A high-resolution study of Holocene paleoclimatic and paleoceanographic changes in the Nordic Seas



[1] High-resolution records from IMAGES core MD95-2011 in the eastern Norwegian Sea provide evidence for relatively large- and small-scale high-latitude climate variability throughout the Holocene. During the early and mid-Holocene a situation possibly driven by consistent stronger westerlies increased the eastward influence of Arctic intermediate and near-surface waters. For the late Holocene a relaxation of the atmospheric forcing resulted in increased influence of Atlantic water. The main changes in Holocene climate show no obvious connection to changing solar irradiance, and spectral analysis reveals no consistent signature for any periodic behavior of Holocene climate at millennial or centennial timescales. There are, however, indications of consistent multidecadal variability.

1. Introduction

[2] The Holocene has been considered a period characterized by relatively stable climate conditions, a view largely based on the stability of the Holocene δ18O from the Greenland ice core compared to the glacial parts [Grootes et al., 1993]. An emerging picture of more variable high northern latitude climates is indicated by a number of recent studies [e.g., Dahl and Nesje, 1996; Bond et al., 1997; Bianchi and McCave, 1999; Bond et al., 2001]. However, the nature of this variability and the possible driving mechanisms are still to a large extent unresolved. It has been debated whether the Holocene climatic and oceanographic changes have the same general characteristics as the millennial scale climate fluctuations of the last glacial period, only with a subdued amplitude, or if other characteristics and driving forces were responsible for the evolution of climate during the current interglacial period. Several papers claim that there is a substantial ∼1500 year periodicity throughout the Holocene, being the source of cold spells documented in different proxy records [Bond et al., 1997; Mayewski et al., 1997; Campbell et al., 1998; Bianchi and McCave, 1999]. Other studies question this result, and evidence in favor of a more prominent 900–1000 and 550 year Holocene periodicity also exists, both from ice core and marine records [Stuvier and Braziunas, 1993; Stuvier et al., 1995; Grootes and Stuiver, 1997; Schulz and Paul, 2002; Chapman and Shackleton, 2000]. Bond et al. [2001] argue that Holocene climate variability is connected to changing solar irradiance, and that the centennial scale variability seen (on the order of 200–500 years) does not reflect any strict periodic climate behavior. Rather, clusters of these changes are argued to produce the previously reported 1500-year variability [Bond et al., 1997].

[3] For the North Atlantic there is evidence of a possible connection between longer term Holocene climate changes and periods of a relatively consistent state of the atmospheric pressure system over the North Atlantic region [Barlow et al., 1997; Keigwin and Pickart, 1999; Schulz and Paul, 2002; Nesje et al., 2000; deMenocal et al., 2000]. Here we provide evidence bearing on these issues from high-resolution foraminiferal δ18O and foraminiferal census records from the Vøring Plateau in the eastern Norwegian Sea (Figure 1).

Figure 1.

Map showing the location of cores JM97-948/2A and MD95-2011. The main current systems of the region are also shown.

2. Material and Methods

[4] The material in this study was obtained from IMAGES piston core MD95-2011 and box core JM97-948/2A (66°58.19N, 07°38.36E and 1048 m water depth) (Figure 1). The cores were spliced to develop a continuous record throughout the Holocene, as MD95-2011 did not recover the youngest part. The last 564 cal. years BP are represented by JM97-948/2A. Box core JM97-948/2A was sampled every half cm from 0–30 cm and MD95-2011 is sampled every cm from 0–745 cm.

2.1. Chronology

[5] The chronology is based on linear interpolation between dated levels. The age control for box core JM97-948/2A is based on nine 210Pb and two Accelerator Mass Spectrometry (AMS) 14C dates. To provide a better fit between the 14C AMS ages of the two cores, 210Pb ages below 10.5 cm in the box core were not used. The decay time of 210Pb (t1/2 = 22.26 years) [Appleby and Oldfield, 1992] also implies that those dates are less certain. The established age model for MD95-2011 rests on twelve 14C AMS dates and the presence of the Vedde ash layer. Two additional AMS 14C dates were not used as they gave inverted ages probably related to resedimentation. One other date, inverted (overlapping age range) compared to the Vedde ash, was also omitted as the Vedde ash age is presumed to be more reliable. The radiocarbon ages were all measured on Neogloboquadrina pachyderma. The AMS 14C ages are corrected for a marine reservoir effect of 400 years and converted to calendar ages by using Calib 4.3 [Stuvier et al., 1998]. All ages are given in calendar years BP 2000 (Figure 2; Table 1). For the late Holocene part of the sequence, we consider the age control to be good. In the earlier parts of the record the accuracy of the age scale is less certain, as fewer dates exist due to a low foraminifera content in the sediment. The possible age ranges of the oldest dates also increase due to the 14C plateaus that characterize this interval (Table 1) [Stuiver et al., 1998]. The average sedimentation rate of the spliced sequence is about 56 cm/kyr.

Figure 2.

(a) Age model for JM97-948/2A. Circles represent 210Pb and 14C dates used in the age model, squares represent 210Pb dates not used in the age model. (b) Age model for MD95-2011. 14C dates used in the age model and the Vedde ash layer are indicated by circles. Triangles represent 14C dates not used.

Table 1. Summary of Tie Points Used to Establish the Age Models for JM97-948/2A and MD95-2011a
IdentifierCoreDepth, cmMaterial DatedSubmitted Sample SizeAge 14C AMS ± 1σCal. Age BP(2000)Age Range ±1σComments
  • a

    The 14C AMS dates were measured at the Leibniz-Labor for Radiometric Dating and Isotope Research, Kiel (KIA), Centre des Faibles Radioactivités, Gif sur Yvette (GifA) and at Laboratoriet for Radiologisk Datering, Trondheim (TUa). The age of the Vedde ash layer is the GRIP age from Grönvold et al. [1995].

  • b

    NPD: N. pachyderma (dex.); NPS: N. pachyderma (sin).

 JM97-948/2A0.25   5 210Pb dated
 JM97-948/2A0.75   9 210Pb dated
 JM97-948/2A1.25   15 210Pb dated
 JM97-948/2A1.75   20 210Pb dated
 JM97-948/2A2.75   30 210Pb dated
 JM97-948/2A3.75   40 210Pb dated
 JM97-948/2A4.75   49 210Pb dated
 JM97-948/2A7.75   68 210Pb dated
 JM97-948/2A10.25   79 210Pb dated
KIA 6285JM97-948/2A21.75NPDb8.24 mg735 ± 40448362–480 
KIA 4800JM97-948/2A30.75NPD8.01 mg940 ± 40579558–601 
GifA96471MD95-201110.5NPD8.94 mg980 ± 60601568–675 
KIA 5600MD95-201124.5NPD8.5 mg1590 ± 4012091150–1237not used
KIA 3925MD95-201130.5NPD8.6 mg1040 ± 40675614–695 
KIA 5601MD95-201147.5NPD8.4 mg1160 ± 30739716–772 
KIA 3926MD95–201170.5NPD8.2 mg1460 ± 501037993–1103 
KIA 6286MD95-201189.5NPD8.76 mg1590 ± 3012091169–1227 
KIA 3927MD95-2011130.5NPD7.2 mg2350 ± 4020121965–2052not used
KIA6287MD95-2011154NPD10.55 mg2335 ± 2519921965–2047 
GifA96472MD95-2011170.5NPD7 mg2620 ± 6023592322–2392 
KIA 10011MD95-2011269.5NPD7.7 mg3820 ± 3538133748–3870 
KIA 463MD95-2011320.5NPD11 mg4330 ± 5044844449–4563 
KIA 464MD95-2011520.5NPD6.9 mg7260 ± 6077477706–7817 
TUa-3315MD95-2011703.5NPSb7.0 mg10775 ± 8512142, 12067, 1201811761–12394not used (Vedde age supposed to be less uncertain)
 MD95-2011709.5   12030 Vedde ash
TUa-3316MD95-2011730.5NPS5.1 mg11875 ± 1401344313114–13825 
KIA 465MD95-2011750.5NPS10 mg12220 ± 901385613496–13889 

2.2. Stable Oxygen Isotopes

[6] Oxygen isotope measurements were performed at the GMS lab at the University of Bergen, using Finnigan MAT 251 and MAT 252 mass spectrometers both equipped with automatic preparation lines (“Kiel device”). Records were obtained for both left and right coiling forms of the planktic foraminifer N. pachyderma and from the benthic foraminifer Cassidulina teretis (normally 12–17 specimens of N. pachyderma and 15–20 specimens of C. teretis; 150–500 μm). Samples were crushed and cleaned with methanol in an ultrasonic bath before being measured. The reproducibility of the analyses is ±0.07‰ for the oxygen isotope measurements, based on replicate measurements of carbonate standards. All results are reported as δ18O in ‰ vs. PDB, using NBS 19. A 5-pt running average was run to smooth the δ18O records. In the text, all reference is to the smoothed δ18O if not otherwise noted. We corrected the δ18O records for the ice volume effect, based on the results of Fairbanks [1989], where 10 m sea level equals 0.11‰ δ18O (Figure 3). Sampling every cm makes the mean temporal resolution 15–27 years.

Figure 3.

(a) Oxygen isotope records in ‰ vs. PDB. Thick black lines show the 5 point running average of the records. (b) Ice volume corrected δ18O. Tie points for the age model are indicated at the bottom of the graph. Green lines indicate events of inversion between the planktic and the benthic signals. Y.D.: Younger Drays. B/A: Bølling-Allerød.

2.3. Foraminiferal Abundance and SST Estimation

[7] The relative abundance of planktic foraminifers is based on census counts of splits of the >150 μm fraction containing approximately 300 planktic foraminifers. Samples were counted at 5 cm intervals for this portion of the study. The planktic foraminifer fauna is characterized by polar, subpolar and transitional species. The dominant species are N. pachyderma (dex) and Globigerina quinqueloba. Other species, including N. pachyderma (sin), Globigerina bulloides, Globigerinita glutinata and Globorotalia inflata, are found in lower abundance in virtually all samples.

[8] Here we present sea surface temperature (SST) estimates for the past 13.6 kyr based on the modern analogue technique (MAT) in conjunction with the ATL916 core-top database (U. Pflaumann, personal communication, 2000). SSTs were calculated based on interpolated modern SSTs at 10-m water depth for August according to Levitus [1994]. The MAT SST estimates are based on the 10 best analogues, using the squared chord distance as the dissimilarity measure. Counting every 5 cm gives a mean temporal resolution of 75–135 years.

2.4. Time Series Analysis

[9] The existence of possible periodic or quasiperiodic behavior of the Holocene climate has been investigated by performing time series analyses on the three oxygen isotope records and on the record of relative abundance of N. pachyderma (sin.). Time series results were obtained by the Blackman-Tukey Method [Blackman and Tukey, 1958] of spectral estimation using the AnalySeries software [Paillard et al., 1996]. The records were initially prepared by removing linear trends and resampling at constant 20-year intervals for the stable isotope records and 100-year intervals for the percentage of N. pachyderma (sin.). The confidence interval was set at 90%. Similar results (not shown) were also obtained using the MultiTaper Method [Thompson, 1982] and the Maximum Entropy Method [Haykin, 1983]. To recognize possible nonstationarity through time, the interval from 11.5 ka BP to the present was investigated first; thereafter 6 ka slices of the records (11.5-6 ka BP, 10-4 ka BP, 8-2 ka BP and 6 ka BP to recent) were run with the same settings.

3. Results

3.1. Planktic δ18O

[10] While an overall comparison of the planktic isotope records shows some individual differences the major trends are clearly related. The records indicate three different intervals of distinct Holocene variability: the time interval ranging from glacial termination to about 8 ka BP, from ∼8 ka BP to ∼3–4 ka BP, and from ∼3–4 ka BP to the present (Figure 3).

[11] At about 11.6 ka BP, the planktic δ18O values stabilized at an interglacial level after a rather rapid transition from the Younger Drays (Figure 3b). After about 10.6 ka BP, there is a gradual trend toward increased δ18O, terminating just before 8 ka BP in the highest values of the entire Holocene. This event is most pronounced in the N. pachyderma (sin.) record (0.70‰ enrichment (raw data)), but has a broader and less distinct expression in the dextral record (Figure 3b). Throughout the early Holocene, the contrast between dextral and sinistral δ18O (planktic Δδ18O) was reduced relative to the younger parts of the Holocene (Figure 4a; Table 2). The mid-Holocene was a more stable period, characterized by small-scale variability superimposed on a slightly increasing δ18O trend (N. pachyderma (dex.) δ18O: 1.1‰ to 1.7‰ (7800-3900 ka BP); N. pachyderma (sin.) δ18O: 1.8‰ to 2.2‰ (7600-3800 ka BP)) (Figure 3b). The planktic Δδ18O was somewhat larger than during the early Holocene (Figure 4a; Table 2). More pronounced variability characterizes the late Holocene (Table 2). Here the generally decreased δ18O was interrupted by shorter events of increased δ18O (Figure 3b). The planktic Δδ18O is somewhat higher than during the mid-Holocene, interrupted by shorter periods of reduced values (Figure 4a).

Figure 4.

(a) Changes in the contrast between δ18O of N. pachyderma (dex.) and N. pachyderma (sin.). (b) Changes in the contrasts between planktic δ18O and benthic δ18O. Y.D.: Younger Drays. B/A: Bølling-Allerød.

Table 2. Mean, Maximum, and Minimum Values (Raw Data) are Presented for Each Time Interval of the Records
 Late Holocene, 0–3.6 ka BPMid-Holocene, 3.6–7.9 ka BP
Mean ±1σMaximum ValueMinimum ValueMean ±1σMaximum ValueMinimum Value
δ18O N. pachyderma (dex)1.27 ± ± 0.171.940.87
δ18O N. pachyderma (sin)1.94 ± 0.212.681.282.02 ±
δ18O C. tereis4.09 ± 0.234.553.394.00 ± 0.214.483.27
Δδ18O (N. pach. (sin)-N. pach. (dex))0.66 ± 0.241.580.140.62 ±
Δδ18O (C. teretis-N. pach. (dex))2.81 ± 0.293.612.112.58 ±
Δδ18O (C. teretis -N. pach. (sin))2.15 ± 0.272.912.341.97 ± 0.232.432.19
% N. pachyderma (sin)14.20 ± 6.7530.533.821.83 ± 8.18448.68
N. pachyderma (dex)/g80.23 ± 71.93387.973.9872.17 ± 56.10308.064.89
N. pachyderma (sin)/g17.49 ± 14.3770.461.3140.41 ± 27.97180.296.51
G. quinqueloba./g28.52 ± 24.20118.551.7252.85 ± 24.27107.466.97
MAT SST11.06 ± 0.7611.938.8810.20 ± 0.7811.518.5
 Early Holocene, 7.9–11.6 ka BPBølling-Allerød/Younger Drays, 11.6–13.8 ka BP
Mean ±1σMaximum ValueMinimum ValueMean ±1σMaximum ValueMinimum Value
δ18O N. pachyderma (dex)1.42 ± ± 0.442.91.14
δ18O N. pachyderma (sin)1.90 ± ±
δ18O C. tereis3.96 ± 0.234.543.33.76 ± 0.314.342.89
Δδ18O (N. pach. (sin)-N. pach. (dex))0.47 ± 0.291.470.030.39 ± 0.502.740.15
Δδ18O (C. teretis-N. pach. (dex))2.55 ± ± 0.572.730.5
Δδ18O (C. teretis-N. pach. (sin))2.08 ± 0.333.292.091.14 ± 0.372.381.04
% N. pachyderma (sin)29.38 ± 15.2680.7810.8494.82 ± 14.2996.8892.67
N. pachyderma (dex)/g36.50 ± 54.50347.117.643.24 ± 1.566.520.87
N. pachyderma (sin)/g39.14 ± 85.73536.164.13104.07 ± 29.94138.5245.06
G. quinqueloba./g20.66 ± 14.6662.242.121.64 ± 0.783.580.58
MAT SST10.22 ± 0.8711.576.273.93 ± 1.044.692.96

3.2. Benthic δ18O

[12] Shifts in the benthic record of δ18O of up to 0.8‰ occur, and 0.3‰ variability is seen frequently, indicating nonstable deep water conditions at the Vøring Plateau during the Holocene. Intervals of decreased benthic δ18O appear to correspond to periods of increased planktic δ18O, and vice versa (Figure 3b). Both planktic-benthic Δδ18O records clearly reflect this relationship (Figure 4b). At about 3.6-2.8 ka BP there is a transition zone after which the planktic and benthic records seem to be in phase, in contrast to the earlier portion of the Holocene.

[13] In the period before 9.8 ka BP, benthic δ18O decreased by more than 0.1‰ compared to modern values (4.1‰). During the Younger Drays, an even more extreme interval of decreased δ18O occurred (Table 2). From 9.8-3.5 ka BP, there was a gradual evolution from slightly increased (4.3‰) to much more decreased (3.6‰) δ18O than today. Increased δ18O values characterize the last 2.8 kyr BP, interrupted by shorter periods of decreased δ18O (Figure 3b).

3.3. Foraminifera and SST Records

[14] Prior to 11.7 ka BP, the N. pachyderma (sin.) content was high and stable at about 95% (Figure 5a). Subsequently, the relative abundance of N. pachyderma (sin.) decreased with the termination of glacial conditions, and after about 10.5 ka BP it stabilized at interglacial values. The average relative abundance between 10.5-3.5 ka BP is 23%, with an increasing percentage marking the cold event at ∼8 ka BP. For the late Holocene, even lower percentages are recorded (Table 2). There is a high degree of correlation between the estimated MAT SST and the relative abundance of N. pachyderma (sin.), the correlation coefficient being 0.89. A slight warming trend observed in the MAT SST during the mid-Holocene might be seen as a contradiction to the other temperature indicators (Figure 5b). This appears to be due to an increase in the abundance of the dextral coiling form of N. pachyderma, a trend not recorded by N. pachyderma (sin.) and G. quinqueloba (Figures 5b–5e). These three species together comprise >88% of the total amount of planktic foraminifers. Low abundances of G. quinqueloba characterize the period prior to 10.5 ka BP. From then on this taxon increases in abundance and maximum values characterize the mid-Holocene. During the late Holocene, the abundances are lower. High-amplitude events are seen frequently on shorter timescales throughout the last ∼9 kyr (Figure 5d; Table 2).

Figure 5.

(a) Percentage of N. pachyderma (sin.). (b) MAT SST estimate (°C). (c) N. pachyderma (dex.)/g sediment. (d) G. quinqueloba/g sediment. (e) N. pachyderma (sin.)/g sediment. Y.D.: Younger Drays. B/A: Bølling-Allerød.

3.4. Time Series Analysis

[15] Time series analysis reveals evidence for weak periodicities on decadal, centennial and millennial timescales. It should be noted that the particular periodicities reported are spectral peaks (local maxima in the power spectra) that may represent broad bands of quasiperiodic variability rather than narrow, well defined periodicities. The N. pachyderma (sin.) δ18O record shows concentrations of variance at 1150- and 550-year periodicities. The 1150-year period is most pronounced in the intervals 8-2 ka BP and 6-0 ka BP, while the 550-year period is better expressed in the intervals 11.5-6 ka BP and 10-4 ka BP. A strong 115-year signal is also seen in the interval 11.5-6 ka BP (Figure 6a). In the record of N. pachyderma (dex.) δ18O, evidence of a 1250-year period is seen. The signal is stronger in the 6-0 ka BP interval than in the 11.5-6, 10-4 and 8-2 ka BP intervals. From 8-2 ka BP, a 417-year period is also seen, while the interval between 11.5-6 ka BP records a relatively strong 115-year signal (Figure 6b). Time series results from % N. pachyderma (sin.) indicate a 1400-year periodic imprint, and for the 11.5-6 ka BP interval a 570-year signal (Figure 6d). The C. teretis δ18O record shows a periodicity of 900 years, but only for the 11.5-6 and the 10-4 ka BP intervals. A significant 260-year signal is also seen in the benthic δ18O. It is very well represented between 10-4 ka BP and also well recorded in the 11.5-6 and the 8-2 ka BP interval. In the 11.5-6 ka BP interval, a 115-year signal exists (Figure 6c). Lower frequency signals can not be reliably distinguished as the length of their periods are too long compared with the length of the investigated intervals. Each δ18O record indicates periodic behavior around 81 years, which appears to be consistent through time, except for the 8-2 ka BP interval in the N. pachyderma (dex.) δ18O, where only a very weak signal is seen.

Figure 6.

Power spectra (Blackman-Tukey method) of (a) δ18O N. pachyderma (sin.), (b) δ18O N. pachyderma (dex.), (c) δ18O C. teretis and (d) % N. pachyderma (sin.). Different studied intervals are represented by different colors: 11.5 ka BP to the present, black; 11.5-6 ka BP, red; 10-4 ka BP, orange; 8-2 ka BP, green; and 6-0 ka BP, blue. BW indicates bandwidth. Note the frequency scale in Figure 6d is different than in Figures 6a–6c.

4. Discussion

4.1. Holocene Paleoclimatic and Paleoceanographic Evolution

[16] Correcting the individual δ18O records for the ice volume effect relates the remaining signal to changes in temperature and salinity. Different salinity-δ18O relationships were considered in order to estimate the relative influence of salinity versus temperature. Traditionally, the North Atlantic mixing line [Craig and Gordon, 1965] has been used as a standard for the Nordic Seas, giving a steep salinity/δ18O gradient due to the negative freshwater end-member (−21‰). As the local hydrological regime of the Nordic Seas may include several sources of freshwater (precipitation, Arctic river influx, melting sea ice) with less negative isotopic compositions, the freshwater end-member will probably not be the same as for the North Atlantic mixing line. Today's situation gives a gradient of 0.6‰ δ18Ow per practical salinity unit (PSU) for the North Atlantic mixing line, compared with 0.2–0.3‰ δ18Ow per PSU for the Nordic Seas mixing line [Østbø, 2000]. Observations from the Svinøy section (NW from 62°N on the Norwegian coast to 64°40N) show that the maximum salinity of the inflowing Atlantic water is about 35.25‰ (100 m depth) [Mork and Blindheim, 2000]. The Arctic intermediate water has a defined salinity between 34.87‰ and 34.91‰ [Blindheim, 1990]. Given these two sources, a possible change in salinity at the studied site is estimated to be in the range of about 0.38‰ PSU for the Holocene, corresponding to 0.08–0.23‰ change in δ18Ow depending on the preferred mixing line. The study site location should be more closely related to the Nordic Seas mixing line, so the lower values are more likely. Thus the main features of the planktic isotope records are more likely related to variations in temperature than salinity.

[17] In Arctic water there is no difference between δ18O values obtained from N. pachyderma (sin.) and N. pachyderma (dex.), while the differences increase going eastward in the Nordic Seas and into warmer Atlantic water masses [Johannessen, 1992]. Thus the varying contrast between the δ18O of N. pachyderma (dex.) and N. pachyderma (sin.) can be used as an indicator of horizontal migration of the Arctic water/Atlantic water interface (Arctic Front). Globigerina quinqueloba is a species whose distribution is closely related to the Arctic front area [Johannessen et al., 1994]. The low planktic Δδ18O and the high G. quinqueloba content indicate that site MD95-2011 was closer to the Arctic front and Arctic water masses during the early and mid-Holocene than during the late Holocene (Figures 4 and 5).

[18] A similar situation has emerged during recent decades because sustained strong westerlies have led to an eastward migration of the subsurface Arctic water [Blindheim et al., 2000]. Based on the combined isotopic and G. quinqueloba evidence, we suggest that the apparently stronger influence of Arctic water in the early and mid-Holocene was the result of more pervasive strong westerlies. Harrison et al. [1992] used an atmospheric general circulation model to simulate the climatic influence of insolation changes throughout the Holocene and found that, during the early and mid-Holocene, the Icelandic Low and the westerly jet stream were displaced further north, producing mild and wet winters in northern Europe. A reconstruction of mean winter precipitation in the Jostedalsbreen area of western Norway indicates higher winter precipitation during early and mid-Holocene [Nesje et al., 2000]. The same period is characterized by cold reconstructed SST and a high trade wind activity at ODP Site 685C off Cap Blanc, west Africa [deMenocal et al., 2000]. An antiphase relationship between temperature anomalies between Greenland and Scandinavia has also been observed during parts of this period [Schulz and Paul, 2002]. All situations might be similar to a positive North Atlantic Oscillation (NAO)-like situation [Hurrell, 1995].

[19] A striking feature of the foraminiferal based records from site MD95-2011 is the lack of evidence for the early Holocene warm optimum, well known from marine and terrestrial records [Karlén, 1988; Koç Karpuz and Jansen, 1992; Koç et al., 1993; Nesje and Dahl, 1993; Lubinski et al., 1999]. However, SST estimates from the same site based on diatom abundance and U37K analysis do record the warmth optimum [Birks, 2001; Calvo et al., 2002]. This discrepancy between phytoplankton and foraminiferal based proxies may be due to the fact that eastward migration of Arctic water, forced by stronger westerlies, tends to preferentially influence the subsurface water, not the uppermost sea surface water [Blindheim et al., 2000]. Diatoms and coccoliths live in the uppermost part of the water column, in the photic zone, and will therefore reflect the imprint of warmer Atlantic surface water and the increased summer insolation of the early to mid-Holocene. The somewhat deeper-dwelling foraminifers should instead reflect a signal affected by the increased influence of Arctic water in the subsurface. In such a situation the foraminifera will be a less reliable SST indicator, as their abundance reflects the near-surface water more than the sea surface.

[20] The late Holocene is characterized by decreased planktic δ18O, increased planktic Δδ18O, a lower content of G. quinqueloba and planktic foraminifers in general, and higher SST values (Figures 3, 4, and 5; Table 2). These changes suggest an increased influence of Atlantic water masses and reduced importance of Arctic water, possibly due to a general relaxation of the atmospheric forcing (weaker and/or more variable westerlies). A change in the subsurface water from Arctic to Atlantic influence explains the apparent shift to locally warmer conditions at about 4 ka BP, and the opposition of this signal to other marine and terrestrial results. A marked change in climatic conditions at about 4 ka BP is well known from the literature. In north Africa, a decrease in monsoon activity, possibly related to the weakening of summer insolation throughout the Holocene, induced a change from wet to dry conditions at approximately 5-4 ka BP [Kutzbach et al., 1998; Lamb et al., 2000; deMenocal et al., 2000]. The vegetation pattern in northern Fennoscandia changed, the deviation of summer temperatures from modern Scandinavian values decreased, and neoglaciation of Fennoscandia, Svalbard and Franz Josef Land area began [Dahl and Nesje, 1996; Svendsen and Mangerud, 1997; Lubinski et al., 1999; Seppä and Hammarlund, 2000].

[21] Decreasing benthic δ18O generally corresponds with increased planktic δ18O at the study location (Figure 3b). If the observed variability is related to temperature, a change of up to ∼3.5°C in deepwater temperature must be explained [Shackleton, 1974]. As the lateral distribution of Arctic water expands, the increased contribution of these cold and fresh subsurface waters might act as a lid, thereby reducing the exchange of bottom water and increasing the deepwater temperature. Over the past few decades, a warming of the Norwegian Sea Deep Water (NSDW) has been registered, apparently connected to the prevailing positive NAO situation [Dickson et al., 1996; Østerhus and Gammelsrød, 1999]. However, the effect of the recent NAO pattern on deepwater temperatures is too small to explain the variability observed in the benthic δ18O. If the vertical distribution of Atlantic waters changes so that the boundary zone between relatively warm Atlantic water (∼6°C) [Blindheim, 1990] and the colder NSDW (∼−0.5°C) [Blindheim, 1990] is forced to greater depths, large changes in temperature might be possible. However, the presumed increase in bottom water temperature is recorded at the same time as our evidence shows Arctic waters to be of increased influence at the subsurface. This makes it unlikely that the warmer Atlantic waters would have penetrated so deep as to influence bottom temperatures at more than 1 km water depth.

[22] A more plausible explanation for the isotope trends might be that benthic δ18O values are influenced by waters carrying a signal that changes due to variable influence of brine water from the Arctic Ocean. The intermediate depth brine water, which forms by salt exclusion during sea ice formation, will follow the predominant currents through the Fram Strait and around the Nordic Seas, and should thereby influence bottom waters at the Vøring Plateau. During transport, the isotopic signature would likely be altered as it mixes with the surrounding water at the interfaces, but it should still give rise to rather decreased values. A positive NAO will increase the freshwater flux to the Arctic Ocean and increase the ice flux through Fram Strait [Kwok and Rothrock, 1999]. Persisting through time such a situation might give rise to the gradually decreased benthic δ18O values reflected in parts of the early Holocene and mid-Holocene (Figure 3b), as formation of new sea ice will increase the rate of brine water formation. During the late Holocene, one might assume decreased discharge of the Russian Arctic rivers and reduced sea ice flux through the Fram Strait. Brine water formation would therefore have decreased, as reflected by increased benthic δ18O (Figure 3b). Driftwood studies from the Arctic region suggest a westward shift of the Transpolar Drift in the last 4 kyr [Dyke et al., 1997], supporting the assumption of reduced ice flux through the Fram Strait. Brine water formation due to deglacial meltwater supply resulted in the most extreme episodes of decreased benthic δ18O, recorded during the Younger Drays (Figure 3b). This event is interpreted as the last imprint of the more extreme variability known from the area during glacial times [Dokken and Jansen, 1999].

4.2. The 8.2 ka BP Event

[23] Around 8100 BP, the planktic δ18O values of N. pachyderma (sin.) culminate in the most extreme cold event of the Holocene (Figure 3b). At the same time, there is a large increase in the amount of foraminifers in the sediment (Figures 5c–5e and 7c). The event is also characterized by an increased percentage of coarser grains, indicating increased winnowing due to stronger bottom current activity or increased ice rafting activity (Figure 7d). The interval of enriched δ18O in N. pachyderma (sin.) is rather abrupt in nature, with the extreme part restricted to a period of about 70 years. During the peak of this cooling, a slight warming recorded in the δ18O of dextral N. pachyderma (sin.) may be the result of different calcification seasons between the two coiling morphotypes. N. pachyderma (dex.) generally calcifies during warmer conditions than the left coiling form, and therefore the most extreme values will not be recorded by this taxon if the cooling is not equally distributed through the year. A shorter but more intense summer season may have been recorded by the dextral N. pachyderma. Similarities are seen between the δ18O from N. pachyderma (sin.) and the cold event at 8.2 ka BP in the δ18O record from GISP2 [Stuiver et al., 1995] (Figures 7a and 7b). The same event is also well known from other terrestrial and marine records from the North Atlantic region [Alley et al., 1997; Bond et al., 1997; Klitgaard-Kristensen et al., 1998; von Grafenstein et al., 1998; Barber et al., 1999; Nesje and Dahl, 2001]. In core MD95-2011, the event represents a cooling of about 3°C, a change somewhat larger than the value estimated by Klitgaard-Kristensen et al. [1998]. An apparent 150-year discrepancy is seen between the timing of our event and the 8.2 ka BP event in GISP2. This is probably the result of uncertainties in the age model, since few dates are available to precisely constrain this interval. The abrupt cooling event at 8.2 ka BP is thought to be related to a disturbance of the thermohaline circulation, as a response to freshwater perturbation after the final outburst of Lakes Agassiz and Ojibway [Barber et al., 1999].

Figure 7.

Details of the 8.2 ka BP event. (a) N. pachyderma (sin.) δ18O in ‰ vs. PDB. (b) Oxygen isotope values from GISP2 [Stuiver et al., 1995]. (c) Total amount of foraminifera/g sediment. (d) Percentage of material <63 μm.

4.3. Holocene Climate Periodicity

[24] The relative abundance of N. pachyderma (sin.) and the N. pachyderma (sin.) δ18O record has been compared to residual Δ14C [Stuiver et al., 1998], and with ice drift events in the North Atlantic reported by Bond et al. [2001] (Figure 8). Both solar activity and the rate of deep water formation influence the Δ14C [Stuiver et al., 1998; Bond et al., 2001]. Bond et al. [2001] argue that Holocene climate variability at millennial and centennial timescales is driven by solar forcing, as the flux of cosmogenic nuclides seems to correlate to the ice drift events. Changes in NADW production are suggested as an additional mechanism for amplifying the solar signals and transmitting them globally. Our records do not show similar results. The correlation between Δ14C and the MD95-2011 records is poor (Δ14C vs. δ18O (N. pachyderma (sin.)) correlation coefficient = 0.23). Major changes recorded in the MD95-2011 record are not always represented by ice rafting events in the Bond et al. [2001] records, and some of the more pronounced ice rafting events are only represented by small changes in the MD95-2011 records (Figure 8). Thus our comparison between the North Atlantic ice drift events (one of the main sources is expected to be the Nordic Seas) and climate proxies from the Nordic Seas do not support the use of North Atlantic ice drift events as regional climate indicators.

Figure 8.

Relative abundance of N. pachyderma (sin.) and N. pachyderma (sin.) δ18O from MD95-2011 plotted against residual Δ14C [Stuiver et al., 1998]. Vertical lines indicate peak ice drift events as recorded by Bond et al. [2001].

[25] Discussions on the periodic behavior of Holocene climate have concentrated on variability centered on 1500, 1000 and 550 years, periods reportedly found in marine and terrestrial ice core records from the North Atlantic region [e.g., Stuiver et al., 1995; Bond et al., 1997; Bianchi and McCave, 1999; Chapman and Shackleton, 2000]. Bond et al. [1997] have argued that Holocene variability is simply a pervasive, subdued expression of the more pronounced millennial scale variability of the last glacial period, possibly related to changes in the thermohaline circulation, while Schulz [2002] has argued against such a possibility. Taking into account uncertainties in the age model and in the exact period of the expected signals, variability in individual proxy records from MD95-2011 may be connected to these cycles (Figure 6). For example the 1400 year-long period seen in the % N. pachyderma (sin.) abundance record is close to the 1500-year cycle, while variability centered on 1250, 1150 and 900 years might be connected to the 1000 year period, or even to the 1500 year cycle. Similarly, 570, 550 and 420 year cycles bear resemblance to the reported 550 years periodicity. However, several aspects of our data contradict such an interpretation. First, the 1400-year period is not recorded as a clear signal, and it is only evident in the lower resolution % N. pachyderma record. The 1500 year period is also absent in the Holocene bidecadal δ18O record from GISP2 [Stuiver et al., 1995]. With increased resolution of North Atlantic ice drift records [Bond et al., 1997], Bond et al. [2001] no longer argue for a strong persistent 1500 year Holocene climate periodicity. Second, all our results are from the same core and are based on the same age model. The internal variations reflect real differences in periodicity and can not be explained by uncertainties in the age model, as can discrepancies between results from two different cores. Third, the strength of the periodic behavior is not consistent through time. Shorter periodicities seem to have dominated in the early and mid-Holocene, while longer periods are more pronounced in later intervals (planktic records). For benthic δ18O, the longer 900-year period is recognized only in the older parts. The 260-year period is only seen in the benthic δ18O, and it is best recorded in the older intervals. A period of 115 year is seen in all δ18O records during the 11.5-6 ka BP interval, and might indicate some sort of consistent behavior during the early part of the Holocene. Climate variability at approximately this frequency is known from several studies [e.g., Stocker and Mysak, 1992; Stuiver et al., 1995; Shabalova and Weber, 1999], but possible driving forces and their climatic significance are not well known. To conclude, the high-resolution records from MD95-2011 do not provide any evidence for a consistent cyclic behavior of Holocene climate at millennial or centennial timescale.

[26] At multidecadal timescales, all δ18O records show increased spectral power at a period of about 81-year (Figures 6a–6c). This 81-year signal seems to be rather consistent through time. Except for the 8-2 ka BP interval of the N. pachyderma (dex.) δ18O record, it is well represented in all investigated intervals. Several papers have reported variability at frequencies of around 80 years, including in tree ring records, ice core records, peat stratigraphy, and historical and instrumental records [Briffa et al., 1992; Schlesinger and Ramankutty, 1994; Mahasenan et al., 1997; Chambers and Blackford, 2001]. Common to these studies is the limited length of the available records, spanning from less than 150 years to about 2 kyr. In contrast, the MD95-2011 δ18O signal is present throughout the last 11.5 kyr. The length of the period (∼80 years) might indicate a possible connection to the solar Gleissberg cycle [e.g., Ribes, 1990; Waple, 1999; Chambers and Blackford, 2001]. Alternative explanations tend to invoke an internal oscillation of the atmospheric-ocean system or a connection to the thermohaline circulation [Stocker and Mysak, 1992; Delworth et al., 1993; Mann et al., 1995; Schlesinger and Ramankutty, 1994; Mahasenan et al., 1997]. The 81-year period cannot explain the large-scale variability during the Holocene, but it may be of importance for understanding some of the underlying small-scale variability.

5. Summary and Conclusions

[27] Based on evidence presented here from IMAGES core MD95-2011, the following characterizations and conclusions can be made. During the early and mid-Holocene, generally cold subsurface conditions occurred due to a location of this site closer to the Arctic front as stronger westerlies gave rise to an eastward migration of subsurface Arctic water. The influence of Atlantic water subsequently increased during the late Holocene due to a general relaxation of the atmospheric forcing. As the atmospheric forcing changed, the changing influence of Arctic intermediate brine water brought about unstable bottom water conditions.

[28] The 8.2 ka BP cold event is well represented in the records from the Vøring plateau. This abrupt cooling event is thought to be related to a disturbance of the thermohaline circulation, as a response to freshwater perturbation after the final outburst of Lakes Agassiz and Ojibway.

[29] Changing solar irradiance can not explain the Holocene climate changes seen in our Nordic Seas records. Neither is there any clear connection between climate changes in the Nordic Seas and North Atlantic ice drift events.

[30] Spectral analysis reveals no consistent signal of periodic climate variability of millennial or centennial timescale throughout the Holocene at this location. On multidecadal timescales, weak evidence exists for consistent variability with a period of approximately 81-year throughout the last 11.5 kyr. This 81-year cycle may be connected to the solar Gleissberg cycle, a result of internal oscillations of the atmospheric-ocean system, or related to changes in the thermohaline circulation.


[31] We thank Dag Inge Blindheim, Odd Hansen and Rune Søraas for laboratory assistance. Trond M. Dokken documented the position of the Vedde ash layer. Martin W. Miles performed MTM and MEM analysis (results not shown) and commented on the manuscript. The study was performed with financial support from The Research Council of Norway and the European Commission as part of the Holocene project. We thank Lloyd Keigwin and Mark Chapman for contributing with constructive reviews.