studies and
the biogeochemical and physical properties of coastal up- highlightsthestressthattheCalCSecosystemcouldperiod-
wellingregions.Inthisstudy,weuseanovel,high-resolution icallyundergoinadditiontoimpactsduetoclimatechange.
globalclimatemodel(GFDL-ESM2.6)toinvestigatethein-
fluenceofwarmandcoldENSOeventsonvariationsinthe
O concentrationandthepHoftheCalCScoastalwaters.An
2
assessment of the CalCS response to six El Niño and seven 1 Introduction
La Niña events in ESM2.6 reveals significant variations in
theresponsebetweenevents.However,thesevariationsover- Ocean deoxygenation (decreasing O2 concentration) and
lay a consistent physical and biogeochemical (O and pH) ocean acidification (decreasing pH) are considered to be
2
response in the composite mean. Focusing on the mean re- major oceanic ecosystem stressors that can severely reduce
sponse,ourresultsdemonstratethatO andpHareaffected habitat suitability in benthic and pelagic ecosystems (e.g.,
2
rather differently in the euphotic zone above ∼100m. The Doneyetal.,2009a;Doney,2010;Gruber,2011;Boppetal.,
strongestO responsereachesuptoseveralhundredsofkilo- 2013;Breitburgetal.,2015).Coastalupwellingecosystems,
2
metersoffshore,whereasthepHsignaloccursonlywithina suchastheCaliforniaCurrentSystem(CalCS),supportsome
∼100kmwidebandalongthecoast.Bysplittingthechanges oftheworld’smostproductivefisheriesduetotheseasonal,
in O and pH into individual physical and biogeochemical wind-driven upwelling of nutrient-rich waters (e.g., Chavez
2
componentsthatareaffectedbyENSOvariability,wefound and Messié, 2009; Messié et al., 2009). The upwelled wa-
thatO variabilityinthesurfaceoceanisprimarilydrivenby ters fuel biological production in the CalCS but are also
2
changes in surface temperature that affect the O2 solubility. characterizedbylowerO2 andlowerpHthanthesurround-
Incontrast,surfacepHchangesarepredominantlydrivenby ing surface waters (e.g., Bograd et al., 2008; Nam et al.,
changesindissolvedinorganiccarbon(DIC),whichinturnis 2011). Although the CalCS is accustomed to seasonal fluc-
tuations in O and pH, several observational studies sug-
2
PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion.
70 G.Turietal.:ResponseofO andpHtoENSOintheCalCSinahigh-resolutionclimatemodel
2
gest that both have decreased in the last decades, particu- event(LaNiña),theseprocessesaretypicallyreversed,with
larly within 100km of the coast (e.g., Bograd et al., 2008; nearshore surface waters being anomalously low in O and
2
Feelyetal.,2008;Nametal.,2011;Bednaršeketal.,2014; pH due to a shoaling of the isopycnal surfaces (Nam et al.,
Bednaršek and Ohman, 2015). Specifically, Bograd et al. 2011).
(2008) found a shoaling of the hypoxic threshold, which AnumberofstudieshavelookedattheinfluenceofENSO
is typically considered to be ∼60µmolkg−1 (Gray et al., on the physics, biogeochemistry, and biology of the CalCS
2002;DiazandRosenberg,2008;Vaquer-SunyerandDuarte, over the past several decades. For instance, the influence of
2008),ofupto90minthesouthernCalCSbetween1984and thestrong1997/1998ElNiñoandsubsequentLaNiñaonthe
2006.Furthermore,Feelyetal.(2008)reportanincidenceof coastal upwelling ecosystem of the US west coast has been
nearshoresurfacewatersbecomingcorrosiveduringtheearly well documented by a variety of observational studies that
upwelling season (May/June) of 2007, with the saturation focus on the physics and hydrography (e.g., Collins et al.,
state of aragonite ((cid:127) ) dropping below 1, corresponding 2002; Kosro, 2002; Lynn and Bograd, 2002; Ryan and No-
arag
topH<7.75(aragoniteisamineralformofcalciumcarbon- ble, 2002; Schwing et al., 2002b), on nutrients and primary
ate,CaCO ,foundintheshellsofcertaincalcifyingmarine production (e.g., Castro et al., 2002; Kahru and Mitchell,
3
organisms). Turi et al. (2016) used a regional ocean model 2000,2002;KudelaandChavez,2002),andonzooplankton
to demonstrate that a recent intensification of upwelling- andhighertrophiclevels(e.g.,Bensonetal.,2002;Hopcroft
favorablewindsislinkedtoadropincoastalpHand(cid:127) .It etal.,2002;Pearcy,2002;Petersonetal.,2002).Ahandful
arag
currentlyremainsunclearwhethertheseobservedandmod- of observational studies have investigated the impact of the
eledchangesintheCalCSoceanbiogeochemistryareongo- 1982/1983ElNiñoonthephysicalregimeoftheCalCS(e.g.,
ingsignalsofanthropogenicclimatechange,andthuscould Simpson,1983;HuyerandSmith,1985)andtheeffectofthe
continue into the future, or whether they are driven by nat- 2002/2003 El Niño on the physics and biology (e.g., Mur-
ural fluctuations in the climate system (e.g., Bakun, 1990; phree et al., 2003; Schwing et al., 2002a). Several regional
Narayanetal.,2010;Bakunetal.,2015;Rykaczewskietal., modelingstudieshaveshowntheconnectionbetweenENSO
2015;Wangetal.,2015). andtheverticaltransport,watercolumndensity,origins,and
TheeasternNorthPacificregionisaffectedbyseveraltele- propertiesofupwelledwater(Jacoxetal.,2014;Jacoxetal.,
connectiveclimatepatterns,includingtheElNiño–Southern 2015b),anddemonstratedtherelativeimportanceofremote
Oscillation(ENSO;e.g.,Alexanderetal.,2002;Mantuaand versuslocalforcingonboththephysicsandthebiogeochem-
Hare, 2002; Di Lorenzo et al., 2008; Alexander, 2010; Di istry of the CalCS (Frischknecht et al., 2015; Jacox et al.,
Lorenzoetal.,2010;Maciasetal.,2012;Franksetal.,2013; 2015a;Frischknechtetal.,2017).Duringthe2010/2011La
Capotondi et al., 2015; Newman et al., 2016). ENSO is a Niña, Nam et al. (2011) observed decreases in O and pH
2
major driver of interannual physical variability and affects in the upwelling region along the coast that were 2–3 times
a range of variables on a local scale in the CalCS, such larger than expected solely due to the cross-shore shoaling
as sea surface temperature, thermocline depth, and the in- oftheisopycnalsurfaces.Theyfoundthattheadditionalre-
tensity and depth of upwelling (e.g., Chavez et al., 2002; duction of O was related to decreased subsurface primary
2
Schwing et al., 2005; Checkley Jr. and Barth, 2009; Jacox production and a short-term strengthened poleward flow of
et al., 2015b), and operates remotely through two distinct theCaliforniaUndercurrent.Inaddition,pHdroppedbelow
pathways:(i)throughtheatmospherebyaffectingtheinten- the critical value of 7.75 both during the upwelling season
sity and location of the Aleutian Low pressure system and (typically April–September)and the two followingLa Niña
the fluxes of momentum, heat, and freshwater through the months. These results suggest that severe low-O and low-
2
surfaceoceanand(ii)throughtheoceanbygeneratingther- pHconditionsmayoccurifLaNiñaconditionsoverlapwith
moclineanomaliesinthetropicalPacificthatpropagateeast- seasonalupwelling.
ward along the Equator and northward as coastally trapped To our knowledge, the study by Nam et al. (2011) is the
waves.DuringatypicalENSOwarmevent(ElNiño),theat- only one to investigate the influence of ENSO on both O
2
mosphericinfluenceistwo-fold:surfaceheatfluxesintothe and pH in the CalCS. It is only recently that CalCS ENSO
oceanareanomalouslyhighandtheAleutianLowintensifies responses have begun to be monitored by modern oceano-
andisdisplacedsoutheastward,causingadecreaseinequa- graphic and biogeochemical measurements. Moreover, we
torward, upwelling-favorable winds in the CalCS. From the have yet to come across any studies investigating responses
oceanicside,thecoastallytrappedwavescauseadepression acrossmultipleENSOevents.Earthsystemmodels(ESMs),
ofthethermo-andnutriclineinthenearshoreregion,leading such as those participating in the Climate Model Intercom-
to reduced upwelling of nutrient-rich, cool waters and thus parisonProjectphase5(CMIP5),provideanopportunityto
limitingbiologicalproductioninthesurfacewaters(Bograd address these limitations. Global ESMs used for multicen-
andLynn,2001).SourcewatersforupwellingintheCalCS tennial climate change simulations, however, simulate the
also tend to be anomalously warm, shallow, and fresh dur- ocean at a fairly coarse horizontal resolution (∼1◦; Stock
ingElNiño(Jacoxetal.,2015b),resultinginsimilarlywarm et al., 2011). Models with such resolutions are challenged
and fresh anomalies at the surface. During an ENSO cold to reproduce the responses of coastal ecosystems to basin-
OceanSci.,14,69–86,2018 www.ocean-sci.net/14/69/2018/
G.Turietal.:ResponseofO andpHtoENSOintheCalCSinahigh-resolutionclimatemodel 71
2
scale ocean variability (Saba et al., 2016). It is commonly the ORNL/CDIAC CO2SYS carbonate chemistry routines
acknowledgedthatatleast0.1◦horizontaloceanresolutionis (LewisandWallace,1998).Diskspacelimitedtheamountof
necessarytoresolvetheRossbyradiusofdeformation(e.g., output that could be saved. Therefore, we analyze all avail-
Fiechter et al., 2014; Dunne et al., 2015), which is around able full-depth profiles of temperature, salinity, O , and the
2
20–60kmintheCalCS(Cheltonetal.,1998).Inthisstudy, hydrogenionconcentration([H+]),fromwhichwecompute
we use a high-resolution (0.1◦), fully coupled global Earth pH,aswellassurfacedissolvedinorganiccarbon(DIC)and
System Model (ESM2.6) developed by NOAA’s Geophys- surfaceALK,onmonthlytimescales.
ical Fluid Dynamics Laboratory (GFDL) to investigate the We analyze a 52-year control simulation with constant
effect of ENSO on O and pH in the CalCS and to address 1990 atmospheric CO forcing. As this simulation was not
2 2
thefollowingquestions: forced by a transient atmospheric CO signal, it lends itself
2
particularly well to the analysis of interannual variability.
1. How consistent is the physical and biogeochemical re- The physical climate of the 52-year control simulation was
sponseoftheCalCSacrossENSOevents?Howdothese initializedfrom1Januaryofyear141ofaCM2.61990con-
responsesdifferbetweendifferentmodelresolutions? trol simulation. The ocean biogeochemistry was initialized
from 1 January of year 105 of a previous development ver-
2. Whataretheprimarydriversandmechanismsaffecting
sionofESM2.61990.ObservedclimatologiesofO ,nitrate,
O andpHintheCalCS? 2
2 phosphate, and silicate were taken from the World Ocean
3. IsthereadifferenceinENSO’sinfluenceonO andpH Atlas 2005 (WOA05; Garcia et al., 2006a, b), and modeled
2
betweenthenearshoreregionasopposedtotheoffshore DICandALKwereinitializedfromtheGlobalDataAnaly-
regionandbetweenthesurfaceasopposedtoatdepth? sisProject(GLODAP;Keyetal.,2004).Theoceanbiogeo-
chemistryinthepreviousESM2.6developmentversionwas
4. How can these results help inform the observational startedfromaspun-upESM2M-COBALT1860controlsim-
communityaboutthelocationandfrequencynecessary ulation(Stocketal.,2014a).
tocaptureENSOsignalsintheirtimeseries?
2.2 Methods
This novel model setup allows us to investigate both
oceanic and atmospheric components of the ENSO forc- We identified model ENSO events through the ±1 stan-
ing on the CalCS, as the horizontal oceanic resolution is
dard deviation of the wintertime (November–December–
highenoughtosimulatecoastallytrappedwavespropagating
January; NDJ) Niño3.4 index, i.e., using area-averaged sea
northalongthecoastandtheatmosphericcomponentallows surface temperatures (SSTs) over 5◦S–5◦N, 170–120◦W.
forarepresentationofbasin-scaleteleconnectionprocesses.
WechosetofocusonNDJratherthanthemorecommonDJF
(December–January–February),asthemaximumENSOsig-
nalobservedinnatureoccursonaverageduringthistimepe-
2 Modeldetailsandmethods
riod.Inaddition,thereisalagintheclimatesystemofseveral
2.1 Modelsetupandsimulation months between the maximum ENSO signal and when this
signal is experienced by the midlatitudes (Alexander et al.,
We use a prototype, fully coupled global ESM2.6 devel- 2002).
opedbyNOAA’sGFDL.IncontrasttoGFDL’spubliclyre- Due to drift issues in the carbonate chemistry at the be-
leased models that undergo years of iterative development ginningofthesimulation,weusedaLanczoshigh-passfilter
and analysis to assure fidelity to a suite of observational with a cutoff frequency of 10 years (121 weights; Duchon,
metrics, ESM2.6 was implemented as a single test simula- 1979).Thedriftaffectsroughlythefirst10to20yearsofthe
tion as proof of concept. ESM2.6 was built upon the high- simulation(seeFig.S1intheSupplement).Weoptednotto
resolution CM2.6 physical climate model (Delworth et al., remove these affected years but rather to filter the data, in
2012; Griffies et al., 2015; Saba et al., 2016). The model’s order to retain as many years as possible, given the relative
ocean component is GFDL’s Modular Ocean Model ver- brevityofthesimulation.Thisprocedureremovedanylong-
sion 5 (MOM5; Griffies, 2012) with a horizontal resolution term trends and decadal variability, and retained variability
of 0.1◦ and 50 vertical layers. The ocean biogeochemistry onaninterannualtimescale(i.e.,thetimescaleofENSOfre-
model is the Carbon, Ocean Biogeochemistry and Lower quency).Usingthisapproach,wewereleftwithsixElNiño
Trophics (COBALT) model as used in Stock et al. (2014a, andsevenLaNiñaeventsinthetime-filtereddata.
b) with modifications as described in Stock et al. (2017). For the majority of the analyses, unless otherwise noted,
COBALT includes 33 prognostic tracers, as well as three we employed standardized anomalies (σ, where a value of
phytoplankton groups and three zooplankton groups to rep- 2 corresponds roughly to a 95% significance as indicated
resent the coupled elemental cycles of carbon, nitrogen, by a Student’s t test) instead of showing absolute values.
phosphorus, silicate, iron, alkalinity (ALK), and lithogenic In order to allow for a more pattern-driven interpretation of
material. Its carbonate chemistry calculation is based on ENSO-relatedsignalsinthefigures,themodeledtimeseries
www.ocean-sci.net/14/69/2018/ OceanSci.,14,69–86,2018
72 G.Turietal.:ResponseofO andpHtoENSOintheCalCSinahigh-resolutionclimatemodel
2
was normalized at each grid point by the interannual stan- signals over the whole North Pacific with positive (nega-
darddeviationderivedfromeachtimeseries.Specifically,a tive) SST anomalies along the Equator and in the Gulf of
monthlymeanclimatologyissubtractedforeachmonth,cre- Alaskaandnegative(positive)SSTanomaliesinthesubtrop-
atingamonthlymeantimeseriesofanomalies.TheLanczos icalgyrearound30◦N(Fig.1b,c,e,f,h,andi).Bothmodels
time filtering is then applied, which removes variability on alsosimulatetheintensification(weakening)oftheAleutian
timescaleslongerthan10years.Fromthis,seasonalaverages LowduringElNiño(LaNiña),thoughthechangesareover-
oftime-filteredanomaliesarecreatedforeachyear.Thus,the estimated and biases in orientation relative to the observed
interannual standard deviation of the seasonal anomalies is record are apparent. It should be noted that both the orien-
computed and used to standardize (i.e., normalize) the sea- tation and position of the Aleutian Low in the observations
sonaltime-filteredanomalies. and in ESM2.6 are likely affected by the limited number of
Figures 6 through 10 show El Niño minus La Niña com- eventspresentinbothtimeseriesandcouldleadtoabiasin
posites. This approach assumes linearity in the signal, i.e., theresults(Deseretal.,2017).
thattheinfluenceofElNiñois−1timestheinfluenceofLa To shed light on the vertical structure of temperature and
Niña. While there are nonlinearities both in the ENSO re- density, we compare four vertical offshore cross sections
gion (i.e., SSTs in the tropical Pacific) and in the response fromtheESM2.6andESM2MmodelstooutputfromaRe-
to the SST anomalies in the extratropics (e.g., Dommenget gional Ocean Modeling System (ROMS) reanalysis of the
et al., 2013; Hoerling et al., 1997), the signal is roughly CalCS (Fig. 2). We chose to use the ROMS reanalysis as a
linear, i.e., a significant portion of the response to El Niño comparison as it has a high spatiotemporal resolution and
events is roughly equal to and opposite of La Niña events combinesthestrengthsofindividualdatasourceswiththose
overtheNorthPacific(DeWeaverandNigam,2002).Inaddi- ofanunconstrained(i.e.,nodataassimilation)oceanmodel
tion,giventhelargeamountofvariabilityinbothElNiñoand to yield a product that is more useful than either of those
La Niña events, and the limited number of ENSO events in alone. The offshore cross sections are located at 44, 40,
observations and in ESM2.6, taking the difference between 36, and 32◦N for the 222km closest to the US west coast.
thetwogreatlyenhancesthesignal-to-noiseratio. The 1980–2010 ROMS reanalysis covers the CalCS at 0.1◦
Thenotationof0,0/1,and1,asusedinFig.5,indicates (∼10km) resolution, assimilates available satellite (SST,
theyearsduringwhichanENSOeventstartsandends.Typi- SSH)andinsitu(temperature,salinity)data(seeNeveuetal.
cally,eventsstartaroundJuneofoneyear(year0)andendin (2016)fordetails),andhasbeenusedextensivelytodescribe
springtosummerofthefollowingyear(year1),withapeak CalCSphysicaldynamics,particularlyinresponsetoENSO
during NDJ (0/1). We employ this terminology as used by variability(Jacoxetal.,2015a,b;Jacoxetal.,2016).ESM2.6
RasmussonandCarpenter(1982). and the ROMS reanalysis show similar results for the com-
positedifferencebetweenwarmandcoldENSOevents,indi-
2.3 ComparisonofmodeledandobservedENSOinthe cating a significantly improved representation of the CalCS
NorthPacific ENSO response in ESM2.6 relative to ESM2M. All three
modelsagreeonthesign ofthetemperatureanomaliesdur-
Toassessthemodel’sperformanceinrepresentingthephys- ing El Niño (although ESM2M underestimates the magni-
ical regime of the North Pacific, we compare the physical tudeoftheseanomalies),andESM2.6andROMShighlight
manifestation of ENSO in ESM2.6 to a suite of observa- thelargesttemperatureanomaliesinthenearshoreregionand
tionaldatasets.Inaddition,weexpandonthisevaluationby above 150m water depth. All three models agree that den-
comparingESM2.6toitsGFDLprecursormodel,ESM2M, sitysurfacestendtodeependuringElNiño(greenlines)and
which has a horizontal oceanic resolution of 1◦. We use a shoalduringLaNiña(bluelines),withadifferenceindepth
500-year pre-industrial ESM2M control run without tran- ofupto30mbetweenmeanElNiñoandLaNiñaconditions.
sient atmospheric CO forcing. Since we analyze normal- Additionally, we compare springtime (February–March–
2
ized values, differences in the magnitude of ENSO and its April;FMA)variabilityofmodeledversusobservedNASA-
impactontheCalCSecosystemwouldbetakenintoaccount SeaWiFS surface chlorophyll (CHL) concentrations along
betweenthetworuns,includingthepossibleinfluenceofdif- theUSwestcoast(Fig.3).Duetoitsbrevity,theCHLrecord
ferentatmosphericCO concentrations. does not lend itself well to an ENSO composite analysis,
2
AcomparisonofclimatologicalSSTfromtheHadleyCen- so we analyze CHL interannual variations via the standard
treSeaIceandSeaSurfaceTemperaturedataset(HadISST) deviation. We find that ESM2.6 represents well the strong
andsealevelpressure(SLP)fromNCEPreanalysisoverthe cross-shore gradient in CHL variability all along the coast,
NorthPacificrevealsthatbothmodelssimulatewellthecli- withhighvariabilityofupto2.5standarddeviationsnearthe
matologicalmeanSSTsignalandrangeaswellastheposi- shore and lower variability offshore. ESM2M, on the other
tion and magnitude of the North Pacific high pressure sys- hand,onlymanagestoreconstructthesamenearshorevalues
tem off the southern US west coast and the Aleutian Low in the region between 40 and 45◦N, but underestimates the
in the Gulf of Alaska (Fig. 1a, d, and g). During El Niño cross-shore gradient significantly to the north and south of
(LaNiña),bothmodelsrepresentthecompositeaverageSST
OceanSci.,14,69–86,2018 www.ocean-sci.net/14/69/2018/
G.Turietal.:ResponseofO andpHtoENSOintheCalCSinahigh-resolutionclimatemodel 73
2
◦
Figure1.February–March–April(FMA)seasurfacetemperature( C,shaded)andDJFsealevelpressure(mb,contour)intheeasternNorth
Pacific:(a,d,andg)climatology(4mbcontourinterval),(b,e,andh)ElNiño/warmhigh-passcomposite(0.5mbcontourinterval),and(c,
f,andi)LaNiña/coldhigh-passcomposite(0.5mbcontourinterval)for(a–c)observations(seasurfacetemperature:HadleyCentreSeaIce
andSeaSurfaceTemperaturedataset(HadISST);sealevelpressure:NCEPreanalysis),(d–f)ESM2.6,and(g–i)ESM2M.
this.BothmodelsoverestimatetheCHLvariabilityoffshore with normalization the biogeochemical imprint of ENSO
ataround36–40◦Nincomparisontotheobservedvariability. eventsislikelyaccentuatedrelativetoothersourcesofvari-
Finally,wecomparemodeledversusobservedNiño3.4in- ation. Furthermore, due to its high horizontal oceanic res-
dices and find that the SST variability in the tropical Pa- olution, ESM2.6 does a particularly good job of reproduc-
cific is overestimated in both models, with the maximum ingthecross-shoregradientsinSSTandCHLalongtheUS
SST variance in ESM2.6 being roughly twice as high as westcoast,withwarmtemperaturesandincreasedCHLvari-
in ESM2M and up to 5 times higher than in the observa- ability close to the coast. ESM2M’s coarse horizontal res-
tions (see Fig. S2). Both models simulate the large-scale olution, on the other hand, does not allow for a represen-
atmospheric and oceanic responses associated with ENSO. tation of finer-scale processes related to coastal upwelling,
However, ENSO events that are similar in magnitude to the and it thus fails to correctly model the cross-shore SST and
strongestonrecordarefarmorecommonandregularinthe CHLgradients.WethusfocusonESM2.6togaininsightinto
modelthantherecenthistoricalrecordsuggests.Thisoffers theregionalbiogeochemicalresponseoftheCalCStostrong
an opportunity to isolate large ENSO signals, though even ENSOevents.
www.ocean-sci.net/14/69/2018/ OceanSci.,14,69–86,2018
74 G.Turietal.:ResponseofO andpHtoENSOintheCalCSinahigh-resolutionclimatemodel
2
Figure2.ESM2.6depth–longitudecrosssectionsofFMAwarm–coldhigh-pass-filteredstandardizedanomaliesoftemperature(unitless)for
(firstcolumn)ESM2.6,(secondcolumn)ESM2M,and(thirdcolumn)ROMS.Thecontoursshowthemeanclimatologicaldensitylines(first
◦ ◦ ◦ ◦
row)44 N,(secondrow)40 N,(thirdrow)36 N,and(fourthrow)32 Nforthe222kmnearesttheUSwestcoast.Theboldlinesshow
thepositionsfortheσ =25level(ESM2.6andESM2M)andtheσ =25.5level(ROMS)fortheclimatology(black),ElNiñoconditions
(green),andLaNiñaconditions(blue),andthethinlinesindicatedensityincrementsof0.25.
3 Results events). The responses of all three variables to the differ-
ent events are highly variable. While some of the SST/SLP
3.1 IndividualrepresentationsofENSOinESM2.6in responses are similar to the signals typical during an East-
theCalCS ernPacific-typeElNiñointheobservationalrecord(seethe
Supplement), the responses to event 2 and event 6 are dras-
Figure 4 highlights the variability in SST (with overlaid tically different from each other and from observed signals.
SLP), O2, and pH between two of the six El Niño events ForSST,O2,andpH,Fig.4a–cshowwidespreadareasofal-
(event 2 and event 6; see Figs. S3–S5 for the other four El ternatingnegativeandpositiveanomalieswithnocleargradi-
NiñoeventsandFig.S5fortheindividualSST/SLPLaNiña entsbetweenthecoastandoffshore.Forevent6,ontheother
OceanSci.,14,69–86,2018 www.ocean-sci.net/14/69/2018/
G.Turietal.:ResponseofO andpHtoENSOintheCalCSinahigh-resolutionclimatemodel 75
2
Figure3.FMAinterannualstandarddeviationofsurfacechlorophyllconcentrations(standardizedbyareaaveragetoproducecomparable
scales;unitless)for(a)1998–2010NASASeaWiFSobservations,(b)ESM2.6,and(c)ESM2M.
Figure4.ESM2.6FMAhigh-pass-filteredstandardizedanomalies(unitless)fortwowarmElNiñoeventsbasedonNDJNiño3.4anoma-
lies: (a–c) event 2 and (d–f) event 6, with panels (a) and (d) displaying sea surface temperature with sea level pressure (0.2σ interval
contours),(b)and(e)surfaceO2,and(c)and(f)surfacepH.
hand,allthreevariablesexhibitstrongeranomalies,covering coastandaveragedover34–44◦N(seeFig.5hforthespec-
broaderareasoffshore(Fig.4d–f). ifiedregion).Weanalyzethesurfaceaswellas100msince
Figure 5a–f show the temporal evolution of surface and the surface is where heat transfer and chemical interactions
subsurface(100m)temperature,O ,andpHforthesixindi- withtheatmospheretakeplace,andthustheatmosphericim-
2
vidualElNiñoandsevenindividualLaNiñaevents,aswell pactofENSOisthegreatest,and100misroughlythedepth
as the composite mean, for a region within 100km of the of the heart of the thermocline. In addition to the modeled
www.ocean-sci.net/14/69/2018/ OceanSci.,14,69–86,2018
76 G.Turietal.:ResponseofO andpHtoENSOintheCalCSinahigh-resolutionclimatemodel
2
Figure5.Temporalevolutionofindividualwarm(red)andcold(blue)events(dashed)andcompositeofallevents(solid)forareaaverages
◦
between34and44 Nand100kmoffshoreoftheUSwestcoastfor(a)ESM2.6seasurfacetemperature,(b)ESM2.6surfaceO2,(c)ESM2.6
surfacepH,(d)ESM2.6100mtemperature,(e)ESM2.6100mO2,(f)ESM2.6100mpH,and(g)HadISSTseasurfacetemperature.The
timeseriesarehigh-passfilteredandseasonallyaveragedandshowtheevolutionofstandardizedanomalies(unitless)duringthepeakENSO
yearsfromJuly–August–September(JAS)beforethepeaktoJune–July–August(JJA)afterthepeak.Theblackoutlineinpanel(h)highlights
theareawithintheCalCSusedforthisanalysis.Thenotationof0/1isasusedinRasmussonandCarpenter(1982).
results, we show SST from the Hadley Center (HadISST; ing La Niña at 100m. As was the case with temperature,
Fig.5g).Atthesurface,temperatureanomaliesareonaver- the largest O signal occurs with σ between 1 and 2 during
2
age positive (negative) during El Niño (La Niña) with val- FMA. This contrasting behavior exhibited by O at the sur-
2
ues of −1σ to 3σ (−2σ to 1σ) for the individual events faceandat100misnotexhibitedbypH,whichliketempera-
(Fig.5aandd).Onaverage,thevariabilityofmodeledSST turedisplaysaconsistentsignalthroughoutthewatercolumn
(Fig. 5a) compares well with the observed SST (Fig. 5g), (Fig.5candf).
both for El Niño and La Niña, and especially for the com- AstheresultsinFigs.4and5show,thevariabilitybetween
posite mean. In the case of O , the coastal surface waters individualeventsisextremelyhigh.Duetothisfact,wefocus
2
experience lower-than-usual O conditions during El Niño ouranalysesonthecompositemeanoverallElNiñoandLa
2
(σ on average around −1), whereas O concentrations tend Niñaeventsfortheremainderofthestudy.
2
to be elevated with σ of up to 1 during La Niña (Fig. 5b).
At100m,thestrongesttemperaturesignaloccursduringthe 3.2 Responseoftemperature,O2,andpHtomean
spring (FMA) following the typical El Niño season (NDJ). ENSOsignal
Furthermore, the CalCS experiences higher-than-usual O
2
Weillustratethemeanresponseofsurfacetemperature,O ,
conditionsduringElNiñoandlowerO concentrationsdur- 2
2
andpHtoENSOinFig.6a–c.SSTanomaliesherearepos-
OceanSci.,14,69–86,2018 www.ocean-sci.net/14/69/2018/
G.Turietal.:ResponseofO andpHtoENSOintheCalCSinahigh-resolutionclimatemodel 77
2
Figure 6. ESM2.6 FMA warm–cold high-pass-filtered standardized anomalies (unitless) for (a) sea surface temperature, (b) surface
O2,(c)surfacepH,(d)100mtemperature,(e)100mO2,and(f)100mpH.
itivebetweenthecoastand∼300–500kmoffshore,aswell is more limited to within a ∼100km wide band along the
asoffshorenorthof40◦Nandsouthof28◦N,withanoma- coast. The mean response of 100m O to ENSO is drasti-
2
liesofupto2σ (Fig.6a).Offshorebetween∼28and40◦N, callydifferentthantheO responseatthesurface,withpos-
2
intheregionofthesubtropicalgyre,SSTanomaliesareneg- itive anomalies within a 100km band along the coast, indi-
ativearound1–2σ andreflectatypicalobservedElNiñosig- cating that two different processes govern the O response
2
nal. The surface O anomalies largely reflect the SST sig- to ENSO at the surface and at depth (Fig. 6e). The 100m
2
nal, albeit with a reversed sign, as would be expected from pH signal looks largely the same as the surface pH signal,
asolubility-drivenresponse.ClosetotheUSwestcoast,O with a slightly broader cross-shore region with positive pH
2
anomaliesarelowerthan−2σ andbecomelessnegativefur- anomaliesofupto2σ (Fig.6f).ThesepositivepHanomalies
ther offshore. Positive O anomalies of around 0.8–2σ are are also more widespread northward and southward along
2
foundoffshoreintheregionofthesubtropicalgyre(Fig.6b). thecoastat100mcomparedtothesurface.Furthermore,the
The surface pH signal differs from the SST and O signals, 100mpHsignalisverysimilartothe100mO signal(com-
2 2
in that positive pH anomalies, with a maximum around 2σ, pareFig.6eandf),suggestingacommonsubsurfaceprocess
are limited to a very narrow band of ∼100km along the influencingbothvariables.
coast (Fig. 6c). Between 100km and around 500km off- InFig.7,weexaminethesameoffshorecrosssectionsas
shore, El Niño minus La Niña pH anomalies are negative inFig.2andfocusontheresponseoftemperature,O ,and
2
around −0.8σ, while further offshore in the region of the pH to ENSO in the vertical plane. Again, we show El Niño
subtropical gyre, pH anomalies are predominantly positive minusLaNiñacompositeanomalies.Thedifferingresponse
again. ofO toENSOatthesurfaceandat100mthatweobserved
2
WeinvestigateENSO-drivenchangesintemperature,O , inFig.6isalsoclearlyvisibleinallfouroffshorecrosssec-
2
andpHinthesubsurface(100m)areainFig.6d–f.Asimi- tionsinFig.7b,e,h,andk.ThepH(Fig.7c,f,i,l)andO
2
larcross-shoregradientintemperatureoccursat100mcom- responsesareverysimilarbetween32and36◦N,suggesting
paredtothesurface,withanomalieslargerthan2σ insome that here they are both affected by the same processes. Be-
regionsalongthecoast(Fig.6d).However,the100msignal tween 40 and 45◦N, on the other hand, El Niño minus La
www.ocean-sci.net/14/69/2018/ OceanSci.,14,69–86,2018
78 G.Turietal.:ResponseofO andpHtoENSOintheCalCSinahigh-resolutionclimatemodel
2
Figure7.ESM2.6depth–longitudecrosssectionsofFMAwarm–coldhigh-pass-filteredstandardizedanomalies(unitless)of(firstcolumn)
◦
temperature,(secondcolumn)O2,and(thirdcolumn)pH(shaded)withmeanclimatologicaldensitylines(contour)along(firstrow)44 N,
◦ ◦ ◦
(secondrow)40 N,(thirdrow)36 N,and(fourthrow)32 Nforthe222kmnearesttheUSwestcoast.Theboldlinesshowthepositions
fortheσ =25levelfortheclimatology(black),ElNiñoconditions(green),andLaNiñaconditions(blue),andthethinlinesindicatedensity
incrementsof0.25.
NiñapHanomaliesaremostlypositivethroughoutthewater 3.3 DriversandprocessesbehindchangesinO and
2
column,indicatingthatinthisregionthereisadifferentpro- pHintheCalCS
cess at play affecting pH in the top ∼100m as opposed to
O2atthesamelatitude. In the surface ocean, O2 is affected by four main pro-
cesses: (i) changes in SST, which affect the O solubility,
2
(ii)wind-drivenvariability,whichdrivestheO gasexchange
2
across the air–sea interface, (iii) primary production, which
reduces CO and increases O in the surface ocean, and
2 2
(iv)oceancirculation,whichaffectshorizontaladvectionand
OceanSci.,14,69–86,2018 www.ocean-sci.net/14/69/2018/