Spatial and temporal variability of sea surface temperature in the Baltic Sea based on 32-years (1982–2013) of satellite data

(1)

ORIGINAL

RESEARCH

ARTICLE

Spatial

and

temporal

variability

of

sea

surface

temperature

in

the

Baltic

Sea

based

on

32-years

(1982

—2013)

of

satellite

data

§

Malgorzata

Stramska

a,b,

*

,

Jagoda

Bia

łogrodzka

a

a_Institute_of_Oceanology_of_the_Polish_Academy_of_Sciences,_Sopot,_Poland b

DepartmentofEarthSciences,SzczecinUniversity,Szczecin,Poland Received21November2014;accepted14April2015

Availableonline19May2015 —235 KEYWORDS Satellitesensing; Seasurface temperature; Physicaloceanography; BalticSea; Annualcycle; Climatechange

Summary Satellitemeasurementsprovidesynopticviewofseasurfacetemperature(SST)and canbeusedtotraceglobalandregionalclimatetrends.Inthisstudy wehaveexaminedthe multiyeartrendsand variabilityoftheBalticSeaSSTusing32-years (1982—2013)of satellite data.Ourresultsindicatethatthereisastatisticallysigni_ficanttrendofincreasingSSTinthe entireBalticSea,withvaluesrangingfrom0.03to0.068Cyear1,dependingonthelocation.SSTs averagedovertheentireBalticSeaincreaseattherateof0.058Cyear1.HighervaluesofSST trendaregenerallypresentinthesummermonths,whiletrendisnotstatisticallysignificantin thewintermonths.TheseasonalcycleofSSTintheBalticSeaischaracterizedbywell-defined winterandsummerseasons.Theaverageamplitude(16_—188C)ofthiscycleissigni_ficantlylarger thanintheNorthSeawaterslocatedatthesamelatitudesastheBalticSea.Theanalyzeddataset alsohighlightsconsiderableinterannualSSTvariability,whichiscoherentindifferentregionsof theBalticSeaandsignificantlycorrelatedwithinterannualvariabilityoftheairtemperature.SST variabilityintheBalticSeainwintercanbelinkedtotheNorthAtlanticOscillationindex. #2015 TheAuthors. Production and hosting byElsevierSp. zo.o.onbehalfof Instituteof OceanologyofthePolishAcademyofSciences.ThisisanopenaccessarticleundertheCC BY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/).

PeerreviewundertheresponsibilityofInstituteofOceanologyofthePolishAcademyofSciences.

§_Financial_support_was_provided_by_the_SatBa_łtyk_project_funded_by_the_European_Union_through_the_European_Regional_Development_Fund

(contractno.POIG.01.01.02-22-011/09entitled'TheSatelliteMonitoringoftheBalticSeaEnvironment').PartialsupportforM.S.comesalso fromthestatutoryfundsoftheInstituteofOceanologyofthePolishAcademyofSciences(IOPAN).

* Correspondingauthorat:InstituteofOceanologyofthePolishAcademyofSciences,Powsta_ńcówWarszawy55,Sopot81-712,Poland. Tel.:+48587311600;fax:+48585512130.

E-mailaddresses:[email protected],[email protected](M.Stramska).

Availableonlineatwww.sciencedirect.com

ScienceDirect

jo u rn al ho m e p age :w w w. el s ev i er. co m / lo c a te /o c e an o

http://dx.doi.org/10.1016/j.oceano.2015.04.004

0078-3234/#2015TheAuthors.ProductionandhostingbyElsevierSp.zo.o.onbehalfofInstituteofOceanologyofthePolishAcademyof Sciences.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/).

(2)

1. Introduction

TheBalticSea(BS)isasemi-enclosedbrackishsealocatedin

NorthernEurope(Fig.1).Theseaisquiteshallowandhasan

averagedepthof only 56m.The hydrography ofthe BSis

stronglyinﬂuencedbythefactthatannuallyabout480km3_of

freshwaterare addedto theBalticSeabyriverrunoffand

atmospheric input. The river runoffis about oneorder of

magnitudelargerthanthenetatmospheric ﬂux

(precipita-tion minus evaporation, Leppäranta and Myrberg, 2009).

Signiﬁcantinﬂuxoffreshwaterfromriverspromotesa

per-manenttwo-layersalinitystructureintheBalticSeawitha

persistent halocline and an estuarine-like water exchange

withtheNorthSea.WaterresidencetimeintheBalticSeais

about30years.LowsalinitywaterscomingoutoftheBaltic

Sea have signiﬁcant inﬂuence on the hydrography of the

North andNorwegianSeas (e.g.,LeppärantaandMyrberg,

2009).Mostofthetime,however,thereisonlylimitedinﬂow

ofdenseandsaltywatertotheBalticSeafromtheNorthSea

(through the Danish Straits, Skagerrak, and Kattegat)

becauseofshallowtopography.Theintensityofthisinﬂow

can sporadically, at irregular intervals of time, increase

signiﬁcantly. Suchevents, calledtheMajorInﬂows depend

onweatherpatterns,whichcontrolthesea-leveldifference

betweentheBalticandtheNorthSeas(e.g.,Gustafssonand

Andersson,2001;Omstedtetal.,2004).The MajorInﬂows

aredifﬁculttopredict,buthaveavitalinﬂuenceonthestate

oftheBalticSea(e.g.,LeppärantaandMyrberg,2009).

Because ofits geographical location, the BSregion has

somearctic characteristics with a pronounced seasonality

(Leppäranta and Myrberg, 2009). At irregular intervals of

time this region is under the inﬂuence of continental or

marineclimateforcing.Ingeneral,thewesterlywinds

dom-inate,but theatmospheric circulationhasastrongannual

cyclecomponent, withmoreintensewesterliesduringthe

autumn and winter seasons. The mean near-surface air

temperature in the Baltic Sea region is usually several

degrees higher than in other geographical regionslocated

atthesamelatitudes.Thereasonforthisisthatatmospheric

circulationandthewarmNorthAtlanticCurrentbringheatto

Western Europe (e.g., Bigg,1996).The transport of warm

watermassesintheNorthAtlantichassigniﬁcanteffectson

climateevenmuchfurthernorth,includingtheArcticOcean

(e.g.,Beszczynska-Mölleretal.,2012;Hollidayetal.,2008).

AcharacteristicfeatureoftheBalticSeahydrographyisthe

seasonalsurfacelayerwithrelativelywarmandlowsalinity

waterembeddedintheupperlayerofwater.Thisseasonal

layer is formed in the spring due to seasonally increased

freshwaterinput,iceandsnowmelt,andmoreefﬁcientsolar

heating. Optical properties of this water layer are closely

linkedtoriverdischargeofwaterwithopticallyactive

com-ponents(dissolvedandsuspendedmatter,e.g.,Babinetal.,

2003;Stramska andŚwirgoń,2014).Indeeperparts ofthe

BalticSeaapermanenthaloclinepersiststhroughoutayearat

depth of about 40—80m. Below it anoxia is common and

interruptedonlybytheMajorInﬂowsoftheNorthSeawater

(LeppärantaandMyrberg,2009).ThehydrographyoftheBaltic

Seaisfurthercomplicatedbythefactthattheseaispartially

coveredbyiceinwinters.TheextentofseaiceintheBalticSea

variesfromyeartoyear.Inaveragewinters,theice-covered

regioninMarchincludestheGulfofBothnia,GulfofFinland,

GulfofRiga,northernpartsoftheBalticProper,andshallow

coastalwaters,includingthoselocatedinthesouthernpartof

thesea.Inextremelycoldwintersalmost alloftheBScan

freeze(LeppärantaandMyrberg,2009).Moredetailsaboutthe

physical oceanography of theBS can befound ina recent

reviewpublishedbyOmstedtetal.(2014).

Scientiﬁceffortsundertakentounderstandpastand

pre-sentchangesintheclimateoftheBalticSearegionhavebeen

intensiﬁedinthepasttwodecades.Ithasbeenshownthat

during the past century the increased frequency of both

anticycloniccirculationandwesterly windshasresulted in

a warmer climate with reduced sea-ice cover. Signiﬁcant

increaseinsurfaceairtemperaturesintheBalticSearegion

hasbeenalsodocumented(e.g.,BACCAuthorTeam,2008;

Rutgerssonetal.,2014).

In the present paper we have focused on satellite sea

surface temperature (SST) data records covering 32-years

(1January,1982to31December,2013).Ourmaingoalisto

analyzelong-termSSTtrendsandtheirregionalandseasonal

variations.Inaddition,ourobjectiveistoreview

character-isticfeaturesoftheannualSSTcycleandpatternsof

inter-annualvariability.Seasurfacetemperatureisanimportant

oceanographic parameter, linked to many processes that

occurintheupperocean.For example,studieson primary

productivity,exchangeofenergywiththeatmosphere,

cli-matechange,oceanandweathermodelingandforecasting

requireinformationabouttheSST.Asaresult,SSThasbeen

classiﬁed as one of the Essential Climate Variables that

supportthe workoftheUN FrameworkConvention on

Cli-mateChange(UNFCCC)andtheIntergovernmentalPanelon

ClimateChange(IPCC,2007).TheinformationaboutSSTin

theBalticSeagainedthroughouranalysiswillcontributeto

improved understanding of the hydrography and climate

trends inthis region. Similar analysesof SSTin the Baltic

Seapublishedbeforehavebeenbasedonsigniﬁcantlyshorter

timeseries.Forexample,Dareckietal.(2008)demonstrated

goodagreementbetweeninsituandsatellitederivedSSTin

Figure1 MapshowingthegeographicallocationoftheBaltic Sea.ThelettersAandBindicatethepositionsofthetimeseries datadisplayedinFig.3,thelettersCandDshowthepositionsof datasetsinFig.4,andtheATisfortheairtemperaturedata presentedinFigs.7,9,and10.

(3)

theBalticSea.Siegeletal.(2006)haveanalyzed15-yearsof

theAdvancedVeryHighResolutionRadiometer(AVHRR)data

and described seasonal and interannual SST variability in

1990—2004. Karagali et al. (2012) carried out an analysis

ofthediurnalSSTcyclebasedonhourlysatelliteSSTs,which

areonlyavailablefrom2004onwards.Bradtkeetal.(2010)

haveperformedathoroughanalysisofspatialand

interann-ualSSTvariationsbasedon20-yearslongdataseries(1986—

2005). Longer datasets were simplynot available at that

time.However,long-termtrendsingeophysicaldatacanbe

detectedwithgoodconﬁdenceonlyifsufﬁcientlylong

obser-vations are available. For example,Beaulieu etal. (2013)

concludedthatapproximately27yearsofcontinuous

satel-lite observations are required to detect global trends in

surface chlorophyll concentrations. Therefore, we believe

thatitisworthwhiletorevisitSSTanalysisintheBalticSea

usingtherecentlyavailable32-yearslongtimeseriesdata.

2. Data

and

methods

Overtheyearssatelliteremotesensinghasestablisheditself

asanindispensabletechniqueofacquiringglobaland

regio-nalinformationaboutoceans.Ourstudyisbasedon32-years

long dataseries (years 1982—2013)known as the National

Oceanic and Atmospheric Administration (NOAA) Optimum

InterpolationSST(OISST)Version2dataset(Reynoldsetal.,

2007).Thesedata havebeen madeavailablebytheNOAA

EarthSystemResearchLaboratoryPhysicalScienceDivision

(ESRL/PSD)throughtheirWebsiteathttp://www.esrl.noaa.

gov/psd/.Theseare dailySSTrecords(onedailyvaluefor

eachpixel),withspatialresolutionof0.2580.258,basedon

the Advanced Very High Resolution Radiometer (AVHRR)

infraredsatellitemeasurements(Pathﬁnderdatain

Septem-ber1981through December2005;operational AVHRRfrom

January 2006). The ﬁnal data set has been derived using

satellite SST retrievals, SST observations from ships and

buoys,andproxySSTsgeneratedfromseaiceconcentrations.

Fulldescriptionofdataprocessingmethodsandcomparisons

betweentheNOAAOISSTSSTandinsitudatacanbefoundin

Reynolds et al. (2007) and at www.ncdc.noaa.gov/sst/ description.php. NOAA OISST SST data are currently the

longest satellite data record that can be used to study

long-termSSTvariabilityandtrends.Thisdatasethasbeen

approvedbyNOAAastheresearchqualitydataset,whichcan

beusedbyoceanographersforfurtherscientiﬁcanalyses.

TheinfraredsatelliteremotesensingSSTalgorithmscan

provide either a skin SST if they are based on radiative

transfermodelsorasubskinSSTifinsituobservationshave

been used to adjust satellite retrievals (Merchant and

LeBorgne, 2004). In the NOAA OI SST Version 2 data set

thebiascorrectionofthesatellitedatahasbeenbasedon

datafromshipsandbuoys,thereforeitcanbeinterpretedas

thebulkSSTatabout0.5mdepth(Reynoldsetal.,2007).In

ordertoapplythecorrectionforbias,thesatellitedatahave

beenclassiﬁedintodaytimeandnighttimebinsandcorrected

separatelyusinginsitudata.Then, allthedatahavebeen

reanalyzedjointlyusingtheoptimuminterpolation(OI)

pro-cedure.TheﬁnaldatarepresentthedailymeanSSTvalues.In

theIOSSTdatasettherearenomissingdata.Missingdata

havebeen interpolated usingOImethod or, iftherewas a

signiﬁcantseaicecoverage(icefractionbetween0.5and1),

theSSTshavebeencalculatedusingstatisticalrelationships

between SSTand sea ice coverage established regionally

(Reynoldsetal.,2007).MinimumSSTsforiceconcentrations

of1havebeenassumedtoequalthefreezingpointofwater

(1.88Cforseawaterand08Cforfreshwater).Inourdataset

intheBaltic Seawe have assumedthatthefreezing

tem-peratureis—0.338C,whichcorrespondstothefreezingpoint

ofseawaterwithsalinityof6PSU(LeppärantaandMyrberg,

2009,page221).Wehavecarriedoutsensitivitytests,which

indicated that our results presented in the ﬁgures and

describedin the Results section would not change

signiﬁ-cantlyifwechangethefreezingtemperatureby50%.

OurmaininterestinthispaperisintheBalticSea(Fig.1),

butmapspresentedintheResultssectionalsocoveravicinity

oftheBalticSea(partsoftheNorthSea) toshow abetter

perspectivefortheanalyzedSSTvariability.Fordiscussionof

SSTvariabilityandtoshowtimeseriesplotswehavealsoused

spatialSSTaveragesrepresentingsmallgeographicalregions.

Forthispurpose wehave subjectively deﬁned four regions

displayedinFig.2a.Theexactpositionsoftheseregionsdonot

haveanyspecialsigniﬁcance,theregionshavebeendeﬁned

mainlytoillustratetemporalpatternsofSSTvariability.The

areaofeachregionisasquareof18by18andtheupperleft

corners of the regions have following geographical

coordi-nates:region 1: 5.6258E and 57.1258N; region 2: 17.1258E

and55.8758N; region 3: 20.1258E and 59.3758N; region 4:

19.8758Eand62.8758N.Forbrevitywewilluseinthispaper

termssuchas'region1averages'.IfwerefertotheBalticSea

averages,weconsidertheseaarealocatedeastfromthe128E.

Inallcasestheregionalaverageswerecalculatedonlyfrom

pixelsrepresentingoceansurface.TheReynoldsSSTdataset

includescomplementaryinformationaboutthetotalSSTerror

derivedfromtherandomsamplingandbiaserror(Reynolds

etal.,2007). Wehaveplotted theaverage errorforyears

1982—2013inFig.2b.Ascanbenoticed,theerrorsassume

largervaluesnearthecoastsincomparisontotheopensea

region. The largest errors are associated with the Gulf of

BothniaandGulfofFinland,mostlikelybecauseoffrequent

cloudydaysandbecauseseasurfaceintheseregionsiscovered

byseaiceoverinthewinters.

ValidatingdatarecordsfromoperationalSSTsatellite

instru-mentsrequiressubstantialvolumesofhighqualityinsitudata.

Suchdata,coveringtheentiretimeperiodofsatellite

observa-tions(1982—2013)werenotavailabletous.Neverthelesswe

havecomparedtheReynoldsSSTdatawithinsitudatathatwe

wereabletoacquire.Theﬁrstdatasetincludesdailywater

temperaturedatafrombuoysandwasobtainedfromtheBaltic

Operational Oceanographic System (www.boos.org). In this

paper we present as an example data from two sensors

deployedat2-mdepth(geographicalpositionsareindicated

inFig.1abylettersAandB,respectively). Thesecond set

consistsofdatacollectedfromships,andasanexamplewe

haveplotteddatafortwostations:BY2andBY5(data

down-loaded from the Swedish Meteorological and Hydrological

Institute,SMHI,http://produkter.smhi.se/pshark).Thesedata

havelowertemporalresolution(onaverageonewater

tem-peratureproﬁlepermonth),buttheycovertheentire32-years

timeperiod(1982—2013)representedintheReynoldsOISST

data.Whencomparingtheinsituandsatellitederivedwater

temperaturesonehastorememberthatbothkindsofwater

temperaturemeasurementsaresubjecttoerrors.Errorsinin

(4)

calibrationsor instrumentmalfunctioning.Errors insatellite

datacanbeduetoerrorsinimperfectatmosphericcorrections.

Forbrevity,inthispaper,werefertoinsituSSTestimatesas

'observation'andtothedifferencesbetweensatellite-derived

andinsituSSTestimatesas'errors'.

Thedifferencesbetweeninsituandsatellite-derivedSSTs

havebeenquantiﬁedasfollows:

-theabsoluteaverageerror(AAE)

AAE¼1

N

XN

i¼1

jOiPij; (1)

-therootmeansquareerror(RMSE)

RMSE¼ 1 N1 XN i¼1 ðPiOiÞ2 " #1=2 ; (2) -thebias(B) B¼1 N XN i¼1 Pi 1 N XN i¼1 Oi¼PiOi; (3)

-themeanabsolutepercentageerror(MPE)

MPE¼1001 N XN i¼1 PiOi Oi ; (4)

where N is the number of measurements, Oi is the water

temperaturefrominsituobservationandPiisthepredicted

value(satelliteSSTdetermination).

We havealso examined meteorological data from the

NOAA-CIRES Climate Diagnostic Center NCEP/NCAR

(Na-tional Centers for Environmental Prediction and National

CenterforAtmosphericResearch)Reanalysis1.The

Reanal-ysis Project employs a state-of-the-art analysis/forecast

systemtoassimilateglobalmeteorologicaldatafrom

vari-ousavailablesourcesfrom1948tothepresent.Thesedata

have coarser spatial resolution than SST data and are

provided on the 2.582.58spatial grid.Inthis paperwe

haveincludedcomparisonsofSSTdatawithdailymean2-m

airtemperature(AT).Wehavealsoanalyzedthenetlatent

andsensibleheatﬂuxes,alongwiththenetlongwaveand

net shortwave radiation estimates, usedto calculate the

netheatﬂuxattheseasurface.Becauseofthelowspatial

resolutionoftheNCEPdataforourcomparisonsofSSTand

NCEPdatawe haveusedthe NCEPdata froma gridpoint

locatedinthemiddleoftheBalticSea(at20.6258E60.08N,

nearregion3).

ToexaminetheinﬂuenceoftheNorthAtlanticOscillation

(NAO)onSSTwehaveusedtheHurrellprincipalcomponent

(PC) based index, obtained from http://climatedataguide.

ucar.edu/guidance/hurrell-north-atlantic-oscillation-nao-index-pc-based. These indices are provided as time series

of the leading Empirical Orthogonal Function (EOF) of the

sea level pressure (SLP) anomalies over the Atlantic

sectorat208—808Nand908W—408E.AnadvantageofthePC

time seriesapproachis thatsuchindicesare more optimal

representations of the spatial patterns of NAO than the

Figure2 (a)The32-yearmeanSST,(b)totalSSTerroraveragedin32years,(c)standarddeviation(STD),and(d)coefﬁcientof variation(CV).AllestimatesarebasedonReynoldsdatafrom1982to2013.Fourstudyregionsdiscussedinthetextareindicatedby blackboxesinpanela.

(5)

station-basedindices(ﬁxedinspace).Station-basedindices

canbeaffectedbysmall-scaleandtransientmeteorological

phenomenanotreallyrelatedtotheNAO.Foramoredetailed

discussionofissuesrelatedtotheNAOindicesseeHurrelletal.

(2003)andHurrellandDeser(2009).

InadditiontoSSTandmeteorologicaldata,wehaveused

satelliteoceancolordatafromMODISAqua(2003—2013)to

showthespatialdistributionofwaterturbidityintheBaltic

Sea. The Baltic Sea waters are optically classiﬁed as

Case 2 waters due to high CDOM and suspended matter

concentrations (e.g.,Babinetal.,2003).InCase2 waters

thestandardNASAalgorithmsgenerallydonotperformwell,

but the performance of the semianalytical algorithm for

the diffuseattenuation coefﬁcient for downwelling

irradi-ance(Kd)isacceptable(Leeetal.,2005a,b).Theaverageof

absolute percentage difference between the in situ

mea-suredandthesemianalyticallyderivedKd(l)intheBalticSea

has been estimates as 14% for l=490nm and 11% for

l=443nm.Therefore,inthispaperwehaveusedthe

semi-analyticallyderivedKd(490),obtainedfromtheNASA'sOcean

ColorWeb(www.oceancolor.gsfc.nasa.gov/)astheannually

averaged Level 3 data (reprocessing version 2013). From

these annualdatawehavecalculated the11yearaverage

Kd(490)distributionintheBalticSea.Recall,thatthe

spec-tralverticalattenuationofdownwellingirradiance,Kd(l,z),

isdeﬁnedas:

Kdðl;zÞ¼d

ln½Edðl;zÞ

dz ; (5)

where Ed(l,z) is the planar downwelling irradiance, z is

thewaterdepth,and lis thelightwavelengthin vacuum

(Mobley,1994).Importantly,Kdinﬂuencestheradiative

heat-ingoftheocean(e.g.,MobleyandBoss,2012;Stramskaand

Dickey,1993;StramskaandZuzewicz,2013;Zaneveldetal.,

1981)andcanhaveasigniﬁcanteffectonSST.

Standardstatisticalmethodshavebeenusedtoexamine

thecorrelationsbetweendifferentquantities,andasimple

linear model for trends has been assumed and tested for

statisticalsigniﬁcance(Ostasiewiczetal.,2006).

3. Results

3.1. AverageSST

Thespatialdistributionofthe32-yearaveraged SSTinthe

Baltic Seais shownin Fig. 2a. Note thatthe geographical

distributionofSSTischaracterizedbyasigniﬁcantgradient

betweenthesouthernandthenorthernparts oftheBaltic

Sea,withthe32-yearaveraged SSTdecreasingfrom about

98CinthesouthernBalticto5—68CintheBothnianBay.The

averageSSTsintheNorthSeashowninFig.2aassumehigher

values than in the Baltic Sea. This can be linked to the

inﬂuenceofadvectionofwarmwaterbytheNorthAtlantic

DriftCurrent,whichisconsideredtobeanimportant

mod-eratingfactorfortheclimateinWesternEurope(e.g.,Bigg,

1996; Moron etal.,1998). In contrast,standard deviation

(STD) calculated from 32-year long time series of SST is

signiﬁcantly largerin theBaltic Seathanin theNorth Sea

(Fig.2c).Coefﬁcientofvariation(CV)calculatedastheratio

of STD and meanSST is also larger in the Baltic Sea and

assumes the largest values in the northeastern regions

(Fig.2d),wheremeanSSTsarethelowest.MostoftheSST

variabilitycharacterizedbySTDandCVdisplayedinFig.2c

anddcanbeattributedtotheannualcycle,whichwillbe

discussedlater.

InFig.3wehavecomparedtheReynoldsOISSTdatawith

insitudatacollectedat2mdepthfromtwooceanographic

moorings.Thelocationoftheﬁrstbuoy(13.878E54.888N)is

indicatedbytheletterAandthelocationofthesecondbuoy

(20.08E59.258N)isindicatedbytheletterBinFig.1.Wewere

unabletoﬁndanypublicdomaindailyinsitudatacovering

exactlythesametime periodastheReynoldsOISSTdata,

thereforetimeseriesshowninFig.3areonlyfewyearslong.

InFig. 4another comparison ispresented using datafrom

ships. These data extend over the same 32 years as the

ReynoldsOISST,buttheyhavenotbeencollectedatregular

time intervals, usually there is only one data point per

month.Whencomparing thedatainFigs.3and4 onehas

tokeepinmindtheissueofmismatchinscalesinthedata

sets.TheReynoldsOISSTdatarepresentmeandailyvalue

averagedinapixelof0.25by0.25degree.Theinsitudata

representwatertemperaturemeasuredat agivenpointin

spaceandvalidforagivenhouroftheday.Inadditionboth

data sets are subject to instrumental errors. Given these

issues,ourcomparisonpresentedinFigs.3and4hastobe

treated with caution and cannot be understood as a true

validationofsatellitedata.Suchvalidationsaredone

routi-nelybyinstitutionallyfundedprograms(NASA, NOAA,ESA)

withdetailedqualitycontrolleddatasets,andareoutofthe

scopeofthispaper.Nevertheless,datashowninFigs.3and4

indicatethatReynoldsOISSTandinsituwatertemperatures

intheBalticSeaagree reasonablywell.Additionalanalysis

(not shown here) indicated that there is no statistically

signiﬁcant trendin thedifferencesbetweenthelong-term

watertemperaturerecords showninFig. 4.Theerror

sta-tisticsfor datadisplayed inFigs.3and4issummarizedin

Table1.Forcomparison,themeandifferencebetweeninsitu

andmonthlysatellitedatapresentedinBradtkeetal.(2010)

was0.588C.

Mapsofthemedianandmaximum SSTsobservedinthe

entire32-yearlongtimeperiod(1982—2013)aredisplayedin

Fig.5.ThemedianSSTsaregenerallylowerandthemaximum

SSTsarehigherintheBalticSeabasinthanthoseobservedin

the North Sea near the entrance to the Baltic Sea. This

suggestsastronger impactofthelocalweatherforcingon

the Baltic Sea SST in comparison to the nearby oceanic

regions,inﬂuenced by advection of water with the North

AtlanticCurrent.IntheBalticSea,thehighest32-yearSSTs

areobservedinthesouthernBalticandintheshallowwaters

oftheGulfofRigaandtheGulfofFinland,withthemaximum

temperaturesexceeding248C(Fig.5b).Generally,the

tem-peratures decrease from the south to the north and the

lowest mean, median and maximum temperatures are

observedinthenorthernregions.

3.2. Seasonalcycle

Variabilityof SST in the Baltic Sea is characterized by an

outstandingseasonal cycle. This isillustrated in Fig. 6.In

Fig. 6a and b we have plotted spatial distribution of the

32-year averaged annual minimum and maximum SST,

respectively.Notethatthecolorscaleisdifferentforevery

(6)

Fig.6canddindicateswhentheminimumandthemaximum

SSTs are observed (year day). The highest minimum and

maximumSSTvalues(2—38Cand188C, respectively)occur

inthesouthernpartofthesea.TheminimumSSTvaluesof

the32-yearaveraged annualcycle areusually observedin

February—April (Fig. 6c), while the maximum values are

generallypresent in July—August(Fig. 6d). In Fig. 6e,the

amplitudeoftheannualcycleispresented.Thisamplitude

hasbeencalculatedasthedifferencebetweenthemaximum

andminimumSSTvaluesplottedinFig.6aandb.Themost

strikingobservationisthattheamplitudeoftheSSTannual

cycleintheBalticSeaisconsiderablylargerthanintheNorth

Sea.ThisisduetosigniﬁcantlylowerannualminimumSSTand

higherannualmaximumSSTintheBalticSeaincomparisonto

theNorthSea(Fig.6aandb).Itisalsostrikingthatevenifthe

maximumandminimumtemperaturescoverabroadrangeof

variability (depending on the geographical location), the

annualSSTamplitudedoesnotchangemuchovertheentire

BalticSea.AscanbeseeninFig.6ethegreatestamplitudeof

the seasonalSSTcycle (morethan178C) occurs inshallow

bays such as the Gulf of Riga and the Gulf of Finland.

Comparison with Fig. 6f, where thespatial distribution of

theverticalattenuationcoefﬁcientfordownwelling

irradi-ance,Kd(490),ispresented,leadsustotheconclusionthat

largeramplitudeofSST canbeassociatedwithregionswhere

watersaremoreturbid.Note,thathigherconcentrationsof

optically active water components, responsible for larger

values of Kd(490), are often associated with considerable

runofffromrivers.Therefore,these aretheregionswhere

waterscanbemorestratiﬁedduetosigniﬁcantamountsof

lowsalinitywatersspreadingonthesurface,andsuch

stra-tiﬁcationlikelyalsoinducesmorepronouncedannualcycle

ofSST.

Tobetterillustratethedevelopmentoftheseasonalcycle

wehaveextractedSSTtimeseriesinthefourlocationsshown

inFig.2a.Comparisonofthe32-yearaveragedannualcycle

ofSST atthesefourlocationsisdisplayedinFig.7a.Fig.7ain

Figure3 ComparisonofReynoldsSSTdatawithwatertemperaturesmeasuredat2mdepthfrombuoysdeployedintheArconabasin (13.878E54.898N,datashowninpanels(a)and(c))andinthenorthernpartoftheBalticProper(21.08E59.258N,datashowninpanels (b)and(d)).ThegeographicalpositionsofthebuoysareindicatedinFig.1bylettersAandB.Thesolidlinesinpanelscanddindicate theY=Xline(seeTable1forerrorstatistics).

Table1 Estimatesofthebias(B),absoluteaverageerror (AAE),meanabsolutepercentageerror(MPE),androotmean squareerror(RMSE)obtainedforcomparisonofReynoldsSST withinsituwatertemperaturemeasuredatsitesA,B,C,and D(seeFig.1).Nisthenumberofobservations.Therelevant datasetsareshowninFigs.3and4.

Sites A B C D B[_8C] _0.05 0.38 0.02 0.09 AAE[8C] 0.42 0.56 0.83 0.8 MPE[%] 7.7 7.4 13.1 10.8 RMSE[8C] 0.57 0.71 1.05 1.05 N 3000 1034 324 309

(7)

additiontoFig.6highlightsasigniﬁcantdifferenceintheSST

cycle in the North Sea (location 1) and in the Baltic Sea

(locations2,3,and4).TheNorthSealocationis

character-izedbysmalleramplitudeoftheannualSSTcyclethanthe

three Baltic Sea locations. This can be at least partly

explainedbythefactthatSSTintheNorthSeaisinﬂuenced

bywateradvectionwithwarmcurrentwhiletheBalticSeais

surroundedbyland.Asaresultoceanicwaterinﬂowintothe

BalticSeaisrestrictedbytopographyandthevariabilityof

SST is mainly due to local air-sea exchange processes. In

addition,seasurfaceintheBalticSeaisexposedrelatively

more often to air masses advected from land than in the

North Sea. Continental air masses typically have a more

pronouncedtemperaturevariabilitycomparedtotheoceanic

airmasses,andthiscanbereﬂectedintheBalticSeaSST

variability.Moreover,theBalticSeaisstronglyinﬂuencedby

thefreshwaterinﬂowfromrivers.Thisinﬂowincreaseswater

columnstabilityandwaterturbidity(StramskaandŚwirgoń,

2014).Increasedwaterturbidityinturnlikelycausesmore

efﬁcientheatingofthesurfacelayerofwater(e.g.,Stramska

andDickey,1993).

Comparisonof SSTand air temperature (AT)data

pre-sented in Fig. 7a shows that the seasonal cycle of AT is

leadingtheseasonal cycleofSSTinwinterwiththe

mini-mum AT occurring in the end of January/beginning of

February. Theminimum AT is on average lower than the

minimumSSTbyabout58C,whilethemaximumSSTandAT

havesimilarvalues and bothare featuredin July/August

(day210—225).

Basedon the32-yeartime series of dailySSTdata,we

havealsodocumented(Fig.7b)acharacteristicannualcycle

oftheSSTstandarddeviation(STD)inourfourgeographical

locations.Fig.7bshowsthatinallfourlocationsthedailySST

standarddeviationsassumehighervalues(1.5—28C)inthe

summerandlowervalues(1.38Candless)inthewinter.The

dailySSTstandarddeviationsinregion1(NorthSea)haveon

averagelowervaluesincomparisontotheBalticSearegions,

whichagainisconsistentwiththefactthatregion1ismore

inﬂuenced by the advection of the oceanic water andair

massesthantheBalticSea.IncomparisontoSST,the

stan-dard deviation of the air temperature (Fig. 7d) has an

oppositeannualcyclewiththelargestSTDvaluesobserved

inthewinter(68C)andsmallerSTD valuesin thesummer.

SuchadifferenceintheSTDpatternsofSSTandATcanbe

explainedbythefactthatduringwinterssurfacewatersin

theBalticSeaundergoefﬁcientwind-inducedandconvective

mixing. Therefore cooling of the water involves large

volumesofwater(notonlysurfacelayers)andthisdecreases

thetemporalvariabilityofSST.InFig.7cspatialpatternsin

the standard deviationof SSTare shown after theannual

cycle has been subtracted at each pixel. This caused a

substantialdecreaseofSTDincomparisontoSTDshownin

Figure4 ComparisonofReynoldsSSTdatawithsurfacewatertemperaturesmeasuredat0.5mdepthfromshipsatstationsBY2 (14.18E55.08N,datashowninpanelsaandc)andBY5(15.998E55.258N,datashowninpanelsbandd).Thegeographicalpositionsof thestationsareindicatedinFig.1bylettersCandD.ThesolidlinesinpanelsaandbindicatetheY=Xline(seeTable1forerror statistics).

(8)

Fig. 2c, where annual variability was included. As can be

seen,STDvaluesinFig.7caregreaterinthesouthernpartof

the Baltic Sea than in the northern part. This can be

explainedbythefactthatwaterinthenorthispartlycovered

byseaiceinthewinter.Inaddition,largerSTDvaluescanbe

associatedwithregionsmoreinﬂuencedbyvariabilityinthe

watercirculation patterns andadvection (Omstedtetal.,

2014).

3.3. Longtermtrendsandinterannualvariability

TimeseriesofannualSSTdatahavebeenusedtoestimate

regionalSSTtrendspresentedinFig.8(statisticallysigni

ﬁ-cantat95%conﬁdencelevel,p<0.05).Trendsofincreasing

annual mean SST have been detected everywhere in the

Baltic Seaand in the North Sea nearthe entrance to the

Baltic Sea. These trends in the Baltic Sea are generally

greater than 0.048Cyear1 and sometimes assume values

ofabout0.068Cyear1,forexampleintheGulfofFinland.

OuranalysisindicatesthatSSTaveragedovertheentireBaltic

Sea increases at the rate of 0.058Cyear1 (or 0.58C per

decade).Similar SSTtrendhasbeen estimatedbyBradtke

etal.(2010),whousedadifferentdataset.Forcomparison,

theglobally averaged SSTtrendcalculated by Goodetal.

(2007) using 20 years of Advanced Very High Resolution

Radiometer Pathﬁnder data (January 1985 to December

2004)hasbeenestimatedas0.188Cand0.178Cperdecade

from daytime and nighttime data,respectively. Thus, the

warmingtrendestimatedbyusfortheBalticSeaisgreater

thantheseglobaltrends.TrendshigherthantheglobalSST

trend werereported before for other coastal regions, for

exampleintheArctic(e.g.,Nghiemetal.,2014;Peterson

et al., 2002).They have been explained by an increasing

water runoff from land, which intensiﬁes water column

stratiﬁcationin theocean (e.g.,Morison etal.,2012)and

suppliesmoreterrigenousmaterial,atthesametime

ampli-fying surface water heating (Fichot et al., 2013). Similar

effectscanbeexpectedtoplayasigniﬁcantroleintheBaltic

Sea.

WehaveobservedlargerangeinSSTvariabilityduetoa

relativelystrongannualcycleintheBalticSea.Theannual

mean SSTs can be signiﬁcantly affected by extreme SST

values. Therefore we have also used annual SST data to

estimate trends for medians, 25th and 75th percentiles

(Fig.8b,c,andd).Notethattheseestimatesarecompletely

independentonourassumptionsaboutthetemperaturesin

the sea ice covered pixels. Our results indicate that the

annualmedianSSTvalueshavebeenincreasingstronger in

thesouthern Balticthan inthe BotnicSea,buttrendsare

statisticallysigniﬁcantintheentireBaltic.The75th

percen-tilesintheannualSSTdatahavebeenincreasingfasterthan

themeanandmedianSST(notethatthescaleinFig.8cis

differentthaninFig.8aandb).Similartrendsforthe25th

percentiles were low and statistically not signiﬁcant (not

shown).ThisindicatesthatthemeanSSTintheBalticSea

increasesmainlyduetochangesoccurringinthewarmseason

of the year. This conclusion is also supported by Fig. 8d,

where trends based on monthly SST data have been

pre-sented. Fig. 8d shows that the multiyear SST trends are

strongestinthesummermonths(July—September).Bradtke

etal.(2010)alsoreportedlargerSSTtrendsinthesummer.

Interestingly, Bradtke et al.(2010) observed negative SST

trendsinthewinter,butsuchnegativetrendshavebeennot

conﬁrmedbyouranalysis.Inadditionwehavecheckedthat

trendsformaximumannualvalueswerestatistically

signiﬁ-cant,buttrendsfortheamplitudeoftheannualcycle(taken

asthedifferencebetweenmaximumandminimumSST)were

notsigniﬁcant,probablybecausetherewasnotrendinthe

minimumSSTvalues.

InadditiontomultiyeartrendstheSSTdataallowedusto

analyze patternsininterannualvariability.As anexample,

time series of annual SSTanomalies averaged in the four

study regions are shown in Fig. 9a. The annual anomalies

weredeﬁnedasthedifferencebetweentheannually

aver-agedSSTminusthe32-yearaverageSSTcalculatedateach

pixel.Next,theanomaliesforallpixelswithineachregion

were averaged to obtain the regional anomalies shown in

Fig. 9a. For comparison, annual ATanomalies at 20.6258E

60.08Narealsoincludedinthisﬁgure.Thedatadisplayedin

Fig.9aindicatethatthelowestannualSSTvaluesinallfour

regionswereobservedin1987.PatternsofinterannualSST

variabilityarecoherentindifferentregionsoftheBalticSea

and in the North Sea in region 1. There is also a good

Figure5 (a)Medianand(b)maximumSSTvaluesinthetime periodbetweenJanuary1,1982andDecember31,2013.Note thatthecolorscaleisdifferentforeachplot.(Figureincoloris providedinthewebversionofthisarticle.)

(9)

correspondence between the 2-m annual air temperature

anomaliesandSSTanomalies.To illustratetherelationship

betweentheATandSSTanomalies,ascatterplotincluding

data from region 3 is shown in Fig. 9b, as an example.

Correlationcoefﬁcient(R)forthesedatais0.88.Analogous

correlationcoefﬁcientsforAT andSST anomaliesinregions1,

2and4(scatterplotsnotshown),are0.69,0.88,and0.89,

respectively.Similarcorrelationwaslowerbutalso

statisti-callysigniﬁcantwhenwecomparedtheanomaliesofSSTand

thenetshortwaveradiationﬂux(R=0.61,datanotshown).

CorrelationbetweenSST andthenetheatﬂuxanomalieswas

notstatisticallysigniﬁcant(R=0.16,datanotshown).

TheNorthAtlanticOscillation(NAO)index(e.g.,Hurrell

and Deser, 2009; Hurrell et al., 2003) is often linked to

climate variability in Europe. Positive NAO index values

areassociatedwithlow-pressureanomaliesovertheIceland

andanomalouslyhigh pressure inthe subtropical Atlantic,

and indicate stronger than average westerlies. The NAO

variabilityhasbeenlinkedforexampletotheprecipitation

(Hurrell,1995)andseasurfacetemperaturepatternsinthe

northeastAtlantic(PlanqueandTaylor,1998),andhasbeen

alsoshowntoinﬂuencethevariabilityintheBalticSea(BACC

AuthorTeam,2008;Hänninenetal.,2000,2003;Kaukerand Meier, 2003). The atmospheric circulation can affect SST

directlythrough processesof air-sea energy exchange and

indirectly byforcing speciﬁc patterns of watercirculation

andadvection. Therefore we haveincluded in Fig. 10the

winteranomalies ofSST (inregion 3)and AT plottedas a

functionofthewinter(DJF)Hurrellprincipalcomponent(PC)

based North Atlantic Oscillation (NAO) index (Hurrell and

Deser,2009;Hurrelletal.,2003).Theresultswere

compar-ableifweplottedinasimilarwaySSTdatafromotherregions

(resultsnotshown).ThecorrelationsbetweenNAOandSSTas

wellasNAOandATinwinterarestatisticallysigniﬁcant,but

similar correlations are not statistically signiﬁcant for the

annualdata(notshown).Thecorrelationscanbeexplained

bythefactthatapositiveNAOindex inwinters is

accom-paniedbystrongwesterlywindsandadvectionofhumidand

warm airmasses from theAtlantic, while a negative NAO

indexisassociatedwiththenortherlywindsadvectingcold

Figure6 Summaryofthe32-yearaveragedannualcycleofSST:(a)minimum,(b)maximumSSTvalues,(c)dayoftheyearwhenthe minimumSSTisobserved,(d)dayoftheyearwhenthemaximumSSTisobserved,(e)amplitudeoftheannualSSTcycle,and(f)the verticalattenuationcoef_ﬁcientKd(490)averagedin11years(2003—2013)basedontheMODISAquadata.(Figureincolorisprovidedin

(10)

arctic air masses. These continental and marine climate

inﬂuencesaremostlikelyreﬂectedinSSTanomalies.

4. Summary

and

conclusions

Ourmaingoalinthispaperwastocharacterize multiyear

SST variability and trends in the Baltic Sea. For this

purpose we have used the NOAA Optimum Interpolation

SST (OISST)Version2 data (Reynoldset al.,2007),which

providelong-term,consistentsynopticviewofthethermal

state of the ocean surface and can be used to assess

regional climate trends and variability. The analysis

of SST variability in the Baltic Sea can lead to a better

understanding of this marine environment, including

Figure8 Trendsintheannual(a)means,(b)mediansand(c)75thpercentilesofSSTinyears1982—2013.Trendsarestatistically signi_ﬁcant(p_<0.05,95%con_ﬁdencelevel).(d)TrendsinthemonthlySSTdataaveragedatthefourstudysitesdisplayedinFig.2a. Figure7 (a)The32-yearaveragedannualcycleofSSTinthefourstudyregions.Thegeographicalpositionofeachregionisindicated inFig.2a.Forcomparison,the32-yearaveragedannualcycleoftheairtemperature(AT)at20.6258E60.08Nisalsoplotted.(b)Annual patternsinthestandarddeviationscalculatedfromdailySST datainthefourstudysites.(c)Spatialdistributionofthestandard deviationscalculatedfromdailySST dataaftersubtractingtheaverageannualSSTcycleateachpixel.(d)Annualpatterninthe standarddeviationoftheairtemperature(AT)at20.625_8E60.0_8N.

(11)

climate related hydrographic and ecosystem trends and dependencies.

Our most important ﬁndings can be summarized as

follows:

The 32-year averaged annual mean and minimum SSTs

intheBalticSea(5to98Cand0.3to38C,respectively)

aresigniﬁcantlylowerthanintheNorthSea(10to118C

and 88C, respectively) near the entrance to the Baltic

Sea.

AnnualmaximumSSTsarehigherintheBalticSeathanin

theNorthSea.Onaverage,themaximum SSTsofabout

198CinthesouthernBalticareobservedinJuly—August,

butinsomeyearsSSTscanexceed248C.Incomparison

32-yearaveragedmaximumannualtemperatureintheNorth

Seaatsimilarlatitudesisabout168C.

Therearelargeregionaldifferencesbetweentheannual

mean,median,maximum,andminimumSSTsinthe

south-ernandnorthernregionsoftheBalticSea.Theycanbe

explainedbythelatitudinalextentoftheBalticSea(from

about538Nto668N).

In spite of this large range of regional SST values the

average amplitudes of the seasonal cycle have similar

value almost everywhere in the Baltic Sea (16—178C).

Theamplitudehasslightlyhighervaluesinshallow bays

and coastal regions, where bathymetry prevents water

mixingandcirculation.Opticalpropertiesofmoreturbid

coastalwatersarealsolikelyanimportantfactorcausing

moreefﬁcientheatingofsurfacewaters.Thiscanbeeven

morereinforcedifwatersarestronglystratiﬁedduetothe

freshwaterrunoff.

AmplitudesoftheannualSSTcyclearesigniﬁcantlylarger

in the Baltic Sea than in the North Sea at the same

latitudes(10—128C).

OverallSSTvariabilityreﬂectedinstandarddeviationsand

coefﬁcientsofvariationisgreaterfortheBalticSeaSSTs

(5.5—68Cand 60—90%,respectively) than for the North

SeaSSTs(48Cand40%,respectively).

Patterns in interannual SST variability are coherent in

differentregionsoftheBaltic SeaandintheNorth Sea

neartheBalticSea.

TheinterannualSST anomaliesaresigniﬁcantlycorrelated

with 2-mairtemperatureanomalies andnetshortwave

radiation.

InterannualSSTvariabilityintheBalticSeainthewinteris

signiﬁcantlycorrelatedwiththeNorthAtlanticOscillation

index.

Wehavedocumentedstatisticallysigniﬁcant32-yearlong

trends (p<0.05, 95% conﬁdence level) of increasing

annual mean, median and 75th percentiles SST. Trend

fortheBalticSeaaveragedannualmeanSSTisestimated

as0.058Cyear1,butdependingontheexactgeographical

positionitsvaluesrangefrom0.03to0.068Cyear1.

Trendsindicatestrongerwarminginthesummermonths.

Trendsarenotstatisticallysigniﬁcantforannualminimum

and25thpercentilesSST.

It is worth noting that our analysis indicates that SST

warmingtrendintheBalticSeaisgreaterthantheglobally

averaged SSTtrend. This ﬁnding is consistent with earlier

resultspublished byBradtkeetal.(2010) andSiegel etal.

(2006).Similartoourﬁndings,Bradtkeetal.(2010)indicated

Figure10 Winteranomaliesof(a)SSTinregion3and(b)air temperature at20.6258E60.08N,plottedasafunctionof the Hurrelwinter(DJF) principalcomponent(PC)basedNorth At-lanticOscillation(NAO)index.

Figure9 (a)TimeseriesoftheannualSSTanomaliesaveraged inthefourstudyregions.Forcomparison,theannualair tem-perature anomalies at 20.6258E 60.08N are also shown. (b) Annual anomalies of SSTin region 3 plotted asa functionof ATanomaliesat20.6258E60.08N.

(12)

thatwarming isstronger inthe summer.However,Bradtke etal.(2010)alsopostulatedanegativeSSTtrendinthewinter

months.Thisnegativetrend hasnotbeenconﬁrmed inour

analysis.Theobservedlargedifferencesbetweenthe

variabil-ityofSSTintheNorthandtheBalticSeascanbeexplainedby

strongerimpactsoftheoceanicwaterandoceanicair

advec-tionontheNorthSeaSSTthanontheBalticSeaSST.Becauseof

thelimitedinﬂowofoceanicwater,SSTintheBalticSeais

mostlyundertheinﬂuence oflocalair-seainteraction

pro-cesses.In addition,atmosphericcirculation bringsat times

continentalairmassesovertheBalticSea,andthis

intermit-tentinﬂuenceofoceanic/continentalairmassesincreasesthe

variabilityofSSTinthissea.Alternatingbetweencontinental

andmarineinﬂuencesistypicalfortheBalticSearegion.Thisis

alsoreﬂectedinthevariabilityoftheNorthAtlanticOscillation

Index.PositiveNAOindexinwintersisassociatedwithstrong

westerlywindsandadvectionofhumidandrelativelywarmair

massesfromtheocean.NegativeNAOindexisassociatedwith

coldwintersduetothenortherlywindsadvectingcoldarctic

airmasses.Ourresultsareinagreementwiththenotionthat

BalticSeaclimate systemmemory is characterizedby two

importanttimescales ofvariability (OmstedtandHansson,

2006a,b).Theﬁrstscaleisassociatedwiththewaterbalance

andsalinityandhasane-foldingtimeofabout30years,while

thesecondscaleisassociatedwiththeheatbalanceandhasan

e-foldingtimeofapproximatelyoneyear.Thismeansthaton

the annual time scale the Baltic Seais almost in thermal

balancewiththeatmosphere.Thisisreﬂectedinthe

variabili-tyofseaiceextentandwatertemperatures.Incontrastwecan

expectthatitwilltakelongerbeforewecandetectchangesin

parameters related to the water balance, such as water

salinity.

IntheBalticSea,thecombinedsystemofrivers,estuaries,

andoceanicwaterinﬂows,createsacomplexhydrodynamic

environment, governed by variable atmospheric forcing

(wind, heat ﬂuxes, seaice processes), irregular river

dis-charge,andvariablebathymetry. Theseaissurroundedby

ninecountries, andthe quality oflife of about 85million

people is affected by the environmental problems in the

BalticSea.Oneoftheimportant aspects oftheBalticSea

researchisthegoalofremediatingtheeutrophicationofthis

marineenvironment(HELCOM,2009).Thisrequiresabetter

understandinghowthissystemreactstoglobalclimateand

humaninduced changes. Progressin oceanographic

under-standingoftheBaltic Seahasbeenhamperedbyalimited

numberoffreeaccessdatabaseswithsystematic,large-scale

in situ measurements. We believe that in this light, the

signiﬁcance ofinformation gainedfrom long-termsatellite

observationscannotbeoverestimated.

Acknowledgments

Authorswouldliketothankallthepeoplewhowereinvolved

intheprogrammesprovidingfreeaccesstothedatasetsused

inthis study. The ocean colour datawere madeavailable

throughtheNationalAeronauticsandSpaceAdministration

(NASA) OceanColor Web (oceancolor.gsfc.nasa.gov/). The

NationalOceanic and Atmospheric(NOAA) Optimum

Inter-polation SST (OISST) Version 2 data set has been made

available by the NOAA Earth System Research Laboratory

Physical Science Division (ESRL/PSD). The meteorological

data were from the National Centers for Environmental

Prediction and National Center for Atmospheric Research

(NCEP/NCAR) Reanalysis Project distributed at http://

www.esrl.noaa.gov/psd/.TheNAOindexdatawereobtained

from the The National Center for Atmospheric Research

(NCAR) (http://climatedataguide.ucar.edu/guidance/

hurrell-north-atlantic-oscillation-nao-index-pc-based). The

Baltic Operational Oceanographic System (www.boos.org)

and the Swedish Meteorological andHydrological Institute

(SMHI, http://produkter.smhi.se/pshark) provided water

temperaturedatafrominsitumeasurements.Wealsothank

Sebastian Meler from the Institute of Oceanology of the

PolishAcademyofSciences(IOPAN),whohelpedwithdata

extraction.

References

Babin,M., Stramski, D.,Ferrari, G.M.,Claustre, H., Bricaud,A., Obolensky,G., etal., 2003.Variationsinthelightabsorption coefﬁcientsofphytoplankton,non-algalparticles,anddissolved organicmatterincoastalwatersaroundEurope.J.Geophys.Res. 108,3211,http://dx.doi.org/10.1029/2001JC000882.

BACC AuthorTeam, 2008.Assessment ofClimate Changefor the BalticSeaBasin.Springer,473.

Beaulieu,C.,Henson,S.A.,Sarmiento,J.L.,Dunne,J.P.,Doney,S.C., Rykaczewski,R.R.,Bopp,L.,2013.Factorschallengingourability todetectlong-termtrendsinoceanchlorophyll.Biogeosciences 10(4),2711_—2724,http://dx.doi.org/10.5194/bg-10-2711-2013. Beszczynska-Möller,A.,Fahrbach,E.,Schauer,U.,Hansen,E.,2012. VariabilityinAtlanticwatertemperatureandtransportatthe entrance to the Arctic Ocean 1997—2010. ICES J. Mar. Sci.,

http://dx.doi.org/10.1093/icesjms/fss056.

Bigg, G.R.,1996. TheOceansand Climate.CambridgeUniversity Press,UK,266.

Bradtke,K.,Herman,A.,Urbański,J.A.,2010.Spatialand interan-nualvariationsofseasonalseasurfacetemperaturepatternsin theBalticSea.Oceanologia52(3),345—362,http://dx.doi.org/ 10.5697/oc.52-3.345.

Darecki,M.,Ficek,D.,Kr_ężel,A.,Ostrowska,M.,Majchrowski,R., Wo_źniak,S.B.,Bradtke,K.,Dera,J.,Wo_źniak,B., 2008. Algo-rithmsfortheremotesensingoftheBalticecosystem (DESAM-BEM).Part2:Empiricalvalidation.Oceanologia50(4),509_—538.

Fichot, C.G., Kaiser,K., Hooker,S.B.,Amon, R.M.W.,Babin, M., Bélanger,S.,Walker,S.A.,Benner,R.,2013.Pan-Arctic distribu-tionsofcontinentalrunoffintheArcticOcean.Sci.Rep.3,1053,

http://dx.doi.org/10.1038/srep01053.

Good,S.A.,Corlett,G.K.,Remedios,J.J.,Noyes,E.J., Llewellyn-Jones,D.T.,2007.TheglobaltrendinSeaSurfaceTemperature from20yearsofAdvancedVeryHighResolutionRadiometerdata. J.Clim.20,1255—1264,http://dx.doi.org/10.1175/JCLI4049.1. Gustafsson,B.G.,Andersson,H.C.,2001.Modelingtheexchangeof theBalticSeafromthemeridionalatmosphericpressure differ-enceacrosstheNorthSea.J.Geophys.Res.106,19731—19744,

http://dx.doi.org/10.1029/2000JC000593.

Hänninen,J.,Vuorinen,I.,Hjelt,P.,2000.Climaticfactorsinthe Atlanticcontroltheoceanographicandecologicalchangesinthe BalticSea.Limnol.Oceanogr.45(3),703_—710.

Hänninen, J., Vuorinen,I., Kornilovs,G., 2003. Atlanticclimatic factorscontroldecadaldynamicsofaBalticSeacopepod,Temora longicornis.Ecography26,672_—678.

HELCOM, 2009.Eutrophication inthe Baltic Sea _—anintegrated thematicassessmentoftheeffectsofnutrientenrichmentand eutrophicationintheBalticSearegion.In:BalticSea Environ-mentProceedings115B.p.148.

Holliday, N.P., Hughes, S.L., Bacon, S., Beszczynska-Möller, A., Hansen,B.,Lavin,A.,Loeng,H., etal.,2008.Reversalofthe

(13)

1960sto1990sfresheningtrendinthenortheastNorthAtlantic andNordicSeas.Geophys.Res.Lett.35,L03614,http://dx.doi. org/10.1029/2007GL032675.

Hurrell,J.W.,1995.DecadaltrendsinthenorthAtlanticoscillation: regionaltemperatureandprecipitation.Science269,676—679.

Hurrell,J.W.,Deser,C.,2009.NorthAtlanticclimatevariability:the roleoftheNorthAtlanticOscillation.J.Mar.Syst.78,28_—41,

http://dx.doi.org/10.1016/j.jmarsys.2008.11.026.

Hurrell,J.W.,Kushnir,Y.,Ottersen,G.,Visbeck,M.(Eds.),2003.The NorthAtlanticOscillation:climatesigni_ﬁcanceand environmen-talimpact,Geophys.Monogr.Ser.,vol.134.279,http://dx.doi. org/10.1029/GM134.

IPCC,2007.In:Solomon,S.,Qin,D.,Manning,M.(Eds.),Climate Change2007:ThePhysicalScienceBasis.ContributionofWorking Group1totheFourthAssessmentReportofthe Intergovernmen-talPanelonClimateChange.IntergovernmentalPanelonClimate Change,Cambridge/NewYork.

Karagali,I.,Høyer,J.,Hasager,C.B.,2012.SSTdiurnalvariabilityin theNorthSea andtheBalticSea.RemoteSens.Environ.121, 159—170,http://dx.doi.org/10.1016/j.rse.2012.01.016. Kauker,F.,Meier,H.E.M.,2003.Modelingdecadalvariabilityofthe

BalticSea:1.Reconstructingatmosphericsurfacedataforthe period1902—1998.J.Geophys.Res.108,3267,http://dx.doi. org/10.1029/2003JC001797.

Lee,Z.-P.,Darecki,M.,Carder,K.L.,Davis,C.O.,Stramski,D.,Rhea, W.J.,2005a.Diffuseattenuationcoef_ﬁcientofdownwelling irra-diance:anevaluationofremotesensingmethods.J.Geophys. Res.110,C02017,http://dx.doi.org/10.1029/2004JC002573. Lee, Z.-P., Du, K.P., Arnone,R.,2005b. A modelfor thediffuse

attenuationcoef_ﬁcientofdownwellingirradiance.J.Geophys. Res.110,C02016,http://dx.doi.org/10.1029/2004JC002275. Leppäranta,M., Myrberg,K.,2009.PhysicalOceanographyofthe

BalticSea.Springer-PraxisBookSeriesinGeophysicalSciences. Springer,Chichester,UK,378.

Merchant,C.J.,LeBorgne,P.,2004.Retrievalofseasurface temper-aturefromspace,basedonmodelingofinfraredradiative trans-fer:capabilitiesandlimitations.J.Atmos.Ocean.Technol.21, 1734—1746.

Mobley,C.D.,1994.LightandWater.RadiativeTransferinNatural Waters.AcademicPress,NewYork.

Mobley,C.D.,Boss,E.,2012.Improvedirradiancesforuseinocean heating,primaryproduction,andphoto-oxidationcalculations. Appl.Opt.51,6549_—6560.

Morison,J.,Kwok,R.,Peralta-Ferriz,C.,Alkire,M.,Rigor,I.,Andersen, R.,Steele,M.,2012.ChangingArcticOceanfreshwaterpathways. Nature481,66_—70,http://dx.doi.org/10.1038/nature10705. Moron,V.,Vautard, R.,Ghil,M.,1998. Trends,inderdecadaland

interannual oscillations in global sea surface temperatures. Clim.Dyn.14,545_—569.

Nghiem, S.V., Hall, D.K.,Rigor, I.G., Li, P.,Neumann, G.,2014. Effects of Mackenzie River discharge and bathymetry on sea ice in the Beaufort Sea. Geophys. Res. Lett. 41, 873—879,

http://dx.doi.org/10.1002/2013GL058956.

Omstedt,A.,Elken,J.,Lehmann,A.,Leppäranta,M.,Meier,H.E.M., Myrberg,K.,Rutgersson,A.,2014.Progressinphysical oceanog-raphy of the Baltic Sea during the 2003—2014 period. Prog. Oceanogr. 128, 139—171, http://dx.doi.org/10.1016/j. pocean.2014.08.010.

Omstedt,A.,Elken,J.,Lehmann,A.,Piechura,J.,2004.Knowledge oftheBalticSeaphysicsgainedduringtheBALTEXandrelated programmes.Prog.Oceanogr.63,1_—28.

Omstedt, A., Hansson, D., 2006a. The Baltic Sea ocean climate systemmemoryandresponsetochangesinthewaterandheat balancecomponents.Cont.ShelfRes.26,236_—251,http://dx. doi.org/10.1016/j.csr.2005.11.003.

Omstedt,A.,Hansson,D.,2006b.Erratumto:'TheBalticSeaocean climatesystemmemoryandresponsetochangesinthewaterand heat balance components'. Cont. Shelf Res. 26, 1685—1687,

http://dx.doi.org/10.1016/j.csr.2006.05.011.

Ostasiewicz,S., Rusnak,Z.,Siedlecka, U., 2006.Statystyka: ele-mentyteoriiizadania.WroclawUniversityofEconomics Publish-ingHouse,Wrocław.

Peterson,B.J.,Holmes,R.M.,Mcclelland,J.W.,Vörösmarty, C.J., Lammers,R.B.,Shiklomanov,A.I.,Shiklomanov,I.A.,Rahmstorf, S.,2002.IncreasingriverdischargetotheArcticOcean.Science 298 (5601), 2171—2173, http://dx.doi.org/10.1126/science. 1077445.

Planque,B.,Taylor,A.H.,1998.Long-termchangesinzooplankton and the climate of the North Atlantic. ICES J. Mar. Sci. 55, 644_—654.

Reynolds,R.W., Smith, T.M., Liu, C.,Chelton, D.B., Casey, K.S., Schlax,M.G., 2007.Daily high-resolution-blendedanalyses for seasurfacetemperature.J.Clim.20,5473_—5496,http://dx.doi. org/10.1175/2007JCLI1824.1.

Rutgersson,A.,Jaagus,J.,Schenk,F.,Stendel,M.,2014.Observed changesandvariabilityofatmosphericparametersintheBaltic Searegionduringthelast200years.Clim.Res.61,177—190.

Siegel,H.,Gerth,M.,Tschersich,G.,2006.Seasurfacetemperature developmentoftheBalticSeaintheperiod1990—2004. Ocea-nologia48(S),119—131.

Stramska, M., Dickey, T.D., 1993. Phytoplankton bloom and the verticalthermalstructureoftheupperocean.J.Mar.Res.51, 819—842.

Stramska,M.,Świrgoń,M.,2014.Inﬂuenceofatmosphericforcing andfreshwaterdischargeoninterannualvariabilityofthevertical diffuse attenuation coef_ﬁcient at 490nm in the Baltic Sea. Remote Sens. Environ. 140, 155_—164, http://dx.doi.org/ 10.1016/j.rse.2013.08.043.

Stramska,M.,Zuzewicz,A.,2013.In_ﬂuenceoftheparameterization ofwateropticalpropertiesonthemodeledseasurface tempera-tureintheBalticSea.Oceanologia55(1),53_—76,http://dx.doi. org/10.5697/oc.55-1.053.

Zaneveld,J.R.,Kitchen,J.C.,Pak,H.,1981.Thein_ﬂuenceofoptical watertypeontheheatingrateofaconstantdepthmixedlayer.J. Geophys.Res.86,6426—6428.