This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/authorsrights Author's personal copy AppliedGeochemistry40(2014)164–179 ContentslistsavailableatScienceDirect Applied Geochemistry journal homepage: www.elsevier.com/locate/apgeochem 234U/238U and d87Sr in peat as tracers of paleosalinity in the Sacramento-San Joaquin Delta of California, USA J.Z. Drexlera,⇑, J.B. Pacesb, C.N. Alpersa, L. Windham-Myersc, L.A. Neymarkb, T.D. Bullenc, H.E. Taylord aU.S.GeologicalSurvey,CaliforniaWaterScienceCenter,6000JStreet,PlacerHall,Sacramento,CA95819,UnitedStates bU.S.GeologicalSurvey,GeosciencesandEnvironmentalChangeScienceCenter,Box25046,DenverFederalCenter,MailStop963,Denver,CO80225,UnitedStates cU.S.GeologicalSurvey,Building15,McKelveyBlg.,MailStop480,345MiddlefieldRoad,MenloPark,CA94025,UnitedStates dU.S.GeologicalSurvey,3215MarineStreet,SuiteE-127,Boulder,CO80303,UnitedStates a r t i c l e i n f o a b s t r a c t Articlehistory: Thepurposeofthisstudywastodeterminethehistoryofpaleosalinityoverthepast6000+yearsinthe Received16May2013 Sacramento-SanJoaquinDelta(theDelta),whichistheinnermostpartoftheSanFranciscoEstuary.We Accepted25October2013 usedacombinationofSrandUconcentrations,d87Srvalues,and234U/238Uactivityratios(AR)inpeatas Availableonline1November2013 proxiesfortrackingpaleosalinity.PeatcoreswerecollectedinmarshesonBrownsIsland,FranksWet- EditorialhandlingbyM.Kersten land,andBaconChannelIslandintheDelta.Coresweredatedusing137Cs,theonsetofPbandHgcon- taminationfromhydraulicgoldmining,and14C.Aproof ofconceptstudyshowedthatthedominant emergentmacrophyteandmajorcomponentofpeatintheDelta,Schoenoplectusspp.,incorporatesSr andUandthattheisotopiccompositionoftheseelementstrackstheambientwatersalinityacrossthe Estuary.ConcentrationsandisotopiccompositionsofSrandUinthethreemainwatersourcescontrib- utingtotheDelta(seawater,SacramentoRiverwater,andSanJoaquinRiverwater)wereusedtocon- structathree-end-membermixingmodel.Deltapaleosalinitywasdeterminedbyexaminingvariations inthedistributionofpeatsamplesthroughtimewithintheareadelineatedbythemixingmodel. TheDeltahaslongbeenconsideredatidalfreshwatermarshregion,butonlypeatsamplesfromFranks WetlandandBaconChannelIslandhaveshownaconsistentlyfreshsignal(<0.5ppt)throughtime.There- fore,theeasternDelta,whichoccursupstreamfromBaconChannelIslandalongtheSanJoaquinRiver anditstributaries,hasalsobeenfreshforthistimeperiod.Overthepast6000+years,thesalinityregime at the western boundary of the Delta (Browns Island) has alternated between fresh and oligohaline (0.5–5ppt). PublishedbyElsevierLtd. 1.Introduction are not common or well distributed in tidal freshwater marshes, which are found in the inner reaches of estuaries. Therefore, a Peatsoilsandwetlandsedimentsarewellknownasarchivesof different proxy is needed to determine paleosalinity in tidal environmental change and contamination (e.g., Shotyk, 1996; freshwatermarshes. Byrne et al., 2001; Charman, 2002). Within these archives there Thenaturalchemistryofpeatmay,inandofitself,beauseful exist numerous proxies that can be used to address questions proxy for paleosalinity, if certain conditions are met. First of all, regarding climate change, hydrological processes, environmental thetracerbeingconsideredmustdifferinconcentrationbetween contamination, and paleosalinity (Ingram et al., 1996; Charman, freshwaterandseawater.Next,thetracerneedstobeconservative 2002; Van der Putten et al., 2012). With regard to paleosalinity, in naturesuch that, oncesequestered,there islittletendencyfor severalproxiessuchaspollen,stratigraphicseeddata,d13Cindia- subsequent mobility in peat, or be present in sufficient quantity tomsandforaminifera,andB,Br,Na,S,Ge,andUconcentrationsin thatexchangewithambientwaterisnegligible.Ideally,thetracer peathavebeenusedtocharacterizesalinityregimesthroughthe should be immune from modification due to changes in state, millennia (Goman and Wells, 2000; Malamud-Roam and Ingram, evaporativeconcentration,or chemicalprecipitationunderambi- 2004;DiRitaetal.,2011).Inaddition,d87Srinfossilbivalvesfound entconditions.Finally,thetracermustchieflyresideintheorganic withinsalineandbrackishestuarinesedimentshasbeenshownto componentofthepeat,whichisderivedfromtheplantsthatgrew beaparticularlyeffectivetracer(IngramandSloan,1992;Ingram in a site under a particular salinity regime (López-Buendia et al., and DePaolo, 1993; Ingram et al., 1996). However, fossil bivalves 1999).Iftheprovenanceofanelementinpeatismostlyfromsed- imentwashedinfromthewatershed(lithogenicsources),theele- mentcannottracesalinityconditionsatthetimetheplantswere ⇑ Correspondingauthor.Tel.:+19162783057;fax:+19162783071. growing.Previousresearchhasshownthatcertainhighlyinsoluble E-mailaddress:[email protected](J.Z.Drexler). 0883-2927/$-seefrontmatterPublishedbyElsevierLtd. http://dx.doi.org/10.1016/j.apgeochem.2013.10.011 Author's personal copy J.Z.Drexleretal./AppliedGeochemistry40(2014)164–179 165 elements,suchastitanium,arepredominantlyfoundinthelitho- fluctuationswould helpmanagersbetterpredict the resilienceof genic fraction of peat and remain highly immobile once incorpo- theDeltasubsequenttotheproposedmajorchangesinwatercon- rated into the peat matrix (Shotyk, 1996; López-Buendia et al., veyance (California Department of Water Resources et al., 2013). 1999; Novak et al., 2011). Such highly insoluble elements have Paleosalinityresearchhasbeencarriedoutinsalineandbrackish beenusedto‘‘normalize’’thesedimentvariabilityinthepeatma- parts of the San Francisco Estuary but in the upstream reaches, trix.Thisallowsforclearidentificationofelementschieflyfoundin only the western periphery of the Delta (at Browns Island) has theorganicfractionofpeat(i.e.,thoselackingarelationshipwith beenstudied(GomanandWells,2000;Goman,2001).Theresults Ti),whichconstitutethepotentialproxiesforpaleosalinity. ofthispreviousresearchshowedthat,between6200and3800cal- Thepurposeofthisstudywastodeterminethehistoryofpale- ibratedyearsbeforepresent(calyrBP),paleosalinityatBrownsIs- osalinityoverthepast6000+yearsintheSacramento-SanJoaquin landwascomparabletopresentvalues(slightlybrackish).Aperiod Delta(theDelta),whichistheinnermostpartoftheSanFrancisco offreshersalinitywasnotedbetween3800and2000calyrBPfol- Estuary.WeusedacombinationofSrandUconcentrations,d87Sr lowedbyareturntotheslightlybrackishconditionsoverthepast values, and 234U/238U activity ratios (AR) in peat as proxies for 2000years.Inordertocharacterizepaleosalinityrangeswithinthe tracking paleosalinity. We chose U and Sr because, similar to Ca, greaterDeltasystem,westudiedpeatfromBrownsIslandaswell Mg,K,andNa,theyarereadilysolubleacrossthesalinityspectrum astwoothersitesonawest-eastgradientwithsalinitiescurrently andhavedistinctconcentrationsinfreshwatervs.seawater(Table ranging from oligohaline (0.5–5 practical salinity units (psu); 1).Furthermore,theyeachhaveheavy,radiogenicisotopecompo- wherepsuistheunitlessPracticalSalinityScaleonwhichseawater sitionsthatmaybeuniquetoeachwatersourceandcanbeused is(cid:2)35)tofresh(<0.5psu). concomitantly as hydrologic tracers. The benefit of this approach isthat,unlikeelementalconcentrations,whichundernear-surface conditionsareaffectedbyavarietyofprocessesandchemicalreac- 2.Studysites tions that can result in complex geochemical behavior, 87Sr/86Sr and 234U/238U are largely immune from fractionation caused by The Sacramento-San Joaquin Delta of California was once a near-surface physical, chemical, and biological processes (Faure 1400km2 region of marshes, channels, and mudflats. Beginning and Powell, 1972; Hart et al., 2004). Furthermore, Sr is plentiful in the mid-1800s, the Delta was largely drained for agriculture inoceanwaterrelativetoriverwaterandSrisotopeshaveprevi- (Thompson, 1957), resulting in its current configuration of more ously been used for tracking paleosalinity (Ingram and Sloan, than100islandsandtractssurroundedby2250kmofman-made 1992;IngramandDePaolo,1993;Ingrametal.,1996).Inaddition, leveesand1130kmofwaterways(Prokopovich,1985).Tidesinthe Srisotopeshavebeensuccessfullyusedasatracerofprovenancein Delta are semidiurnal and microtidal, with a normal range of organicmaterials(Shandetal.,2007;vonCarnap-Bornheimetal., approximately one meter (Shlemon and Begg, 1975; Atwater, 2007). Uranium, which is commonly present at much lower 1980).TheclimateintheDeltaischaracterizedasMediterranean concentrationsthanSr,hastheextrabenefitofbeingtakenupvery withcoolwintersandhot,drysummers(Thompson,1957).Mean efficientlyandheldtightlybythepeatmatrixunderreducingcon- annualprecipitationisapproximately36cm,butactualyearlypre- ditions(Szalay,1964;Idizetal.,1986;Johnsonetal.,1987;Owen cipitationvariesfromhalftoalmostfourtimesthisamount.Over et al., 1992; Zielinski et al., 2000; Novak et al., 2011). Therefore, 80% of precipitation occurs from November through March 234U/238U isotopic composition of peat can be used as a tracer (Thompson,1957). andindependentcheckonSrisotopebehavior,whichmaybemore Currently,thewaterflowintheDeltaishighlyregulatedsothe susceptibletopost-depositionalmobility. saltwater wedge originating in the San Francisco Bay has little We applied our approach using peat cores collected in tidal chancetotravelmuchfurthereastthanthewesternboundaryof marshesintheDelta.Wefirstconductedaproofofconceptstudy theDelta.However,priortoconstructionofdamsonnearlyallma- in the greater San Francisco Estuary to determine whether living jor tributaries to the Delta during the 1940s to 1970s, salinity plants,whichultimatelybecomethebulkofthepeatmatrix,track incursionsintotheDeltacouldextendasfarastheeasternDelta, the salinity signal with regard to these tracers. Then we con- especiallyduringadryyear(Ingebritsenetal.,2000).Furthermore, structedathree-componentmixingmodelforwaterusingacom- inanygivenyear,salinitylikelyrangedfromfreshduringtherainy bination of Sr and U concentrations, d87Sr values, and 234U/238U seasontoslightlybrackishorverybrackishduringthedryseason, activityratios(AR)inthethreehydrologicend-members:seawa- dependingontheamountofannualprecipitation.Currently,how- ter, Sacramento River water, and San Joaquin River water. The ever, salinity is regulated for the purpose of maintaining water resultingmixingmodelwascomparedtothed87Srand 234U/238U qualityaswellasmanagingaparametercalledX ,whichisthedis- 2 compositions of marsh peat samples, nearly all of which plotted tancefromthemouthoftheestuaryattheGoldenGateuptheaxis withinthe‘‘mixingtriangle’’delineatedbytheendmembers.Delta towherethetidally-averagedbottomsalinityis2psu(Kimmerer, paleosalinitywasdeterminedbyexaminingpatternsinthedistri- 2002). butionofpeatsamplesthroughtimewithinthemixingtriangle. Marshstudysiteswerechosenalongthehistoricfloodplainof TodaytheSacramento-SanJoaquinDeltaisclassifiedasatidal theSacramentoRiveraswellastheglacialoutwashareaalongthe freshwater delta, yet whether this has been the case throughout SanJoaquinRiver(Fig.1).Inaddition,siteswereselectedfromhigh its (cid:2)6700-year history (Drexler et al., 2007) remains largely energyenvironmentssuchastheconfluenceoftheSacramentoand unknown. Knowledge of the extent and timing of past natural SanJoaquinRiverstomorequiescentenvironmentssuchasdistrib- Table1 Concentrations(mgl(cid:3)1)ofthemajorcationsCa,Mg,K,andNa,andminorcations,SrandUaswellaspHinfreshwaterfromtheSacramentoandSanJoaquinRiversandseawater. RiverdatawereretrievedfromtheNationalWaterInformationSystemoftheU.S.GeologicalSurveyfortheperiodfrom1950to2010andrepresentmedianconcentrationsofall measurements,whichincludeallchemicalspecies.SeawaterconcentrationsarefromTurekian(1968). Ca Mg K Na Sr U pH SanJoaquinRiver 32 15 3 69 0.389 0.005 7.7 SacramentoRiver 12 6.1 1.2 9.1 0.091 0.00013 7.7 Seawater 400 1290 392 10800 8.1 0.0033 7.9–8.2 Author's personal copy 166 J.Z.Drexleretal./AppliedGeochemistry40(2014)164–179 3.Methods 3.1.Samplecollection In the summer of 2005, peat cores were collected from each marsh site using a modified 5-cm-diameter Livingstone corer (Wright, 1991). At Browns Island, the first core, BRIC4, unlike all theothercoresinthestudy,didnothavegoodrecoverynearthe surface due to a particularly dense root mat. In addition, BRIC4 didnotreachtheunderlyingmineralsubstrate,eventhoughmuch clay was already present below 700cm. Therefore, an additional core(BRIC5)thatincludedtheentirepeatcolumnof922cmwas collectedinMarch2007within2moftheBRIC4site.Asoilmono- lithofapproximately50(cid:4)50(cid:4)50cm(BRIP)wasexcavatedfrom thesurfacetoimproverecoveryoverthatachievedwiththeLiving- stone corer. Further details about coring can be found in Drexler etal.(2009a). Real-timekinematicgeographicpositioningwasusedtoestab- lish the elevations and coordinates of the coring locations. Tidal benchmarkLSS13(NationalOceanicandAtmosphericAdministra- tiontidalstation9415064locatednearAntioch,CA)withastatic surveyed ellipsoid height of (cid:3)28.75 meters was used to adjust the elevations of the coring sites to local mean sea level. Details concerning peat coring and survey procedures can be found in Drexleretal.(2009b).Overallcharacteristicsofpeatateachmarsh siteaswellasmarshsurfaceelevationsareincludedinTable2. DuringNovemberandDecember2009,weconductedaproofof conceptstudytoevaluatewhethermarshplantsreflectthesalinity andisotopicsignalofthehydrogenicversuslithogeniccomponents withintheirambientgrowthenvironment.Wechose4marshsites intheSanFranciscoEstuaryrangingfrombrackishtofresh:Gali- Fig.1. Mapofpeat coringandplantcollectionsites(triangles)andriverwater nasCreekinSanPabloBay,RushRanchinGrizzlyBay,andBrowns collectionsites(circles)intheSanFranciscoEstuaryanditswatershed,California, IslandandFranksWetlandintheDelta(Fig.1).Wecollectedentire USA. specimens of Schoenoplectus californicus (dominant in Galinas utariesoftheSanJoaquinRiver.Thethreesites,BrownsIsland(BRI, Creek)orS.acutus(dominantintheothersites)atthreedifferent 268ha,highenergy),FranksWetland(FW,28ha,lowenergy),and places within each marsh. These plant species were chosen be- Bacon Channel Island (BACHI, 10ha, low energy) are relatively cause(1)theyareubiquitousinmostbrackishandfreshwatertidal undisturbed marshes, which, unlike those in the vast majority of marshes in the San Francisco Estuary, and (2) they are emergent the region, were not converted for agriculture (Drexler et al., macrophytes with large roots and rhizomes, which are the main 2009a).Vegetationonthesenaturalmarshislandsishighlyproduc- organiccontributorstothepeatmatrix.Entireplantsampleswere tiveanddominatedbytheemergentmacrophytesSchoenoplectus collected as close as possible to water quality gauges for which americanus(Americanbulrush),S.acutus(hardstembulrush),Phrag- long-termsalinitydatawereavailable.Plantsampleswererinsed mitesaustralis(commonreed),andTyphaspp.(cattail)andshrub- in the field with ambient water to remove sediment and placed scrub wetland species including Salix lasiolepis (arroyo willow) inlabeledplasticbags. andCornussericea(red-osierdogwood)(Drexleretal.,2009b).Peat WatersampleswerecollectedattheU.S.GeologicalSurveyriv- depositsatthesesitesrangefromseventooverninemetersindepth ergaugesontheSacramentoRiveratFreeportandtheSanJoaquin (Table2)andcontainabundantorganicmaterialincludingdecaying RiveratVernalis(Fig. 1)onMarch15and16,2011,respectively. plantroots,rhizomes,andstems,someofwhichareverywellpre- Sampleswerefilteredwitha0.45lmfilterandstoredin500-ml, served(GomanandWells,2000;Drexler,2011).Suchpreservation pre-cleaned high-density polyethylene bottles until analyzed for oforganicmaterialdemonstratesthatreducingconditionsarepre- SrandUconcentrationsandisotopiccompositions. valentinthepeatmatrix,causingdecompositiontoproceedalong slow,anaerobicpathways(GambrellandPatrick,1978).Reducing 3.2.Coreprocessing conditionsaretypicalintidalmarshsoilsduetothehighwaterta- ble,whichismaintainedbytheperiodicityofthetides(particularly In the laboratory, core stratigraphy was documented, cores inmicrotidalregionssuchastheDelta)aswellasthelowratesof weresplitlengthwiseandphotographed,andonelongitudinalhalf hydraulic conductivity that keep soils at or near saturation, even ofthecorewasarchivedforfutureuse.Bulkdensitywasobtained duringebbtides(Rabenhorst,2001). by sectioning cores into 2-cm-thick blocks, measuring each Table2 ElevationsandbasicdescriptionsofpeatcoresfromBACHI,BRI,andFW. Core Elevationoftopofcore(mMSL) Depthofpeatcolumn(cm) Meanbulkdensity(gcm(cid:3)3,±sd) Mean%organicmatter(±sd) BACHI 0.21 726 0.12±0.05 76±16 BRI 0.51 922 0.31±0.15 40±18 FW 0.27 718a 0.14±0.08 72±19 a Onlythetop608cmofcorewereusedbecausedeeperpeatlayerscontainedinversionsinradiocarbondates,precludingaccuratedating. Author's personal copy J.Z.Drexleretal./AppliedGeochemistry40(2014)164–179 167 dimension,obtainingwetweightofthesample,dryingovernight Concentrationsof Sr, U, Ti, Al, and Zr in freeze-driedpeat and at 105(cid:2)C, and then weighingagain to obtain dry weight (Givelet modernplantrootsweredeterminedbyinductivelycoupledplas- et al., 2004). Core data were examined for compression and/or mamassspectrometry(ICP–MS)usingaPerkinElmerElanModel expansion,butnomathematicalcorrectionswereneeded.Thesoil 6000andinductivelycoupledplasmaatomicemissionspectrome- monolithremovedfromthesurfaceofBrownsIslandwascorrected try (ICP-AES) using a Perkin Elmer Optima Model 3300DV. For forexpansion. theseanalyses,approximately100mgofsolidmaterialwascom- pletelydissolvedinanHCl–HNO –HFacidmixtureusingamicro- 3 wave total-digestion procedure (Roth et al., 1997; Barber et al., 3.3.Plantprocessing 2003; Hart et al., 2005). The digested samples were diluted at 1:10 (volume:volume, digest:water) with 18MXcm deionized To remove as much sediment as possible, live roots and rhi- water and were preserved with distilled nitric acid. Aerosols of zomes,whicharethemainorganiccontributorstopeatformation, acidified aqueous samples were introduced into both spectrome- were rinsed briefly in deionized water and blotted dry. Further ters using a high-solids Burgener pneumatic nebulizer. Multiple cleaning involved immersing rinsed samples in deionized water internal standards covering most of the mass range (In, Ir, and and placing them in a vibrating disrupter/homogenizer (Vortex). Rh)wereusedtocorrectforsystemdrift.Standardreferencemate- This process was repeated until the deionized water ran clear. rials(SRM)fromtheNationalInstituteofStandardsandTechnol- Balchtubescontainingthesampleswerethenfilledwithdeionized ogy (NIST) (Buffalo River Sediment, NIST SRM 8704 and Marine water,placedinasonicbathandsonicatedat51kHzfor20min. SedimentNISTSRM2702)weredigestedandanalyzedwitheach Afterparticlessettledfor1h,samplesweregentlyretrievedfrom analyticalbatch.Additionaldetailsregardingthespecificanalysis thetubesandexaminedat40(cid:4)magnificationforthepresenceof techniques,procedures,andinstrumentalsettingsforICP-MSanal- mineral particles. When particles were no longer observed, sam- yses can be found in Garbarino and Taylor (1996) and Taylor ples were freeze-dried for 12–24h until all liquid was removed. (2001);andforICP–AESinGarbarinoandTaylor(1979),Bossand Driedsampleswerestoredinsealedvialsuntilanalysis. Fredeen(1999),andHartetal.(2005). All concentrations were determined in triplicate on a single 3.4.Watersalinitydata digestionofeachfreeze-dried,homogenizedsample.Thestandard deviation(sd)forthetriplicateanalyseswasdeterminedforeach Surface-watersalinitydataforeachofthemarshsitesusedfor digest. In addition, replicate digestions were done every 10 sam- theplantstudywereobtainedfromlong-termgaugestationsmon- plestotesthomogeneity. itoredbytheU.S.GeologicalSurvey,theCaliforniaDepartmentof WaterResources,andtheNationalEstuarineResearchReservePro- 3.6.SrandUisotopeanalyses gram. Mean annual surface salinity values, data sources, years available,andlong-termsalinitytrendsareprovidedinTable3. IsotopiccompositionsofSrandUweredeterminedintheUSGS DenverRadiogenicIsotopeLaboratorybythermalionizationmass 3.5.Chemicalanalyses spectrometryonasubsetofthepeatandplantdigestionsusedfor elementalanalysis,and onsamplesof river water.Both Srand U Thepercentageoforganicmatterindriedbulkdensitysamples were chemically purified using ion chromatography (AG1(cid:4)8 for was determined by standard loss on ignition procedures (Heiri U;Sr-Spec™forSr),loadedontorheniumfilaments(doubleassem- etal.,2001)at4-cmintervalsnearthetopandbottomofeachcore bliesforU;singlefilamentsloadedalongwithtantalumoxidefor andotherwiseat10-cmintervals.Onaverage,duplicateswererun Sr),andrunonaThermoFinniganTritonmulti-collector,thermal every9–10samples.Reproducibilityofduplicateanalysesranged ionization,isotoperatiomassspectrometer. from0%to4.8%withanaveragevalueof0.74%. Srisotoperatiosweredeterminedineitherstaticormulti-dy- Two-cmsectionsofcorewerefrozenondryiceandshippedto namicanalyticalmode,and87Sr/86Sratomicratioswerecorrected theUSGSlaboratoryinBoulder,Coloradowheretheywerefreeze- for mass fractionation using 88Sr/86Sr measured in the same run. dried and milled in polyethylene vials with acrylic balls using a Reported 87Sr/86Sr values are also corrected for inter-laboratory SPEX(cid:3) Model 8000 mixer mill. In the monolithic block from the biasbynormalizingtothemeanvalueobtainedforSrisotopestan- BRI surface, inorganic (clay lenses) and organic material (‘‘rela- dardNIST987analyzedalongwithsamplesandassumingavalue tivelypurepeat’’)in4contiguous2-cmsectionsrangingindepth of 0.710248 (McArthur et al., 2001). Replicate analyses of Sr in from0.220to0.171mMSLwereseparatedusingaceramicknife modern-marine carbonate standard, EN-1, run during this time to compare the chemistry of these different components of the gaveameanvalueof0.709176±0.000014(2(cid:4)sd;N=48)which peatmatrix. is within error of the accepted value of 0.709174±0.000002 for Table3 Surfacewatersalinitydataforgauges(deployedbelowmeanlowerlowwaterata15-mintimestep)intheSanFranciscoEstuary.Alldatawerecollectedassalinity,exceptfor BrownsIslandsiteswhichwerecollectedasspecificconductanceandcorrectedtosalinitybymultiplyingby0.0006364(Ganjuetal.,2005). Site Meansalinity Source Yearsavailable (psu,(sd)) GalinasCreek(SanPabloBay) 23.2(4.2) SanFranciscoBayNationalEstuarineResearchReserve 5/1/2008–12/31/2009 underNOAA’sNationalEstuarineResearchReserve System’s(NERRS)NationalMonitoringProgram RushRanch(GrizzlyBay) 6.1(2.0) NERRSNationalMonitoringProgram 5/20/2008–12/31/2009 BrownsIsland(theDelta) 1.8(1.8) TwoUSGSgauges:sitesdescribedinGanjuetal.(2005) 3/27/2002–5/20/2002;11/9/2002–12/17/2002; 3/12/2003–4/16/2003;9/5–28/2005;10/12–26/2005; 1/11/2006–2/1/2006 FranksWetland(theDelta) 0.7(0.4) SJOgaugeoftheCaliforniaDepartmentofWaterResources 1/1/2007–12/31/2008 Author's personal copy 168 J.Z.Drexleretal./AppliedGeochemistry40(2014)164–179 modernseawater(McArthuretal.,2006). Deltavaluesin permil (Hornbergeretal.,1999;AlpersandHunerlach,2000;Domagalski, (‰)usedinthisreportarecalculatedwiththefollowingequation: 2001;Bouseetal.,2010).Therefore,weidentifiedthiscriticaltime d87Sr=((87Sr/86Sr /87Sr/86Sr )(cid:3)1)(cid:4)1000, using the periodinCaliforniahistorybysearchingformajorincreasesinPb sample seawater 87Sr/86Sr value given above. Analytical errors determined and/or Hg in the peat record. By normalizing these elements to seawater by replicate analyses of standards and unknowns are better than Ti to account for variable sediment content (Alpers et al., 2008), ±0.00002(=±0.03‰ford87Sr)andrepresent95%confidencelevels wefoundrapidincreasesinPbcontaminationbeginningatapprox- (2(cid:4)sd). imately(cid:3)0.153(±0.12)mMSLatBRI,(cid:3)0.29(±0.04)mMSLatFW, U isotope ratios were determined in multi-dynamic peak- and(cid:3)0.17mMSLatBACHI. hoppingmodeusingasinglesecondaryelectronmultiplier.Ratios Radiocarbondatingofachenes(Schoenoplectusfruitingbodies), were correctedfor mass fractionation using theknown236U/233U seeds, charcoal, and other terrestrial macrofossils were used to isotope ratio in an added double spike. Replicate analyses of estimate ages greater than 250years. Radiocarbon samples were U-isotope standard (NIST4321B run between January 2009 and analyzedbyacceleratormassspectrometryattheCenterforAccel- July2010) yieldeda mean234U/235Uatomic ratioof0.0072921± erator Mass Spectrometry, Lawrence Livermore National Labora- 0.0000135 (=±0.19% based on 2(cid:4)sd; N=134), which is within tory in Livermore, California. Ages were calibrated using CALIB analytical uncertainty of the accepted value (0.007294± (version5.0.1;StuiverandReimer,1993)withtheINTCAL04curve 0.000028).Correctionsforinstrumentdriftweremadebynormal- (Reimeretal.,2004).Age-depthsplinefitmodelsforthecomplete izing234U/235Uvaluesmeasuredforunknownsbythesamefactor peat profiles (all 2-cm sections) were constructed following the needed to correct measured 234U/235U ratios for the SRM 4231B procedureinHeegaardetal.(2005).Additionalinformationabout standardruninthesamemagazine.Measured234U/235Uatomicra- the 14C dating of the peat cores, including all data, is reported tios were converted to 234U/238U activity ratios (AR) using decay elsewhere(Drexleretal.,2009a). constantspublishedbyChengetal.(2000)(k =2.8262(cid:4)10(cid:3)6yr(cid:3)1) 234 andJaffeyetal.(1971)(k =1.55125(cid:4)10(cid:3)10yr(cid:3)1),assumingthat 238 all U has a 238U/235U composition of 137.88 (Steiger and Jäger, 3.8.Three-componentisotoperatiomixingmodels 1977). Replicate analyses of a solution of 69 million year old U Water in the Sacramento-San Joaquin Delta is derived from ore (Ludwig et al., 1985) assumed to be in radioactive secular freshwater inputs from the Sacramento and San Joaquin Rivers equilibrium were typically within analytical uncertainty of 1.000. However, the long-term average 234U/238U AR value of andseawaterfromtheSanFranciscoEstuary.Eachofthesesources hasadistinctchemicalandisotopicfingerprint.Therefore,thedis- 0.9983±0.0025(2(cid:4)sd;N=93)indicatesasmallsystematicbias, solvedionloadinthewatercolumnatanygivenplacewillmostly implying that the ore ‘‘standard’’ may not completely satisfy the consist of a mixture of these three sources, which reflect the ion assumptionofclosed-systemevolution.Analyticalerrorsformea- sured 234U/238U AR values are given at the 95% confidence level concentrations of each end member. The d87Sr and 234U/238U AR compositionsofthemixturedependonboththeisotopiccomposi- (2(cid:4)sd) and include contributions from within-run uncertainties tionsandtheSrandUconcentrationsofendmembers.Becausethe (counting statistics) plus uncertainties propagated from blank, isotopic compositions of these constituents are not strongly spike,andmassfractionationcorrections,aswellasexternalerror affected by near-surface chemical reactions or processes, simple derivedfrommultipleanalysesoftheUisotopestandard. binary mixing equations were used to describe the resulting d87Sr and 234U/238U AR compositions of the mixed water. The 3.7.Peatdating three-componentmixingmodelsemployedinthisstudyarebased Peat profiles were analyzed for 137Cs in order to estimate the ontheoryandequationsgiveninFaure(1986,chapter9)andFaure depth of the 1963 layer. The activities of 137Cs in core sections andMensing(2005,chapter16). Inasystemconsistingoftwoend-membercomponents,Xand withinthetopmeterofpeatwerecountedusingagammadetector Y,theconcentrationofanyelementinthemixture,C ,isdepen- (low-background germanium detector) and a multi-channel ana- M dentontheconcentrationsofthatelementinbothendmembers, lyzeratUSGSlaboratoriesinDenver,COandMenloPark,CA.Sam- C andC ,andthefractionofmixing,f,suchthat ples were analyzed for 24–48h to achieve the desired precision. X Y Thismethodhasbeenusedextensivelytodatepeatandsediments C ¼C f þC ð1(cid:3)f Þ ð1Þ inwetlandsandlakes(ArmentanoandWoodwell,1975;DeLaune M X X Y X et al., 1983; Ritchie and McHenry, 1990; Van Metre and Fuller, wheref =X/(X+Y)andvariesfrom0to1.Mixturesofthetwocom- 2009). 137Cs is a product of atmospheric fall-out from nuclear X ponentswillresultinpointsthatdefineastraightlinebetweenthe weapons testing and power plant accidents. Significant levels of end-membercompositionsonplotsofC versusC thataredirectly this isotope first appeared in the atmosphere in the early 1950s, X Y proportionaltothevalueoff(SupplementalFig.1a).However,the withpeakquantitiesdetectedin1963.Thepeatorsedimentlayer isotopiccompositionofthemixture,R ,dependsonboththeiso- from 1963 can be identified based on the maximum activity of M tope ratios of end members, R and R , as well as their elemental 137Csinaprofile.Itisimportanttonotethatthereissomeuncer- X Y concentrationssuchthat taintyaboutthelong-termimmobilityof137Csinpeatduetopro- cessessuchasdiffusionandadvectionthroughporewater,uptake R ¼R C f þR C ð1(cid:3)f Þ ð2Þ byvegetation,andflushingbywatersofhighersalinity(Turetsky M X X x Y Y X et al., 2004; Foster et al., 2006). However, because research has Because values of R are most strongly influenced by the M shown lower mobility of 137Cs in peat containing clay minerals componentwiththehigherconcentration,mixingcurvesnolonger (MacKenzie et al., 1997), which is the case in the Delta, and definestraightlinesonplotsofR versusC ,butinsteadafamily M M becausewearemostinterestedinvariationsonamillennialtime ofhyperboliccurveswhosedegreeofcurvaturedependsonthedif- scale, we decided that using 137Cs dating to estimate the 1963 ference in concentration between C and C (Faure and Mensing, X Y horizonsatisfiestheneedsofthisstudy.Weestimatedthelocation 2005, Fig. 16.5). Furthermore, mixing fractions defined by f are ofthe1850peathorizon,whichrepresentstheinitiationoftheCal- no longer distributed proportionally along the mixing curve, but ifornia Gold Rush period, by examining peat geochemistry. arecompressedtowardtheendmemberwiththehigherconcen- Hydraulic gold mining, which began in 1852, resulted in a sharp tration (Faure and Mensing, 2005, Fig. 16.6). Binary mixtures of rise in Pb and/or Hg contamination in the San Francisco Estuary two elements (Sr and U) with different isotope ratios can be Author's personal copy J.Z.Drexleretal./AppliedGeochemistry40(2014)164–179 169 treatedinthesamewayusingtwosetsofmixingequations.Values (a) ofd87Sr and234U/238UAR foranygivenvalueoffwillplotalong M M thecurvedependingonvaluesofSr ,Sr ,U ,andU . X Y X Y Three-component mixing of d87Sr and 234U/238U AR between components X, Y, and Z, can be treated as a series of separate twocomponentmixtures.Ifallthreecomponentsarepresent,mix- tureswillplotwithinthepolygonalspacedefinedbythree2-com- ponent mixing curves between X–Y, X–Z, and Y–Z. To quantify proportions of each component, mixing webs can be calculated using the two-component mixing curves between intermediate endmembersconsistingoftheR andC determinedat0.1inter- M M valsoffformixturesalongallthreetwo-endmembercurves. Forthisstudy,d87Srand234U/238Uendmembersconsistofcom- positionsobservedinseawater,theSacramentoRiver,andtheSan JoaquinRiver(SupplementalFig.1b).Valuesfortheseawaterend memberswereobtainedfromMcArthuretal.(2006)andDelanghe (b) etal.(2002).DatafortheSacramentoRiverendmemberwereob- tainedfromwatersamplescollectedforthisstudyanddataavail- able from the U.S. Geological Survey National Water Information System(NWIS).DatafortheSanJoaquinRiverendmemberwere obtained from water samples collected for this study, NWIS, and Westcot et al. (1992). Values of d87Sr for river water determined in this study are similar to those reported by Ingram and Sloan (1992). For the purposesof this study,we assume that d87Sr and 234U/238U AR values of all water sources within the Estuary have remained relatively constant throughout the Holocene, which is justified as a first approximation because the lithogenic compo- nents of the watershed have not changed for well over 100,000years(IngramandSloan,1992). 4.Results Fig.2. Meansurfacewatersalinityvs.(a)d87Srand(b)234U/238Uactivityratio(AR) determinedinlivingSchoenoplectusroots/rhizomescollectedfromfoursitesinthe 4.1.Proofofconceptstudy SanFranciscoEstuaryrepresentingasalinitygradientfromfreshtosalinemarshes. DataarefromSupplementalTable1.Simplebinarymixingmodelsareshownfor The results of the proof of conceptstudy illustrate that the Sr mixturesofseawaterandSacramentoRiverwater(dottedlines)andSanJoaquin Riverwater(solidlines)usingdatafromSupplementalTable3.Italicizednumbers isotopiccompositionsofmodernplantsreliablyintegratethesalin- in(b)areUconcentrationsinroot/rhizometissuefromSupplementalTable1. itysignalthroughouttheEstuary.SrandUconcentrationsandiso- tope data for Schoenoplectus roots/rhizomes are provided in Supplemental Table 1. d87Sr values in Schoenoplectus roots/rhi- Sr concentrations (median value for 11 samples in Supplemental zomescompriseanarrowrangeateachsite,butawiderrangebe- Table1is15lgg(cid:3)1forSr),makingthemmoresusceptibletoinflu- tween sites that is consistent with location within the Estuary encebyfactorsotherthansimplemixing. (Fig. 2a). Seawater dominates d87Sr compositions at salinities Unlike Sr isotopes, U isotopes can be fractionated as a conse- aboveafewpsu.Althoughthed87Srcompositionofwaterpresent quence of recoil processes associated with alpha-decay of 238U. ateachsitewasnotmeasured,dataforplantstendtofollowthe Duringthisprocess,234Uisphysicallydisplacedfromthelocation mixing relationship expected between seawater and freshwater oftheparental238Uatomduetotheejectionofthealphaparticle represented by Sacramento and San Joaquin River water (solid (Kigoshi, 1971). The resulting 234U decay products are more anddashedcurvesinFig.2a).Thehyperbolicnatureofthemixing susceptible to mobilization than 238U due to factors such as curvesistheresultofthenon-lineareffectswhenlargedifferences radiation-induced ionization and damage of the crystal lattice inSrconcentrationarepresentbetweenthemixingendmembers. (Gascoyne, 1992; Chabaux et al., 2008; Porcelli, 2008). Direct Theobservedpatternofd87Srversussalinityinplantsisverysim- implantationofrecoil234Uis particularlynoticeableinsituations ilar to a plot of surface-water salinity and d87Sr measurements where adjacent phases have high and low U concentrations shownbyHobbsetal.(2010). (Neymark and Amelin, 2008), such as the case here with low U Incontrast,asimilarplotof234U/238UARversussalinityshows root/rhizometissuegrowinginhigherUpeat.Directimplantation that compositions of roots/rhizomes from Franks Wetland and of recoil 234U into root/rhizome tissue may be further enhanced GalinasCreekarebroadlyconsistentwithamixingmodelinvolv- because recoil distances increase substantially in low density ingseawaterandfreshwaterfromtheSacramentoRiver.However, materials (Chabaux et al., 2008). Roots/rhizomes from BRI and samples from Browns Island and Rush Ranch have much greater RushRanchareparticularlysusceptibletoinfluencebythisprocess variationsthatdeviatetohighervaluesthanthoseexpectedfrom becauseoftheirlowUconcentrationscomparedtoroot/rhizomes simplemixingofseawaterandfreshwater(Fig.2b).Wenotethat analyzedfromFWandGalinasCreek.Itislikelythatallroots/rhi- roots/rhizomeswiththehighest234U/238UARvaluestendtohave zomesincorporatealimitedamountofrecoil-generated234Ufrom substantially lower U concentrations ((cid:2)0.11lgg(cid:3)1 or less) than thepeatinwhichtheygrowinadditiontosolubleUderivedfrom samples that conform to the salinity mixing pattern (typically theoverlyingwatercolumn.However,atlowconcentrations, the >0.2lgg(cid:3)1) or samples of peat from the Delta (median value of additionofrecoil234Umayhaveanoticeableeffectontheisotopic 3.8lgg(cid:3)1 for 150 analyses). We also note that U concentrations compositionofUincorporatedintothelivingroot/rhizometissue. inroots/rhizomeswithelevated234U/238UARaremuchlowerthan Consequently, root samples with very low U concentrations are Author's personal copy 170 J.Z.Drexleretal./AppliedGeochemistry40(2014)164–179 likelytohave234U/238UARvaluesthatarehigherthanvaluesinthe water. However, starting around 350BCE, 234U/238U AR values in overlyingwatercolumn. peatsamplesfromBRIgraduallydecreasetowardsvaluesobserved inseawaterandSanJoaquinRiverwater,reachingalowof(cid:2)1.15at around1850–1900CE.Duringthistimeperiod,peatsamplesfrom 4.2.Depthprofilesinchemicalandisotopedata FWandBACHIshiftsubtlytowardSacramentoRivercomposition intermsofboth234U/238UARandd87Sr. Profiles of Sr and U concentrations show a large amount of scatterwithtime(ordepth)atindividualsites(Fig.3aandb).This is particularly the case at Browns Island, which is located at the 4.3.RelationshipswithTicontent confluenceofthetworivers.Correlationofshort-termfluctuations isnotreadilyapparentbetweensites.Furthermore,bothSrandU ThereisastrongrelationshipbetweenTiandashcontentinthe concentrations show substantial overlap between different loca- peatfromallthreesitesintheDelta(Fig.4).Thisstrongrelation- tionswithintheDeltaanddonotclearlyreflectasalinitygradient ship justifies the use of Ti normalization to account for varying fromBRItoBACHI. amounts of sediment being incorporated into the peat through Incontrast,profilesofSrandUisotopiccompositionsshowless time. scatterovertimeandgreaterseparationbetweensites(Fig.3cand Plots of various elements vs. Ti suggest different processes by d). All peat samples have d87Sr compositions that range between which elements have become incorporated into the peat matrix valuesmeasuredforSacramentoRiverwater((cid:3)3.83‰)andseawa- (Fig. 5). Because a pure organic component could not be isolated ter(0‰).ValuesforBRIpeatsamplesaretypicallyequaltoorsub- from peat samples prior to analyses, a means of discriminating stantially higher than values observed in FW or BACHI samples, between contributionsfrom lithogenic and hydrogenic sources is whichisconsistentwiththegreatercontributionsfromseawater needed. A number of elements are highly insoluble in typical inthewesternDeltathanfurtherupstream. surfacewaterincludingTi,Al,andZr.Thesourcefortheseelements Unliked87Sr,234U/238UARvaluesfortwoofthethreesourcesof is the inorganic sediment incorporated into the peat. This is water (San Joaquin River and seawater) are nearly identical evident in plots of Al and Zr versus Ti from peat samples at BRI (Fig.3d).Valuesof234U/238UARforolderpeatsamplesclearlydis- (Fig.5aandb)whereallthreeelementsdefinelithogenicmixing tinguishBRIfromthetwofreshersites,BACHIandFW(Fig.3d).FW lines that pass through the origin as well as through clay-rich andBACHIsamplesplotclosertothevalue forSanJoaquinRiver samples from either near the top of the profile or the basal clay. waterandBRIsamplesplotclosertovaluesforSacramentoRiver Noadditionalsourcefortheseelementsisrequired. Fig.3. PlotsofSrconcentrations(a),Uconcentrations(b),d87Srvalues(c),and234U/238UAR(d)versuselevationrelativetomeansealevel(andpeatageforBRI)inpeat samples. Author's personal copy J.Z.Drexleretal./AppliedGeochemistry40(2014)164–179 171 100 digestionsofpeatsamples(U ).ValuesforU rangefromnear0% T N to 89%; however, median values of 55% and 73% are obtained dependingon whetherbasal clay or clay-richlayers near thetop 80 of the profile are used to represent the lithogenic component. TheseresultsindicatethatmostoftheUpresentinpeatsamples %) 60 isderivedfromnon-lithogenicsources. wt. UnlikeU,Srconcentrationsinpeatsamplesshownorelationto h ( 40 FW: y = 171.91x + 14.003 lithogenicmixingmodelsandcanhavevalueseitherlowerorhigh- s A R2 = 0.7617, p < 0.0001 er than those expected based on Ti concentrations (Fig. 5d). Fur- BRI: y = 119.91x + 26.274 thermore, peat samples show a crude negative correlation 20 R2 = 0.8309, p < 0.0001 betweenSrandTi(graybandinFig.5d).Theserelationsindicate BACHI: y = 174.41x + 2.7613 that lithogenic Sr does not provide a significant contribution to R2 = 0.8421, p < 0.0001 0 peatsamplesdespitesimilarconcentrationsinpeatandclay-rich 0 0.1 0.2 0.3 0.4 0.5 0.6 samples.Instead,samplesofclay-richmaterialsmaybecontami- Ti (wt. %) natedwithsmallamountsoforganicmatterthatcontributemuch oftheSrinthebulkanalysis. BRI FW ThesedataindicatethatSrandUinpeatsamplesarenotpartic- BACHI Linear (BRI) ularlysensitivetocontributionsfrominorganicsedimentincluded Linear (FW) Linear (BACHI) inthepeatmatrix.UnlikeTi,bothUandSraresolubleinoxidizing surfacewatersincludingseawaterandriverwater.Therefore,itis Fig.4. RelationshipsbetweenTiconcentrationsandashcontentofpeatsamplesat likelythatvariationsobservedintheseelementsareduetoshifts BRI,FW,andBACHI. incontributionsfrommultiplehydrogenicsources.Minimalinflu- ence by lithogenic components in Delta peat samples is further The lithogenic source also contains some U and Sr; however, supported by isotope data. Four of the five samples of clay-rich neither of these elements has a strong correlation with Ti as do materialfromBRIhavemeasured234U/238UARvaluesthatarelow- Al and Zr (Fig. 5c and d). U concentrations of most peat samples erthanthoseobservedinpeat(Fig.6a).Theseresultsareconsis- aregreaterthanvaluespredictedbycontributionsfromthelitho- tent withthe conceptthat most lithogenic source material is old genic source based on Ti concentrations (i.e., they fall above the enough to have reached radioactive secular equilibrium lithogenicmixinglinesinFig.5c).TheproportionofUderivedfrom (234U/238U AR=1.0). However, the fact that all clay-rich samples non-lithogenicsources(U )canbecalculatedbysubtractingtheU have 234U/238U AR values greater than 1 indicates a strong N attributabletolithogenicsources(U)fromtheUmeasuredintotal likelihoodthatthefinesedimentconstitutingthelithogenicfaction L Fig.5. RelationshipsbetweenTiconcentrationsandaluminum(a),zirconium(b),uranium(c),andstrontium(d)atBRI.Lithogenicmixinglinesareshownforplotsof concentrationandarebasedontheassumptionthatclay-richmaterialsrepresentthelithogeniccomponentofpeatsamplesandthatconcentrationsattributabletothe lithogeniccomponentarezerowhenTiiszero.R2andp-valuesprovidedforsimplelinearregressionsfororganic-richpeat(thickgraylines).SeetextforexplanationofUN andULin(c). Author's personal copy 172 J.Z.Drexleretal./AppliedGeochemistry40(2014)164–179 includesUadsorbedfromthewaterinwhichitwastransported. component drops to less than about 70%, 234U/238U AR in mixed Thesorptioncapacityoffineclays,ferrihydroxidesandamorphous water becomes dominated by U from the San Joaquin River, silica,canvaryfromminimaltosubstantialdependingonmineral- obscuringfurtherquantification. ogy and grain size as well as physical conditions and chemical Because Sr and U present in peat samples are dominated by compositions of the fluids (Borovec, 1981; Ames et al., 1983; hydrogenic rather than lithogeniccomponents,isotopic composi- Anderssonet al., 2001;Chabaux etal., 2003, 2008). U adsorption tionsofwholepeatdigestionsshouldreflectmixturesofthethree onto surfaces of fine, inorganic particles during transport further mainwatersourceswithintheDelta.Nearlyallpeatsamplesrep- decouples the connection between lithogenic and hydrogenic resenting different time periods in the Delta fall within or very sourcesforU.Consequently,valuescalculatedforU likelyinclude closetothedistorted3-membermixingtriangle(Fig.7).Samples L asubstantialcomponentofU whose234U/238UARismorelikelyto that fall outside the limits of the mixing model consist of one of N reflect the composition of the hydrogenic source rather than the 37 peat samples from BRI (from the (cid:2)1850 to 1963 period), all original lithogenic source from which Ti is derived. This scenario the clay samples from BRI (which are dominated by lithogenic is supported by the lack of a correlation between 234U/238U AR ratherthanhydrogeniccomponents),and4of23FWpeatsamples andTiinpeatsamplesthatwouldberequiredifthelithogenicfrac- (Fig.7bandc).PeatsamplesfromBACHIshowthetightestcluster tionmadeasignificantcontributiontoU (Fig.6a).Likewise,d87Sr ofd87Srand234U/238UARvalues,probablybecausethissiteisgeo- T valuesinpeatdonotshowanoverallcorrelationwithTi,although graphically the least likely to be influenced by hydrologic inputs samples do show an intriguing V-shaped pattern that reflects from either seawater or Sacramento River water. Peat samples changesinbothTiandd87Srwithtime(Fig.6b).Variationsinthese fromFWshowsubstantiallygreaterscatter,consistentwithitspo- constituentsshift more or less systematicallythroughouttheen- sitionclosertobothseawaterandSacramentoRiversources.d87Sr tire 6000+year period of Delta evolution (see profile in Fig. 3c). compositions indicate that seawater was a minor constituent Thus,theycannotreflectsimplemixingtrendsbetweenlithogenic remaining<(cid:2)5% throughout the last 6000years. Variations in U andnon-lithogeniccomponentsatanygiventimeandmorelikely isotopes in FW peat indicate that freshwater sources fluctuated reflectevolvingdepositionalandenvironmentalconditions. greatly, alternating between domination by the San Joaquin vs. the Sacramento River (Fig. 7c). Peat samples from BRI show the 4.4.PaleosalinitymixingmodelbasedonSr–Udata greatestamountofvariabilitybetweenfreshwatersourcesbecause of its position at the confluence of both rivers and the greatest TheresultsofsimplemixinginSr–Uconcentrationspacecanbe influence from seawater due to its location at the western represented by an ordinary ternary plot (Supplemental Fig. 1a), (seaward)boundaryoftheDelta(Fig.7b). however in d87Sr–234U/238U AR space, the ternary plot becomes Over time, there is considerable variability in d87Sr and warpedandstretchedduetothelargedifferencesinSrandUcon- 234U/238U AR in BRI peat samples (Fig. 7b), which can be related centrations between end members and the non-linear nature of tochangingpatternsofpaleosalinityintheDelta.Theoldestpeat mixing relations (Supplemental Fig. 1b). Mixtures of the three samples(4350–1550BCEshownbyblacksymbols)havethelow- end members are greatly influenced by addition of seawater be- estd87Srvaluesandlikelyrepresentconditionswiththeleastsea- causeofitsrelativelyhighconcentrationofSr.Additionofaslittle water (a mean of approximately 1%, which is equivalent to as10%seawatercausesd87Srtoincreasefromlowvaluesinfresh- (cid:2)0.35psu).Thesesamplesshowawiderangeof234U/238UARval- watersources((cid:3)2.62‰to(cid:3)3.84‰forSanJoaquinandSacramento ues and may represent frequent shifts between the San Joaquin Riverwater,respectively,SupplementalFig.1a)tovaluesashighas andSacramentoRiversprovidingthedominantsourceofdissolved (cid:3)0.35‰.Furthermore,thelargedifferenceinUconcentrationbe- ionstothesite.Duringtheperiodfrom1550to350BCE(darkblue tweenthetworivers(0.13and5.0lgl(cid:3)1forSacramentoandSan symbols),samplesclusteraroundd87Srof(cid:3)2.0,whichisveryclose JoaquinRiverwater,respectively,SupplementalFig.1b)resultsin tothemeansalinityoftheoldest(black)groupofsamples.These substantial variation in 234U/238U AR values even if water in the samples also have 234U/238U AR values between 1.30 to 1.35 and Delta remains dominated by freshwater sources. Mixtures of up plot close to the mixing line between seawater and Sacramento to 20% San Joaquin River and 80% Sacramento River waters can River water, indicating that the Sacramento River dominated the be readily discriminated; however, once the Sacramento River freshwater component during this time. Peat deposited during (a) (b) Fig.6. RelationshipsthroughtimebetweenTiconcentrationsand234U/238Uactivityratio(a)andd87Sr(b)inpeatsamplesfromBRI.