Benthic foraminiferal biodiversity response to a changing Arctic palaeoclimate in the last years

Transcription

1 Available online at Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007) Benthic foraminiferal biodiversity response to a changing Arctic palaeoclimate in the last years Jutta E. Wollenburg a,, Andreas Mackensen a, Wolfgang Kuhnt b a Alfred Wegener Institute for Polar and Marine Research, Columbusstrasse, Bremerhaven, Germany b Department of Geosiences, Christian-Albrecht-University, Kiel, Germany Received 5 October 2006; received in revised form 17 April 2007; accepted 4 May 2007 Abstract Investigations on the benthic foraminiferal fauna of sediment cores from the Eurasian Basin (Arctic Ocean) continental margin, reveal well correlated biodiversity maxima and minima during the last 24 kyrs. The temporal variable biodiversities observed at the core sites, on a large scale, reflect the palaeoclimatic evolution of the high northern latitudes. Herein sediments deposited during interglacials and interstadials reveal highest species richness, whereas, glacials and stadials are documented by lower species numbers. In high-resolution core PS2837 different periodicities in species richness in sediments from the Late Weichselian and Holocene can be detected. Biodiversity periodicities of 1.57 kyrs and 0.76 kyrs characterize sediments of the Late Weichselian, whereas, with the retreat of glacial ice sheets and shorter seasonal ice coverage, Holocene sediments reveal shorter periodicities of 1.16 kyrs and 0.54 kyrs. With the establishment of modern hydrographic conditions at about 4000 BP, 4 kyrs, a significant increase in the amplitudes of species richness can be observed. Although sediments deposited during warmings reveal highest species richness, we doubt that water temperature determines the abundance and distribution of most benthic foraminifera in the Arctic Ocean. We rather suggest that the temporal variabilities in species richness reflect changes in the availability of food, which, in the Arctic Ocean, mainly depends on extend and duration of seasonal sea-ice retreat. In the study area the latter one, besides the seasonal varying insolation, is essentially determined by the advection-rates and temperature of Atlantic water, entering the Arctic Ocean via the West Spitsbergen Current. Because both, flowrate and temperature of the West Spitsbergen Current were increased during interstadials and interglacials, a predominantly indirect influence of palaeotemperature on the biodiversity of benthic foraminifera is suggested. Yet, rare species, like Atlantic species' may indeed have a temperature sensitive metabolism Elsevier B.V. All rights reserved. Keywords: Foraminifera; Diversity; Arctic ocean; Climate; Weichselian; Holocene 1. Introduction 1.1. Environmental conditions in the Arctic Ocean Corresponding author. Tel.: ; fax: address: jwollenburg@awi-bremerhaven.de (J.E. Wollenburg). Today the Arctic Ocean is a semi-enclosed well-stratified ocean, perennially ice-covered in the centre and seasonally ice free on the Arctic shelves. As a continuation of /$ - see front matter 2007 Elsevier B.V. All rights reserved. doi: /j.palaeo

2 196 J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007) the North Atlantic Gulf Stream, temperate, saline Atlantic water enters the Arctic Ocean via the St. Anna Trough (Barents Sea Branch, BSB) and through the eastern side of Fram Strait, first as the West Spitsbergen Current (WSC) then as the North Spitsbergen Current (NSC) and Yermak Slope Current (YSC) (Fig. 1). The Atlantic water then Fig. 1. Overview of the study area showing locations of sediment cores, general circulation pattern of advected Atlantic waters, and the GS-2c Svalbard Barents Sea ice sheet extent.

3 J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007) flows anti-clockwise along the continental margin, and as it flows becomes less saline and colder, as it encounters and melts sea-ice (Rudels et al., 1999, 2004). Insolation and sea-ice cover primarily determine Arctic primary production. During the summer, the advection of warm Atlantic water via the WSC, NSC and YSC enhances the seasonal sea-ice retreat and thus primary production off Spitsbergen and along the northwestern Barents Sea slope (e.g., Smith et al., 1987, Hibler, 1989; Clough et al., 1997). Due to cooling, net precipitation and melting sea-ice, the BSB looses much of its original heat content in the Barents Sea and thus has only a minor effect on the seasonally sea-ice retreat of the Siberian shelves. Based on the distribution of planktonic microfossils and ice rafted debris (IRD), the CLIMAP Project Members (1981) concluded that the warm Atlantic water advection was interrupted and a more or less permanent sea-ice cover extended from the Arctic Ocean into the Nordic Seas during the Last Glacial Maximum (LGM kyrs BP; Greenland Stadial 2c 2b (GS-2c to GS-2b); Björck et al., 1998). However, during the last decades a wealth of new marine data has been assembled, which radically changed this view. It is now widely accepted that Atlantic water entered the Arctic Ocean via Fram Strait even for extended periods of the LGM (GS-2c) (Hebbeln et al., 1994; Forman et al., 1995; Dokken and Hald, 1996; Hebbeln and Wefer, 1997). The advection of heat and moisture influenced the waxing and waning of shelf-based ice sheets, and the distribution of seasonally ice-free waters in the Arctic Ocean (Stein et al., 1994; Mangerud et al., 1998; Nørgaard-Pedersen et al., 1998; Knies et al., 1999, 2000; Wollenburg et al., 2001; Nørgaard-Pedersen et al., 2003; Wollenburg et al., 2004). With the final decay of shelfbased ice sheets, interglacial environmental conditions were established at kyrs BP (Polyak and Mikhailov, 1996; Hald and Aspeli, 1997; Hald et al., 1999) Deep-sea biodiversity In the 1960s to 1970s it was generally accepted that high deep-sea biodiversities are the result of long-term environmental stability that allows the evolution of many highly specialised species within narrow niches that coexist at competitive equilibrium ( stability-time hypothesis ; Sanders et al., 1965; Hessler and Sanders, 1967; Sanders, 1968; Sanders and Hessler, 1969; Buzas and Gibson, 1969; Gibson and Buzas, 1973; Hessler, 1974; Hessler and Jumars, 1974; Paul and Menzies, 1984). However, modern studies visualize a different, rather dynamic deep-sea environment in which temperature and food vary seasonally, over time scales of days or hours (Billett et al., 1983; Graf, 1989; Thiel et al., 1990; Graf, 1992; Rice et al., 1994; Beaulieu and Smith, 1998). These modern studies on the deep-sea environment revealed large-scale patterns in regional and historical taxonomic richness, that have stimulated a wealth of explanations and hypotheses (reviewed by Willig et al., 2003). Faced with the likelihood that many environmental and climatical processes influence patterns of biodiversity, it is difficult to reject most hypotheses. Using the fossil record of benthic foraminifera over the last years, we can ignore hypotheses regarding speciation and/or extinction rates (e.g. Currie et al., 2004 for an overview). This leaves two prominent climate-based hypotheses for broad-scale biodiversity patterns in the deep-sea: the physiological tolerance and the energy-richness hypotheses (Currie et al., 2004). The physical tolerance hypothesis implies that patterns of diversity reflect attributes of the physical environment that influence the outcome of species interactions, therefore diversity should be strongly correlated with the physical aspects of the environment (Schluter and Ricklefs, 1993). Today, the energy-richness hypothesis, which claims that energy availability generates and maintains richness gradients, is the leading contender of ecological hypotheses explaining large-scale patterns of taxonomic richness (Allen et al., 2002; Hawkins et al., 2003; Brown et al., 2004; Currie et al., 2004; Hunt et al., 2005). In its standard form this hypothesis postulates that warmer and more productive environments have more individuals and thus more species (Hutchinson, 1959). Many data for endotherms support this assumption. However, studies of ectotherms and birds indicate that the average energy flux of populations is temperature invariant, such that these organisms will decrease the specimens number and increase the taxonomic richness with increasing temperature (Allen et al., 2002; Storch, 2003). Therefore, taxonomic richness of many ecto- and some endotherms is dependent on (i) metabolic processes, such as the uptake, transformation, and allocation of energetic and material resources in organisms, and (ii) the effect of temperature on the kinetics of biochemical reactions and ecological interactions (Allen et al., 2003). Whether limits on the broad-scale taxonomic richness of animals are mainly set by the energy flowing through food webs or by the total energy entering a geographic area is still a matter of debate (Cronin and Raymo, 1997; Snelgrove and Smith, 2002; Storch, 2003; Hawkins et al., 2003; Hunt et al., 2005). In many regions of the world's oceans benthic faunas reveal a parabolic bathymetric biodiversity pattern with maximum biodiversities at the lower continental slope (e.g. Lutze and Coulborn, 1984; Rex et al., 1993, 1997; Rogers, 2000) or,forthearctic Ocean, the shelf edge (Wollenburg and Kuhnt, 2000). Because the amount of food reaching the benthic

4 198 J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007) Table 1 Core position, water depth, and applied stratigraphic model for the analyzed cores Core Locality Latitude Longitude Water depth (m) Author of age model PS Yermak Plateau N 2 23 E 1023 Nørgaard-Pedersen et al. (2003) PS Yermak Plateau N 2 25 E 1028 Nørgaard-Pedersen et al. (2003) PS Yermak Plateau slope N E 2550 Vogt (1997) PS Barents Sea slope N E 995 Matthiessen et al. (2001) PS Laptev Sea slope N E 981 Spielhagen et al. (2005) PS Fram Strait N E 1522 Hebbeln (1992) community decreases with increasing water depth and there is no general parabolic temperature distribution in the water column of non ice-covered oceans, at least regional biodiversity patterns can not be attributed to the energy flux only. As Gaston (2000) suggested, no single mechanism adequately explains a given biodiversity pattern, and observed patterns vary due to local, and particularly due to regional or ocean wide mechanisms. Additional controls on deep-sea biodiversity include deep-water oxygenation, hydrodynamic regimes, sediment heterogeneity, catastrophic or patchy physical disturbance, and competition (Hermelin and Shimmield, 1990; Gooday et al., 2000; den Dulk et al., 2000; Levin et al., 2001; Snelgrove and Smith, 2002; Heinz and Hemleben, 2003; Schmiedl and Leuschner, 2005; Singh and Gupta, 2005). In particular, basin-tobasin differences in biodiversity are moreover defined by dispersal and gene flow (Rex et al., 1993, Gray, 2002; Gage, 2004) Biodiversity of benthic foraminifera in the Arctic Ocean At high northern latitudes the deep-sea benthos lives in a fast changing environment (Walsh et al., 2003; ACIA, 2004; Renaud et al., 2006). The influences of spatiotemporal varying environmental parameters are documented in the distribution of species, faunal associations, and biodiversity (Lagoe, 1977; Schröder-Adams et al., 1990; Scott and Vilks, 1991; Hunt and Corliss, 1993; Ishman and Foley, 1996; Ishman et al., 1996; Wollenburg and Mackensen, 1998a,b; Wollenburg and Kuhnt, 2000; Wollenburg et al., 2001, 2004; Polyak et al., 2004, for the modern and past Arctic Ocean). Whereas the patchy distribution of faunal associations usually are an image of interfering environmental processes, the regional biodiversity pattern of abyssal Arctic benthic foraminifera can be attributed to presumably only one or two regulating environmental parameters (Lagoe, 1976; Wollenburg and Kuhnt, 2000). Due to the sea-ice coverage and the elementary hydrography, physical environmental parameters, with the exception of temperature, should have no significant influence on the variance of abyssal Arctic foraminifera biodiversity. Therefore, the species richness of abyssal benthic foraminifera may be expected to vary along energy-gradients, gradients in temperature, or the availability of food. In their study on modern benthic foraminiferal faunas of the Eurasian Basin and their ecological controls, Wollenburg and Kuhnt (2000) ruled out a significant influence of temperature, and suggested that the number of niches and thus biodiversity are essentially controlled by the availability of food. In the Arctic Ocean the spatial faunal density pattern of benthic foraminifera and bacteria is comparable, because the former one can be correlated to regional varying carbon fluxes, this may also apply for bacteria (Wollenburg and Mackensen, 1998a; Kröncke et al., 2000). Therefore, for extended areas of the Arctic Ocean, two of the main important food sources of benthic foraminifera, phytodetritus and bacteria (e.g. Goldstein, 1999), are obviously linked to primary production. Besides the insolation, especially the different strength and temperature of inflowing Atlantic Water during glacial/stadial and interglacial/interstadial intervals, determined the seasonal ice-retreat and thereby palaeoproductivity. Thus a thorough knowledge on past spatial temporal variabilities in biodiversity at four continental slope sites, may provide information on palaeoclimatic induced changes in palaeoproductivity, sea-ice coverage and, rather qualitative, sea surface temperature. 2. Materials and methods 2.1. Material, preparation and age models Thecoresweretakenat1000to2500mwaterdepth, below the core of inflowing Atlantic water, from the Fram Strait, Yermak Plateau, northern Barents Sea continental slope and as far as the Laptev Sea (Table 1, Fig. 1). This study uses data of the upper 320, 402, 100, and 620 cm of sediment in cores PS1290, PS2837, PS2212, and PS2458, respectively. We used published age models (Table 1), however, C 14 ages older than N 12 kyrs BP were corrected with the new calibration program of Fairbanks et al. (2005). Ages between stratigraphic tie points are linearly interpolated,

5 J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007) Fig. 2. Down core gradients of Fisher α, H, ln(s), ln(e), carbon flux, WBFAR, agglutinated species, Atlantic species, phytodetritus species, M. zaandami and δ 18 O values of N. pachyderma (Nørgaard-Pedersen et al., 2003) in core PS2837. The shading marks core sections with b 100 specimens per sample, horizontal lines indicate the termination between the last Glacial and the Holocene, stadials, interstadials, and vice versa. GS=Greenland Stadial, GI =Greenland Interstadial. 199

6 200 J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007) Table 2 Sample spacing, mean age resolution, mean sample volume and mean dry weight for the analyzed cores PS PS PS PS PS Sample spacing (cm) (upper 50 cm, and terminations) 2 ( cm) 1 4 (below 100 cm) Mean age resolution (years) Mean sample volume (cm 3 ) Mean dry weight (g) except for a minor age adjustment (within the age uncertainties of Fairbanks et al., 2005) regarding the onset of the Younger Dryas in high-resolution core PS2837 (Fig. 2). This study uses further the event stratigraphy based on the δ 18 O isotopic record of the GRIP Greenland ice core (INTIMATE group; Björck et al., 1998). One-centimeter thick samples, taken at distances as noted in Table 2, were wet sieved after freeze-drying using a 63-μm sieve. Where possible, at least 300 specimens were counted from the size fraction N 63 μm, following the taxonomic concept of Wollenburg and Mackensen (1998a) 1. However, samples from deglacial periods are often affected by carbonate dissolution. In such samples foraminiferal numbers are low. Samples containing less than 40 specimens were excluded from statistical analyses. Because we are aware that specimens numbers b100 are critical for H analyses (e.g. Murray, 1991), core sections containing less than 100 specimens are highlighted in the figures. Shallow water foraminifera are incorporated in drifting sea-ice by the formation of frazil ice, the initial stage of seaice formation, or by ice gouging. They comprise approximately 7% of the modern benthic foraminiferal thanatocoenosis and may obscure the original biodiversity signal (Wollenburg, 1995; Wollenburg and Kuhnt, 2000). We therefore excluded species confined to a shallow water habitat (e.g., Cribroelphidium hallandense, C. albiumbilicatulum, Eoeponidella pulchella, Verneuilinulla arctica) from biodiversity analyses Data processing Biodiversity There is a log or semi-log relationship between the number of specimens and the number of species observed (Hayek and Buzas, 1997). If we were always to count the exact same number of specimens per sample, environmental biodiversity would equal the species number. 1 Tables of foraminiferal counts, like all other tables of the first author, are available electronically via jwollenburg. However, it is impossible to split-up samples to a constant specimens number. Therefore, a number of indices have evolved to describe species biodiversity from variable specimens counts. Which one is applied depends to a large extent on individual preferences. Here, we characterize biodiversity trends by the two most widely used biodiversity measurements, the information function H (Buzas and Gibson, 1969) with its decomposition equation ln(s) and ln(e)(buzas and Hayek, 1996), and the Fisher α Index (Fisher et al., 1943) Fisher α (Fisher et al., 1943) The Fisher α index (Fisher et al., 1943) postulates that the species number (S) increases with logarithmic increasing specimens numbers (N): α +αx 2 /2+ +αx n / n= αln(1 x). Given N and S the parameter α can be calculated by N/S=(e S /α 1)/(S/α). The constant x is calculated by x=n/(n+α) (Hayek and Buzas, 1997). In contrast to H, the Fisher α Index is generated from specimens and species numbers and thus is robust and reliable even when specimens and species numbers are low. However when we aim to predict species numbers from Fisher α indices, we have to ensure an underlying lognormal increase in species richness with increasing specimens numbers, otherwise the species numbers will be underestimated. Since our intention is to reveal past biodiversity trends, the Fisher α Index should be an adequate measurement, although we are unable to test the underlying species/specimens distribution Information Function (H) (Shannon, 1948) The Information Function (H) (Shannon, 1948), the out coming E (Buzas and Gibson, 1969), ln(s) and ln(e) analyses (Buzas and Hayek, 1996) measure the number of species and their proportions with no assumption to an underlying distribution. H ¼ Rp i lnp i ðequation 1Þ where p i is the proportion of the i th species The amount each species contributes to the values of H depends on its proportion in the assemblage p i and is

7 J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007) Table 3 PS2837 number of species, specimens counted, biodiversity indices Fisher α, H, ln(s), and ln(e) Depth (cm) Age (kyr BP) Species (no.) Specimens (no.) Fisher α H ln (S) ln(e) Table 3 (continued) Depth (cm) Age (kyr BP) Species (no.) Specimens (no.) Fisher α H ln (S) ln(e) (continued on next page)

8 202 J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007) Table 3 (continued) Depth (cm) Age (kyr BP) Species (no.) Specimens (no.) Fisher α H ln (S) ln(e) Table 3 (continued) Depth (cm) Age (kyr BP) Species (no.) Specimens (no.) Fisher α highest between 17 and 61% (Murray, 1991). E is a measure of equitability or dominance: E=e h /S (equation 2). The biodiversity index H has been criticized because it does not reveal how much of its final value is due to the number of species and due to species proportions. This problem can be solved by taking the natural logarithms of each side of equation 2 H = ln(s) + ln(e). With this decomposition equation H can be analyzed for species richness and evenness Spectral analysis For spectral analysis the Fisher α record of core PS2837 was resampled at equally spaced 100-year intervals. Two methods were used within the ANALYSERIES software package (Paillard et al., 1996): 1. Blackman and Tukey (1958) for its high confidence of the results; 2. The maximum entropy method (e.g. Haykin, 1983) for its highresolution Faunal parameters (benthic foraminiferal accumulation rate, carbon flux, Atlantic species, phytodetritus species) We will discuss the biodiversity trends of the analyzed sediment cores versus some general parameters of the benthic foraminiferal fauna listed below, which have been published and discussed for cores PS2212-3, PS and PS2837 (Wollenburg et al., 2001, 2004). For cores PS and PS (and the other cores) tables of these parameters are available electronically via pangaea.de/home/jwollenburg. H ln (S) ln(e)

9 J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007) Fig. 3. Down core gradients of Fisher α, H, ln(s), ln(e), carbon flux, WBFAR, agglutinated species, Atlantic species, phytodetritus species, M. zaandami and δ 18 O values of N. pachyderma (Hebbeln, 1991) in core PS

10 204 J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007) Table 4 PS1290 number of species, specimens counted, biodiversity indices Fisher α, H, ln(s), and ln(e) Depth (cm) Age (kyr BP) Species (no.) Specimens (no.) Fisher α Weight benthic foraminiferal accumulation rate (WBFAR) is defined as number of specimens per 10 g dry sediment per 1 kyrs, and calculated from the number of individuals per 10 g dry sediment using the mass accumulation rates H ln (S) ln(e) (MAR) of Nørgaard-Pedersen et al. (2003), Spielhagen et al. (2005), Vogt (1997), Knies et al. (1999), andhebbeln (1991). Carbon fluxes are calculated from correspondence analyses factor values (CA) of benthic foraminiferal counts (core and modern surface samples), and the relation of modern CA to published primary production data (see Wollenburg et al., 2004 for details). Epistominella exigua, E. pusilla, Islandiella helenae, and I. norcrossi are regarded as phytodetritus species (Gooday, 1994; Smart et al., 1994; Thomas et al., 1995; Wollenburg et al., 2001), and are described by their relative abundance. This group lumps species with an observed explosive reproduction triggered by phytoplankton bloom-derived fluff deposits (Epistominella spp.; Gooday, 1994; Smart et al., 1994), and species for which such a reproduction is deduced from their spatiotemporal modern densities (pers. observation Wollenburg). In the modern Arctic Ocean such annual fluff accumulations are as the distribution of these species restricted to areas with prolonged sea-ice retreat (Wollenburg and Kuhnt, 2000). The abundance of Melonis zaandami is positively related to high fluxes of slightly altered organic matter (Caralp, 1989; Korsun and Polyak, 1989; Corliss, 1991; Mackensen et al., 1995; Wollenburg et al., 2004), and is described by relative abundance. Atlantic species sensu Rasmussen et al. (1996, not Rasmussen et al., 2002) and Wollenburg et al. (2001, 2004) are comprised by E. pusilla, Pullenia osloensis, P. bulloides, Discorbinella berthelothi, Sigmoilopsis schlumbergeri and Eggerella bradyi, and are described by their relative abundance. Most Atlantic species are calcareous species that in the modern Arctic Ocean are restricted to water depths b1200 m under the influence of the advected warm Atlantic water. The agglutinated species S. schlumbergeri and E. bradyi are absent in the modern Arctic Ocean and Norwegian Greenland Seas, but show a late Weichselian occurrence (Sejrup et al., 1984; Jansen and Erlenkeuser, 1985). 3. Results 3.1. Biodiversity High-resolution core PS2837 The short Greenland Interstadial 2 (GI-2, kyrs BP in the GISP2 ice core; Bender et al., 1994) is represented by only two samples with values of 4.9, 5.2 and 1.83, 1.85 for Fisher α and H, respectively. During the Greenland Stadial 2c (GS-2c, kyrs BP) biodiversity dropped to values of (Fisher α) and

11 J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007) Fig. 4. Down core gradients of Fisher α, H, ln(s), ln(e), carbon flux, WBFAR, agglutinated species, Atlantic species, phytodetritus species, M. zaandami and δ 18 O values of N. pachyderma (Vogt, 1997) in core PS

12 206 J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007) (H) thereafter, biodiversity increased to a maxima of 7.0 (Fisher α) and 2.14 (H) around the middle of GS-2b ( kyrs BP). Subsequent decreasing values culminated in a well-documented low biodiversity epoch ( Fisher α, H) at the termination GS-2a/GI-1e (14.7 kyr BP). Large variations in both indices ( Fisher α, H) characterize the Greenland Interstadial I (GI- 1, Bølling Allerød), with higher biodiversities during the warmer periods GI-1e ( kyrs BP), GI-1c ( kyrs BP), and GI-1a ( kyrs BP). A period of stable moderate Fisher α ( ) and H ( ) values extended from the GS-1 (Younger Dryas, kyrs BP) to 10.8 ka. The Holocene is characterized by fluctuating high biodiversities of and for Fisher α and H, respectively. Evenness and thus ln(e) are rather constant and relatively high (close to 0) in the analyzed samples, with only episodic low equitabilities at 17.4, 11.9, and kyrs BP. Because H=ln(S)+ln(E), the down core variance in H values are thus mainly a reflection of species richness (ln(s)) (Fig. 2, Table 3) Core PS1290 In core PS1290 moderate Fisher α values ( ) characterize GS-2, followed by a low biodiversity interval ( ) at the termination GS-2a/GI-1e. Sediments from GI-1 are characterized by large variations in Fisher α values ( ), with higher biodiversities during the warmer substages GI-1e, GI-1a and perhaps GI-1c (Fisher α values are critical due to low specimens numbers), whereas the lowest core biodiversities ( and for Fisher α and H, respectively) are revealed from the cold substage GI-1b. Holocene sediments are characterizedbyanincreaseinfisherα values until 9.5 ka BP, thereafter, species richness remains at high values ( ), despite significant variabilities (Fig. 3, Table 4). Table 5 PS2212 number of species, specimens counted, biodiversity indices Fisher α, H, ln(s), and ln(e) Depth (cm) Age (kyr BP) Species (no.) Specimens (no.) Fisher α H ln (S) ln(e) Where the specimens number per sample is N 70, the biodiversity trend revealed by the information function match that of the Fisher α, this is the case for GS-2 and the Holocene. However, the Holocene increase in species richness is concealed by very low ln(e) in the calculation of H. Thus, in contrast to Fisher α, the mean Holocene H values differ only slightly from those of GS-2. Generally, sediments of the GI-1 and GS-1 ( kyrs BP) are characterized by very low specimens numbers, are these specimens distributed rather evenly on moderate species numbers, the value of H is essentially determined by high ln(e) values. In such sediments, as in GI-1c, the terminations GS-2a/GI-1e, and especially GI-1a/GS-1, the relative biodiversity values of H and Fisher α differ significantly. Evenness and thus ln(e) was low during the GS-2, even lower during the Holocene, yet, often high in GI-1 to GS-1. As mentioned above, high ln(e) values are usually revealed from specimen-poor sediments of moderate species richness; in such core sections H reflects equitability rather than species richness Core PS2212 In samples of core PS2212 Fisher α values are low ( ) before the onset of the Holocene, and thereafter steadily increases towards the core top ( ). The information function shows a reversed pattern with a steady decrease in H. An examination of ln(s) and ln(e) reveals that the increasing species richness is suppressed by the decreasing evenness in the information function (Fig. 4, Table 5) Core PS2458 Samples from the GI-1 and Younger Dryas (GS-1) of core PS2458 reveal low biodiversities (Fisher α and H). Thereafter, the Fisher α index indicates a steady biodiversity increase until ka BP, a period of decreasing values until 8.3 ka BP, and a second increase until the present day. A biodiversity increase until ka BP is also shown by the information function. From 10.4 to 3.5 kyrs BP, H mainly reflects varying evenness with a steep decrease in H and ln(e)valuesuntil 9.23 ka BP, an increase in values until 7.2 ka BP, again a decrease in values until 4.8 ka BP, and a final biodiversity increase until the present day (Fig. 5, Table 6) Power spectral analysis of high-resolution core PS2837 To remove the long homogenous slope displayed at lower frequencies, the power spectral analysis of Fisher α values from sediments of core PS2837 was restricted to

13 Fig. 5. Down core gradients of Fisher α, H, ln(s), and ln(e), carbon flux, WBFAR, agglutinated species, Atlantic species, phytodetritus species, M. zaandami and δ 18 O values of N. pachyderma (Spielhagen et al., 2005) in core PS2458. J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007)

14 208 J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007) Table 6 PS2458 number of species, specimens counted, biodiversity indices Fisher α, H, ln(s), and ln(e) Depth (cm) Age (kyr BP) frequencies N0.2. Power spectral analysis reveals cyclicities of 4.12, 1.57, 1.18, 0.76, and 0.54 kyrs above the 95% confidence level (Fig. 6). Separate analyses of the Holocene (0 11kyr BP), and glacial (11 24 kyr BP) core section show that the 1.57 and 0.76 kyr cyclicities are confined to the glacial, and the 4.1, 1.18, and 0.54 kyr cyclicities occur in the Holocene interval. In contrast to the 0.5 and 1.1 kyr cyclicities there is no repetitive drop in Fisher α gradients at intervals of 4.1 kyr. The peak at 4.1 kyr may thus indicate a multiple or combination of tones of the 1.18 and 0.54 kyr periodicities (see Rial and Analcerio, 2000; Esper et al., 2004). 4. Discussion Species (no.) Specimens (no.) Fisher α 4.1. Reliability of the biodiversity signal: reproduction and taphonomic controls Like everywhere in the world, Arctic benthic foraminiferal thanatocoenosis is no one-to-one record of the living fauna (Wollenburg and Mackensen, 1998a,b; Wollenburg and Kuhnt, 2000). Because dead assemblages H ln (S) ln(e) give a time-averaged record over many years (decades to centuries at the sites of this study) their species number occasionally exceed that of the living faunas at the same sites. However, taphonomic effects must be taken into account, e.g. addition of species into the area and loss of species through in situ dissolution of calcareous tests or breakdown of fragile agglutinated forms (e.g. Murray and Alve, 1999; Sanders, 2003; Murray, 2003). We already mentioned the incorporation of allochthonous ice-rafted benthic foraminifera in foraminiferal thanatocoenosis, and that we tried to minimize this problem by the exclusion of species restricted to a shallow water habitat. However, there remains the problem of species that are distributed from shallow to deeper water realms like Cassidulina reniforme (10 N 3000 m, Schröder-Adams et al., 1990; Wollenburg and Mackensen, 1998a,b; Korsun and Hald, 2000). It is impossible to determine how many specimens of such species are of autochthonous or allochthonous origin. Yet, typical ice-rafted benthic foraminifera (see Wollenburg, 1995) are rare in the analyzed cores (0 b 2%), as should be the allochthonous proportion of such problematic species. Approximately 180 agglutinated species constitute 20 to 85% of the modern Arctic benthic foraminifera fauna in the study area (Wollenburg and Mackensen, 1998a,b; Wollenburg and Kuhnt, 2000). Most of these species have no or only low fossil potential (Wollenburg and Kuhnt, 2000). However, the most abundant agglutinated species at modern sites of m water depth, are common faunal components of the analyzed core samples. In the modern Arctic Ocean the regional proportion of agglutinants is positively related to carbon flux and/or the presumed seasonal or permanent distribution of corrosive bottom waters. It is suggested that the permanent influence of such corrosive bottom waters not only lead to taphonomic loss in calcareous specimens, but also prevents the secretion of calcareous tests, favouring the distribution of agglutinated species (Scott and Vilks, 1991; Wollenburg and Kuhnt, 2000). In the analyzed cores, samples with high proportions ( 60%) of agglutinated foraminifera reveal low WBFAR (spec./10 g dry sediment/ 1 kyr, see Section 2.2.5), usually lack pristine calcareous specimens, and small sized and thin-walled calcareous species are typically absent. Such samples provide evidence of short or long-lasting intervals during GI-1 and/or GS-1 (all cores), and the last 8.5 kyrs of cores PS2138 and PS2458, (Figs. 2 5) with prevailing carbonate aggressive conditions. It is difficult to qualify and quantify the taphonomic loss even for the transfer from modern bio- to taphocoenosis, due to the lack of year-round observations. Yet, for site PS2837, PS2458 and PS2212 the biodiversity of living

15 J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007) Fig. 6. Blackman Tuckey spectrogram of Fisher α frequencies above the 95% confidence level in core PS2837. a) Spectrogram of the interval 0 24 kyrs given by 0.488bDP/Pb3.08. Solid line Blackmand Tukey, dashed line maximum entroy method. b) Spectrogram of the glacial interval kyrs given by 0.546bDP/Pb Solid line Blackmand Tukey, dashed line maximum entroy method. c) Spectrogram of the last 11 kyrs given by 0.545bDP/Pb2.53. Solid line Blackmand Tukey, dashed line maximum entroy method. foraminifera approximates the core top thanatocoenosis. Furthermore, although the core samples may suffer taphonomic loss of agglutinated and/or calcareous species, none of the cores reveals a negative relation of biodiversity with either WBFAR or the proportion of agglutinated species. In contrary, species richness is usually high in samples accumulated under prevailing carbonate aggressive conditions, despite the potential loss (Figs. 3 5). We thus suggest that taphonomic controls had only a modest effect on the observed variabilities of species richness per analyzed unit of approximately 300 specimens Comparison of biodiversity measurements In this study we discuss two different measures of biodiversity, the Fisher α index and the information function.

16 Fig. 7. Comparison of Fisher α down core gradients in cores PS1290, PS2837, PS2212 and PS2458 with δ 18 O values in GISP2 ice core (Stuiver and Grootes, 2000), and δ 18 O values of N. pachyderma in core MD (Dreger, 1999). 210 J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007)

17 J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007) Fig. 8. Down core gradients of Fisher α and IRD in core PS2837 versus IRD records of cores GIK (N75 N, 14 E; Sarnthein et al., 2003), Hole (49.87 N, W; Bond et al., 1992) and biogenic silica (BSi) record of Arolik Lake (59 28 M, W; Hu et al., 2003). Horizontal lines mark cold events indicated by the biodiversity and BSi record, shading marks periods of good correlations between both data sets. Hereby, results are comparable to most other biodiversity studies. The Fisher α indices, as function of the species richness, at all core sites reveal correlatable temporal variabilities (see inter-core-correlation coefficients, Table 7). In contrast, there is no constant spatiotemporal variability of H. In cores of high species richness and low equitability (high correlation coefficients of core ln(s) and(h), as in cores PS2837 and PS2458, the general temporal variability of H matches the Fisher α. A counter-example is core PS2212, because its H values are almost exclusively determined by equitability (correlation coefficient ln(e) and H=0.94). We suggest that Fisher α is a suitable measure of palaeobiodiversity in the Arctic Ocean, because all core sites reveal correlatable temporal variabilities. The validity of the information function and equitability strongly depends on the sample size and the need to recognise every species of the environment (e.g. Buzas, 1979; Murray, 1991, Table 1). Although the cores of this study were sampled with constant surface areas and volumes, the WBFAR is far from constant but varies from 0 to specimens per sample (PS2837). It is thus impossible to avoid problems in the application of H and E.Incontrastto the spatiotemporal correlatable species richness values

18 Table 7 Intra-core/inter-core correlation coefficients of the distinct records of biodiversity parameters, correlation coefficients of biodiversity records with δ 18 O records of the GISP2 ice core, with δ 18 O records of N. pachyderma from the same core and core MD PS2837 PS1290 PS2212 PS2458 Fisher α H ln(s) ln(e) Fisher α H ln(s) ln(e) Fisher α H ln(s) ln(e) Fisher α H ln(s) ln(e) Gisp2 δ 18 O MD δ 18 O N. pachyd PS PS PS PS PS2837 Fisher α PS2837 H PS2837 ln(s) PS2837 ln(e) PS1290 Fisher α PS1290 H PS1290 ln(s) PS1290 ln(e) PS2212 Fisher α PS2212 H PS2212 ln(s) PS2212 ln(e) 0.14 PS2458 Fisher α PS2458 H PS2458 ln(s) J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007)

19 J.E. Wollenburg et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 255 (2007) (correlation coefficients of Fisher α and ln(s) 0.67), the down core equitability patterns of the core sites differ significantly (correlation coefficients of 0.5 to 0.6). In samples from the last deglaciation low WBFAR are often attended by rapid shifts in equitability and thus H (Figs. 2 5; see also Wollenburg and Kuhnt, 2000). Hereby, the H values of sediments from the GI-1 to GS-1 are more variable than the Fisher α values, and the meaning of H switches between an image of species richness to a measure of equitability. Except for site PS2837, the Holocene increase in species richness is concomitant by low equitabilities. Therefore, at most core sites the Holocene H values differ only insignificantly from glacial H values, derived from species poor samples of high equitability (e.g. Fig. 3, PS1290) Mechanisms driving benthic foraminifera biodiversity during the last 24 kyr Once considered to be a constant, spatially uniform, and isolated environment, the deep-sea is now recognized as a dynamic, richly textured environment that is inextricably linked to the global biosphere (Levin et al., 2001). Thus, regional deep-sea biodiversity patterns are controlled by disturbance, heterogeneity, biotic interactions (competition predation), temperature, the availability of food, and oxygen concentrations (Cronin and Raymo, 1997; Gooday and Rathburn, 1999; van der Zwaan et al., 1999; Levin et al., 2001; Lambshead et al., 2002; Snelgrove and Smith, 2002). The cores of this study were taken from 1000 m and 2550 m water depth and separated by more than 2300 km. Because their temporal variability in species richness is comparable (inter-core correlation coefficients rn0.6 for correlations with core PS2458, rn0.8 remaining cores), biodiversity maxima and minima should largely be the consequences of supra-regional processes: oceanography, climate, or climate-forced changes in environmental conditions like changes in sea-ice retreat and palaeoproductivity. In higher latitudes sea-surface temperatures, wind and current systems besides nutrients determine primary production. Due to a rapid and efficient benthic pelagic coupling, primary production and the high latitudinal deep-sea fauna respond to surface ocean processes within days (e.g. Gooday and Turley, 1990; Graf, 1992; Linke et al., 1995; Drazen et al., 1998; Boetius and Damm, 1998; Heinz, 1999; Wollenburg and Kuhnt, 2000; Gooday and Hughes, 2002). Despite this fact, the observed decrease in species richness with increasing latitude is assumed to be mainly a reflection of seasonal versus stable carbon fluxes (Rex et al., 1993; Thomas and Gooday, 1996; Hayek and Buzas, 1997; Lambshead et al., 2000; Culver and Buzas, 2000). Other workers suggested that biodiversity is positively related to the availability of food under oligotrophic to mesotrophic conditions (e.g. Gooday and Rathburn, 1999; Kurbjeweit et al., 2000; Culver and Buzas, 2000; Lambshead et al., 2002). However, this correlation becomes negative when high carbon fluxes cause a severe depletion in bottom water oxygen contents (e.g. Hermelin and Shimmield, 1990; Gooday et al., 2000; den Dulk et al., 2000; Heinz and Hemleben, 2003; Schmiedl and Leuscher, 2005; Singh and Gupta, 2005). This finding is partly corroborated by data from the modern Arctic Ocean, which showed that biodiversity of benthic foraminifera is positively related to carbon flux values b 7gC/m 2 /yr, yet negatively beyond this value (Wollenburg and Kuhnt, 2000). However, due to the lack of oxygen measurements from pore water and the bottom-pore water interface, the actual cause of the inverse correlation at high carbon fluxes remains speculative. In the analyzed cores species richness and the δ 18 O values of Neogloboquadrina pachyderma 2 show a moderate to strong negative correlation (correlation coefficients r= 0.5 for core PS2458, 0.62 to r= 0.81 remaining cores; Table 7). Although not all samples with low planktonic oxygen isotope values reveal a highly diverse benthic fauna, the relation between heavy δ 18 O values of N. pachyderma and low biodiversity faunas is strict. This also applies for the faunal parameters Atlantic species and phytodetritus species. These parameters do not allow to estimate diversity; yet, low biodiversity faunas characterize samples with vanishing amounts of Atlantic and phytodetritus species. The reverse, to deduce low biodiversities from low portions of Atlantic and phytodetritus species is impossible, because, apart from one agglutinated Atlantic species, both groups are comprised by thin-shelled calcareous species with low resistance versus carbonate aggressive conditions, that prevailed for extended periods of the last deglaciation and Holocene (see Section 4.1.). During glacial and deglacial periods (GI-2 to GS-1) low biodiversity faunas can be deduced from low carbon fluxes; however, highly productive intervals not necessarily generated a high species richness, and any relation between carbon flux and biodiversity was lost during the Holocene. Regarding the temporal biodiversity fluctuations against the background of the 2 In contrast to the other time-equivalent oxygen isotope records of the Arctic Ocean, the record of PS2458 reflects salinity changes which were influenced by variable freshwater runoff and a growing marine influence during the postglacial transgression of the Laptev Sea shelf (Spielhagen et al., 2005).