freshwater pearl mussels from northern sweden serve as ...multi-decadal records of 18o signals in...

Hydrol Earth Syst Sci 24 673ndash696 2020httpsdoiorg105194hess-24-673-2020copy Author(s) 2020 This work is distributed underthe Creative Commons Attribution 40 License

Freshwater pearl mussels from northern Sweden serve as long-termhigh-resolution stream water isotope recordersBernd R Schoumlne1 Aliona E Meret2 Sven M Baier3 Jens Fiebig4 Jan Esper5 Jeffrey McDonnell6 andLaurent Pfister7

1Institute of Geosciences University of Mainz Mainz 55128 Germany2Naturhistoriska riksmuseet Stockholm 114 18 Sweden3Agilent Technologies Sales amp Services GmbH amp Co KG Frankfurt am Main 60528 Germany4Institute of Geosciences J W Goethe University Frankfurt am Main 60438 Germany5Department of Geography University of Mainz Mainz 55128 Germany6Global Institute for Water Security University of Saskatchewan Saskatoon SK S7N 3H5 Canada7University of Luxembourg Faculty of Science Technology and Medicine 2 Avenue de lrsquoUniversiteacute4365 Esch-sur-Alzette Luxembourg

Correspondence Bernd R Schoumlne (schoenebuni-mainzde)

Received 29 June 2019 ndash Discussion started 29 July 2019Revised 27 November 2019 ndash Accepted 16 January 2020 ndash Published 17 February 2020

Abstract The stable isotope composition of lacustrine sed-iments is routinely used to infer Late Holocene changes inprecipitation over Scandinavia and ultimately atmosphericcirculation dynamics in the North Atlantic realm Howeversuch archives only provide a low temporal resolution (ca15 years) precluding the ability to identify changes on inter-annual and quasi-decadal timescales Here we present anew high-resolution reconstruction using shells of freshwa-ter pearl mussels Margaritifera margaritifera from threestreams in northern Sweden We present seasonally to annu-ally resolved calendar-aligned stable oxygen and carbon iso-tope data from 10 specimens covering the time interval from1819 to 1998 The bivalves studied formed their shells nearequilibrium with the oxygen isotope signature of ambientwater and thus reflect hydrological processes in the catch-ment as well as changes albeit damped in the isotope sig-nature of local atmospheric precipitation The shell oxygenisotopes were significantly correlated with the North AtlanticOscillation index (up to 56 explained variability) suggest-ing that the moisture that winter precipitation formed fromoriginated predominantly in the North Atlantic during NAO+

years but in the Arctic during NAOminus years The isotope sig-nature of winter precipitation was attenuated in the streamwater and this damping effect was eventually recorded bythe shells Shell stable carbon isotope values did not show

consistent ontogenetic trends but rather oscillated around anaverage that ranged from ca minus1200 to minus1300 permil amongthe streams studied Results of this study contribute to animproved understanding of climate dynamics in Scandinaviaand the North Atlantic sector and can help to constrain eco-hydrological changes in riverine ecosystems Moreover longisotope records of precipitation and streamflow are pivotalto improve our understanding and modeling of hydrologicalecological biogeochemical and atmospheric processes Ournew approach offers a much higher temporal resolution andsuperior dating control than data from existing archives

1 Introduction

Multi-decadal records of δ18O signals in precipitation andstream water are important for documenting climate changeimpacts on river systems (Rank et al 2017) improving themechanistic understanding of water flow and quality control-ling processes (Darling and Bowes 2016) and testing Earthhydrological and land surface models (Reckerth et al 2017Risi et al 2016 see also Tetzlaff et al 2014) However thecommon sedimentary archives used for such purposes typi-cally do not provide the required ie at least annual tem-poral resolution (eg Rosqvist et al 2007) In such stud-

Published by Copernicus Publications on behalf of the European Geosciences Union

674 B R Schoumlne et al Freshwater pearl mussels as long-term high-resolution stream water isotope recorders

ies the temporal changes in the oxygen isotope signaturesof meteoric water are encoded in biogenic tissues and abio-genic minerals formed in rivers and lakes (eg Teranes andMcKenzie 2001 Leng and Marshall 2004) In fact manystudies have determined the oxygen isotope composition ofdiatoms ostracods authigenic carbonate and aquatic cellu-lose preserved in lacustrine sediments to reconstruct LateHolocene changes in precipitation over Scandinavia and ul-timately atmospheric circulation dynamics in the North At-lantic realm (eg Hammarlund et al 2002 Andersson et al2010 Rosqvist et al 2004 2013) With their short residencetimes of a few months (Rosqvist et al 2013) the hydro-logically connected through-flow lakes of northern Scandi-navia are an ideal region for this type of study Their isotopesignatures ndash while damped in comparison with isotope sig-nals in precipitation ndash directly respond to changes in the pre-cipitation isotope composition (Leng and Marshall 2004)However even the highest available temporal resolution ofsuch records (15 years per sample in sediments from LakeTibetanus Swedish Lapland Rosqvist et al 2007) is stillinsufficient to resolve inter-annual- to decadal-scale variabil-ity ie the timescales at which the North Atlantic Oscilla-tion (NAO) operates (Hurrell 1995 Trouet et al 2009) TheNAO steers weather and climate dynamics in northern Scan-dinavia and determines the origin of air masses from whichmeteoric waters form While sediment records are still of vi-tal importance for century-scale and millennial-scale varia-tions new approaches are needed for finer-scale resolutionon the 1ndash100-year timescale

One underutilized approach for hydroclimate reconstruc-tion is the use of freshwater mussels as natural stream waterstable isotope recorders (Dettman et al 1999 Kelemen et al2017 Pfister et al 2018 2019) Their shells can provide sea-sonally to annually resolved chronologically precisely con-strained records of environmental changes in the form ofvariable increment widths (which refers to the distance be-tween subsequent growth lines) and geochemical properties(eg Nystroumlm et al 1996 Schoumlne et al 2005a Geist et al2005 Black et al 2010 Schoumlne and Krause 2016 Geezaet al 2019 2020 Kelemen et al 2019) In particular sim-ilar to marine (Epstein et al 1953 Mook and Vogel 1968Killingley and Berger 1979) and other freshwater bivalves(Dettman et al 1999 Kaandorp et al 2003 Versteegh etal 2009 Kelemen et al 2017 Pfister et al 2019) Margar-itifera margaritifera forms its shell near equilibrium with theoxygen isotope composition of the ambient water (δ18Ow)(Pfister et al 2018 Schoumlne et al 2005a) If the fractiona-tion of oxygen isotopes between the water and shell carbon-ate is only temperature-dependent and the temperature dur-ing shell formation is known or can be otherwise estimated(eg from shell growth rate) reconstruction of the oxygenisotope signature of the water can be carried out from thatof the shell CaCO3 (δ18Os) Freshwater pearl mussels Mmargaritifera are particularly useful in this respect becausethey can reach a life span of 60 years (Pulteney 1781) or

80 years (von Hessling 1859) to over 200 years (Ziuganovet al 2000 Mutvei and Westermark 2001) offering an in-sight into long-term changes in freshwater ecosystems at anunprecedented temporal resolution

Here we present the first absolutely dated annually re-solved stable oxygen and carbon isotope record of freshwaterpearl mussels from three different streams in northern Swe-den covering nearly 2 centuries (1819ndash1998) We test theability of these freshwater pearl mussels to reconstruct theNAO index and associated changes in precipitation prove-nance using shell oxygen isotope data (δ18Os) We evalu-ate how shell oxygen isotope data compare to δ18O valuesin stream water (δ18Ow) and precipitation (δ18Op) as wellas limited existing environmental data (such as stream wa-ter temperature) we also evaluate how these variables relateto each other We hypothesize that δ18Ow and δ18Op valuesare positively correlated with one another as well as withthe North Atlantic Oscillation index In other words duringpositive NAO years oxygen isotope values in stream wa-ter and shells are higher than during negative NAO yearswhen shell and stream water oxygen isotope values tend tobe more depleted in 18O In addition we explore the phys-ical controls on shell stable carbon isotope signatures Pre-sumably these data reflect changes in the stable carbon iso-tope value of dissolved inorganic carbon which in turn issensitive to changes in primary production We leverage pastwork in Sweden that has shown that the main growing seasonof M margaritifera occurs from mid-May to mid-Octoberwith the fastest growth rates occurring between June and Au-gust (Dunca and Mutvei 2001 Dunca et al 2005 Schoumlne etal 2004a b 2005a) We use changes in the annual incre-ment width of M margaritifera to infer water temperaturebecause growth rates are faster during warm summers andresult in broader increment widths (Schoumlne et al 2004a b2005a) Because specimens of a given population react sim-ilarly to changes in temperature their average shell growthpatterns can be used to estimate climate and hydrologicalchanges Consequently increment series of specimens withoverlapping life spans can be crossdated and combined toform longer chronologies covering several centuries (Schoumlneet al 2004a b 2005a)

2 Material and methods

We collected 10 specimens of the freshwater pearl musselM margaritifera from one river and two creeks in Nor-rland (Norrbotten County) northern Sweden (Figs 1 2Table 1) Bivalves were collected between 1993 and 1999and included nine living specimens and one found dead andarticulated (bi-valved ED-GJ-D6R) Four individuals weretaken from the stream Nuortejaurbaumlcken (NJB) two fromstream Grundtraumlsktjaumlrnbaumlcken (GTB) and four from Goumlrjearingn(GJ) River (Fig 1) Because M margaritifera is an endan-gered species (Moorkens et al 2018) we refrained from col-

Hydrol Earth Syst Sci 24 673ndash696 2020 wwwhydrol-earth-syst-scinet246732020

B R Schoumlne et al Freshwater pearl mussels as long-term high-resolution stream water isotope recorders 675

lecting additional specimens that could have covered the timeinterval between the initial collection and the preparation ofthis paper instead we relied on bivalves that we obtained ndashwith permission ndash for a co-authorrsquos (AEM formerly knownas Elena Dunca) postdoctoral project and another co-authorrsquos(SMB) doctoral thesis

The bedrock in the catchments studied is dominated byorthogneiss and granodiorite The vegetation at GTB (ca90 m asl above sea level) and GJ (ca 200 m asl) consistedof a mixed birch forest whereas conifers shrubs and bushesdominated at NJB (ca 400 m asl) Thus the streams stud-ied were rich in humin acids The streams studied were fedby small upstream open (flow-through) lakes

21 Sample preparation

The soft tissues were removed immediately after collectionand shells were then air-dried One valve of each specimenwas wrapped in a protective layer of WIKO metal epoxyresin no 5 and mounted to a Plexiglas cube using GluetecMultipower plastic welder no 3 Shells were then cut perpen-dicular to the growth lines using a low-speed saw (BuehlerIsomet) equipped with a diamond-coated (low-diamond con-centration) wafering thin blade (400 microm thickness) Onespecimen (ED-NJB-A3R) was cut along the longest axiswhereas all of the others were cut along the height axisfrom the umbo to the ventral margin (Fig 2a) From eachspecimen two ca 3 mm thick shell slabs were obtained andmounted onto glass slides with the mirroring sides (the por-tions that were located to the left and right of the saw bladeduring the cutting process) facing upward This method fa-cilitated the temporal alignment of isotope data measured inone slab to growth patterns determined in the other shell slabThe shell slabs were ground on glass plates using suspen-sions of 800 and 1200 grit SiC powder and subsequently pol-ished with Al2O3 powder (grain size of 1 microm) on a BuehlerG-cloth Between each grinding step and after polishing theshell slabs were ultrasonically cleaned with water

22 Shell growth pattern analysis

For growth pattern analysis one polished shell slab was im-mersed in Mutveirsquos solution for 20 min at 37ndash40 C underconstant stirring (Schoumlne et al 2005b) After careful rins-ing in demineralized water the stained sections were air-dried under a fume hood Dyed thick-sections were thenviewed under a binocular microscope (Olympus SZX16) thatwas equipped with sectoral dark-field illumination (SchottVisiLED MC1000) and were photographed using a CanonEOS 600D camera (Fig 2b) The widths of the annual incre-ments were determined to the nearest ca 1 microm with imageprocessing software (Panopea copy Peinl and Schoumlne) Mea-surements were completed in the outer portion of the outershell layer (oOSL consisting of prismatic microstructure)from the boundary between the oOSL and the inner por-

tion of the outer shell layer (iOSL consisting of nacrous mi-crostructure) perpendicularly to the previous annual growthline (Fig 2c) Annual increment width chronologies weredetrended with stiff cubic spline functions and standardizedto produce dimensionless measures of growth (standardizedgrowth indices ie SGI values ndash σ ) following standard scle-rochronological methods (Helama et al 2006 Butler et al2013 Schoumlne 2013) Briefly for detrending measured an-nual increment widths were divided by the data predicted bythe cubic spline fit From each resulting growth index wesubtracted the mean of all growth indices and divided theresult by the standard deviation of all of the growth indicesof the respective bivalve specimen This transformation re-sulted in SGI chronologies Due to low heteroscedasticityno variance correction was needed (Frank et al 2007) Un-certainties in annual increment measurements resulted in aSGI error of plusmn006σ

23 Stable isotope analysis

The other polished shell slab of each specimen was usedfor stable isotope analysis To avoid contamination of theshell aragonite powders (Schoumlne et al 2017) the curedepoxy resin and the periostracum were completely removedprior to sampling A total of 1551 powder samples (32ndash128 microg) were obtained from the oOSL by means of mi-cromilling (Fig 2c) under a stereomicroscope at 160times mag-nification An equidistant sampling strategy was applied iethe milling step size was held constant within each annualincrement (Schoumlne et al 2005c) We used a cylindricaldiamond-coated drill bit (1 mm diameter KometGebr Bras-seler GmbH and Co KG model no 835 104 010) mountedon a Rexim Minimo drill While the drilling device was af-fixed to the microscope the sample was handheld duringsampling In early ontogenetic years up to 16 samples wereobtained between successive annual growth lines In the lat-est ontogenetic portions of specimens ED-NJB-A2R (the lastyear of life) and ED-GJ-D6R (the last 9 years of life) eachisotope sample represented 2ndash3 years

Stable carbon and oxygen isotopes were measured at theInstitute of Geosciences at the JW Goethe University ofFrankfurtMain (Germany) Carbonate powder samples weredigested in He-flushed borosilicate Exetainer vials at 72 Cusing a water-free phosphoric acid The released CO2 gaswas then measured in continuous-flow mode with a Ther-moFisher MAT 253 gas source isotope ratio mass spectrom-eter coupled to a GasBench II Stable isotope ratios were cor-rected against an NBS-19 calibrated Carrara marble (δ13C=+202 permil δ18O=minus176 permil) Results are expressed as partsper thousand (permil) relative to the Vienna Pee Dee Belem-nite (VPDB) scale The long-term accuracy based on blindlymeasured reference materials with known isotope compo-sition is better than 005 permil for both isotope systems Notethat no correction was applied for differences in fractionationfactors of the reference material (calcite) and shells (arag-

wwwhydrol-earth-syst-scinet246732020 Hydrol Earth Syst Sci 24 673ndash696 2020


Figure 1 Maps showing the sample sites in northern Sweden (a) Topographic map of Scandinavia (b) An enlargement of the red boxin panel (a) showing Norrbotten County (yellow) a province in northern Sweden and localities where bivalve shells (Margaritifera mar-garitifera) were collected and isotopes in rivers and precipitation were measured The shell collection sites (filled circles) are coded asfollows NJB represents the Nuortejaurbaumlcken GTB represents the Grundtraumlsktjaumlrnbaumlcken and GJ represents Goumlrjearingn River Sk representsthe Skellefte River (near Slagnaumls) a GNIR site The GNIP sites of Racksund and Arjeplog are represented by Rs and Ap respectivelyThe base map in panel (a) is sourced from TUBS and used under a Creative Commons license httpscommonswikimediaorgwikiFileSweden_in_Europe_(relief)svg (last access 5 February 2020) The base map in panel (b) is sourced from Erik Frohne (redrawn by Sil-verkey) and used under a Creative Commons license httpscommonswikimediaorgwikiFileSweden_Norrbotten_location_mapsvg (lastaccess 5 February 2020)

Table 1 Shell of M margaritifera from three streams in northern Sweden used in the present study for isotope and growth pattern analysisThe last hyphenated section of the specimen ID represents whether bivalves were collected alive (A) or dead (D) the specimen number andwhich valve was used (R denotes right and L denotes left)

Stream name Specimen ID Coordinates and Agea Alive during years (CE) No isotope sampleselevation (years) (CE) (coverage of yearsc)

Nuortejaurbaumlcken ED-NJB-A6R 6542prime1322primeprime N 22 1972ndash1993 175 (1ndash22b)ED-NJB-A4R 01902prime3101primeprime E 27 1967ndash1993 154 (2ndash27b)ED-NJB-A2R ca 400 m asl 48 1946ndash1993 78 (2ndash48b)ED-NJB-A3R 24 1970ndash1993 50 (1ndash24b)

Grundtraumlsktjaumlrnbaumlcken ED-GTB-A1R 6602prime5998primeprime N 51 1943ndash1993 368 (2ndash49)ED-GTB-A2R 02205prime0225primeprime E 51 1943ndash1993 315 (3ndash49)

ca 90 m asl

Goumlrjearingn ED-GJ-A1L 6620prime3077primeprime N 80 1916ndash1997 56 (25ndash80b)ED-GJ-A2R 02030prime1502primeprime E 82 1918ndash1997 76 (1ndash78)ED-GJ-A3L ca 200 m asl 123 1875ndash1997 110 (29ndash122)ED-GJ-D5L 181 1819ndash1999 169 (1ndash180)

a Minimum estimate of life span b Last sampled year incomplete c Add 10 years to these values to obtain approximate ontogenetic years



Figure 2 Sclerochronological analysis of Margaritifera margaritifera (a) The left valve of a freshwater pearl mussel The cutting axis isindicated by a white line Note the erosion in the umbonal shell portion (b) A Mutvei-immersed shell slab showing the outer and inner shelllayers (OSL and ISL respectively) separated by the myostracum (white line) The OSL is further subdivided into an outer and inner portion(oOSL and iOSL respectively) The ISL and iOSL consist of a nacreous microstructure and the oOSL consists of a prismatic microstructure(c) An enlargement of panel (b) shows the annual growth patterns The annual increment width measurements (yellow) were completed asperpendiculars from the intersection of the oOSL and iOSL toward the next annual growth line The semitransparent red and orange boxesschematically illustrate the micromilling sampling technique

onite verified by Raman spectroscopy) because the pale-othermometry equation used below (Eq 2) also did not con-sider these differences (Fuumlllenbach et al 2015) Howeverthe correction of minus038 permil would be required if δ18O val-ues of shells and other carbonates were compared with eachother

24 Instrumental data sets

Shell growth and isotope data were compared to a set ofenvironmental variables including the station-based winter(DJFM) NAO index (obtained from httpsclimatedataguideucaredu last access 9 April 2019) as well as oxygen iso-tope values of river water (δ18Ow) and weighted (correctedfor precipitation amounts) oxygen isotope values of precip-itation (δ18Op) Data on monthly river water and precipita-tion were sourced from the Global Network of Isotopes inPrecipitation (GNIP) and the Global Network of Isotopesin Rivers (GNIR) available at the International Atomic En-ergy Agencyhttpsnucleusiaeaorgwiserindexaspx (lastaccess 1 April 2019) Furthermore monthly air temperature(Ta) data came from the station Stensele and are availableat the Swedish Meteorological and Hydrological Institute

httpswwwsmhise (last access 5 February 2020) Fromthese data the monthly stream water temperature (Tw) wascomputed using the summer airndashstream water temperatureconversion by Schoumlne et al (2004a) and was supplementedby the standard errors of the slope and intercept

Tw = 088plusmn 005times Taminus 086plusmn 049 (1)

25 Weighted annual shell isotope data

Because the shell growth rate varied during the growing sea-son ndash with the fastest biomineralization rates occurring dur-ing June and July (Dunca et al 2005) ndash the annual growthincrements are biased toward summer and powder samplestaken from the shells at equidistant intervals represent dif-ferent amounts of time To compute growing season av-erages (henceforth referred to as ldquoannual averagesrdquo) fromsuch intra-annual shell isotope data (δ18Os δ13Cs) weighted(henceforth denoted with an asterisk) annual means are thusneeded ie δ18Olowasts and δ13Clowasts values (Schoumlne et al 2004a)The relative proportion of time of the growing season rep-resented by each isotope sample was computed from a pre-viously published intra-annual growth curve of juvenile Mmargaritifera from Sweden (Dunca et al 2005) For exam-



ple if four isotope samples were taken between two winterlines at equidistant intervals the first sample would represent2238 of the time of the main growing season durationand the second third and fourth would represent 2028 2447 and 3287 of the time of the main growing sea-son respectively (Table 2) Accordingly the weighted annualmean isotope values (δ18Olowasts δ13Clowasts ) were calculated by mul-tiplying these numbers (weights) by the respective δ18Os andδ13Cs values and dividing the sum of the products by 100 (seeSupplement) The four isotope samples from the exampleabove comprise the time intervals from 23 May to 22 June23 June to 21 July 22 July to 25 August and 26 Augustto 12 October respectively Missing isotope data due to lostpowder machine error air in the Exetainer etc were filled inusing linear interpolation in 20 instances We assumed thatthe timing and rate of seasonal growth remained nearly un-changed throughout the lifetime of the specimens and in thestudy region (see also Sect 4)

26 Reconstruction of oxygen isotope signatures ofstream water on annual and intra-annualtimescales

To assess how well the shells recorded δ18Ow values oninter-annual timescales the stable oxygen isotope signatureof stream water (δ18Olowastwr) during the main growing season(ldquoannualrdquo δ18Olowastwr) was reconstructed from δ18Olowasts data andthe arithmetic average of (monthly) stream water temper-atures Tw during the same time interval ie 23 Mayndash12 October Using this approach the effect of temperature-dependent oxygen isotope fractionation was removed fromthe δ18Olowasts data For this purpose the paleothermometry equa-tion of Grossman and Ku (1986 corrected for the VPDBndashVSMOW scale difference following Gonfiantini et al 1995)was solved for δ18Olowastwr Eq (2)

δ18Olowastwr =1943minus 434times δ18Olowasts minus Tw

minus434 (2)

Because air temperature data were only available from 1860onward Tw values prior to that time were inferred from age-detrended and standardized annual growth increment data(SGI values) using a linear regression model similar to thatintroduced by Schoumlne et al (2004a) In the revised modelSGI data of 25 shells from northern Sweden (15 publishedchronologies provided in the article cited above and 10 newchronologies from the specimens studied in the present pa-per) were arithmetically averaged for each year and then re-gressed against weighted annual water temperature hereafterreferred to as annual T lowastw The annual T lowastw data consider vari-ations in the seasonal shell growth rate A total of 629 2549 2452 2192 1688 and 490 of the an-nual growth increment was formed in each month betweenMay and October respectively The values were multipliedby Tw of the corresponding month and the sum of the prod-ucts was divided by 100 to obtain the annual T lowastw data The

revised (shell growth vs temperature) model is as follows

T lowastw = 145plusmn 019timesSGI+ 842plusmn 008 (3)

For coherency purposes we also applied this model to post-1859 SGI values and computed stream water temperaturesthat were subsequently used to estimate δ18Olowastwr(SGI) values

To assess how well the shells recorded δ18Ow values atintra-annual timescales we focused on two shells from NJB(ED-NJB-A4R and ED-NJB-A6R) which provided the high-est isotope resolution of 1ndash2 weeks per sample during the fewyears of overlap between the GNIP and GNIR data Note that(only for this bivalve sampling locality) monthly instrumen-tal oxygen isotope data were available from the GNIP andGNIR data sets (data by Burgman et al 1981) The δ18Owdata were measured in the Skellefte River near Slagnaumls ca40 km SW of NJB (6534prime5950primeprime N 01810prime3912primeprime E) andcovered the time interval from 1973 to 1980 The δ18Op datacame from Racksund (6602prime6000primeprime N 01737prime6000primeprime E ca75 km NW of NJB) and covered the time interval from 1975to 1979 Because precipitation amounts were not availablefrom Racksund we computed average monthly precipitationamounts from data recorded at Arjeplog (6602prime6000primeprime N01753prime6000primeprime E) from 1961 to 1967 (see Supplement) Ar-jeplog is located ca 65 km NW of NJB and ca 12 km W ofRacksund Equation (2) was used to calculate δ18Olowastwr valuesfrom individual δ18Olowasts data and water temperature that ex-isted during the time when the respective shell portion wasformed Intra-annual water temperatures were computed asweighted averages T lowastw from monthly Tw considering sea-sonal changes in the shell growth rate For example if fourpowder samples were taken from the shell at equidistant in-tervals within one annual increment 629 of the first sam-ple was formed in May and 1863 was formed in June (sumca 25 ) The average temperature during that time intervalis computed using these numbers as follows (Tw of May times00629+ Tw of June times 01863)25 A total of 686 of thesecond sample from that annual increment formed in Juneand 1797 formed in July Accordingly the average tem-perature was (Tw of June times00686+Tw of July times1747)25Note that annual δ18Olowastwr values can also be computed fromintra-annual δ18Olowastwr data but this approach is much moretime-consuming and complex than the method described fur-ther above However both methods produce nearly identicalresults (see Supplement)

27 Stable carbon isotopes of the shells

Besides the winter and summer NAO index weighted an-nual stable carbon isotope data of the shells δ13Clowasts valueswere compared to shell growth data (SGI chronologies) Be-cause the δ13Clowasts values could potentially be influenced by on-togenetic effects the chronologies were detrended and stan-dardized (δ13Clowasts(d)) following methods typically used to re-move ontogenetic age trends from annual increment width



Table 2 Weights for isotope samples of Margaritifera margaritifera Due to variations in the seasonal shell growth rate each isotope sampletaken at equidistant intervals represents different amounts of time To calculate seasonal or annual averages from individual isotope data therelative proportion of time of the growing season contained in each sample must be considered when weighted averages are computed Theduration of the growing season comprises 143 d and covers the time interval from 23 May to 12 October

Number of isotope Weight of nth isotope sample () within an annual increment direction of growth to the right (increasing numbers)

samples per annual 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th 11th 12th 13th 14th 15th 16thincrement

1 100002 4266 57343 2797 3147 40564 2238 2028 2447 32875 1818 1539 1888 2027 27286 1538 1259 1469 1678 1818 22387 1329 1188 1119 1329 1399 1608 20288 1159 1079 909 1119 1258 1189 1469 18189 1049 979 769 909 1049 1189 1049 1329 167810 979 839 769 770 909 978 980 1049 1189 153811 909 769 770 559 769 839 979 840 1049 1048 146912 839 699 700 559 699 770 839 839 769 1049 909 132913 769 630 699 559 560 629 769 840 699 769 979 839 125914 769 560 629 559 490 629 630 699 699 770 628 910 839 118915 629 630 559 560 419 560 559 699 630 699 629 700 838 770 111916 629 560 559 490 419 490 559 560 629 490 769 559 770 699 769 1049

chronologies (see eg Schoumlne 2013) Detrending was car-ried out with cubic spline functions capable of removing anydirected trend toward higher or lower values throughout thelifetime

3 Results

The lengths of the annual increment chronologies of M mar-garitifera from the three streams studied (the Nuortejaur-baumlcken Grundtraumlsktjaumlrnbaumlcken and Goumlrjearingn) ranged from21 to 181 years and covered the time interval from 1819to 1999 CE (Table 1) Because the umbonal shell portionswere deeply corroded and the outer shell layer was missingndash a typical feature of long-lived freshwater bivalves (Schoumlneet al 2004a Fig 2a) ndash the actual ontogenetic ages of thespecimens could not be determined and may have been up to10 years higher than the ages listed in Table 1

31 Shell growth and temperature

The 10 new SGI series from NJB GTB and GJ were com-bined with 15 published annual increment series of M mar-garitifera from the Paumlrlaumllven Paumlrlskalsbaumlcken and Boumlls-manaringn streams (Schoumlne et al 2004a b 2005a) to form arevised Norrland master chronology During the 50-year cal-ibration interval from 1926 to 1975 (the same time intervalwas used in the previous study by Schoumlne et al 2004a b2005a) the chronology was significantly (p lt 005 noteall p values of linear regression analyses in this paper areBonferroni-adjusted) and positively correlated (R = 074R2= 055) with the weighted annual stream water temper-

ature (T lowastw) during the main growing season (Fig 3) These

values were similar to the previously published coefficientof determination for a stacked record using M margaritiferaspecimens from streams across Sweden (R2

= 060 Schoumlneet al 2005a note that this number is for SGI vs an arith-metic annual Tw a regression of SGI against weighted an-nual Tw returns an R2 of 064)

32 Shell stable oxygen isotope data

The shell oxygen isotope curves showed distinct seasonaland inter-annual variations (Figs 4 5) The former were par-ticularly well developed in specimens from GTB and NJB(Fig 4) which were sampled with a very high spatial resolu-tion of ca 30 microm (ED-GTB-A1R ED-GTB-A2R ED-NJB-A4R and ED-NJB-A6R) In these shells up to 16 sampleswere obtained from single annual increments translating intoa temporal resolution of 1ndash2 weeks per sample Typicallythe highest δ18Os values of each cycle occurred at the winterlines and the lowest values occurred about half way betweenconsecutive winter lines (Fig 4) The largest seasonal δ18Osamplitudes of ca 220 permil were measured in specimens fromGTB (minus868 permil tominus1091 permil) and ca 170 permil was measuredin shells from NJB (minus863 permil to minus1031 permil)

Weighted annual shell oxygen isotope (δ18Olowasts ) values fluc-tuated on decadal timescales (common period of ca 8 years)with amplitudes larger than those occurring on seasonalscales ie ca 250 permil and 300 permil in shells from NJB(minus863 permil to minus1110 permil) and GTB (minus784 permil to minus1085 permil)respectively (Fig 5a b) The chronologies from GJ also re-vealed a century-scale variation with minima in the 1820sand 1960s and maxima in the 1880s and 1990s (Fig 5c) Theδ18Olowasts curves of specimens from the same locality showednotable agreement in terms of absolute values and visual



Figure 3 (a) Time series and (b) cross-plot of the age-detrended and standardized annual shell growth rate (SGI values) and water temper-ature during the main growing season (23 Mayndash12 October) Water temperatures were computed from monthly air temperature data using apublished transfer function and considering seasonally varying rates of shell growth The gray box in panel (a) denotes the 50-year calibrationinterval from which the temperature model (b) was constructed As seen from the cross-plot in panel (b) 55 of the variation in annualshell growth was highly significantly explained by water temperature Higher temperature resulted in faster shell growth

agreement (running similarity) specifically specimens fromNJB and GTB (Fig 5a b) However the longest chronologyfrom GJ only showed slight agreement with the remainingthree series from that site (Fig 5c) The similarity amongthe series also changed through time (Fig 5a b c) In someyears the difference between the series was less than 020 permilat NJB (N = 4) and GTB (N = 2 1983) and 010 permil at GJ(N = 4 1953) whereas in other years the differences variedby up to 082 permil at NJB and 100 permil at GTB and GJ Averageshell oxygen isotope chronologies of the three streams stud-ied exhibited a strong running similarity (passed the ldquoGleich-laumlufigkeitstestrdquo by Baillie and Pilcher 1973 for p lt 0001)and were significantly positively correlated with each other(the R2 value of NJB vs GTB was 034 NJB vs GJ was040 and GTB vs GJ was 036 ndash all at p lt 00001)

33 Shell stable oxygen isotope data and instrumentalrecords

At NJB ndash the only bivalve sampling site for which measuredstream water isotope data were available from nearby locali-ties ndash the MayndashOctober ranges of reconstructed and instru-mental stream water δ18O values between 1973 and 1980(excluding 1977 due to missing δ18Ow data) were in closeagreement (shells were 283 and 319 permil vs stream waterwhich was 320 permil Fig 6a) During the same time intervalarithmetic means plusmn 1 standard deviation of the shells wereminus1248plusmn 074 permil (ED-NJB-A6R N = 79) and minus1245plusmn066 permil (ED-NJB-A4R N = 44) whereas the stream watervalue wasminus1233plusmn076 permil (Skellefte RiverN = 42) Whencomputed from growing season averages (N = 7) shell val-ues were minus1248plusmn 029 permil and minus1242plusmn 034 permil respec-tively and the stream water value wasminus1230plusmn032 permil Ac-

cording to nonparametric t tests these data sets are statisti-cally indistinguishable Furthermore the inter-annual trendsof δ18Olowastwr and δ18Ow values were similar (Fig 6a) val-ues declined by ca 100 permil between 1973 and 1977 fol-lowed by a slight increase of ca 050 permil until 1980 In con-trast to the damped stream water signal (the average sea-sonal range during the 4 years ndash 1975 1976 1978 and1979 ndash for which both stream water and precipitation datawere available wasminus150plusmn057 permil) δ18Op values exhibitedmuch stronger fluctuations at the seasonal scale (on aver-age minus937plusmn 281 permil extreme monthly values of minus421 permiland minus1760 permil N = 46 station Racksund Fig 6b) andon inter-annual timescales (unweighted annual averages ofminus1141 permil to 1368 permil weighted DecemberndashSeptember av-erages of minus954 permil to 1316 permil)

Despite the limited number of instrumental data season-ally averaged δ18Olowastwr data showed some ndash although not al-ways statistically significant ndash agreement with δ18Ow andweighted δ18Op data (corrected for precipitation amounts)respectively both in terms of correlation coefficients and ab-solute values (Table 3) These findings were corroborated bythe regression analyses of instrumental δ18Op values againstδ18Ow values (Table 3) For example the oxygen isotope val-ues of summer (JunendashSeptember) precipitation were signif-icantly (Bonferroni-adjusted p lt 005) and positively corre-lated with those of shell carbonate precipitated during thesame time interval (98 of the variability was explainedin both specimens but only at p lt 005 in ED-NJB-A6R)Likewise δ18Ow and δ18Op values during summer werepositively correlated with each other (R = 091) althoughless significantly (p = 0546) Strong relationships werealso found for δ18Olowastwr and δ18Ow values during the maingrowing season as well as annual δ18Olowastwr and Decemberndash



Table 3 Relationship between the stable oxygen isotope values in precipitation (amount-corrected δ18Op) river water and shells of Margar-itifera margaritifera from Nuortejaurbaumlcken during different portions of the year (during the 4 years for which data from shells water andprecipitation were available 1975 1976 1978 and 1979 hence N = 4) The arithmetic mean δ18O values for each portion of the year arealso given The rationale behind the comparison of δ18O values of winter precipitation and spring (MayndashJune) river water or shell carbonateis that the isotope signature of meltwater may have left a signal in the water Statistically significant values (Bonferroni-adjusted p lt 005)are marked in bold Isotope values next to months represent multiyear averages

δ18Op (Racksund) δ18Ow (Skellefte River)

Season Dectminus1 to Sept Jun to Sep Dectminus1 to Febt May to Oct Jun to Sep May to Juneminus1139 permil minus1098 permil minus1418 permil minus1246 permil minus1239 permil minus1308 permil

δ18OwSkellefte River

MayndashOctminus1246 permil

R = 100R2 = 100p = 0006

JunndashSepminus1239 permil

R = 091R2= 083

p = 0546MayndashJunminus1308 permil

R = 095R2= 090

p = 1000

δ18OlowastwrED-NJB-A6R


R = 098R2= 096

p = 0134

R = 099R2= 097

p = 0065JunndashSepminus1244 permil

R = 099R2 = 098p = 0045

R = 086R2= 075


R = 046R2= 021

p = 1000

R = 064R2= 041

p = 1000



R = 099R2 = 098p = 0035

R = 099R2 = 098p = 0034


R = 099R2= 098

p = 0070

R = 095R2= 091


R = 076R2= 058

p = 1000

R = 089R2= 080

p = 0484

September δ18Op values The underlying assumption for thelatter was that the δ18Olowastwr average value reflects the com-bined δ18Op of snow precipitated during the last winter (re-ceived as meltwater during spring) and rain precipitated dur-ing summer Instrumental data supported this hypothesis be-cause stream water δ18O values during the main growingseason were highly significantly and positively correlatedwith DecemberndashSeptember δ18Op data (Table 3) Converselychanges in the isotope signal of winter (DecemberndashFebruary)snow were only weakly and not significantly mirrored bychanges in stream water oxygen isotope values during thesnowmelt period (MayndashJune) or in δ18Olowastwr values from shellportions formed during the same time interval (Table 3) Dur-ing the 4 years under study (1975 1976 1978 and 1979)measured and reconstructed δ18Ow values were nearly iden-tical during the main growing season (δ18Ow of minus1246 permil

δ18Olowastwr of minus1257 permil and minus1246 permil) and during summer(δ18Ow of minus1239 permil δ18Olowastwr of minus1244 permil and minus1243 permil)(Table 3) In contrast isotopes in precipitation and river wa-ter showed larger discrepancies (see the text above Fig 6band Table 3)

34 Shell stable oxygen isotope data and synopticcirculation patterns (NAO)

Site-specific annual δ18Olowastwr (and δ18Olowastwr(SGI)) chronolo-gies (computed as arithmetic averages of all chronologiesat a given stream) were significantly (Bonferroni-adjustedplt 005) positively correlated with the NAO indices (Fig 7Table 4) In NAO+ years the δ18Olowastwr (and δ18Olowastwr(SGI)) val-ues were higher than during NAOminus years The strongest cor-relation existed between the winter (DecemberndashMarch) NAO



Figure 4 Shell stable oxygen and carbon isotope chronologiesfrom four specimens of Margaritifera margaritifera from Nuorte-jaurbaumlcken and Grundtraumlsktjaumlrnbaumlcken that were sampled with veryhigh spatial resolution and from which the majority of the isotopedata were obtained (Table 1) (a) ED-NJB-A6R (b) ED-NJB-A4R(c) ED-GTB-A1R and (d) ED-GTB-A2R Individual isotope sam-ples represent time intervals of a little as 6 d to 2 weeks in ontoge-netically young shell portions and up to one full growing season inthe last few years of life Red vertical lines represent annual growthlines Because the umbonal shell portions are corroded the exactontogenetic age at which the chronologies start cannot be providedAssuming that the first 10 years of life are missing sampling inpanel (a) started in year 11 in panels (b) and (c) in year 12 and inpanel (d) in year 13 (see also Table 1)

and δ18Olowastwr (and δ18Olowastwr(SGI)) values at NJB (44 to 49 of the variability is explained) At GTB the amount of vari-ability explained ranged between 24 and 27 whereasat GJ only 16 to 18 of the inter-annual δ18Olowastwr (andδ18Olowastwr(SGI)) variability was explained by the winter NAO

(wNAO) index Between 1947 and 1991 (the time interval forwhich isotope data were available for all sites) the R2 val-ues were more similar to each other and ranged between 027and 046 (Table 4) All sites reflected well-known features ofthe instrumental NAO index series such as the recent (1970ndash2000) positive shift toward a more dominant wNAO whichdelivered isotopically more positive (less depleted in 18O)winter precipitation to our region of interest (Fig 7a b c)The correlation between δ18Olowastwr (and δ18Olowastwr(SGI)) values andthe summer (JunendashAugust) NAO index was much lower thanfor the wNAO but likewise positive and sometimes signifi-cant at p lt 005 (Table 4) Between 1947 and 1991 7 to43 of the inter-annual oxygen isotope variability was ex-plained by the summer NAO index

We have also computed an average δ18Olowastwr(SGI) curve forthe entire study region (Fig 8a b c) Because the level(absolute values) of the three streams differed from eachother (average δ18Olowastwr values of NJB GTB and GJ from1947 to 1992 were minus1251 permil minus1221 permil and minus1416 permilrespectively) the site-specific series were standardized andthen arithmetically averaged The resulting chronologyδ18Olowastwr(Norrland) was strongly positively and statistically sig-nificantly (Bonferroni-adjusted p value below 005) corre-lated with the wNAO index (56 of the variability ex-plained Fig 8a) Despite the limited instrumental data setδ18O values of river water and precipitation were stronglypositively correlated with the wNAO index (R2 values of072 and 084 respectively Fig 8d e) but the Bonferroni-adjusted p values exceeded 005 (note the uncorrected p val-ues were 007 and 003 respectively)

35 Shell stable carbon isotope data

Shell stable carbon isotope (δ13Cs) data showed less distinctseasonal variations than δ18Os values but the highest valueswere also often associated with the winter lines and the low-est values occurred between subsequent winter lines (Fig 4)The largest seasonal amplitudes of ca 390 permil were observedin specimens from NJB (minus821 permil tominus1210 permil) and ca 1 permilsmaller ranges at GTB (minus1097 permil to minus1388 permil)

Weighted annual δ13Clowasts curves varied greatly from eachother in terms of change throughout the lifetime of theorganism among localities and even at the same locality(Fig 5d e f) Note that all curves started in early ontogeny(below the age of 10) except for ED-GJ-A1L and ED-GJ-A3L that began at a minimum age of 25 and 29 respectively(Table 1) Whereas two specimens from NJB (ED-NJB-A6R and ED-NJB-A4R) showed strong ontogenetic δ13Clowaststrends from ca minus870 permil to minus1250 permil weaker trends to-ward more negative values were observed in ED-NJB-A2R(ca minus1000 permil to minus1170 permil) and shells from GTB (caminus1150 permil to minus1300 permil) Opposite ontogenetic trends oc-curred in ED-GJ-A1L and ED-GJ-A2R (ca minus1500 permil tominus1200 permil) but no trends at all were found in ED-NJB-A3R ED-GJ-A3L and ED-GJ-D6R (fluctuations around



Figure 5 Annual shell stable oxygen and carbon isotope chronologies of the specimens of Margaritifera margaritifera studied Data werecomputed as weighted averages from intra-annual isotope data ie growth rate-related variations were taken into consideration Panels(a) (d) and (g) represent the stream Nuortejaurbaumlcken panels (b) (e) and (h) represent the stream Grundtraumlsktjaumlrnbaumlcken and panels (c) (f)and (i) represent Goumlrjearingn River (andashc) Oxygen isotopes (dndashf) carbon isotopes and (gndashi) detrended and standardized carbon isotope valuesare also shown

Table 4 Site-specific annual isotope chronologies of Margaritifera margaritifera shells linearly regressed against winter and summer NAO(wNAO and sNAO respectively) as well as the detrended and standardized shell growth rate (SGI) δ18Olowastwr data were computed from shelloxygen isotope data and temperature data were computed from instrumental air temperatures whereas in the case of δ18Olowastwr(SGI) datatemperatures were estimated from a growth-temperature model See text for details Statistically significant values (Bonferroni-adjustedp lt 005) are marked in bold

δ18Olowastwr δ18Olowastwr(SGI) δ13Clowasts(d)

NJB GTB GJ NJB GTB GJ NJB GTB GJ

wNAO(DJFM)

R = 067R2 = 044p lt 00001

R = 049R2 = 024p = 00011

R = 039R2 = 016p lt 00001

R = 070R2 = 049p lt 00001

R = 052R2 = 027p = 00005

R = 042R2 = 018p lt 00001

R =minus018R2= 003

p = 10000

R =minus031R2= 010

p = 01911

R =minus010R2= 001

p = 10000

wNAO(DJFM)1947ndash1991

R = 065R2 = 043p lt 00001

R = 052R2 = 027p = 00008

R = 060R2 = 036p lt 00001

R = 068R2 = 046p lt 00001

R = 056R2 = 031p = 00002

R = 065R2 = 042p lt 00001

R =minus017R2= 003

p = 10000

R =minus030R2= 009

p = 02657

R = 014R2= 002

p = 10000

sNAO (JJA) R = 038R2 = 014p = 00293

R = 040R2 = 016p = 00138

R = 020R2= 004

p = 00704

R = 029R2= 009

p = 01451

R = 034R2= 011

p = 00593

R = 002R2= 000

p = 10000

R = 012R2= 001

p = 10000

R = 001R2= 000

p = 10000

R = 004R2= 000

p = 10000

sNAO (JJA)1947ndash1991

R = 065R2 = 043p lt 00001

R = 040R2 = 016p = 00212

R = 038R2 = 014p = 00333

R = 027R2= 007

p = 02172

R = 032R2= 010

p = 00985

R = 026R2= 007

p = 02581

R = 013R2= 002

p = 10000

R = 010R2= 001

p = 10000

R = 015R2= 002

p = 10000

SGI R =minus028R2= 008

p = 03812

R =minus023R2= 005

p = 06938

R = 008R2= 001

p = 10000

SGI1947ndash1991

R =minus027R2= 007

p = 04202

R =minus022R2= 005

p = 09238

R = 010R2= 001

p = 10000



Figure 6 Intra-annual stable oxygen isotope values (1973ndash1980)(a) Monthly isotopes measured in the Skellefte River (MayndashOctober) and weighted seasonal averages (δ18Olowastwr) of two shells(Margaritifera margaritifera) from Nuortejaurbaumlcken (see Fig 1)According to nonparametric t tests instrumental and reconstructedoxygen isotope data are statistically indistinguishable Also notethat inter-annual changes are nearly identical (b) Comparison ofmonthly oxygen isotope data in stream water (Skellefte River MayndashOctober) and precipitation (Racksund whole year)

minus1200 permil) All curves were also overlain by some decadalvariability (typical periods of 3ndash6 13ndash16 and 60ndash80 years)Even after detrending and standardization (Fig 5g h i) nostatistically significant correlation at p lt 005 was found be-tween the average δ13Clowasts(d) curves of the three sites (NJBndashGTB R =minus011 R2

= 001 NJBndashGJ R =minus017 R2=

003 GTBndashGJ R = 010 R2= 001) However at each

site individual curves revealed reasonable visual agreementspecifically at NJB and GTB (Fig 5g h) At GJ the agree-ment was largely limited to the low-frequency oscillations(Fig 5i)

The detrended and standardized annual shell stable carbonisotope (δ13Cs(d)) curves showed no statistically significant(Bonferroni-adjusted p lt 005) agreement with the NAO in-dices or shell growth rate (SGI values) (Fig 7 Table 4) A

weak negative correlation (10 explained variability) onlyexisted between δ13Clowasts(d) values and the wNAO at NJB Somevisual agreement was apparent between δ13Cs(d) values andSGI in the low-frequency realm For example at NJB fastergrowth during the mid-1950s 1970s 1980s and 1990s felltogether with lower δ13Cs(d) values (Fig 7g) Likewise atGTB faster shell growth seemed to be inversely linked toδ13Cs(d) values (Fig 7h)

4 Discussion

41 Advantages and disadvantages of using bivalveshells for stream water δ18O reconstructioncomparison with sedimentary archives

Our results have shown that shells of freshwater pearl mus-sels from streams in northern Scandinavia (fed predomi-nantly by small open lakes and precipitation) can serveas a long-term high-resolution archive of the stable oxy-gen isotope signature of the water in which they lived Be-cause δ18Ow values have a much lower seasonal amplitudethan δ18Op values (ie δ18Ow signals are damped relativeto δ18Op data as a result of the water transit times throughthe catchment of the stream) the observed and reconstructedstream water isotope signals mirror the seasonal and inter-annual variability in the δ18Op values The NAO and subse-quent atmospheric circulation patterns determine the originof air masses and subsequently the δ18O signal in precipita-tion

Compared with lake sediments which have traditionallybeen used for similar reconstructions at nearby localities(eg Hammarlund et al 2002 Andersson et al 2010Rosqvist et al 2004 2013) this new shell-based archive hasa number of advantages

The effect of temperature-dependent oxygen isotope frac-tionation can be removed from δ18Os values so that the sta-ble oxygen isotope signature of the water in which the bi-valves lived can be computed This is possible by solving thepaleothermometry equation of Grossman and Ku (1986) forδ18Olowastwr (Eq 2) and computing the oxygen isotope values ofthe water from those of the shells and stream water temper-ature The stream water temperature during shell growth canbe reconstructed from shell growth rate data (Eq 3 Schoumlneet al 2004a b 2005a) or the instrumental air temperature(Eq 1 Morrill et al 2005 Chen and Fang 2015) Howeversimilar studies in which the oxygen isotope composition ofmicrofossils or authigenic carbonate obtained from lake sed-iments were used to infer the oxygen isotope value of thewater merely relied on estimates of the temperature variabil-ity during the formation of the diatoms ostracods and abio-genic carbonates among others as well as how these temper-ature changes affected reconstructions of δ18Ow values (egRosqvist et al 2013) In such studies it was impossible toreconstruct the actual water temperatures from other proxy



Figure 7 Site-specific weighted annual δ18Olowastwr (andashf) and δ13Clowasts(d) (gndashi) curves of Margaritifera margaritifera compared to the winter (andashc)and summer (dndashf) North Atlantic Oscillation indices as well as the detrended and standardized shell growth rate (gndashi) Panels (a) (d) and (g)show Nuortejaurbaumlcken panels (b) (e) and (h) show Grundtraumlsktjaumlrnbaumlcken and panels (c) (f) and (i) show Goumlrjearingn

archives Moreover at least in some of these archives suchas diatoms the effect of temperature on the fractionation ofoxygen isotopes between the skeleton and the ambient wateris still debated (Leng 2006)

M margaritifera precipitates its shell near oxygen isotopeequilibrium with the ambient water and shell δ18O valuesreflect stream water δ18O data This may not be the case in allof the archives that have previously been used For exampleostracods possibly exhibit vital effects (Leng and Marshall2004)

The shells can provide seasonally to inter-annually re-solved data In the present study each sample typically rep-resented as little as 1 week up to one full growing season(1 ldquoyearrdquo mid-May to mid-October Dunca et al 2005) Invery slow growing shell portions of ontogenetically old spec-imens individual samples occasionally covered 2 or in ex-ceptional cases 3 years of growth which resulted in a reduc-tion of variance If required a refined sampling strategy andcomputer-controlled micromilling could ensure that time-averaging consistently remains below 1 year Such high-resolution isotope data can be used for a more detailed anal-ysis of changes in the precipitationndashrunoff transformationacross different seasons Furthermore the specific samplingmethod based on micromilling produced uninterrupted iso-tope chronologies ie no shell portion of the outer shelllayer remained un-sampled Due to the high temporal reso-lution bivalve shell-based isotope chronologies can provideinsights into inter-annual- and decadal-scale paleoclimatic

variability With the new precisely calendar-aligned data itbecomes possible to test hypotheses brought forward in pre-vious studies according to which δ18O signatures of meteoricwater are controlled by the winter andor summer NAO (egRosqvist et al 2007 2013)

Each sample taken from the shells can be placed in a pre-cise temporal context The very season and exact calendaryear during which the respective shell portion formed canbe determined in shells of specimens with known dates ofdeath based on the seasonal growth curve and annual incre-ment counts Existing studies suffer from the disadvantagethat time cannot be precisely constrained neither at seasonalnor annual timescales (unless varved sediments are avail-able) However isotope results can be biased toward a par-ticular season of the year or a specific years within a decadeSuch biases can be avoided with sub-annual data provided bybivalve shells

In summary bivalve shells can provide uninterruptedseasonally to annually resolved precisely temporally con-strained records of past stream water isotope data that enablea direct comparison with climate indices and instrumentalenvironmental data In contrast to bivalve shells sedimentaryarchives come with a much coarser temporal resolution Eachsample taken from sediments typically represents the averageof several years and the specific season and calendar yearduring which the ostracods diatoms authigenic carbonatesetc grew remains unknown Conversely the time intervalscovered by sedimentary archives are much larger and can re-



Figure 8 Oxygen isotope data compared to the winter NAO index (a) Standardized δ18Olowastwr chronology of the study region compared to thewinter NAO index between 1950 and 1998 (b) Same as in panel (a) but only when the East Atlantic Pattern (EA) index has the same sign(EQ) as the winter NAO (c) Same as in panel (a) but only for cases when the EA index is in the opposite (OP) mode to the winter NAO(d) δ18Ow values of the Skellefte River (during the growing season of the mussels ndash from May to October) in comparison with the winterNAO index (1975ndash1980) (e) δ18O values of precipitation (DecemberndashSeptember) measured at Racksund in comparison with the winterNAO index (1975ndash1979)

veal century-scale and millennial-scale variations with muchless effort than sclerochronology-based records As such thetwo types of archives could complement each other perfectlyand increase the understanding of past climatic variabilityFor example once the low-frequency variations have beenreconstructed from sedimentary archives a more detailed in-sight into seasonal to inter-annual climate variability can beobtained from bivalve shells As long as the date of death ofthe bivalves is known such records can be placed in abso-lute temporal context (calendar year) Although the same iscurrently impossible with fossil shells each absolutely dated(radiocarbon and amino acid racemization dating) shell of along-lived bivalve species can open a seasonally to annuallyresolved window into the climatic and hydrological past of aregion of interest

42 M margaritifera shell δ18O values reflect streamwater δ18O values

Unfortunately complete high-resolution and long-termrecords of δ18Ow values of the streams studied were notavailable Such data are required for a direct comparisonwith those reconstructed from shells (δ18Olowastwr or δ18Olowastwr(SGI)

values) and to determine if the bivalves precipitated theirshells near oxygen isotope equilibrium with the ambient wa-ter However one of the study sites (NJB) is located closeto the Skellefte River where δ18Ow values were irregularlyanalyzed between 1973 and 1980 (Fig 6a) by the Water Re-sources Programme (GNIR data set) It should be noted thatthe δ18Ow data of GNIR merely reflect temporal snapshotsnot actual monthly averages In fact the isotope signatureof meteoric water can vary significantly on short timescales(eg Darling 2004 Leng and Marshall 2004 Rodgers etal 2005) In addition for some months no GNIR data wereavailable In contrast shell isotope data represent changes inthe isotope composition of the water over coherent time in-tervals ranging from 1 week to 1 year (and in few cases 2 or3 years) Due to the micromilling sampling technique unin-terrupted δ18Os time-series were available Thus it is com-pelling how well the ranges of intra-annual δ18Olowastwr data com-pared to instrumental oxygen isotope data of the SkellefteRiver (Fig 6a) and that summer averages as well as grow-ing season averages of shells and GNIR were nearly identi-cal (Table 3) Furthermore in each stream studied individ-ual δ18Olowastwr series agreed strongly with each other (Fig 5)



All of these aspects strongly suggest that shell formation oc-curred near equilibrium with the oxygen isotope compositionof the ambient water and M margaritifera recorded changesin stream water δ18O values Our conclusions are in agree-ment with previously published results from various differ-ent freshwater mussels (eg Dettman et al 1999 Kaandorpet al 2003 Versteegh et al 2009) and numerous marinebivalves (eg Epstein et al 1953 Mook and Vogel 1968Killingley and Berger 1979)

43 Site-specific and synoptic information recorded inshell oxygen isotopes

Although individual chronologies from a given stream com-pared well to each other with respect to absolute values thethree sites studied differed by almost 200 permil (the averageδ18Olowastwr values between 1947 and 1992 were minus1251 permil atNJB minus1221 permil at GTB and minus1416 permil at GJ Figs 5 7) Ifour interpretation is correct and δ18Os values of the margar-itiferids studied reflect the oxygen isotope signature of thewater in which they lived then these numbers reflect hydro-logical differences in the upstream catchment that are con-trolled by a complex set of physiographic characteristicscatchment size and elevation transit times upstream lakesize and depth controlling the potential for evaporative de-pletion in 16O stream flux rates stream width and depth hu-midity wind speed groundwater influx differences in melt-water influx an so on (Peralta-Tapia et al 2014 Geris etal 2017 Pfister et al 2017) However detailed monitoringwould be required to identify and quantify the actual rea-son(s) for the observed hydrological differences Thus werefrain from speculation

Despite the site-specific differences described above theδ18Olowastwr chronologies of the three streams were signifi-cantly positively correlated with each other suggesting thatcommon environmental forcings controlled isotope changesthroughout the study region Previous studies suggest thatthese environmental forcings may include changes in the iso-topic composition of precipitation specifically the amountorigin and air mass trajectory of winter snow and summerrain the timing of snowmelt as well as the condensationtemperature (Rosqvist et al 2013) The latter is probablythe most difficult to assess because no records are availabledocumenting the temperature height and latitude at whichthe respective clouds formed Moreover we cannot confi-dently assess the link between the isotope signature of pre-cipitation and stream water because only limited and inco-herent data sets are available from the study region In ad-dition data on precipitation amounts were taken from an-other locality and another time interval However it is wellknown that precipitation in northern Scandinavia particu-larly during winter originates from two different sourcesthe Atlantic and arcticpolar regions (Rosqvist et al 2013)and that the moisture in these air masses is isotopically dis-tinct (Araguaacutes-Araguaacutes et al 2000 Bowen and Wilkinson

2002) During NAO+ years the sea level pressure differ-ence between the Azores High and the Iceland Low is par-ticularly large resulting in mild wet winters in central andnorthern Europe with strong westerlies carrying heat andmoisture across the Atlantic Ocean toward higher latitudes(Hurrell et al 2003) During NAOminus years however wester-lies are weaker and the Polar Front is shifted southward al-lowing arctic air masses to reach northern Scandinavia Pre-cipitation originating from the North Atlantic is isotopicallyheavier (δ18Op of minus500 permil to minus1000 permil) than precipita-tion from subarctic and polar regions (δ18Op of minus1000 permilto minus1500 permil) Furthermore changes in air mass propertiesover northern Europe are controlled by atmospheric pressurepatterns in the North Atlantic particularly the NAO duringwinter (Hurrell 1995 Hurrell et al 2003) The positive cor-relation between δ18Olowastwr chronologies of the three streamsstudied and the wNAO index (Table 4 Figs 7a b c 8a) sug-gests that the shell isotopes recorded a winter precipitationsignal and this can be explained as follows A larger pro-portion of arctic air masses carried to northern Scandinaviaduring winter resulted in lower δ18Op values whereas thepredominance of North Atlantic air masses caused the oppo-site In NAO+ years strong westerlies carried North Atlanticair masses far northward so that winter precipitation in north-ern Sweden had significantly higher δ18Op values than dur-ing NAOminus years When the NAO was in its negative stateprecipitation predominantly originated from moisture fromthe polar regions which is depleted in 18O and hence haslower δ18Op values The specific isotope signatures in thestreams were controlled by the snowmelt in spring Essen-tially the bivalves recorded the (damped) isotope signal ofthe last winter precipitation ndash occasionally mixed with springand summer precipitation ndash in their shells This hypothesis issupported by the correlation of the few available GNIP andGNIR data with the wNAO index (Fig 8d e) Rosqvist etal (2007) hypothesized that the summer NAO strongly in-fluences δ18Op values and thus the δ18Ow signature of theopen through-flow lakes in northern Scandinavia Howeverour data did not support a profound influence of the summerNAO index on δ18Olowastwr values (Fig 7d e f) This conclusionis consistent with other studies suggesting that the summerNAO has a much weaker influence on European climate thanthe NAO during winter (eg Hurrell 1995)

Following Baldini et al (2008) and Comas-Bru etal (2016) northern Sweden is not the ideal place to conductoxygen-isotope-based wNAO reconstructions Their modelspredicted only a weak negative correlation or no correlationbetween δ18Op values and the wNAO index in our study re-gion (Baldini et al 2008 Fig 1 Comas-Bru et al 2016Fig 3a) One possible explanation for this weak correlationis the limited and temporally incoherent GNIP data set innorthern Sweden from which these authors extracted theδ18Op data that were used to construct the numerical mod-els In contrast δ18O data of diatoms from open lakes innorthern Sweden revealed a strong link to the amount of



precipitation and δ18Op values which reportedly are bothcontrolled by the predominant state of the NAO (Hammar-lund et al 2002 Andersson et al 2010 Rosqvist et al2004 2007 2013) Findings of the present study substanti-ated these proxy-based interpretations Furthermore we pre-sented for the first time oxygen isotope time-series with suf-ficient temporal resolution (annual) and the precise temporalcontrol (calendar years) required for a year-to-year compari-son with the NAO index time-series

As Comas-Bru et al (2016) further suggested the rela-tionship between δ18Op values and the wNAO index is sub-ject to spatial nonstationarities because the southern poleof the NAO migrates along a NEndashSW axis in response tothe state of another major atmospheric circulation mode inthe North Atlantic realm known as the East Atlantic Oscil-lation or the East Atlantic Pattern (EA) (Moore and Ren-frew 2012 Moore et al 2013 Comas-Bru and McDermott2014) Like the NAO the EA is most distinct during win-ter and describes atmospheric pressure anomalies betweenthe North Atlantic west of Ireland (low) and the subtropi-cal North Atlantic (high) Through the interaction of thesecirculation patterns the correlation between the wNAO andδ18Op values can weaken at times in certain regions Forexample when both indices are in their positive state thejet stream shifts poleward (Woolings and Blackburn 2012)and the storm trajectories that enter Europe in winter take amore northerly route (Comas-Bru et al 2016) The δ18Opvalues will then be lower than during NAO+EAminus years Toidentify whether this applies to the study region in questionwe followed Comas-Bru et al (2016) and tested if the rela-tionship between the wNAO and reconstructed stream wateroxygen isotope data remained significant during years whenthe signs of both indices were the same (EQ) and duringyears when they were opposite (OP) (Note that the EA in-dex is only available from 1950 onward) As demonstratedin Fig 8b and c the correlations between the region-wideshell-based oxygen isotope curve (δ18Olowastwr(Norrland)) and thewNAO (EQ R = 083 R2

= 069 p lt 00001) as well asthe wNAO (OP R = 065 R2

= 042 p = 00021) remainpositive and significant above the Bonferroni-adjusted 95 confidence level Hence the relationship between the wNAOand δ18Olowastwr values in the study region is not compromised bythe EA thus δ18Olowastwr values serve as a faithful proxy for thewNAO index

44 Damped stream water oxygen isotope signals

Compared with the large isotope difference between winterprecipitation sourced from SW or N air masses the hugeseasonal spread and inter-annual fluctuations of δ18Op val-ues (seasonal fluctuation of minus421 permil to minus1760 permil Fig 6binter-annual unweighted DecemberndashJanuary averages ofminus1018 permil to 1464 permil weighted DecemberndashSeptember av-erages of minus954 permil to minus1410 permil Fig 8e) as well as the pre-dicted seasonal variance of δ18Ow values in the study region

(Waterisotopes Database 2019 httpwwwwaterisotopesorg last access 25 May 2019 minus870 permil to 1730 permil)the observed and shell-derived variance of the stream wa-ter δ18O values was notably small and barely exceeded200 permil both on seasonal (Fig 6) and inter-annual timescales(Fig 5a b c) This figure agrees well with seasonal ampli-tudes determined in other streams at higher latitudes in theNorthern Hemisphere (Halder et al 2015) and can broadlybe explained by catchment damping effects due to water col-lection mixing storage and release processes in upstreamlakes and groundwater from which these streams were fedThe catchment mean transit time (MTT) determined via asimple precipitation vs stream flow isotope signal amplitudedamping approach (as per de Walle et al 1997) is approx-imately 6 months ndash corroborating the hypothesis of a mixedsnowmelt and precipitation contribution to the stream waterδ18O signal during the growing season

The attenuated variance on inter-annual timescales canpossibly be explained ndash amongst others ndash by inter-annualchanges in the amount of winter precipitation and the tim-ing of snowmelt Colder spring temperatures typically re-sulted in a delayed snowmelt so that lower oxygen isotopesignatures still prevailed in the stream water when the maingrowing season of the bivalves started However winter pre-cipitation amounts remained below average in NAOminus yearsmeaning that the net effect on δ18Ow values in spring wasless severe than the isotope shift in δ18Op values In con-trast the amount of snow precipitated during NAO+ yearswas larger but milder spring temperatures resulted in an ear-lier and faster snowmelt thus the effect on the isotope signa-ture of stream water at the beginning of the growing seasonof the mussels likely remained moderate

45 Sub-annual dating precision and relative changesin the seasonal shell growth rate

The precision with which the time that is represented by in-dividual isotope samples can be determined depends on thevalidity of the seasonal growth model We assumed that thetiming of seasonal shell growth was similar to published dataof M margaritifera and remained the same in each year andeach specimen This may not be entirely correct becausethe timing and rate of seasonal shell growth can potentiallyvary between localities among years and among individu-als however in M margaritifera the seasonal timing ofshell growth is remarkably invariant across large distances(Dunca et al 2005) A major dating error exceeding 4 weeksseems unlikely because the oxygen isotope series of individ-ual specimens at each site were in good agreement Presum-ably the timing of seasonal shell growth is controlled by ge-netically determined biological clocks which serve to main-tain a consistent duration of the growing season (Schoumlne2008) Although shells grew faster in some years and slowerin others the relative seasonal changes in shell growth rateslikely remained similar and consisted of a gradual increase as



the water warmed and more food became available in springand summer followed by a gradual decline as temperaturesdropped in fall It was further assumed that the timing ofshell growth has not significantly changed through the life-time of the specimens studied In fact if ontogenetic changesin seasonal growth traits had occurred it would be impossi-ble to crossdate growth curves from young and old individu-als and construct master chronologies (Schoumlne et al 2004ab 2005a Helama et al 2006 Black et al 2010) Based onthese arguments seasonal dating errors were likely minor

46 Shell stable carbon isotopes

Our results are consistent with previous studies using long-lived bivalves (Beirne et al 2012 Schoumlne et al 2005c2011) where δ13Cs chronologies of M margaritifera didnot show consistent ontogenetic trends but rather oscillatedaround an average value (ca minus1200 permil to minus1300 permil) Thetime series of NJB were too short to reject the hypothesisof directed trends throughout the lifetime of the organismhowever we propose here that the δ13Cs values of shellsfrom that stream would also average out at ca minus1250 permilas at the other two studied sites if longer chronologies wereavailable If a contribution of metabolic CO2 to the shellcarbonate exists in this species (which we cannot precludebecause no δ13C values of the dissolved inorganic carbonDIC data are available for the streams studied) it likely re-mains nearly constant through the lifetime of the organismas it does in other long-lived bivalve mollusks (Schoumlne et al2005c 2011 Butler et al 2011 Reynolds et al 2017) Ob-served stable carbon isotope signatures in the mussel shellsare within the range of those expected and observed in streamwaters of northern Europe (minus1000 permil to minus1500 permil Lengand Marshall 2004)

Seasonal and inter-annual changes in δ13Cs values couldbe indicative of changes in primary production food com-position respiration and the influx of terrestrial detritusHowever in the absence of information on how the envi-ronment of the streams that were studied changed throughtime we can only speculate about possible causes of tempo-ral δ13CDIC variations For example increased primary pro-duction in the water would not only have propelled shellgrowth rate but would also have resulted in a depletion of12C in the DIC pool and thus higher δ13CDIC and δ13Csvalues However just the opposite was observed on seasonaland inter-annual timescales The highest δ13Cs values oftenoccurred near the annual growth lines ie during times ofslow growth and although not statistically significant an-nual δ13Clowasts(d) values at NJB and GTB were inversely relatedto the shell growth rate (Fig 7g h Table 4) Accordinglyδ13Clowasts(d) values do not seem to reflect phytoplankton dynam-ics Another possibility is that a change in the compositionof mussel food occurred which changed the shell stable car-bon isotope values without a statistically significant effect onshell growth rate Because the isotope signatures of potential

food sources differ from each other (eg Gladyshev 2009) achange in the relative proportions of phytoplankton decom-posing plant litter from the surrounding catchment vegeta-tion bacteria particulate organic matter derived from higherorganisms etc could have left a footprint in the δ13Clowasts(d) val-ues Furthermore seasonal and inter-annual changes in res-piration or the influx of terrestrial detritus may have changedthe isotope signature of the DIC pool and thus the shellsSupport for the latter comes from the weak negative correla-tion between δ13Clowasts(d) values and the wNAO (Table 4 with-out Bonferroni correction p values remained below 005)After wet (snow-rich) winters (NAO+ years) stronger ter-restrial runoff may have flushed increased amounts of lightcarbon into the streams which lowered δ13CDIC values Totest these hypotheses data on the stable carbon isotope sig-nature of digested food and DIC would be required which isa task for subsequent studies

47 Error analysis and sensitivity tests

To test the robustness of the findings presented in Ta-bles 3 and 4 as well as their interpretation we have prop-agated all uncertainties associated with measurements andmodeled data and randomly generated δ18Olowastwr δ

18Olowastwr(SGI)δ18Olowastwr(Norrland) and δ13Clowasts(d) chronologies (via Monte Carlosimulation) A brief overview of the errors and simulationprocedures are provided below

Water temperature estimates (Eq 1) were associated withan error (1 standard deviation) ofplusmn207 C Amongst othersthis large uncertainty results from the combination of tem-perature data of four different streams which all varied withrespect to the average temperature and year-to-year variabil-ity The error exceeds the inter-annual variance (1 standarddeviation of plusmn090 C) of the instrumental water tempera-ture average (864 C) by more than 2 times Instead of re-constructing Tw from Ta with an uncertainty of plusmn207 Cwe could have assumed a constant water temperature valueof 864 C with an uncertainty of only plusmn090 C Howeverour goal was to improve the δ18Olowastwr reconstructions by takingthe actual year-to-year temperature variability into accountTo simulate the effect of different temperature uncertaintieswe randomly generated 1000 T lowastw chronologies with an errorof plusmn090 C as well as 1000 chronologies with an error ofplusmn207 C Both sets of simulated T lowastw time-series were usedin subsequent calculations Errors involved with shell growthpatterns include the measurement error (plusmn1 microm equivalent toan SGI error of plusmn006 units) and the variance of crossdatedSGI data In different calendar years the standard error ofthe mean of the 25 SGI chronologies ranged between plusmn003and plusmn066 SGI units The measurement and crossdating un-certainties were propagated and 1000 new SGI chronologieswere randomly generated and regressed against simulated T lowastwchronologies The uncertainty of the new SGI vs T lowastw model(standard error of plusmn135 C) was propagated in subsequentcalculations of δ18Olowastwr(SGI) values using Eq (2) A third set



Table 5 Results of sensitivity tests To test the robustness of statistically significant correlations presented in Tables 3 and 4 uncertainties(one of them the error associated with the reconstruction of stream water temperatures Tw from air temperatures Ta) were propagatedand used to randomly generate δ18Olowastwr(SGI) chronologies which were subsequently regressed against the winter North Atlantic Oscillation(wNAO) indices Simulations were computed with propagated T lowastw values of 207 and 090 C See text for details Statistically significantvalues (Bonferroni-adjusted p lt 005) are marked in bold

T lowastw uncertainty =plusmn207 C T lowastw uncertainty =plusmn090 C

Norrland NJB GTB GJ Norrland NJB GTB GJ

wNAO1950ndash1998

R = 060R2 = 036p = 00007

R = 065R2 = 042p lt 00001

wNAO (EQ)1950ndash1998

R = 070R2 = 051p = 00001

R = 076R2 = 058p lt 00001

wNAO (OP)1950ndash1998

R = 045R2= 022

p = 00710

R = 050R2 = 026p = 00256

wNAO R = 062R2 = 038p = 00001

R = 046R2 = 022p = 00075

R = 035R2 = 013p = 00008

R = 066R2 = 043p lt 00001

R = 049R2 = 024p = 00028

R = 038R2 = 015p = 00001

wNAO1947ndash1991

R = 060R2 = 036p = 00003

R = 048R2 = 024p = 00088

R = 051R2 = 027p = 00067

R = 064R2 = 041p lt 00001

R = 051R2 = 027p = 00026

R = 057R2 = 033p = 00007

of uncertainties was associated with isotope measurements(analytical precision error 1 standard deviation=plusmn006 permil)the calculation of site-specific annual averages from contem-poraneous specimens (plusmn011 permil to plusmn015 permil for δ18O on av-erage plusmn037 permil to plusmn042 permil for δ13C on average) and thecalculation of the Norrland average All errors were prop-agated and new δ18Olowastwr δ

18Olowastwr(SGI) δ18Olowastwr(Norrland) and

δ13Clowasts(d) chronologies were simulated (1000 representationseach) The chronologies simulated were regressed againstNAO and SGI chronologies (results of sensitivity tests forthe regressions of δ18Olowastwr(SGI) and δ18Olowastwr(Norrland) values vswNAO indices are given in Table 5)

According to the complex simulation experiments the ob-served links between reconstructed stream water oxygen iso-tope values and the wNAO largely remained statistically ro-bust irrespective of which T lowastw error was assumed (Table 5)This outcome is not particularly surprising given that eventhe annual δ18Os chronologies of the studiesrsquo specimenswere strongly coherent and values fluctuated at timescalessimilar to that of the wNAO (Fig 4) Apparently decadal-scale atmospheric circulation patterns indeed exert a strongcontrol over the isotope signature of stream water in the studyarea However none of the correlations between shell isotopedata and the sNAO were statistically significant at the prede-fined value of p le 005 The importance of summer rainfallseems much less important for the isotope value of streamwater than winter snow As before the relationship betweenstable carbon isotope data of the shells and climate indices as

well as the shell growth rate remained weak and were statisti-cally not significant Inevitably the propagated errors specif-ically the uncertainty associated with the reconstruction ofthe stream water temperature from air temperature resulted ina notable drop in the amount of variability explained and inthe statistical probability (Table 5) The use of instrumentalwater temperatures could greatly improve the reconstructionof δ18Olowastwr values as the measurement error would be of theorder of 01 C instead of 207 or 090 C Thus future cali-bration studies should be conducted in monitored streams

5 Summary and conclusions

Stable oxygen isotope values in shells of freshwater pearlmussels M margaritifera from streams in northern Swedenmirror stream water stable oxygen isotope values Despitea well-known damping of the precipitation signal in streamwater isotope records these mollusks archive local precipi-tation and synoptic atmospheric circulation signals specifi-cally the NAO during winter Stable carbon isotope data ofthe shells are more challenging to interpret but they seemto record local environmental conditions such as changes inDIC andor food composition Future studies should be con-ducted in streams in which temperature DIC and food levelsare closely monitored to further improve the reconstructionof stream water δ18O values from δ18Os data and better un-derstand the meaning of δ13Cs fluctuations



The bivalve shell oxygen isotope record presented hereextends back to 1819 CE but there is the potential to de-velop longer isotope chronologies via the use of fossil shellsof M margaritifera collected in the field or taken from mu-seum collections With suitable material and by applying thecrossdating technique the existing chronology could prob-ably be extended by several centuries back in time Streamwater isotope records may shed new light on pressing ques-tions related to climate change impacts on river systems themechanistic understanding of water flow and quality con-trolling processes calibration and validation of flow andtransport models climate and Earth system modeling timevariant catchment travel time modeling and so on Longerand coherent chronologies are essential to reliably iden-tify multidecadal-scale and century-scale climate dynamicsEven individual radiocarbon-dated fossil shells that do notoverlap with the existing master chronology can providevaluable paleoclimate information because each M margar-itifera specimen can open a seasonally to annually resolvedmultiyear window into the history of streams



Appendix A

Table A1 Overview of abbreviations used in the paper

Streams studied

GJ GoumlrjearingnGTB GrundtraumlsktjaumlrnbaumlckenNJB Nuortejaurbaumlcken

Sclerochronology

iOSL Inner portion of the outer shell layeroOSL Outer portion of the outer shell layerSGI values Standardized growth indices

Climate indices and environmental data sets

EA East Atlantic Oscillation (superscript plus and minus denotes if the EA is in its positiveor negative state)

GNIP Global Network of Isotopes in PrecipitationGNIR Global Network of Isotopes in RiversNAO North Atlantic Oscillation (superscript plus and minus denotes if the NAO is in its

positive or negative state)sNAO North Atlantic Oscillation during summer (JunendashSeptember)wNAO North Atlantic Oscillation during winter (DecemberndashMarch)wNAO(EQ) Winters during which the NAO and EA have the same signwNAO(OP) Winters during which the NAO and EA have opposite signs

Stable carbon isotopes

δ13Cs Stable carbon isotope value of the shell carbonateδ13Clowasts Weighted (considering variations in seasonal shell growth rate) δ13Cs value annual

δ13Clowasts refers to the growing season mean valueδ13Clowasts(d) Detrended and standardized weighted ldquoannualrdquo (ie growing season) δ13Cs mean value

Stable oxygen isotopes

δ18Op Stable oxygen isotope value of precipitationδ18Os Stable oxygen isotope value of the shell carbonateδ18Olowasts Weighted (considering variations in seasonal shell growth rate) δ18Os value annual

δ18Olowasts refers to the growing season mean valueδ18Ow Stable oxygen isotope value of the water in which the bivalve livedδ18Olowastwr Stable oxygen isotope value of the water reconstructed from δ18Olowasts and Twδ18Olowastwr(Norrland) δ18Olowastwr(SGI) average of all studied specimens in Norrbotten Countyδ18Olowastwr(SGI) Stable oxygen isotope value of the water reconstructed from δ18Olowasts and T lowastw

Temperature

Ta Instrumental air temperatureTw Stream water temperature reconstructed from TaT lowastw Weighted (considering variations in seasonal shell growth rate) stream water temper-

ature reconstructed from SGI and Tw annual T lowastw refers to the growing season meanvalue



Code and data availability All data and code used in this studyare available from the authors upon request Additional supplemen-tary files are available at httpswwwpaleontologyuni-mainzdedatasetsHESS_2019_337_supplementszip (last access 5 February2020)

Sample availability Bivalve shell samples are archived and storedin the paleontological collection of the University of Mainz

Supplement The supplement related to this article is available on-line at httpsdoiorg105194hess-24-673-2020-supplement

Author contributions BRS designed the study performed the anal-yses and wrote the paper AEM and SMB conducted the field workand collected samples SMB sampled the shells and temporallyaligned the isotope data JF isotopically analyzed the shell powderLP conducted MTT calculations All authors jointly contributed tothe discussion and interpretation of the data

Competing interests The authors declare that they have no conflictof interest

Acknowledgements We thank Denis Scholz and Erika Pietronirofor constructive discussions We are grateful for comments and sug-gestions provided by two anonymous reviewers that greatly im-proved the quality of this article This study has been made possiblethrough a research grant by the Deutsche Forschungsgemeinschaft(DFG) to BRS (grant no SCHO7931)

Financial support This research has been supported by theDeutsche Forschungsgemeinschaft (grant no SCHO7931)

This open-access publication was fundedby Johannes Gutenberg University Mainz

Review statement This paper was edited by Brian Berkowitz andreviewed by two anonymous referees

References

Andersson S Rosqvist G Leng M J Wastegard S andBlaauw M Late Holocene climate change in central Swedeninferred from lacustrine stable isotope data J Quaternary Sci25 1305ndash1316 httpsdoiorg101002jqs1415 2010

Araguaacutes-Araguaacutes L Froehlich K and Rozanski K Deu-terium and oxygen-18 isotope composition of precipitationand atmospheric moisture Hydrol Process 14 1341ndash1355httpsdoiorg1010021099-1085(20000615)148lt1341AID-HYP983gt30CO2-Z 2000

Baillie M G L and Pilcher J R A simple crossdating programfor tree-ring research Tree-ring Bull 33 7ndash14 1973

Baldini L M McDermott F Foley A M and Baldini J UL Spatial variability in the European winter precipitation δ18O-NAO relationship Implications for reconstructing NAO-modeclimate variability in the Holocene Geophys Res Lett 35L04709 httpsdoiorg1010292007GL032027 2008

Beirne E C Wanamaker Jr A D and Feindel S C Experimen-tal validation of environmental controls on the δ13C of Arctica is-landica (ocean quahog) shell carbonate Geochim CosmochimAc 84 395ndash409 httpsdoiorg101016jgca2012010212012

Black B A Dunham J B Blundon B W Raggon MF and Zima D Spatial variability in growth-incrementchronologies of long-lived freshwater mussels Implicationsfor climate impacts and reconstructions Eacutecosci 17 240ndash250httpsdoiorg10298017-3-3353 2010

Bowen G J and Wilkinson B Spatial distribu-tion of δ18O in meteoric precipitation Geol-ogy 30 315ndash318 httpsdoiorg1011300091-7613(2002)030lt0315SDOOIMgt20CO2 2002

Burgman J O Eriksson E and Westman F Oxygen-18 varia-tion in river waters in Sweden Avd Hydrol Unpublished Re-port Uppsala Univ Naturgeogr Inst Uppsala Sweden 42 p1981

Butler P G Wanamaker Jr A D Scourse J D Richardson CA and Reynolds D J Long-term stability of δ13C with respectto biological age in the aragonite shell of mature specimens of thebivalve mollusk Arctica islandica Palaeogeogr Palaeocl 30221ndash30 httpsdoiorg101016jpalaeo201003038 2011

Butler P G Wanamaker Jr A D Scourse J D Richard-son C A and Reynolds D J Variability of ma-rine climate on the North Icelandic Shelf in a 1357-year proxy archive based on growth increments in the bi-valve Arctica islandica Palaeogeogr Palaeocl 373 141ndash151httpsdoiorg101016jpalaeo201201016 2013

Chen G and Fang X Accuracy of hourly water temperaturesin rivers calculated from air temperatures Water 7 1068ndash1087httpsdoiorg103390w7031068 2015

Comas-Bru L and McDermott F Impacts of the EA and SCApatterns on the European twentieth century NAOwinter cli-mate relationship Q J Roy Meteor Soc 140 354ndash363httpsdoiorg101002qj2158 2014

Comas-Bru L McDermott F and Werner M The ef-fect of the East Atlantic pattern on the precipitation δ18O-NAO relationship in Europe J Clim Dyn 47 2059ndash2069httpsdoiorg101007s00382-015-2950-1 2016

Darling W G Hydrological factors in the interpretationof stable isotopic proxy data present and past a Eu-ropean perspective Quaternary Sci Rev 23 743ndash770httpsdoiorg101016jquascirev200306016 2004

Darling W G and Bowes M J A long-term study of stable iso-topes as tracers of processes governing water flow and qual-ity in a lowland river basin Hydrol Process 30 2178ndash2195httpsdoiorg101016jquascirev200306016 2016

Dettman D L Reische A K and Lohmann K C Controlson the stable isotope composition of seasonal growth bandsin aragonitic fresh-water bivalves (unionidae) Geochim Cos-



mochim Ac 63 1049ndash1057 httpsdoiorg101016S0016-7037(99)00020-4 1999

DeWalle D R Edwards P J Swistock B R Aravena R andDrimmie R J Seasonal isotope hydrology of three Appalachianforest catchments Hydrol Process 11 1895ndash1906 1997

Dunca E and Mutvei H Comparison of microgrowth pattern inMargaritifera margaritifera shells from south and north SwedenAm Malacol Bull 16 239ndash250 2001

Dunca E Schoumlne B R and Mutvei H Freshwater bi-valves tell of past climates But how clearly do shells frompolluted rivers speak Palaeogeogr Palaeocl 228 43ndash57httpsdoiorg101016jpalaeo200503050 2005

Epstein S Buchsbaum R Lowenstam H A and Urey HC Revised carbonate-water isotopic temperature scale GeolSoc Am Bull 64 1315ndash1326 httpsdoiorg1011300016-7606(1953)64[1315RCITS]20CO2 1953

Frank D Esper J and Cook E R Adjustment forproxy number and coherence in a large-scale temper-ature reconstruction Geophys Res Lett 34 L16709httpsdoiorg1010292007GL030571 2007

Fuumlllenbach C S Schoumlne B R and Mertz-Kraus RStrontiumlithium ratio in shells of Cerastoderma ed-ule (Bivalvia) ndash A new potential temperature proxyfor brackish environments Chem Geol 417 341ndash355httpsdoiorg101016jchemgeo201510030 2015

Geeza T J Gillikin D P Goodwin D H Evans S D Wat-ters T and Warner N R Controls on magnesium man-ganese strontium and barium concentrations recorded in fresh-water mussel shells from Ohio Chem Geol 526 142ndash152httpsdoiorg101016jchemgeo201801001 2019

Geeza T J Gillikin D P McDevitt B Van Sice K and WarnerN R Accumulation of Marcellus Formation oil and gas wastew-ater metals in freshwater mussel shells Environ Sci Tech-nol 52 10883ndash10892 httpsdoiorg101021acsest8b027272020

Geist J Auerswald K and Boom A Stable carbon isotopes infreshwater mussel shells Environmental record or marker formetabolic activity Geochim Cosmochim Ac 69 3545ndash3554httpsdoiorg101016jgca200503010 2005

Geris J Tetzlaff D McDonnell J J and Soulsby C Spatial andtemporal patterns of soil water storage and vegetation water usein humid northern catchments Sci Total Environ 595 486ndash493httpsdoiorg101016jscitotenv201703275 2017

Gladyshev M I Stable isotope analyses in aquatic ecology (a re-view) Journal of Siberian Federal University ndash Biology 4 381ndash402 httpsdoiorg10175161997-1389-0220 2009

Gonfiantini R Stichler W and Rozanski K Standards and inter-comparison materials distributed by the International Atomic En-ergy Agency for stable isotope measurements (IAEA-TECDOC-825) International Atomic Energy Agency (IAEA) ViennaAustria 13ndash29 available at httpswww-pubiaeaorgMTCDpublicationsPDFte_825_prnpdf (last access 5 February 2020)1995

Grossman E L and Ku T-L Oxygen and carbon isotope frac-tionation in biogenic aragonite temperature effects Chem GeolIsot Geosci Sect 59 59ndash74 httpsdoiorg1010160168-9622(86)90057-6 1986

Halder J Terzer S Wassenaar L I Araguaacutes-Araguaacutes L Jand Aggarwal P K The Global Network of Isotopes in Rivers

(GNIR) integration of water isotopes in watershed observationand riverine research Hydrol Earth Syst Sci 19 3419ndash3431httpsdoiorg105194hess-19-3419-2015 2015

Hammarlund D Barnekow L Birks H J B Buckardt Band Edwards T W D Holocene changes in atmospheric cir-culation recorded in the oxygen-isotope stratigraphy of lacus-trine carbonates from northern Sweden Holocene 12 339ndash351httpsdoiorg1011910959683602hl548rp 2002

Helama S Schoumlne B R Black B A and Dunca E Construct-ing long-term proxy series for aquatic environments with abso-lute dating control using a sclerochronological approach intro-duction and advanced applications Mar Freshw Res 57 591ndash599 httpsdoiorg101071MF05176 2006

Hurrell J W Decadal trends in the North Atlantic Oscillation re-gional temperatures and precipitation Science 269 676ndash679httpsdoiorg101126science2695224676 1995

Hurrell J W Kushnir Y Ottersen G and Visbeck MAn overview of the North Atlantic Oscillation in TheNorth Atlantic Oscillation Climatic Significance and Environ-mental Impact edited by Hurrell J W Kushnir Y Ot-tersen G and Visbeck M Geophysical Monograph Amer-ican Geophysical Union Washington DC USA 134 1ndash35httpsdoiorg101029134GM01 2003

Kaandorp R J G Vonhof H B Del Busto C Wesselingh F PGanssen G M Marmoacutel A E Romero Pittman L and vanHinte J E Seasonal stable isotope variations of the modernAmazonian freshwater bivalve Anodontites trapesialis Palaeo-geogr Palaeocl 194 339ndash354 httpsdoiorg101016S0031-0182(03)00332-8 2003

Kelemen Z Gillikin D P Graniero L E Havel H Darcham-beau F Borges A V Yambeacuteleacute A Bassirou A and Bouil-lon S Calibration of hydroclimate proxies in freshwater bivalveshells from Central and West Africa Geochim Cosmochim Ac208 41ndash62 httpsdoiorg101016jgca201703025 2017

Kelemen Z Gillikin D P and Bouillon S Relationship be-tween river water chemistry and shell chemistry of two tropicalAfrican freshwater bivalve species Chem Geol 526 130ndash141httpsdoiorg101016jchemgeo201804026 2019

Killingley J S and Berger W H Stable isotopes in a molluskshell detection of upwelling events Science 205 186ndash188httpsdoiorg101126science2054402186 1979

Leng M L Isotopes in Palaeoenvironmental Research Dev Pale-oenviron Res 10 1ndash307 httpsdoiorg1010071-4020-2504-1 2006

Leng M L and Marshall J D Palaeoclimate in-terpretation of stable isotope data from lake sedi-ment archives Quaternary Sci Rev 23 811ndash831httpsdoiorg101016jquascirev200306012 2004

Mook W G and Vogel J C Isotopic equilibrium be-tween shells and their environment Science 159 874ndash875httpsdoiorg101126science1593817874 1968

Moore G W K and Renfrew I A Cold European winters inter-play between the NAO and the East Atlantic mode Atmos SciLett 13 1ndash8 httpsdoiorg101002asl356 2012

Moore G W K Renfrew I A and Pickart R S Multidecadalmobility of the North Atlantic Oscillation J Climate 26 2453ndash2466 httpsdoiorg101175JCLI-D-12-000231 2013

Moorkens E Cordeiro J Seddon M B von Proschwitz T andWoolnough D Margaritifera margaritifera (errata version pub-



lished in 2018) The IUCN Red List of Threatened Species 2018eT12799A128686456 httpsdoiorg102305IUCNUK2017-3RLTST12799A508865en 2018

Morrill J C Bales R C and Conklin M H Estimatingstream temperature from air temperature Implications forfuture water quality J Environ Engineer 131 139ndash146httpsdoiorg101061(ASCE)0733-9372(2005)1311(139)2005

Mutvei H and Westermark T How environmental informationcan be obtained from naiad shells Ecol Stud 145 367ndash379httpsdoiorg101007978-3-642-56869-5_21 2001

Nystroumlm J Dunca E Mutvei H and Lindh U Environ-mental history as reflected by freshwater pearl mussels in theriver Vramsaringn southern Sweden Ambio 25 350ndash355 availableat httpswwwjstororgstable4314490 (last access 5 February2020) 1996

Peralta-Tapia A Sponseller R A Tetzlaff D Soulsby C andLaudon H Connecting precipitation inputs and soil flow path-ways to stream water in contrasting boreal catchments HydrolProcess 29 3546ndash3555 httpsdoiorg101002hyp103002014

Pfister L Martiacutenez-Carreras N Hissler C Klaus J Stew-art M K and McDonnell J J Bedrock geology controls oncatchment storage mixing and release a comparative analy-sis of 16 nested catchments Hydrol Process 31 1828ndash1845httpsdoiorg101002hyp11134 2017

Pfister L Thielen F Deloule E Valle N Lentzen E Grave CBeisel J-N and McDonnell J J Freshwater pearl mussels as astream water stable isotope recorder Ecohydrol 2018e e2007httpsdoiorg101002eco2007 2018

Pfister L Grave C Beisel J-N and McDonnell J J Aglobal assessment of freshwater mollusk shell oxygen isotopesignatures and their relation to precipitation and stream waterSci Rep 9 4312 httpsdoiorg101038s41598-019-40369-02019

Pulteney R A General View of the Writing of Linnaeus Payneand White London UK httpsdoiorg105962bhltitle968851781

Rank D Wyhlidal S Schott K Weigand S and Oblin A Tem-poral and spatial distribution of isotopes in river water in Cen-tral Europe 50 years of experience with the Austrian networkof isotopes in rivers Isotop Environ Health Stud 54 115ndash136httpsdoiorg1010801025601620171383906 2017

Reckerth A Stichler W Schmidt A and Stumpp CLong-term data set analysis of stable isotopic com-position in German rivers J Hydrol 552 718ndash731httpsdoiorg101016jjhydrol201707022 2017

Reynolds D J Hall I R Scourse J D Richardson C A Wana-maker A D and Butler P G Biological and climate controlson North Atlantic marine carbon dynamics over the last millen-nium Insights from an absolutely dated shell-based record fromthe North Icelandic shelf Global Biogeochem Cy 31 1718ndash1735 httpsdoiorg1010022017GB005708 2017

Risi C Ogeacutee J Bony S and Kurz Besson C The water isotopicversion of the land-surface model ORCHIDEE Implementationevaluation sensitivity to hydrological parameters Hydrol Cur-rent Res 7 258 httpsdoiorg1041722157-758710002582016

Rodgers P Soulsby C Waldron S and Tetzlaff D Using stableisotope tracers to assess hydrological flow paths residence timesand landscape influences in a nested mesoscale catchment Hy-drol Earth Syst Sci 9 139ndash155 httpsdoiorg105194hess-9-139-2005 2005

Rosqvist G Jonsson C Yam R Karlen W and ShemeshA Diatom oxygen isotopes in pro-glacial lake sedimentsfrom northern Sweden a 5000 year record of atmo-spheric circulation Quaternary Sci Rev 23 851ndash859httpsdoiorg101016jquascirev200306009 2004

Rosqvist G C Leng M J and Jonsson C North At-lantic region atmospheric circulation dynamics inferredfrom a late-Holocene lacustrine carbonate isotope recordnorthern Swedish Lapland Holocene 17 867ndash873httpsdoiorg1011770959683607080508 2007

Rosqvist G C Leng M J Goslar T Sloane H J BiglerC Cunningham L Dadal A Bergman J Berntsson AJonsson C and Wastegaringrd S Shifts in precipitation dur-ing the last millennium in northern Scandinavia from la-custrine isotope records Quaternary Sci Rev 66 22ndash34httpsdoiorg101016jquascirev201210030 2013

Schoumlne B R The curse of physiology ndash challenges and opportuni-ties in the interpretation of geochemical data from mollusk shellsGeo-Mar Lett 28 269ndash285 httpsdoiorg101007s00367-008-0114-6 2008

Schoumlne B R Arctica islandica (Bivalvia) A unique pa-leoenvironmental archive of the northern North At-lantic Ocean Global Planet Change 111 199ndash225httpsdoiorg101016jgloplacha201309013 2013

Schoumlne B R and Krause R A Retrospective en-vironmental biomonitoring ndash Mussel Watch ex-panded Global Planet Change 144 228ndash251httpsdoiorg101016jgloplacha201608002 2016

Schoumlne B R Dunca E Mutvei H and Norlund U A 217-yearrecord of summer air temperature reconstructed from freshwaterpearl mussels (M margarifitera Sweden) Quaternary Sci Rev23 1803ndash1816 httpsdoiorg101016jquascirev2004020172004a

Schoumlne B R Dunca E Mutvei H and Norlund UCorrigendum to ldquoA 217-year record of summer air tem-perature reconstructed from freshwater pearl mussels (Mmargarifitera Sweden)rdquo [Quaternary Science Reviews23 (2004) 1803ndash1816] Quaternary Sci Rev 23 2057httpsdoiorg101016jquascirev200407005 2004b

Schoumlne B R Dunca E Mutvei H Baier S and Fiebig J Scan-dinavian climate since the late 18th century reconstructed fromshells of bivalve mollusks Z Dt Ges Geowiss 156 501ndash515httpsdoiorg1011271860-180420050156-0501 2005a

Schoumlne B R Dunca E Fiebig J and Pfeiffer M Mutveirsquossolution an ideal agent for resolving microgrowth structuresof biogenic carbonates Palaeogeogr Palaeocl 228 149ndash166httpsdoiorg101016jpalaeo200503054 2005b

Schoumlne B R Fiebig J Pfeiffer M Gleszlig R Hickson JJohnson A L A Dreyer W and Oschmann W Cli-mate records from a bivalved Methuselah (Arctica islandicaMollusca Iceland) Palaeogeogr Palaeocl 228 130ndash148httpsdoiorg101016jpalaeo200503049 2005c

Schoumlne B R Wanamaker Jr A D Fiebig J TheacutebaultJ and Kreutz K J Annually resolved δ13Cshell chronolo-



gies of long-lived bivalve mollusks (Arctica islandica) re-veal oceanic carbon dynamics in the temperate North Atlanticduring recent centuries Palaeogeogr Palaeocl 302 31ndash42httpsdoiorg101016jpalaeo201002002 2011

Schoumlne B R Schmitt K and Maus M Effects of sample pre-treatment and external contamination on bivalve shell and Car-rara marble δ18O and δ13C signatures Palaeogeogr Palaeocl484 22ndash32 httpsdoiorg101016jpalaeo201610026 2017

Teranes J L and McKenzie J A Lacustrine oxygenisotope record of 20th-century climate change in cen-tral Europe evaluation of climatic controls on oxygenisotopes in precipitation J Paleolimnol 26 131ndash146httpsdoiorg101023A1011175701502 2001

Tetzlaff D Buttle J Carey S K McGuire K Laudon H andSoulsby C Tracer-based assessment of flow paths storage andrunoff generation in northern catchments a review Hydrol Pro-cess 29 3475ndash3490 httpsdoiorg101002hyp10412 2014

Trouet V Esper J Graham N E Baker A Scourse J D andFrank D C Persistent positive North Atlantic Oscillation modedominated the Medieval Climate Anomaly Science 324 78ndash80httpsdoiorg101126science1166349 2009

Versteegh E A A Troelstra S R Vonhof H B and Kroon DOxygen isotope composition of bivalve seasonal growth incre-ments and ambient water in the rivers Rhine and Meuse Palaios24 497ndash504 httpsdoiorg102110palo2008p08-071r 2009

von Hessling T Die Perlmuscheln und ihre Perlen naturwis-senschaftlich und geschichtlich mit Beruumlcksichtigung derPerlengewaumlsser Bayerns Engelmann Leipzig Germanyhttpsdoiorg105962bhltitle47047 1859

Woollings T and Blackburn M The North Atlantic jet streamunder climate change and its relation to the NAO and EA pat-terns J Climate 25 886ndash902 httpsdoiorg101175JCLI-D-11-000871 2012

Ziuganov V San Miguel E Neves RJ Longa A Fernaacutendez CAmaro R Beletsky V Popkovitch E Kaliuzhin S and John-son T Life span variation of the freshwater pearl shell A modelspecies for testing longevity mechanisms in animals Ambio 29102ndash105 httpsdoiorg1015790044-7447-292102 2000


Abstract

Introduction

Material and methods

Sample preparation

Shell growth pattern analysis

Stable isotope analysis

Instrumental data sets

Weighted annual shell isotope data

Reconstruction of oxygen isotope signatures of stream water on annual and intra-annual timescales

Stable carbon isotopes of the shells

Results

Shell growth and temperature

Shell stable oxygen isotope data

Shell stable oxygen isotope data and instrumental records

Shell stable oxygen isotope data and synoptic circulation patterns (NAO)

Shell stable carbon isotope data

Discussion

Advantages and disadvantages of using bivalve shells for stream water 18O reconstruction comparison with sedimentary archives

M margaritifera shell 18O values reflect stream water 18O values

Site-specific and synoptic information recorded in shell oxygen isotopes

Damped stream water oxygen isotope signals

Sub-annual dating precision and relative changes in the seasonal shell growth rate

Shell stable carbon isotopes

Error analysis and sensitivity tests

Summary and conclusions

Appendix A

Code and data availability

Sample availability

Supplement

Author contributions

Competing interests

Acknowledgements

Financial support

Review statement

References


ies the temporal changes in the oxygen isotope signaturesof meteoric water are encoded in biogenic tissues and abio-genic minerals formed in rivers and lakes (eg Teranes andMcKenzie 2001 Leng and Marshall 2004) In fact manystudies have determined the oxygen isotope composition ofdiatoms ostracods authigenic carbonate and aquatic cellu-lose preserved in lacustrine sediments to reconstruct LateHolocene changes in precipitation over Scandinavia and ul-timately atmospheric circulation dynamics in the North At-lantic realm (eg Hammarlund et al 2002 Andersson et al2010 Rosqvist et al 2004 2013) With their short residencetimes of a few months (Rosqvist et al 2013) the hydro-logically connected through-flow lakes of northern Scandi-navia are an ideal region for this type of study Their isotopesignatures ndash while damped in comparison with isotope sig-nals in precipitation ndash directly respond to changes in the pre-cipitation isotope composition (Leng and Marshall 2004)However even the highest available temporal resolution ofsuch records (15 years per sample in sediments from LakeTibetanus Swedish Lapland Rosqvist et al 2007) is stillinsufficient to resolve inter-annual- to decadal-scale variabil-ity ie the timescales at which the North Atlantic Oscilla-tion (NAO) operates (Hurrell 1995 Trouet et al 2009) TheNAO steers weather and climate dynamics in northern Scan-dinavia and determines the origin of air masses from whichmeteoric waters form While sediment records are still of vi-tal importance for century-scale and millennial-scale varia-tions new approaches are needed for finer-scale resolutionon the 1ndash100-year timescale

One underutilized approach for hydroclimate reconstruc-tion is the use of freshwater mussels as natural stream waterstable isotope recorders (Dettman et al 1999 Kelemen et al2017 Pfister et al 2018 2019) Their shells can provide sea-sonally to annually resolved chronologically precisely con-strained records of environmental changes in the form ofvariable increment widths (which refers to the distance be-tween subsequent growth lines) and geochemical properties(eg Nystroumlm et al 1996 Schoumlne et al 2005a Geist et al2005 Black et al 2010 Schoumlne and Krause 2016 Geezaet al 2019 2020 Kelemen et al 2019) In particular sim-ilar to marine (Epstein et al 1953 Mook and Vogel 1968Killingley and Berger 1979) and other freshwater bivalves(Dettman et al 1999 Kaandorp et al 2003 Versteegh etal 2009 Kelemen et al 2017 Pfister et al 2019) Margar-itifera margaritifera forms its shell near equilibrium with theoxygen isotope composition of the ambient water (δ18Ow)(Pfister et al 2018 Schoumlne et al 2005a) If the fractiona-tion of oxygen isotopes between the water and shell carbon-ate is only temperature-dependent and the temperature dur-ing shell formation is known or can be otherwise estimated(eg from shell growth rate) reconstruction of the oxygenisotope signature of the water can be carried out from thatof the shell CaCO3 (δ18Os) Freshwater pearl mussels Mmargaritifera are particularly useful in this respect becausethey can reach a life span of 60 years (Pulteney 1781) or

80 years (von Hessling 1859) to over 200 years (Ziuganovet al 2000 Mutvei and Westermark 2001) offering an in-sight into long-term changes in freshwater ecosystems at anunprecedented temporal resolution

Here we present the first absolutely dated annually re-solved stable oxygen and carbon isotope record of freshwaterpearl mussels from three different streams in northern Swe-den covering nearly 2 centuries (1819ndash1998) We test theability of these freshwater pearl mussels to reconstruct theNAO index and associated changes in precipitation prove-nance using shell oxygen isotope data (δ18Os) We evalu-ate how shell oxygen isotope data compare to δ18O valuesin stream water (δ18Ow) and precipitation (δ18Op) as wellas limited existing environmental data (such as stream wa-ter temperature) we also evaluate how these variables relateto each other We hypothesize that δ18Ow and δ18Op valuesare positively correlated with one another as well as withthe North Atlantic Oscillation index In other words duringpositive NAO years oxygen isotope values in stream wa-ter and shells are higher than during negative NAO yearswhen shell and stream water oxygen isotope values tend tobe more depleted in 18O In addition we explore the phys-ical controls on shell stable carbon isotope signatures Pre-sumably these data reflect changes in the stable carbon iso-tope value of dissolved inorganic carbon which in turn issensitive to changes in primary production We leverage pastwork in Sweden that has shown that the main growing seasonof M margaritifera occurs from mid-May to mid-Octoberwith the fastest growth rates occurring between June and Au-gust (Dunca and Mutvei 2001 Dunca et al 2005 Schoumlne etal 2004a b 2005a) We use changes in the annual incre-ment width of M margaritifera to infer water temperaturebecause growth rates are faster during warm summers andresult in broader increment widths (Schoumlne et al 2004a b2005a) Because specimens of a given population react sim-ilarly to changes in temperature their average shell growthpatterns can be used to estimate climate and hydrologicalchanges Consequently increment series of specimens withoverlapping life spans can be crossdated and combined toform longer chronologies covering several centuries (Schoumlneet al 2004a b 2005a)

2 Material and methods

We collected 10 specimens of the freshwater pearl musselM margaritifera from one river and two creeks in Nor-rland (Norrbotten County) northern Sweden (Figs 1 2Table 1) Bivalves were collected between 1993 and 1999and included nine living specimens and one found dead andarticulated (bi-valved ED-GJ-D6R) Four individuals weretaken from the stream Nuortejaurbaumlcken (NJB) two fromstream Grundtraumlsktjaumlrnbaumlcken (GTB) and four from Goumlrjearingn(GJ) River (Fig 1) Because M margaritifera is an endan-gered species (Moorkens et al 2018) we refrained from col-




















ca 90 m asl



















minus434 (2)













1 100002 4266 57343 2797 3147 40564 2238 2028 2447 32875 1818 1539 1888 2027 27286 1538 1259 1469 1678 1818 22387 1329 1188 1119 1329 1399 1608 20288 1159 1079 909 1119 1258 1189 1469 18189 1049 979 769 909 1049 1189 1049 1329 167810 979 839 769 770 909 978 980 1049 1189 153811 909 769 770 559 769 839 979 840 1049 1048 146912 839 699 700 559 699 770 839 839 769 1049 909 132913 769 630 699 559 560 629 769 840 699 769 979 839 125914 769 560 629 559 490 629 630 699 699 770 628 910 839 118915 629 630 559 560 419 560 559 699 630 699 629 700 838 770 111916 629 560 559 490 419 490 559 560 629 490 769 559 770 699 769 1049


3 Results

























R = 100R2 = 100p = 0006


R = 091R2= 083


R = 095R2= 090

p = 1000



R = 098R2= 096

p = 0134

R = 099R2= 097


R = 099R2 = 098p = 0045

R = 086R2= 075


R = 046R2= 021

p = 1000

R = 064R2= 041

p = 1000



R = 099R2 = 098p = 0035

R = 099R2 = 098p = 0034


R = 099R2= 098

p = 0070

R = 095R2= 091


R = 076R2= 058

p = 1000

R = 089R2= 080

p = 0484




















wNAO(DJFM)

R = 067R2 = 044p lt 00001

R = 049R2 = 024p = 00011

R = 039R2 = 016p lt 00001

R = 070R2 = 049p lt 00001

R = 052R2 = 027p = 00005

R = 042R2 = 018p lt 00001

R =minus018R2= 003

p = 10000

R =minus031R2= 010

p = 01911

R =minus010R2= 001

p = 10000


R = 065R2 = 043p lt 00001

R = 052R2 = 027p = 00008

R = 060R2 = 036p lt 00001

R = 068R2 = 046p lt 00001

R = 056R2 = 031p = 00002

R = 065R2 = 042p lt 00001

R =minus017R2= 003

p = 10000

R =minus030R2= 009

p = 02657

R = 014R2= 002

p = 10000

sNAO (JJA) R = 038R2 = 014p = 00293

R = 040R2 = 016p = 00138

R = 020R2= 004

p = 00704

R = 029R2= 009

p = 01451

R = 034R2= 011

p = 00593

R = 002R2= 000

p = 10000

R = 012R2= 001

p = 10000

R = 001R2= 000

p = 10000

R = 004R2= 000

p = 10000


R = 065R2 = 043p lt 00001

R = 040R2 = 016p = 00212

R = 038R2 = 014p = 00333

R = 027R2= 007

p = 02172

R = 032R2= 010

p = 00985

R = 026R2= 007

p = 02581

R = 013R2= 002

p = 10000

R = 010R2= 001

p = 10000

R = 015R2= 002

p = 10000


p = 03812

R =minus023R2= 005

p = 06938

R = 008R2= 001

p = 10000

SGI1947ndash1991

R =minus027R2= 007

p = 04202

R =minus022R2= 005

p = 09238

R = 010R2= 001

p = 10000










4 Discussion

























































wNAO1950ndash1998

R = 060R2 = 036p = 00007

R = 065R2 = 042p lt 00001


R = 070R2 = 051p = 00001

R = 076R2 = 058p lt 00001


R = 045R2= 022

p = 00710

R = 050R2 = 026p = 00256

wNAO R = 062R2 = 038p = 00001

R = 046R2 = 022p = 00075

R = 035R2 = 013p = 00008

R = 066R2 = 043p lt 00001

R = 049R2 = 024p = 00028

R = 038R2 = 015p = 00001

wNAO1947ndash1991

R = 060R2 = 036p = 00003

R = 048R2 = 024p = 00088

R = 051R2 = 027p = 00067

R = 064R2 = 041p lt 00001

R = 051R2 = 027p = 00026

R = 057R2 = 033p = 00007













Appendix A


Streams studied


Sclerochronology












Temperature














References

























































































Abstract

Introduction


Sample preparation







Results






Discussion









Appendix A


Sample availability

Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References



















ca 90 m asl



















minus434 (2)













1 100002 4266 57343 2797 3147 40564 2238 2028 2447 32875 1818 1539 1888 2027 27286 1538 1259 1469 1678 1818 22387 1329 1188 1119 1329 1399 1608 20288 1159 1079 909 1119 1258 1189 1469 18189 1049 979 769 909 1049 1189 1049 1329 167810 979 839 769 770 909 978 980 1049 1189 153811 909 769 770 559 769 839 979 840 1049 1048 146912 839 699 700 559 699 770 839 839 769 1049 909 132913 769 630 699 559 560 629 769 840 699 769 979 839 125914 769 560 629 559 490 629 630 699 699 770 628 910 839 118915 629 630 559 560 419 560 559 699 630 699 629 700 838 770 111916 629 560 559 490 419 490 559 560 629 490 769 559 770 699 769 1049


3 Results

























R = 100R2 = 100p = 0006


R = 091R2= 083


R = 095R2= 090

p = 1000



R = 098R2= 096

p = 0134

R = 099R2= 097


R = 099R2 = 098p = 0045

R = 086R2= 075


R = 046R2= 021

p = 1000

R = 064R2= 041

p = 1000



R = 099R2 = 098p = 0035

R = 099R2 = 098p = 0034


R = 099R2= 098

p = 0070

R = 095R2= 091


R = 076R2= 058

p = 1000

R = 089R2= 080

p = 0484




















wNAO(DJFM)

R = 067R2 = 044p lt 00001

R = 049R2 = 024p = 00011

R = 039R2 = 016p lt 00001

R = 070R2 = 049p lt 00001

R = 052R2 = 027p = 00005

R = 042R2 = 018p lt 00001

R =minus018R2= 003

p = 10000

R =minus031R2= 010

p = 01911

R =minus010R2= 001

p = 10000


R = 065R2 = 043p lt 00001

R = 052R2 = 027p = 00008

R = 060R2 = 036p lt 00001

R = 068R2 = 046p lt 00001

R = 056R2 = 031p = 00002

R = 065R2 = 042p lt 00001

R =minus017R2= 003

p = 10000

R =minus030R2= 009

p = 02657

R = 014R2= 002

p = 10000

sNAO (JJA) R = 038R2 = 014p = 00293

R = 040R2 = 016p = 00138

R = 020R2= 004

p = 00704

R = 029R2= 009

p = 01451

R = 034R2= 011

p = 00593

R = 002R2= 000

p = 10000

R = 012R2= 001

p = 10000

R = 001R2= 000

p = 10000

R = 004R2= 000

p = 10000


R = 065R2 = 043p lt 00001

R = 040R2 = 016p = 00212

R = 038R2 = 014p = 00333

R = 027R2= 007

p = 02172

R = 032R2= 010

p = 00985

R = 026R2= 007

p = 02581

R = 013R2= 002

p = 10000

R = 010R2= 001

p = 10000

R = 015R2= 002

p = 10000


p = 03812

R =minus023R2= 005

p = 06938

R = 008R2= 001

p = 10000

SGI1947ndash1991

R =minus027R2= 007

p = 04202

R =minus022R2= 005

p = 09238

R = 010R2= 001

p = 10000










4 Discussion

























































wNAO1950ndash1998

R = 060R2 = 036p = 00007

R = 065R2 = 042p lt 00001


R = 070R2 = 051p = 00001

R = 076R2 = 058p lt 00001


R = 045R2= 022

p = 00710

R = 050R2 = 026p = 00256

wNAO R = 062R2 = 038p = 00001

R = 046R2 = 022p = 00075

R = 035R2 = 013p = 00008

R = 066R2 = 043p lt 00001

R = 049R2 = 024p = 00028

R = 038R2 = 015p = 00001

wNAO1947ndash1991

R = 060R2 = 036p = 00003

R = 048R2 = 024p = 00088

R = 051R2 = 027p = 00067

R = 064R2 = 041p lt 00001

R = 051R2 = 027p = 00026

R = 057R2 = 033p = 00007













Appendix A


Streams studied


Sclerochronology












Temperature














References

























































































Abstract

Introduction


Sample preparation







Results






Discussion









Appendix A


Sample availability

Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References
















minus434 (2)













1 100002 4266 57343 2797 3147 40564 2238 2028 2447 32875 1818 1539 1888 2027 27286 1538 1259 1469 1678 1818 22387 1329 1188 1119 1329 1399 1608 20288 1159 1079 909 1119 1258 1189 1469 18189 1049 979 769 909 1049 1189 1049 1329 167810 979 839 769 770 909 978 980 1049 1189 153811 909 769 770 559 769 839 979 840 1049 1048 146912 839 699 700 559 699 770 839 839 769 1049 909 132913 769 630 699 559 560 629 769 840 699 769 979 839 125914 769 560 629 559 490 629 630 699 699 770 628 910 839 118915 629 630 559 560 419 560 559 699 630 699 629 700 838 770 111916 629 560 559 490 419 490 559 560 629 490 769 559 770 699 769 1049


3 Results

























R = 100R2 = 100p = 0006


R = 091R2= 083


R = 095R2= 090

p = 1000



R = 098R2= 096

p = 0134

R = 099R2= 097


R = 099R2 = 098p = 0045

R = 086R2= 075


R = 046R2= 021

p = 1000

R = 064R2= 041

p = 1000



R = 099R2 = 098p = 0035

R = 099R2 = 098p = 0034


R = 099R2= 098

p = 0070

R = 095R2= 091


R = 076R2= 058

p = 1000

R = 089R2= 080

p = 0484




















wNAO(DJFM)

R = 067R2 = 044p lt 00001

R = 049R2 = 024p = 00011

R = 039R2 = 016p lt 00001

R = 070R2 = 049p lt 00001

R = 052R2 = 027p = 00005

R = 042R2 = 018p lt 00001

R =minus018R2= 003

p = 10000

R =minus031R2= 010

p = 01911

R =minus010R2= 001

p = 10000


R = 065R2 = 043p lt 00001

R = 052R2 = 027p = 00008

R = 060R2 = 036p lt 00001

R = 068R2 = 046p lt 00001

R = 056R2 = 031p = 00002

R = 065R2 = 042p lt 00001

R =minus017R2= 003

p = 10000

R =minus030R2= 009

p = 02657

R = 014R2= 002

p = 10000

sNAO (JJA) R = 038R2 = 014p = 00293

R = 040R2 = 016p = 00138

R = 020R2= 004

p = 00704

R = 029R2= 009

p = 01451

R = 034R2= 011

p = 00593

R = 002R2= 000

p = 10000

R = 012R2= 001

p = 10000

R = 001R2= 000

p = 10000

R = 004R2= 000

p = 10000


R = 065R2 = 043p lt 00001

R = 040R2 = 016p = 00212

R = 038R2 = 014p = 00333

R = 027R2= 007

p = 02172

R = 032R2= 010

p = 00985

R = 026R2= 007

p = 02581

R = 013R2= 002

p = 10000

R = 010R2= 001

p = 10000

R = 015R2= 002

p = 10000


p = 03812

R =minus023R2= 005

p = 06938

R = 008R2= 001

p = 10000

SGI1947ndash1991

R =minus027R2= 007

p = 04202

R =minus022R2= 005

p = 09238

R = 010R2= 001

p = 10000










4 Discussion

























































wNAO1950ndash1998

R = 060R2 = 036p = 00007

R = 065R2 = 042p lt 00001


R = 070R2 = 051p = 00001

R = 076R2 = 058p lt 00001


R = 045R2= 022

p = 00710

R = 050R2 = 026p = 00256

wNAO R = 062R2 = 038p = 00001

R = 046R2 = 022p = 00075

R = 035R2 = 013p = 00008

R = 066R2 = 043p lt 00001

R = 049R2 = 024p = 00028

R = 038R2 = 015p = 00001

wNAO1947ndash1991

R = 060R2 = 036p = 00003

R = 048R2 = 024p = 00088

R = 051R2 = 027p = 00067

R = 064R2 = 041p lt 00001

R = 051R2 = 027p = 00026

R = 057R2 = 033p = 00007













Appendix A


Streams studied


Sclerochronology












Temperature














References

























































































Abstract

Introduction


Sample preparation







Results






Discussion









Appendix A


Sample availability

Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References





1 100002 4266 57343 2797 3147 40564 2238 2028 2447 32875 1818 1539 1888 2027 27286 1538 1259 1469 1678 1818 22387 1329 1188 1119 1329 1399 1608 20288 1159 1079 909 1119 1258 1189 1469 18189 1049 979 769 909 1049 1189 1049 1329 167810 979 839 769 770 909 978 980 1049 1189 153811 909 769 770 559 769 839 979 840 1049 1048 146912 839 699 700 559 699 770 839 839 769 1049 909 132913 769 630 699 559 560 629 769 840 699 769 979 839 125914 769 560 629 559 490 629 630 699 699 770 628 910 839 118915 629 630 559 560 419 560 559 699 630 699 629 700 838 770 111916 629 560 559 490 419 490 559 560 629 490 769 559 770 699 769 1049


3 Results

























R = 100R2 = 100p = 0006


R = 091R2= 083


R = 095R2= 090

p = 1000



R = 098R2= 096

p = 0134

R = 099R2= 097


R = 099R2 = 098p = 0045

R = 086R2= 075


R = 046R2= 021

p = 1000

R = 064R2= 041

p = 1000



R = 099R2 = 098p = 0035

R = 099R2 = 098p = 0034


R = 099R2= 098

p = 0070

R = 095R2= 091


R = 076R2= 058

p = 1000

R = 089R2= 080

p = 0484




















wNAO(DJFM)

R = 067R2 = 044p lt 00001

R = 049R2 = 024p = 00011

R = 039R2 = 016p lt 00001

R = 070R2 = 049p lt 00001

R = 052R2 = 027p = 00005

R = 042R2 = 018p lt 00001

R =minus018R2= 003

p = 10000

R =minus031R2= 010

p = 01911

R =minus010R2= 001

p = 10000


R = 065R2 = 043p lt 00001

R = 052R2 = 027p = 00008

R = 060R2 = 036p lt 00001

R = 068R2 = 046p lt 00001

R = 056R2 = 031p = 00002

R = 065R2 = 042p lt 00001

R =minus017R2= 003

p = 10000

R =minus030R2= 009

p = 02657

R = 014R2= 002

p = 10000

sNAO (JJA) R = 038R2 = 014p = 00293

R = 040R2 = 016p = 00138

R = 020R2= 004

p = 00704

R = 029R2= 009

p = 01451

R = 034R2= 011

p = 00593

R = 002R2= 000

p = 10000

R = 012R2= 001

p = 10000

R = 001R2= 000

p = 10000

R = 004R2= 000

p = 10000


R = 065R2 = 043p lt 00001

R = 040R2 = 016p = 00212

R = 038R2 = 014p = 00333

R = 027R2= 007

p = 02172

R = 032R2= 010

p = 00985

R = 026R2= 007

p = 02581

R = 013R2= 002

p = 10000

R = 010R2= 001

p = 10000

R = 015R2= 002

p = 10000


p = 03812

R =minus023R2= 005

p = 06938

R = 008R2= 001

p = 10000

SGI1947ndash1991

R =minus027R2= 007

p = 04202

R =minus022R2= 005

p = 09238

R = 010R2= 001

p = 10000










4 Discussion

























































wNAO1950ndash1998

R = 060R2 = 036p = 00007

R = 065R2 = 042p lt 00001


R = 070R2 = 051p = 00001

R = 076R2 = 058p lt 00001


R = 045R2= 022

p = 00710

R = 050R2 = 026p = 00256

wNAO R = 062R2 = 038p = 00001

R = 046R2 = 022p = 00075

R = 035R2 = 013p = 00008

R = 066R2 = 043p lt 00001

R = 049R2 = 024p = 00028

R = 038R2 = 015p = 00001

wNAO1947ndash1991

R = 060R2 = 036p = 00003

R = 048R2 = 024p = 00088

R = 051R2 = 027p = 00067

R = 064R2 = 041p lt 00001

R = 051R2 = 027p = 00026

R = 057R2 = 033p = 00007













Appendix A


Streams studied


Sclerochronology












Temperature














References

























































































Abstract

Introduction


Sample preparation







Results






Discussion









Appendix A


Sample availability

Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References















R = 100R2 = 100p = 0006


R = 091R2= 083


R = 095R2= 090

p = 1000



R = 098R2= 096

p = 0134

R = 099R2= 097


R = 099R2 = 098p = 0045

R = 086R2= 075


R = 046R2= 021

p = 1000

R = 064R2= 041

p = 1000



R = 099R2 = 098p = 0035

R = 099R2 = 098p = 0034


R = 099R2= 098

p = 0070

R = 095R2= 091


R = 076R2= 058

p = 1000

R = 089R2= 080

p = 0484




















wNAO(DJFM)

R = 067R2 = 044p lt 00001

R = 049R2 = 024p = 00011

R = 039R2 = 016p lt 00001

R = 070R2 = 049p lt 00001

R = 052R2 = 027p = 00005

R = 042R2 = 018p lt 00001

R =minus018R2= 003

p = 10000

R =minus031R2= 010

p = 01911

R =minus010R2= 001

p = 10000


R = 065R2 = 043p lt 00001

R = 052R2 = 027p = 00008

R = 060R2 = 036p lt 00001

R = 068R2 = 046p lt 00001

R = 056R2 = 031p = 00002

R = 065R2 = 042p lt 00001

R =minus017R2= 003

p = 10000

R =minus030R2= 009

p = 02657

R = 014R2= 002

p = 10000

sNAO (JJA) R = 038R2 = 014p = 00293

R = 040R2 = 016p = 00138

R = 020R2= 004

p = 00704

R = 029R2= 009

p = 01451

R = 034R2= 011

p = 00593

R = 002R2= 000

p = 10000

R = 012R2= 001

p = 10000

R = 001R2= 000

p = 10000

R = 004R2= 000

p = 10000


R = 065R2 = 043p lt 00001

R = 040R2 = 016p = 00212

R = 038R2 = 014p = 00333

R = 027R2= 007

p = 02172

R = 032R2= 010

p = 00985

R = 026R2= 007

p = 02581

R = 013R2= 002

p = 10000

R = 010R2= 001

p = 10000

R = 015R2= 002

p = 10000


p = 03812

R =minus023R2= 005

p = 06938

R = 008R2= 001

p = 10000

SGI1947ndash1991

R =minus027R2= 007

p = 04202

R =minus022R2= 005

p = 09238

R = 010R2= 001

p = 10000










4 Discussion

























































wNAO1950ndash1998

R = 060R2 = 036p = 00007

R = 065R2 = 042p lt 00001


R = 070R2 = 051p = 00001

R = 076R2 = 058p lt 00001


R = 045R2= 022

p = 00710

R = 050R2 = 026p = 00256

wNAO R = 062R2 = 038p = 00001

R = 046R2 = 022p = 00075

R = 035R2 = 013p = 00008

R = 066R2 = 043p lt 00001

R = 049R2 = 024p = 00028

R = 038R2 = 015p = 00001

wNAO1947ndash1991

R = 060R2 = 036p = 00003

R = 048R2 = 024p = 00088

R = 051R2 = 027p = 00067

R = 064R2 = 041p lt 00001

R = 051R2 = 027p = 00026

R = 057R2 = 033p = 00007













Appendix A


Streams studied


Sclerochronology












Temperature














References

























































































Abstract

Introduction


Sample preparation







Results






Discussion









Appendix A


Sample availability

Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References















wNAO(DJFM)

R = 067R2 = 044p lt 00001

R = 049R2 = 024p = 00011

R = 039R2 = 016p lt 00001

R = 070R2 = 049p lt 00001

R = 052R2 = 027p = 00005

R = 042R2 = 018p lt 00001

R =minus018R2= 003

p = 10000

R =minus031R2= 010

p = 01911

R =minus010R2= 001

p = 10000


R = 065R2 = 043p lt 00001

R = 052R2 = 027p = 00008

R = 060R2 = 036p lt 00001

R = 068R2 = 046p lt 00001

R = 056R2 = 031p = 00002

R = 065R2 = 042p lt 00001

R =minus017R2= 003

p = 10000

R =minus030R2= 009

p = 02657

R = 014R2= 002

p = 10000

sNAO (JJA) R = 038R2 = 014p = 00293

R = 040R2 = 016p = 00138

R = 020R2= 004

p = 00704

R = 029R2= 009

p = 01451

R = 034R2= 011

p = 00593

R = 002R2= 000

p = 10000

R = 012R2= 001

p = 10000

R = 001R2= 000

p = 10000

R = 004R2= 000

p = 10000


R = 065R2 = 043p lt 00001

R = 040R2 = 016p = 00212

R = 038R2 = 014p = 00333

R = 027R2= 007

p = 02172

R = 032R2= 010

p = 00985

R = 026R2= 007

p = 02581

R = 013R2= 002

p = 10000

R = 010R2= 001

p = 10000

R = 015R2= 002

p = 10000


p = 03812

R =minus023R2= 005

p = 06938

R = 008R2= 001

p = 10000

SGI1947ndash1991

R =minus027R2= 007

p = 04202

R =minus022R2= 005

p = 09238

R = 010R2= 001

p = 10000










4 Discussion

























































wNAO1950ndash1998

R = 060R2 = 036p = 00007

R = 065R2 = 042p lt 00001


R = 070R2 = 051p = 00001

R = 076R2 = 058p lt 00001


R = 045R2= 022

p = 00710

R = 050R2 = 026p = 00256

wNAO R = 062R2 = 038p = 00001

R = 046R2 = 022p = 00075

R = 035R2 = 013p = 00008

R = 066R2 = 043p lt 00001

R = 049R2 = 024p = 00028

R = 038R2 = 015p = 00001

wNAO1947ndash1991

R = 060R2 = 036p = 00003

R = 048R2 = 024p = 00088

R = 051R2 = 027p = 00067

R = 064R2 = 041p lt 00001

R = 051R2 = 027p = 00026

R = 057R2 = 033p = 00007













Appendix A


Streams studied


Sclerochronology












Temperature














References

























































































Abstract

Introduction


Sample preparation







Results






Discussion









Appendix A


Sample availability

Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References









4 Discussion

























































wNAO1950ndash1998

R = 060R2 = 036p = 00007

R = 065R2 = 042p lt 00001


R = 070R2 = 051p = 00001

R = 076R2 = 058p lt 00001


R = 045R2= 022

p = 00710

R = 050R2 = 026p = 00256

wNAO R = 062R2 = 038p = 00001

R = 046R2 = 022p = 00075

R = 035R2 = 013p = 00008

R = 066R2 = 043p lt 00001

R = 049R2 = 024p = 00028

R = 038R2 = 015p = 00001

wNAO1947ndash1991

R = 060R2 = 036p = 00003

R = 048R2 = 024p = 00088

R = 051R2 = 027p = 00067

R = 064R2 = 041p lt 00001

R = 051R2 = 027p = 00026

R = 057R2 = 033p = 00007













Appendix A


Streams studied


Sclerochronology












Temperature














References

























































































Abstract

Introduction


Sample preparation







Results






Discussion









Appendix A


Sample availability

Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References




















































wNAO1950ndash1998

R = 060R2 = 036p = 00007

R = 065R2 = 042p lt 00001


R = 070R2 = 051p = 00001

R = 076R2 = 058p lt 00001


R = 045R2= 022

p = 00710

R = 050R2 = 026p = 00256

wNAO R = 062R2 = 038p = 00001

R = 046R2 = 022p = 00075

R = 035R2 = 013p = 00008

R = 066R2 = 043p lt 00001

R = 049R2 = 024p = 00028

R = 038R2 = 015p = 00001

wNAO1947ndash1991

R = 060R2 = 036p = 00003

R = 048R2 = 024p = 00088

R = 051R2 = 027p = 00067

R = 064R2 = 041p lt 00001

R = 051R2 = 027p = 00026

R = 057R2 = 033p = 00007













Appendix A


Streams studied


Sclerochronology












Temperature














References

























































































Abstract

Introduction


Sample preparation







Results






Discussion









Appendix A


Sample availability

Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References





Appendix A


Streams studied


Sclerochronology












Temperature














References

























































































Abstract

Introduction


Sample preparation







Results






Discussion









Appendix A


Sample availability

Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References











References

























































































Abstract

Introduction


Sample preparation







Results






Discussion









Appendix A


Sample availability

Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References








































































Abstract

Introduction


Sample preparation







Results






Discussion









Appendix A


Sample availability

Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

freshwater pearl mussels from northern sweden serve as ...multi-decadal records of 18o signals in...

Documents