Marine reservoir age variability and water mass distribution in the Iceland Sea

Transcription

1 Quaternary Science Reviews 23 (2004) Marine reservoir age variability and water mass distribution in the Iceland Sea Jo n Eirı ksson a,, Gudru n Larsen a, Karen Luise Knudsen b, Jan Heinemeier c, Leifur A. Sı monarson a a Science Institute, University of Iceland, IS-101 Reykjavík, Iceland b Department of Earth Sciences, University of Aarhus, DK-8000 Århus C, Denmark c AMS 14 C Dating Laboratory, Institute of Physics and Astronomy, University of Aarhus, DK-8000 Århus C, Denmark Abstract Lateglacial and Holocene tephra markers from Icelandic source volcanoes have been identified in five sediment cores from the North Icelandic shelf and correlated with tephra layers in reference soil sections in North Iceland and the GRIP ice core. Land-sea correlation of tephra markers, that have been radiocarbon dated with terrestrial material or dated by documentary evidence, provides a tool for monitoring reservoir age variability in the region. Age models developed for the shelf sediments north of Iceland, based on offshore tephrochronology on one hand and on calibrated AMS 14 C datings of marine molluscs on the other, display major deviations during the last 4500 years. The inferred temporal variability in the reservoir age of the regional water masses exceeds by far the variability expected from the marine model calculations. The observed reservoir ages are generally considerably higher, by up to 450 years, than the standard model ocean. It is postulated that the intervals with increased and variable marine reservoir age reflect incursions of Arctic water masses derived from the East Greenland Current to the Iceland Sea and the North Icelandic shelf. r 2004 Elsevier Ltd. All rights reserved. 1. Introduction Palaeoclimatic studies that extend beyond historical and instrumental records must rely on dating techniques which enable reconstruction of time series for climatevariability related proxies. Marine sediment cores from high resolution sedimentary basins on the North Icelandic shelf provide archives of Holocene palaeoceanographic changes in the vicinity of the oceanic Polar Front (Fig. 1). The purpose of this study is to construct reliable age models for palaeo-climatic data from the marine environment in the region and to interpret the discrepancies between the AMS 14 C datings and the tephrochronological age models in terms of changes in the water mass distribution. Corresponding author. Tel.: ; fax: address: jeir@rhi.hi.is (J. Eiríksson). The present position of the Polar Front separates Arctic surface water of the East Greenland Current and the East Icelandic Current frombranches of the North Atlantic Current to the west and north of Iceland. The Arctic surface water of the East Icelandic Current is partly derived fromthe East Greenland Current (Polar water) and partly fromwesterly eddies of the Norwegian Atlantic Current (Atlantic water) (e.g. Stefa nsson, 1962; Hansen and Østerhus, 2000). The strong gradients both in the ocean and in the atmosphere make this region extremely sensitive to climatic changes. The chronology for recent studies of the North Icelandic shelf sediments has been based on radiocarbon dates (Andrews et al., 2000, 2001a, b; Andrews and Giraudeau, 2003) or on combined tephrochonology and AMS 14 C datings of either molluscs or benthic or planktonic foraminifera (Eirı ksson et al., 2000a, b; Jiang et al., 2002; Knudsen and Eirı ksson, 2002; Larsen et al., 2002). The application of tephra-based age models has /$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi: /j.quascirev

2 2248 J. Eiríksson et al. / Quaternary Science Reviews 23 (2004) Fig. 1. The regional modern surface circulation around Iceland and the position of the marine Polar Front. Depth contour intervals 1000 m (modified after Knudsen and Eiríksson, 2002, and Hurdle, 1986, Map A8). TFZ=Tjörnes Fracture Zone. The inset location map shows the study area off North Iceland (depth contour intervals 100 m). revealed that AMS 14 C dates frommolluscs show variable deviation fromtephrochronological age models, indicating that the ocean reservoir age at the coring sites has varied with time (Eirı ksson et al., 2000a, b; Knudsen and Eirı ksson, 2002; Larsen et al., 2002). The results of these studies demonstrate a real need for independent control on 14 C dating of marine sediment cores obtained froman oceanographic boundary region such as the Polar Front separating the Atlantic and Arctic water masses. The study area on the North Icelandic shelf has the advantage of being close to numerous source volcanoes of Holocene tephras (Fig. 2). Tephra markers that can be traced fromvolcanic source regions (see Fig. 2) into the marine depositional environment can provide control on radiocarbon dates fromthat environment. The ages of Icelandic tephra markers used in this study are based on historical records fromiceland for the last 900 years, on correlation to the Greenland ice-core chronology and on radiocarbon dates of terrestrial material. It was demonstrated by Larsen et al. (2002) that a high resolution tephra stratigraphy can be used to link chronologies in the terrestrial North Iceland and a high resolution marine sediment core (MD992275). The investigation presented a detailed land-sea correlation of the regional terrestrial tephrochronology with the

3 J. Eiríksson et al. / Quaternary Science Reviews 23 (2004) Fig. 2. The most relevant volcanic systems for the tephrochronology of northern Iceland and the North Icelandic shelf. marine record. This comparison of a tephrochronological age model and a radiocarbon one makes it possible to study local reservoir age problems. The pre-bomb reservoir age of the coastal waters of Iceland was determined by Ha kansson (1983) to yr. This value is 35 years lower than the 400 yr reservoir age correction used by Andersen et al. (1989) for Icelandic coastal samples and the ca. 400 yr correction conventionally applied to marine 14 C ages (Stuiver et al., 1998a, b). In discussing reservoir age variability in the region, Austin et al. (1994) noted that at present there is no surface 14 C gradient between 401 and 70 1N in the North Atlantic surface waters. The coastal and shelf surface waters around Iceland today are dominated by the Irminger Current (Fig. 1), which is derived fromthe North Atlantic Current and thus has the same affinity as the waters off western Norway, the Faroe Islands and the British Isles. In contrast, the reservoir age of the East Greenland Current north of the Polar Front, has been determined to 530 yr (Tauber and Funder, 1975) and to yr (Hjort, 1973; Ha kansson, 1983). This means that the recent, pre-bomb apparent age difference across the Polar Front is about yr. It is suggested that discrepancies between the two age models are related to palaeoceanographic changes in the region and resulting changes in the reservoir age of the water masses on the North Icelandic shelf. The Holocene Climatic Optimum in this area is reflected by a pronounced influence of Atlantic waters, brought to the North Icelandic shelf by the Irminger Current already at about 10,300 cal. BP (Knudsen et al., 2004b). In that context it is notable that Eirı ksson et al. (2000b) did show that a 400 year reservoir correction is applicable at the level of the Saksunarvatn ash (10,200 cal. BP). During Lateglacial conditions, however, Arctic water masses prevailed, and a reservoir correction of years is required at the level of the Lateglacial Vedde Ash in this region (Eirı ksson et al., 2000b; Haflidason et al., 2000). Results from the Late Holocene record of a series of shelf cores, analysed with a time resolution of up to 5 yr/cm, are included in this study. The investigations are organised jointly by the University of Iceland and the University of Aarhus, Denmark, as the umbrella project PANIS (Palaeoenvironments of the North Icelandic Shelf). Results fromfive of these cores are treated in this paper.

4 2250 J. Eiríksson et al. / Quaternary Science Reviews 23 (2004) Materials and methods Five sediment records from the Tjo rnes Fracture Zone (TFZ) on the shelf north of Iceland provide data on palaeoceanographic variability in the region (Fig. 1). Samples for the present study were obtained from box cores, gravity cores and CALYPSO piston cores acquired on two research cruises (see inset map, Fig. 1). The gravity cores HM and HM and box cores at the same sites were obtained on the 1995 BIOICE cruise on RV Haakan Mosby. The piston cores MD992271, MD and MD were obtained on the 1999 GINNA cruise on RV Marion Dufresne, organised by the IMAGES programme. The Tjo rnes Fracture Zone is topographically expressed as a series of north south trending extensional troughs and ridges offsetting the Kolbeinsey Ridge spreading axis, which extends onto the northernmost part of the shelf eastwards to O xarfjo rdur, where the spreading axis continues southwards across north Iceland. Two of the cores, HM and MD (464 and 665 mwater depth) are located in Eyjafjardara ll, a deep extensional trough south of Kolbeinsey Ridge. Cores MD and HM are located in shallower water (315 and 400 mdepth) to the west of Eyjafjardara ll. The easternmost core, MD (440 m depth), is located in the Skja lfandadju p trough, which separates two of the TFZ volcanically active shallow submarine ridges. Age models based on major Lateglacial and Holocene tephra markers indicate sustained high and relatively uniformsedimentation rates at the CALYPSO coring sites (Fig. 3), and the same is true for the gravity cores covering the upper Holocene. All cores were subsampled by extracting 1 cmthick slices of a split core. The various sedimentological, micropalaeontological and stable isotope parameters have generally been obtained at a time resolution of years. Every cm of each core has been checked for available material for AMS 14 C dating. The sample levels for dating have mostly been chosen with respect to possible dating of important stratigraphic levels such as biozone boundaries and tephra layers. Major tephra markers were identified by visual inspection of split cores, and on X-ray films. These were subsampled and sieved on a 63 mmsieve for the preparation of polished thin sections. Other tephra Fig. 3. Age-depth diagrams for cores MD992271, MD and MD (for details in the Late Holocene part, see Figs. 5, 8 and 9) showing tephra marker horizons back to the level of the Vedde Ash. The age models are based only on tephra marker horizons (tephrochronological age models). Some of the tephra layers are historically dated, others are dated on terrestrial material. The core depth is plotted on the x-axis. The position of each tephra layer in the cores is shown as a vertical line. The calibrated age is plotted on the y- axis and the age of each tephra layer is shown by a horizontal line. Cross indicates a 14 C age with one s error bars.

5 J. Eiríksson et al. / Quaternary Science Reviews 23 (2004) layers were identified as grain size anomalies represented by peaks in the sand fraction. Major element geochemical analysis of volcanic glass shards (Table 1A E) were carried out by standard wavelength dispersal technique on an ARL-SEMQ microprobe at the Geological Institute, University of Bergen, Norway, with an accelerating voltage of 15 kv, a beamcurrent of 10 na, and a defocused beamdiameter of 6 12 mm. Natural and synthetic minerals and glasses were used as standards (see also Larsen et al., 2002). Tephrochronological age models have been constructed for all the cores. The age models are constrained by first appearance depth of each marker, and the historical or terrestrial radiocarbon date of the marker. Between markers, linear interpolation has been used to estimate age of samples (straight line age model segments on Figs. 3 and 5 9). The calibrated radiocarbon dates are shown with 1 s error bars on Figs The deviation between each calibrated radiocarbon age and the tephrochronological age of the same level, termed Dr, is also shown with 1 s calibration error bars. The reservoir age offset (DR, Fig. 10) fromthe standard model ocean has been calculated for each radiocarbon date by means of the calibration curves. Although DR is the relevant physical parameter, directly related to the ocean 14 C concentration, we have chosen to show Dr because its value can be read directly off the graphs. As seen from Fig. 10, the two parameters vary almost identically. Post-settlement tephra markers are dated to an exact or an approximate year AD. For pre-settlement tephra markers, there is an additional error associated with the terrestrial radiocarbon dating error. This affects the markers Snæfellsjo kull 1 ( C yr BP, Larsen et al., 2002), Hekla 3 ( C yr BP, Dugmore et al., 1995), and Hekla 4 ( C yr BP, Dugmore et al., 1995). Because these errors are relatively small compared to the deviations, they have not been incorporated in the graphs presented here. Most of the tephra markers display a northerly dispersal pattern and were deposited as airfall tephra, both in the terrestrial and marine environment. Sea-ice cover may occasionally have delayed deposition on the sea-floor, but due to summer melting it is considered unlikely that this delay has ever exceeded one season in the upper Holocene records. The foraminiferal samples, each representing a 1 cm sediment slice, were analysed at 1 5 cm intervals. They were washed through 1000, 125 and 63 mmsieves according to the methods described by Feyling-Hanssen et al. (1971) and Knudsen (1998). Total benthic foraminiferal assemblages were analysed in the 125 mm fraction by counting at least 300 specimens of benthic foraminifera, when possible. The palaeoecological interpretations are based on comparisons with modern faunal distributions in the North Atlantic region. Modern foraminiferal faunal distribution on the North Icelandic shelf has been described by Rytter et al. (2002) and Weiner et al. (1999) Marine dating material The 14 C datings of marine samples (Table 2) were carried out at the AMS 14 C Dating Laboratory at the University of Aarhus, Denmark. The dates have been corrected for natural isotopic fractionation by normalisation to 13 C= 25% VPDB, and calibrated with CALIB 4.1 (Stuiver et al., 1998a), using the marine model calibration curve (Stuiver et al., 1998b). The marine model is a calculated smooth calibration curve for the mixed layer of a model ocean. To accommodate local effects in the actual oceans, the quantity DR is defined as the difference between the measured sample age and the global model curve. For a given region, DR is expected to be constant in time to a first approximation, to the extent that the regional reservoir, from which the sample is taken, parallels the global marine model. Since the reservoir age is the difference between the strongly fluctuating atmospheric calibration curve and the slowly responding smooth ocean curve, it is expected to vary (typically up to 100 yr) with time for a given location (water mass). A standard reservoir correction of about C years (DR ¼ 0) is built into this model (see also Andersen et al., 1989). In this paper we simply refer to calibrated years BP (before 1950) as years. The dating material selected for AMS 14 C dating consists of a range of carbonate shells frommarine organisms. Available dating material did not permit the use of one single species throughout. Most of the dates are frommolluscs, which have the advantage of being less prone to reworking by bottomcurrents than e.g. foraminifera (Heier-Nielsen et al., 1995a). AMS 14 C dating of benthic and planktonic foraminifera at the same level as molluscs in the research area, however, yielded no significant age difference (Eirı ksson et al., 2000b). However, recent studies of shelf and shallow marine sediment dating in areas close to calcareous bedrock (Heier-Nielsen et al., 1995b; Dyke et al., 2002) and in sediments containing glacially eroded carbon (Forman and Polyak, 1997) have indicated anomalously high radiocarbon ages, specifically when dating deposit feeding molluscs. Consequently, the possible dating errors related to burrowing depths, feeding habits, and possible contamination by old carbon must be evaluated when reservoir problems are investigated. If molluscs incorporate old carbon from pore waters that is not derived directly fromthe sea-water, this may affect the radiocarbon dates and lead to too high age values. Therefore, species which feed upon phytoplankton, smaller forms of zooplankton and detritus in suspension were selected whenever possible in the

6 2252 J. Eiríksson et al. / Quaternary Science Reviews 23 (2004) Table 1 Representative microprobe glass analyses of tephra horizons relevant to this study from the five cores MD (A), HM (B), HM (C), MD (D) and MD (E) SiO 2 TiO 2 Al 2 O 3 FeO MnO MgO CaO Na 2 O K 2 O P 2 O 5 Sum (A) MD992271/52 Hekla n.d n.d n.d n.d n.d MD992271/ n.d Hekla n.d n.d n.d n.d MD992271/ Saksunarvatn MD992271/ Vedde (B) HM107-03/37 KOL (C) HM107-01/48 KOL n.d n.d n.d n.d n.d HM107-01/ n.d Hekla n.d n.d n.d n.d (D) MD992273/168 V n.d n.d n.d n.d

7 J. Eiríksson et al. / Quaternary Science Reviews 23 (2004) Table 1 (continued ) SiO 2 TiO 2 Al 2 O 3 FeO MnO MgO CaO Na 2 O K 2 O P 2 O 5 Sum n.d MD992273/ n.d Hekla n.d n.d n.d n.d n.d n.d MD992273/ n.d Hekla n.d n.d n.d n.d MD992273/ n.d Hekla n.d n.d n.d n.d MD992273/ n.d Hekla n.d n.d n.d n.d (E) MD992275/101 V n.d n.d n.d n.d n.d MD992275/1552 Hekla MD992275/2560 Saksunarvatn n.d n.d n.d n.d n.d MD992275/3405 Vedde n.d n.d n.d n.d n.d n.d n.d n.d n.d n.d In most cases five analyses with sums between 97.5% and 100.5% are presented for each sample. The exceptions are Vedde tephra from cores MD and MD where both the rhyolitic and basaltic component is presented, KOL 1372 where 10 analyses from core HM are presented (locus typicus), and Hekla 1300 from core MD where 7 analyses demonstrate the range of composition. The analyses do not show the complete range of compostion of each tephra layer. Depth refers to core barrel depths.

8 Table 2 Tephra marker horizons and radiocarbon dates in cores MD992271, HM (including top 10 cmfrombox core), HM (including top 10 cmfrombox core), MD and MD992275, calibrated with CALIB4 using the marine98 dataset (Stuiver et al., 1998a, b) 2254 Core Depth (cm) Lab no. Material 14 C age BP71 s Cal. yr BP71 s Cal. range (s), 1 s q 13 C Expected tephrochronological age Dr, cal. BP deviation DR, 14 C deviation MD AAR-7697 Siphonodentalium lobatum MD Hekla MD AAR-7698 cf. Lunatia pallida, Thyasira cf. equalis MD AAR-7699 Bathyarca glacialis MD AAR-7800 Yoldiella lenticula MD AAR-7701 Lunatia pallida MD AAR-7768 Thyasira equalis, Cuspidaria obesa, Thyasira cf. gouldi MD AAR-7769 Thyasira equalis MD Hekla MD AAR-7770 Yoldiella sp., Yoldiella intermedia, Thyasira equalis MD AAR-7771 Thyasira equalis, cf. Arctinula greenlandica MD AAR-7772 Bathyarca glacialis MD AAR-7773 Yoldiella sp MD Saksunarvatn ash MD Vedde Ash HM AAR-3698 Bathyarca glacialis HM KOL HM Hekla HM AAR-5105 Propebela cf. reticulata, Yoldiella sp., Oenopota sp. HM AAR-4029 Siphonodentalium lobatum HM AAR-5059 Thyasira sp., Arctinula greenlandica HM AAR-5747 Foraminifera, total benthic fauna HM AAR-2930 Yoldiella intermedia HM AAR-5746 Arctinula greenlandica, Yoldiella lenticula, Thyasira sp. HM AAR-3699 Thyasira cf. gouldi HM Hekla HM AAR-6081 Foraminifera, total benthic fauna HM AAR-3781 Bathyarca cf. glacialis Yoldiella sp., Bathyarca glacialis J. Eiríksson et al. / Quaternary Science Reviews 23 (2004)

9 HM AAR-3700 Yoldiella sp HM AAR-5744 Yoldiella fraterna HM Hekla HM AAR-5041 Yoldiella lenticula HM AAR-5106 Yoldiella fraterna HM AAR-2931 Thyasira sp., Yoldiella sp. HM AAR-3779 Siphonodentalium lobatum HM KOL HM Hekla HM KIA17226 Siphonodentalium lobatum HM AAR-4194 Foraminifera, total benthic fauna HM AAR-4196 Foraminifera, total benthic fauna HM AAR-5193 Siphonodentalium lobatum HM Hekla HM AAR-5104 Thyasira equalis HM KIA17227 Siphonodentalium lobatum HM AAR-3701 Bathyarca glacialis HM AAR-2929 Thyasira equalis MD AAR-8426 Thyasira equalis MD AAR-8337 Thyasira equalis MD AAR-8428 Thyasira equalis MD AAR-8427 Thyasira equalis MD V MD AAR-8330 Thyasira equalis MD AAR-8331 Thyasira equalis MD AAR-8412 Thyasira equalis MD AAR-8333 Thyasira equalis MD Hekla MD AAR-8334 Thyasira gouldi MD AAR-8429 Siphonodentalium lobatum MD Hekla MD AAR-8335 Thyasira equalis, Thyasira sp. MD AAR-8562 Thyasira equalis MD AAR-8563 Thyasira equalis MD AAR-8564 Cuspidaria glacialis MD AAR-8565 Thyasira equalis MD Hekla MD Hekla MD AAR-7702 Lunatia pallida MD AAR-7116 Thyasira equalis MD AAR-7117 Thyasira cf. equalis MD V MD AAR-6089 Siphonodentalium lobatum MD AAR-7118 Thyasira equalis J. Eiríksson et al. / Quaternary Science Reviews 23 (2004)

10 Table 2 (continued) 2256 Core Depth (cm) Lab no. Material 14 C age BP71 s Cal. yr BP71 s Cal. range (s), 1 s q 13 C Expected tephrochronological age Dr, cal. BP deviation DR, 14 C deviation MD AAR-7119 Nuculana sp MD V MD V MD AAR-7120 Thyasira equalis, Thyasira sp. MD Hekla MD AAR-7121 Thyasira equalis MD Hekla MD AAR-6931 Thyasira equalis MD Settlement tephra (V871/877) 1080 MD AAR-7122 Siphonodentalium lobatum MD AAR-6932 Siphonodentalium lobatum MD AAR-6933 Bathyarca glacialis MD AAR-7703 Thyasira equalis MD Snæfellsjo kull MD AAR-6934 cf. Dentalium entalis MD AAR-6935 cf. Siphonodentalium lobatum MD AAR-7123 Thyasira equalis MD AAR-6936 Thyasira sp MD AAR-7124 Thyasira cf. equalis MD Hekla MD AAR-6937 Siphonodentalium lobatum MD AAR-6938 Bathyarca glacialis MD AAR-6088 Siphonodentalium lobatum MD AAR-7125 Siphonodentalium lobatum MD Hekla MD Saksunarvatn ash MD Vedde Ash A standard reservoir correction of about 400 years (DR=0) is built into the radiocarbon model. The three right-hand columns indicate the expected tephrochronological age of each dated sample, the deviation between each calibrated radiocarbon age and the tephrochronological age of the same level (Dr) and the calculated DR value for each radiocarbon date (additional reservoir age correction to the 400 years). 1=assumed standard q 13 C value. J. Eiríksson et al. / Quaternary Science Reviews 23 (2004)

11 J. Eiríksson et al. / Quaternary Science Reviews 23 (2004) present study. Previous research has shown that molluscs build their carbonate shells directly from CO 2 in the ocean or frompore water in the burrowed sediment (Mook, 1971; Eisma et al., 1976), indicating that in the case of deposit feeders, burrowing depth is more important than feeding habits. The dated species fall into the three following groups with regard to feeding habit and food: (1) scavengers (carnivores), (2) deposit feeders and (3) suspension feeders. The scaphopods Dentalium entalis and Siphonodentalium lobatum are among the scavengers, living in muddy or sandy sediment, only the small posterior aperture projects above the surface (Muus, 1959). They feed on microscopic organisms, especially foraminifera in the surrounding sediment, and are physically adapted to selecting and capturing food particles fromthe environment. There is a well-developed radula in the mouth and even small jaws. The gastropod species Lunatia pallida, Propebela cf. reticulata and Oenopota sp. also have a well-developed radula in the mouth and feed upon other animals, living or carrion (Thorson, 1944). In East Greenland and around Jan Mayen and Iceland the two bivalve species Cuspidaria obesa and C. glacialis feed by swallowing smaller animals (Ockelmann, 1958). Five bivalve species used for AMS 14 C dating in the present project belong to the deposits feeders, feeding on organic particles deposited on and in the sediment of the bottom together with associated microorganisms. This group comprises Nuculana sp., Yoldiella lenticula, Y. intermedia, Y. fraterna and Yoldiella sp. (Ockelmann, 1958). The remaining four mollusc species are first and foremost suspension feeders, viz. Bathyarca glacialis, Arctinula greenlandica, Thyasira equalis and T. gouldi (Ockelmann, 1958; Bernard, 1979). T. equalis is the most frequent species in the samples analysed in the present study. In East Greenland and around Jan Mayen as well as Iceland it lives as a suspension feeder (Ockelmann, 1958), and in the Beaufort Sea it is also a filter feeder, but shows some modifications towards macrophagy (Bernard, 1979). Dando (2001) found it, however, in the UK part of the North Sea to live both as a suspension and deposit feeder. On the continental shelf west of central Norway, Myhrvold et al. (2004) reported it as feeding on deposits on the sediment surface. Such sediment is more or less water saturated and exchanges water with the water mass above. The sulphide mining of the species T. equalis is noteable. It uses the extensional foot (up to 13.9 cm) to gain sulphur to symbiotic bacteria living on their gills (Defour and Felbeck, 2003). However, this sulphide mining tendency cannot have affected our dates very much, as far as we can see. To test this, six pairs of molluscs from the same or nearly the same level in cores MD and MD were AMS 14 C dated, with T. equalis forming one half of each pair (Table 3). The q 13 C values measured on T. equalis are generally very negative ( 7 to 8%), and possibly some fractionation of carbon isotopes is associated with the bacterial symbiosis. The source of the negative values may relate to the animal s metabolism, or possibly to emanations of gas containing negative q 13 C values. As the values are consistently negative at all coring sites, it is considered less likely that the gas origin is the primary source. The dating results (Table 3, Fig. 4) show overlap in all cases and no significant age difference between T. equalis and other species, which include scavengers and deposit feeders. However, the T. equalis samples show a tendency towards higher ages in four cases out of six, the exceptions being the deposit feeding Yoldiella cf. Table 3 Radiocarbon dates and calibration results for mollusc pairs with differing feeding habits, dated at the same or nearly the same level in cores MD and MD One of the species in each pair is Thyasira equalis, which displays very low q 13 C values. For discussion see text and Table 2 Core Depth (cm) Lab no. Material 14 C age BP71 s Cal. yr BP71 s Cal. range (s), 1 s q 13 C MD AAR-7784 cf. Dentalium entalis MD AAR-7783 Thyasira equalis MD AAR-7785 Siphonodentalium lobatum MD AAR-7786 Thyasira equalis MD AAR-7788 Yoldiella cf. lenticula MD AAR-7789 Thyasira equalis MD AAR-7792 Siphonodentalium lobatum MD AAR-7793 Thyasira equalis MD AAR-7676 Oenopota sp MD AAR-7677 Thyasira equalis MD AAR-7770 Yoldiella sp MD AAR-7769 Thyasira equalis

12 2258 J. Eiríksson et al. / Quaternary Science Reviews 23 (2004) Fig. 4. Calibrated ages of pairwise dating of six samples at identical or close to identical depths in cores MD and MD992271, where one of the mollusc species is T. equalis, the other being scavenger or deposit feeder species. Probability plots and one s as well as two s error bars are shown. For discussion see text. lenticula, and Yoldiella sp. In high precision dating, this apparent tendency needs to be investigated further to test if the observed age difference tendency is significant or has arisen by chance. As there appears to be no significant difference in the age of the suspension feeder Thyasira equalis and deposit feeders or scavengers when dated pairwise at the same level in the sediment record, it is considered

13 J. Eiríksson et al. / Quaternary Science Reviews 23 (2004) highly unlikely that the so-called Portlandia effect (Dyke et al., 2002), i.e. contamination of the shells with old carbon due to deposit feeding, is a significant factor in deviations between tephrochronological age models and AMS 14 C based age models in the North Icelandic shelf cores. The deviations are up to 5 10 times larger than the observed (but not significant) species related age difference tendency. All dated occurrences of Yoldiella and Nuculana (deposit feeders) have been particularly checked, and there is no tendency for anomalous high ages. Total absence of detrital carbonate in the shelf sediments (Knudsen et al., 2004b) reflects the fact that there are no calcareous rocks in Iceland, which is the provenance region for the shelf sediments. Extremely high sedimentation rates in all the studied sediment cores are also considered to reduce the possibility of uptake of old organic carbon by burrowing animals. 3. Results 3.1. The tephra layers A total of thirteen different tephra layers, spanning the period 10, C BP to AD 1717 have been identified so far in the five cores presented here (from west to east MD992271, HM107-03, HM107-01, MD and MD992275). Selected intervals of these cores have been processed and the most distinct tephra horizons were sampled and analysed. In core MD992275, eleven tephra layers have been identified (Fig. 3), while three to five tephra layers were identified in the other cores (Figs. 3 and 5 9). Two out of the thirteen tephra layers occur in all five cores, i.e. a tephra layer fromthe AD 1104 Hekla eruption (Hekla 1104, Thorarinsson, 1967) and Hekla 3, C BP (Dugmore et al., 1995), both of silicic composition. The tephra stratigraphy has already been described for two of the cores, i.e. core HM by Eirı ksson et al. (2000a) and core MD by Larsen et al. (2002). Three tephra layers, previously undetected in the marine sediments off North Iceland, have been identified. Only these are treated in detail here. In core HM107-03, a basaltic tephra layer with poorly vesiculated grains, over 2 mm in longest diameter, is found at 37 cmdepth, 16 cmabove the Hekla 1104 tephra (Fig. 6 and Table 1B). No matching tephra has been identified on land in North Iceland. This tephra is preliminarily attributed to a submarine eruption northwest of the Grı msey island, described in a contemporary annal for Fig. 5. Age-depth diagramfor the last 4500 calibrated years in core MD showing tephra marker horizons as well as calibrated AMS 14 C datings of molluscs (shown with7one standard deviation). A marine reservoir age of 400 years has been applied. The age model is based only on tephra marker horizons (tephrochronological age model). The Hekla 1104 is historically dated, while Hekla 3 is dated on terrestrial material from Iceland. The right-hand diagramshows the deviation (Dr) of each calibrated radiocarbon date fromthe tephrochronological age of the same level.

14 2260 J. Eiríksson et al. / Quaternary Science Reviews 23 (2004) Fig. 6. Age-depth diagramfor the last 4500 calibrated years in core HM (see also Eiríksson et al., 2000a; Knudsen and Eiríksson, 2002) showing tephra marker horizons as well as calibrated AMS 14 C datings of molluscs or foraminifera (shown with7one standard deviation). A marine reservoir age of 400 years has been applied. The age model is based only on tephra marker horizons (tephrochronological age model). The KOL 1372 and Hekla 1104 are historically dated, while Hekla 3 and Hekla 4 are radiocarbon dated on terrestrial material from Iceland. The right-hand diagram shows the deviation (Dr) of each calibrated radiocarbon date fromthe tephrochronological age of the same level. the year This tephra layer, KOL 1372, is also distinct in core HM (Fig. 7 and Table 1C). In the easternmost core, MD992275, a basaltic tephra with the chemical characteristics of the Veidivo tn volcanic systemoccurs at a depth of 101 cm(fig. 9 and Table 1E). It is correlated with a tephra layer found in North Iceland and attributed to an eruption on that volcanic systemin AD 1717 (Thorarinsson, 1950; Larsen, 1982). This tephra, V 1717, has also been identified in core MD (Fig. 8 and Table 1D). At 1552 cmin MD992275, a silicic tephra layer with the chemical signature of the Hekla volcanic system is found (Fig. 9 and Table 1C). Its glass composition matches that of the Hekla 5 tephra (Sigurdsson, 1982). Peat enclosing the Hekla 5 has been dated to and Cyr BP (Vilmundardo ttir and Kaldal, 1982). The oldest two tephra layers presented here are found in cores MD and MD (Fig. 3). The upper one is 6 8 cmthick and occurs at 602 and 2560 cm, respectively. It has the chemical signature of the Grı msvo tn volcanic system(tables 1A and E) and is correlated to the Skagi-Saksunarvatn tephra that is widely found in North Iceland (Hjort et al., 1985; Bjo rck et al., 1992; Pe tursson and Larsen, 1992; Principato and Geirsdo ttir, 2002) and off North Iceland (e.g. Andrews et al., 2002; Geirsdo ttir and O lafsdo ttir, 2002). The lower of the two tephra layers is 1 8 cmthick and occurs at 702 cmand 3405 cmdepth, respectively. The highly silicic and the basaltic glass (Table 1) match the composition previously published for the Sko gar-vedde glass (Bjo rck et al., 1992; Norddahl and Haflidason, 1992; Lacasse et al., 1995). In the following sections, we compare the radiocarbon and tephra based age models for the lower part of cores MD992271, MD and MD992275, and we describe the construction of age models for the last 4500 years in all the five cores along an east west transect and compare the radiocarbon and tephra based age models Age models for the lower parts of cores MD992271, MD and MD The early Holocene segments of the age models for the CALYPSO piston cores MD and MD (Fig. 3) are based on linear interpolation fromthe lowest marker down to the early Holocene Saksunarvatn ash, and then again on linear interpolation down to the

15 J. Eiríksson et al. / Quaternary Science Reviews 23 (2004) Fig. 7. Age-depth diagramfor the last 4500 calibrated years in core HM showing tephra marker horizons as well as calibrated AMS 14 C datings of molluscs and foraminifera (shown with7one standard deviation). A marine reservoir age of 400 years has been applied. The age model is based only on tephra marker horizons (tephrochronological age model). The KOL 1372 and Hekla 1104 are historically dated, while Hekla 3 is radiocarbon dated on terrestrial material from Iceland. The right-hand diagram shows the deviation (Dr) of each calibrated radiocarbon date from the tephrochronological age of the same level. Lateglacial Vedde Ash. In core MD992273, a linear interpolation between Hekla 3 and Hekla 4 is used for the lowest relevant segment, and the age model is then extended downcore to a basal radiocarbon date (AA , BP; 6899 ( ) cal. BP) Age models for the last 5000 years Core MD Two tephra markers have been found within the 4500 year record in core MD992271, the Hekla 1104 and the Hekla 3 (Fig. 5). The lowest part of the age model below Hekla 3 is based on a linear interpolation down to the level of the early Holocene Saksunarvatn ash (Fig. 3). The age of the top of the core is estimated to 250 cal. BP. This is based on a comparison of the benthic foraminifera biostratigraphy of core MD with other cores in the region. The mean Holocene sedimentation rate is over 50 cm/ka. Except for two, the calibrated AMS 14 C dates yield ages that are higher than the tephrochronological ages of corresponding samples. Increasing deviation is notable fromjust before 4000 cal. BP up to 2700 cal. BP, followed by a decreasing trend, which is apparently reversed around 700 cal. BP. Core HM The tephrochronological age model for core HM (Fig. 6) is based on three markers fromhekla: Hekla 4, Hekla 3 and Hekla 1104, and the local tephra correlated with the AD 1372 eruption northwest of Grı msey (KOL 1372). The sedimentation rate averages at over 75 cm/ka. Fifteen AMS 14 C datings are available. Fromthe base of the core up to ca cal. BP the radiocarbon and tephrochronological age models are similar. The deviation increases sharply close to 3000 cal. BP at the level of Hekla 3 and rises to more than 400 calibrated years. At 1000 cal. BP, the deviation is around 200 years (but with high uncertainty), while negative deviations, observed at 850 and 500 cal. BP, indicate reservoir ages similar to or slightly lower than today. Core HM The tephrochronological age model for core HM is based on Hekla 3, Hekla 1104 and KOL 1372 (Fig. 7). Nine AMS 14 C datings are available. Below Hekla 3, a linear extension of the interval between Hekla 1104 and Hekla 3 has been used. High, positive deviations between the radiocarbon and

16 2262 J. Eiríksson et al. / Quaternary Science Reviews 23 (2004) Fig. 8. Age-depth diagramfor the last 4500 calibrated years in core MD showing tephra marker horizons as well as calibrated AMS 14 C datings of molluscs (shown with7one standard deviation). A marine reservoir age of 400 years has been applied. The age model is based only on tephra marker horizons (tephrochronological age model). The V 1717, Hekla 1300 and Hekla 1104 are historically dated, while the levels for Hekla 3 and Hekla 4 are based on radiocarbon datings of terrestrial material from Iceland. The right-hand diagram shows the deviation (Dr) of each calibrated radiocarbon date fromthe tephrochronological age of the same level. tephra age models are observed immediately above the Hekla 3 tephra. There is no deviation at 1600 cal. BP, while positive values of ca. 200 years are observed again at 1000 and at 450 cal. BP. The 1000 cal. BP value agrees well with the HM value at that level, while the 450 cal. BP value may indicate either a rapid change between 500 (as registered in HM107-03) and 450 cal. BP (HM107-01), or that the reservoir ages were different at the two sites. Core MD Numerous tephra layers have been observed in the uppermost 600 cm of core MD In this study, however, we include only three of them, Hekla 1104, Hekla 1300 and V 1717 (Figs. 3 and 8), as they can be correlated with the tephra stratigraphy on land. Further work is needed before reliable correlations can be presented for the other tephras, although preliminary results so far indicate that they fit very well with the predicted age model. A total of 15 AMS 14 C dates are available fromthe Late Holocene interval, covering ca cal. yrs. The top of core MD is assumed to date from AD 1950 or 0 cal. BP. The lowest deviation value between the calibrated radiocarbon dates and the tephrochronological age model (Dr) occurs at 700 cal. BP (Medieval WarmPeriod), with increasing although slightly fluctuating values up to 100 cal. BP, after which the deviation values decrease. The youngest values are not very reliable due to lack of constraint on the exact age of the top of the core. Core MD The age model for the Late Holocene part of core MD is based on 6 wellconstrained dates of tephra layers younger than 1130 years (Fig. 9), which is the time span since the settlement of Iceland, and on radiocarbon dates of the two older tephra horizons, the Snæfellsjo kull 1 and Hekla 3 (for further discussion, see Larsen et al., 2002). A total of 22 AMS 14 C dates are available fromthe interval. There is a relatively high positive deviation between cal. BP. After a decrease to a low positive value just before 3000 cal. BP, maximum values are reached at cal. BP. This is followed by a continuous decreasing trend towards a value of close to zero at 1850 cal. BP. Between 1700 and 150 cal. BP, the deviations remain at a level around calibrated years, with maximum deviations at , and cal. BP. A final decrease is observed towards modern time, i.e. to a 400 year reservoir age as found today in this area.

17 J. Eiríksson et al. / Quaternary Science Reviews 23 (2004) Fig. 9. Age-depth diagramfor the last 4500 calibrated years in core MD (see also Larsen et al., 2002) showing tephra marker horizons as well as calibrated AMS 14 C datings of molluscs (shown with7one standard deviation). A marine reservoir age of 400 years has been applied. The age model is based only on tephra marker horizons (tephrochronological age model). The tephra layers V 1717, V 1477 and V 1410, Hekla 1300 and Hekla 1104 are historically dated, the Settlement layer is dated on the basis of correlation with the GRIP ice-core, while Snæfellsjo kull 1 and Hekla 3 are radiocarbon dated on terrestrial material from Iceland. The right-hand diagram shows the deviation (Dr) of each calibrated radiocarbon date fromthe tephrochronological age of the same level. 4. Discussion If the reservoir age offset, DR, in waters off North Iceland had remained constant throughout the Holocene, a tephrochronological age model and a calibrated 14 C age model should fit closely, assuming that the marine calibration data set truthfully reflects the oceanic mixing of 14 C variability in the atmosphere. The significant discrepancies observed in the area between the two age models may have several causes. In the present study, we examine the possibility that the discrepancies reflect changes in reservoir age caused by reconfiguration of the oceanic Polar Front, separating predominantly Arctic water with a high apparent age, and Atlantic water with normal apparent age. Other possible causes include deficiencies in the marine calibration data set for the region, or changes in the global thermohaline circulation that might affect the apparent age of the regional water masses. In view of the tests presented in this study and with respect to ecology of the dated organisms, differences in feeding habits cannot explain the large observed deviations. Fig. 10 shows a summary of the deviations (Dr) between all available calibrated radiocarbon dates from the North Icelandic shelf and the tephrochronological age models, as well as a smoothed curve for the deviations. It is evident that the local reservoir age has generally been higher than 400 years and that we need additional reservoir age correction to fit the tephrochronological age models. In general, negative Dr values are observed around 4000 cal. BP, followed by an increasing trend that is reversed at ca cal. BP. A dramatic deviation peak observed at 2650 cal. BP is the culmination of a trend that started at close to 3500 cal. BP. The next minimum is observed at 2000 cal. BP, and low values are also apparent at 800 cal. BP and perhaps during the last 150 years. It is appropriate to compare this distinct variability with the available archives of palaeoceanographic variability during the last 4500 years. As an example, the percentage distributions of two selected benthic foraminiferal species from core HM (Fig. 10) show a clear pattern in the assemblage changes through the last 4500 cal. years. Cassidulina neoteretis (Seidenk-

18 2264 J. Eiríksson et al. / Quaternary Science Reviews 23 (2004) Fig. 10. The percentage distrubution of two benthic foraminiferal species (Cassidulina neoteretis and Islandiella norcrossi) and the total benthic foraminiferal flux values (no/cm 2 /year) for core HM are compared with the deviation of all available calibrated radiocarbon dates (Dr in calibrated years), as well as the calculated additional reservoir age (DR in 14 C years) of each radiocarbon date that can be compared with constrained tephrocronological age models on the North Icelandic shelf. The Dr and DR values are plotted with one s error bands. A separate smoothed curve for the Dr data was produced by means of distance weighted least squares smoothing with 10% of the data points in each running window. rantz), which is an indicator of the relatively warm, high-salinity Atlantic water masses of the Irminger Current (see e.g. Jennings and Helgado ttir, 1994; Seidenkrantz, 1995; Hald and Steinsund, 1996), is frequent in the lower part of the core. There is a general decrease after 3000 cal. BP, but with temporary increases again around 1000 cal. BP and in modern time. The arctic water species Islandiella norcrossi (Cushman) (see e.g. Nørvang, 1945; Hald and Steinsund, 1996), however, rises fromvalues around 10% in the lowermost part up to its maximum values of 20 40% that have prevailed since about 3500 cal. BP. Prominent peaks in IRD flux, observed at about 3000 cal. BP, both in HM and HM107 01, correspond to a period