Evolution of iceberg melting, biological productivity, and the record of Icelandic volcanism in the Irminger basin since 630 ka

Marine Geology 212 (2004) 133 152 www.elsevier.com/locate/margeo Evolution of iceberg melting, biological productivity, and the record of Icelandic volcanism in the Irminger basin since 630 ka Kristen St. John a, *, Benjamin P. Flower b,1, Lawrence Krissek c,2 a Department of Geology, Appalachian State University, Boone, NC, USA b College of Marine Sciences, University of South Florida, St. Petersburg, FL, USA c Department of Geological Sciences, Ohio State University, Columbus, OH, USA Received 8 September 2003; received in revised form 27 August 2004; accepted 17 September 2004 Abstract Planktic y 18 O and y 13 C records and point count records of biogenic, volcanic, and nonvolcanic terrigenous [ice-rafted debris (IRD)] sediment components from Hole 919A in the Irminger basin, northern North Atlantic provide a comprehensive dataset from which a paleoceanographic reconstruction for the last 630 kyr has been developed. The paleoceanographic evolution of the Irminger basin during this time contains both long-term patterns and significant developmental steps. One long-term pattern observed is the persistent deposition of hematite-stained ice-rafted debris. This record suggests that the modern and late Pleistocene discharges of icebergs from northern redbed regions to the Irminger Sea lie in the low end of the range observed over the last 630 kyr. In addition, Arctic front fluctuations appear to have been the main controlling factor on the long-term accumulation patterns of IRD and planktic biogenic groups. The Hole 919A sediment record also contains a long-term association between felsic volcanic ash abundances and light y 18 O excursions in both interglacial and glacial stages, which suggests a causal link between deglaciations and explosive Icelandic eruptions. A significant developmental step in the paleoceanographic reconstruction based on benthic evidence was for diminished supply of Denmark Strait Overflow Water (DSOW) beginning at ~380 ka, possibly initiated by the influx of meltwater from broad-scale iceberg discharges along the east Greenland coast. There is also planktic evidence of a two-step cooling of sea surface conditions in the Irminger basin, first at ~338 309 ka and later at ~211 190 ka, after which both glacials and interglacials were colder as the Arctic front migrated southeast of Site 919. In addition to offering * Corresponding author. New address: Department of Geology and Environmental Science, MSC 7703, James Madison University, Harrisonburg, VA, USA. Fax: +1 540 568 8058. E-mail addresses: stjohnke@appstate.edu (K. St. John)8 bflower@seas.marine.usf.edu (B.P. Flower)8 krissek@mps.ohio-state.edu (L. Krissek). 1 Fax: +1 727 553 1189. 2 Fax: +1 614 292 1496. 0025-3227/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.margeo.2004.09.004

134 K. St. John et al. / Marine Geology 212 (2004) 133 152 these findings, this reconstruction provides a longer-term geologic context for the interpretation of more recent paleoceanographic events and patterns of deposition from this region. D 2004 Elsevier B.V. All rights reserved. Keywords: Irminger basin; Pleistocene; marine sediments; ice-rafting; foraminifera; volcanic ash 1. Introduction A key question in understanding North Atlantic paleoceanography is whether the climatic events and cycles of deposition of the last glaciation were typical or atypical of the Pleistocene. While it is established that an interplay of orbital and suborbital variables largely controlled marine sedimentation since at least the last glacial maximum (e.g., Bond et al., 1997; Shackleton et al., 2000; Elliot et al., 2001), the longerterm paleoceanographic record of the Pleistocene is not as well developed, and therefore the broader context is lacking when interpreting recent millennial scale climate variability. This study aims to fill in some of that context by reconstructing the paleoceanographic and paleoclimatic history of the western Irminger basin since 630 ka through a comprehensive examination of the isotopic and biogenic, volcanic, and nonvolcanic terrigenous [ice-rafted debris (IRD)] sediment records from Ocean Drilling Program Hole 919A (Fig. 1). With high sedimentation rates of icebergtransported debris, biogenic remains, and occasional volcaniclastic material and due to its subpolar setting south of the Denmark Strait and near the Arctic front, the Irminger basin is sensitive to climate change and well suited for paleoceanographic studies. Fig. 1. Map showing the location of ODP Site 919 in the western Irminger basin, the general locations of potential IRD source areas, major modern surface currents, and the position of the modern Arctic front.

K. St. John et al. / Marine Geology 212 (2004) 133 152 135 Important and relevant paleoceanographic results have already come from Irminger basin sediment records and from other nearby localities. For example, Bond and Lotti (1995) used a 35-kyr record of provenance-distinct IRD (hematite-stained grains) from the Denmark Strait to support the conclusion that synchronous ice-rafting events in the North Atlantic correlated to the 2 3 kyr Dansgaard Oeschger (D O) temperature oscillations in Greenland ice cores during marine isotope stage (MIS) 2 3. Elliot et al. s (1998, 2001) high-resolution studies of the iceberg discharges to the western Irminger basin since 60 ka further established that sediment input to the Irminger basin was affected by two oscillating and interacting systems: Heinrich-type deposition due to the release of massive iceberg armadas from continental ice sheets every 5 10 kyr and more frequent D O ice discharge events from northern coastal ice sheets. In addition, van Kreveld et al. (2000) made the argument that D O cycles were driven by the internal dynamics of the east Greenland ice sheet based on surface and deepwater paleoclimate records from the Irminger Sea. Other relevant Irminger basin studies consisted of longer-term (Plio Pleistocene) but lower resolution paleoceanographic reconstructions, in which generally only single proxies were used. This primarily includes those studies related to Ocean Drilling Program Leg 152 (Saunders et al., 1998), such as Spezzaferri s, (1998) planktic foraminifera biostratigraphy, Flower s (1998) development of oxygen and carbon isotope stratigraphies for Hole 919A, and Koç and Flower s (1998) tracking of the Arctic front through diatom abundance and preservation records. In addition, Krissek and St. John (2002) used the IRD mass accumulation record of Hole 919A since 960 ka to identify temporal covariations of provenance-distinctive grain types; this covariation suggested that glaciated areas in the Precambrian basement of SE Greenland and in the Tertiary flood basalts south of Scoresby Sund experienced similar iceberg release histories during the Pleistocene. By examining the abundances of multiple sediment components at an average sample resolution of 1.3 kyr over the last 630 kyr, the dataset described here improves our understanding of the paleoceanographic evolution of the Irminger basin. This study provides a longer temporal record than that of Elliot et al. (1998, 2001) (630 vs. 60 kyr), and it has better age control, has more detailed temporal resolution, and is more comprehensive than previous studies of long-term paleoclimatic records from this area (e.g., Flower, 1998; Koç and Flower, 1998; Spezzaferri, 1998; Krissek and St. John, 2002). In addition, no previous studies of the Irminger Sea examined the long-term sediment input records of hematite-stained grains or dispersed (nonlayered) volcanic ash. In sum, this study positions us to address the question of how the last glacial cycle fits in longer-term patterns of Pleistocene climate variability. 2. Methods 2.1. Age model Oxygen and carbon isotope data from Hole 919A were generated on Neogloboquadrina pachyderma (s) from the 150 250-Am size fraction (Flower, 1998; Koç et al., 2001; this study). Data are reported on a meters composite depth (mcd) scale based on splicing Holes 919A and 919B using shipboard magnetic susceptibility (Shipboard Scientific Party, 1994). This procedure accounts for known gaps between cores. Below core 919A 3H (29.02 mcd), no material is available to splice between cores, so successive cores were placed below previous cores on the mcd scale after sediment expansion. N. pachyderma (s) oxygen isotope data were correlated to the orbitally tuned benthic y 18 O composite of cores V19 30 (Shackleton and Pisias, 1985) and ODP Site 677 (Shackleton et al., 1990), updated to a retuned time scale (Shackleton, 2000). This time scale is similar but not identical to that found on Shackleton s Delphi website (http://www.delphi.esc. cam.ac.uk/coredata/v677846.html). For Hole 919A, transitions between oxygen isotope stages were used exclusively to minimize bovertuningq the record and introducing artificial sedimentation rate changes, following Shackleton (2000). It is difficult to identify all the known isotopic stages and substages because of coring gaps, insufficient data resolution, and overprinting by isotopically light meltwater. For example, the marine isotope stage (MIS) 5/4 boundary (5.0 in

136 K. St. John et al. / Marine Geology 212 (2004) 133 152 the nomenclature of Prell et al. (1986)) is subject to interpretation. The MIS 5/4 boundary (74 ka) is interpreted to lie at 15.0 mcd but could lie at 10.95 mcd. However, the latter placement is inconsistent with the interpretation of Ash Zone 2 (~55 ka) at 11.19 mcd (Lacasse et al., 1998). Furthermore, a well-known minimum in planktic y 13 C occurs during MIS 4 in the northwest Atlantic (Labeyrie and Duplessy, 1985; Elliot et al., 1998). The presence of this minimum at ~14 mcd supports our interpreted MIS 5/4 boundary at 15.0 mcd. The data do not allow unambiguous definition of the MIS 4/3 and 3/2 boundaries. The resultant sedimentation rates range from 7.4 during MIS 9 to 27.6 cm/kyr during MIS 13, with a mean of 15.4 cm/kyr. Overall, the existing data allow adequate age control for defining orbital-scale climate variability in the Irminger Sea (Table 1). Foraminiferal assemblage data and ice-rafted debris (IRD) data allowed refinement of the age model at glacial terminations by identifying when interglacial conditions commenced relative to N. pachyderma (s) y 18 O. This is necessary because low-salinity meltwater clearly influenced surface waters in the subpolar North Atlantic during some terminations, leading to an early y 18 O decrease relative to global ice volume (e.g., Lehman et al., 1993; Raymo et al., 1998; Oppo et al., 2001; Table 1 Age-control points and calculated linear sedimentation rates (LSR) for Hole 919A Event Age (ka) Depth (mcd) LSR (m/kyr) 2.0 13 1.7 0.13 5.0 74 15.0 0.22 6.0 132 20.3 0.09 7.0 190 27.6 0.13 8.0 249 35.4 0.13 9.0 305 46.4 0.20 10.0 336 48.7 0.07 11.0 364 52.6 0.14 12.0 425 57.7 0.08 13.0 480 65.9 0.15 14.0 526 78.6 0.28 15.0 567 89.6 0.27 16.0 622 94.7 0.09 The calculated LSR is to be read from the table as follows: from 0 to 13 ka, the LSR is 0.13 m/ky, from 13 to 74 ka, the LSR is 0.22 m/ ky, and so on. Wright and Flower, 2002). In Hole 919A, early y 18 O decreases (to values b3.0x) are often accompanied by high % IRD and high % N. pachyderma (s), indicative of glacial conditions. Interglacial conditions commenced when % IRD decreased below about 25% and/or when % N. pachyderma (s) decreased below about b75%. Accordingly, we considered that glacial terminations were marked by transitions toward minimum values in % IRD and % N. pachyderma (s) after y 18 O reached minimum values (b3.0x). For example, at the MIS 12/11 transition, N. pachyderma (s) y 18 O decreased to 3.0x at 58 mcd, but % IRD and % N. pachyderma (s) reached interglacial levels at 57.7 mcd, indicating that MIS 12.0 should be placed at 57.7 mcd. Similar relationships at terminations I VII allowed refinement of the age model found in Flower (1998). 2.2. Sediment counts Sediment counts were completed every 20 cm downcore to 96 mcd, yielding a total of 450 samples and an average sampling interval of 1.3 kyr. For each of these counts, the N150-Am grain size fraction was split to about 500 grains, weighed, and examined under a binocular microscope. Specimens of each planktic foraminifera species, quartz mineral grains, hematite-stained mineral grains, mafic and felsic crystalline rock fragments, sedimentary rock fragments, dark (brown and black) and colorless volcanic ash (i.e., tephra), benthic foraminifera, pyritized burrows, radiolarians, and diatoms were counted; the results of which are discussed here. Because volcanic ash has multiple transport modes, IRD was defined on an ash-free basis in this study; IRD includes the nonvolcanic nonauthigenic terrigenous mineral and rock fragments. 3. Results 3.1. Stable isotopes Oxygen isotope data exhibit a series of variations with amplitudes of 0.5x to 1.3x between 0 and 48 mcd (331 ka), and individual values range between 2.59x and 4.67x (Figs. 2 and 3). Higher

K. St. John et al. / Marine Geology 212 (2004) 133 152 137 Fig. 2. Total volcanic ash counts, nonvolcanic terrigenous (IRD) counts, N. pachyderma (s) carbon and oxygen isotope variations, the % N. pachyderma (s) record, and calculated linear sedimentation rates for Hole 919A plotted versus depth in meters composite depth (mcd). All sediment counts were on the N150-Am grain size fraction. Isotope variations are expressed as per mil. amplitude excursions are more common in the record below 48 mcd and include high-amplitude light isotope excursions of 1.41x to 3.2x. The decrease at 96 95 mcd also defines the maximum range of y 18 O values in the 0 96 mcd record, from 2.03x to 5.2x.

138 K. St. John et al. / Marine Geology 212 (2004) 133 152 Fig. 3. Hole 919A planktic foraminifera faunal records versus age. Due to their generally low individual relative abundances, percentages of Globoconella inflata, G. scitula, G. glutinata, and Orbulina species are grouped together as % of other nonpolar foraminifera groups. For reference, marine isotope interglacial stages are shaded and numbered.

K. St. John et al. / Marine Geology 212 (2004) 133 152 139 Carbon isotope data exhibit a series of variations with amplitudes of 0.5x to 1.28x between 0 and 96 mcd (631 ka), and individual values range from 1.0x to +0.28x (Figs. 2 and 4). Below 76 mcd (519 ka), y 13 C values are negative, with the exception of a positive excursion of 0.17x at 82.3 mcd (539.9 ka). Between 75 and 36.6 mcd (255.1 ka), y 13 C values display sharp high-amplitude variations. Above 36.6 mcd, carbon isotopic variations are generally more gradual and less negative than those observed between 75 and 36.6 mcd. The greatest negative excursion in the carbon isotope record ( 1.0x) coincides with a sharp decrease in y 18 O at 59.6 57.9 mcd (438 426.5 ka). 3.2. Sediment counts The biogenic fraction was dominated by planktic foraminifera, varying in abundance from 0 to N176,000 counts/gram (Fig. 3). Neogloboquadrina (s) was the most abundant planktic foraminifera species, comprising N88% of the total planktic Fig. 4. Hole 919A biogenic counts of siliceous planktic taxa, benthic foraminifera, and pyritized burrows versus age. Marine isotope interglacial stages are shaded and numbered.

140 K. St. John et al. / Marine Geology 212 (2004) 133 152 Fig. 5. Hole 919A volcanic and nonvolcanic terrigenous (IRD) counts versus age and IRD component relative abundances. Marine isotope interglacial stages are shaded and numbered.

K. St. John et al. / Marine Geology 212 (2004) 133 152 141 foraminifera specimens in the overall record and commonly forming N95% of the total foraminifera deposited since 190.8 ka. However, N. pachyderma (s) abundances exhibit high-amplitude fluctuations between 190.8 and 485 ka and smaller amplitude fluctuations between 486 and 630 ka. Other than the brief disappearance of N. pachyderma (s) at 190.8 ka, which coincided with a drop in the total foraminifera Fig. 6. Generalized stratigraphic column for Hole 919A, highlighting depth horizons of visually identified ash beds and the results of the total ash counts and the relative percentages of colorless ash of the total ash input.

142 K. St. John et al. / Marine Geology 212 (2004) 133 152 abundance, all of the decreases in N. pachyderma (s) coincided with increases in one or more nonpolar foraminifera groups, including N. pachyderma (d), Turborotalia quinqueloba, Globigerina bulloides, Globoconella inflata, Globorotalia scitula, Globigerinita glutinata, and Orbulina species. Collectively, there was a distinct change in the planktic foraminifera record near the end of marine isotope stage (MIS) 7, which is marked by a significant decrease in foraminiferal species-level diversity. Diatoms, radiolarians, benthic foraminifera, and pyritized burrows were most abundant and diverse during MIS 11 14 but became rare or absent in younger sediments (Fig. 4). The nonvolcanic terrigenous component (IRD) varied in abundance from 0 to 122,000 counts/gram, with a mean value of 11,438 counts/gram for the last 630 kyr (Figs. 2 and 5). High IRD events (defined as N50,000 counts/gram) were most common during glacial stages, particularly during MIS 14, 12, 8, and 6. High IRD events also occurred during interglacials, including those in MIS 11 and 9. An extended interval of high IRD input occurred since the onset of MIS 4. Quartz is the dominant IRD component, forming an average of 84% of the total IRD. Some of these quartz grains are hematite-stained. Such grains vary widely in abundance (0 100%); on average, however, 25% of the total IRD input was hematite-stained. Crystalline felsic rock fragments, mafic rock fragments, and sedimentary rock fragments comprised the remainder of the IRD input to Hole 919A, averaging 6%, 3%, and 7%, respectively, of the total IRD. Sedimentary rock fragments and crystalline mafic rock fragments displayed similar patterns of accumulation; they were both more abundant between 630 and 255 ka than in sediments younger than 255 ka. Crystalline felsic rock fragments were most abundant before 540 ka and also within MIS 6 and 5. Volcanic ash varied in abundance from 0 to 82,143 counts/gram and had a mean value of 5391 counts/ gram for the last 630 kyr (Figs. 2 and 5). High volcanic ash input (N25,000 counts/gram) did not occur in this record until the end of MIS 11 (381.5 ka). The majority (82% on average) of the ash was brown or black. While colorless ash was less abundant overall, it periodically comprised ~60% to 100% of the total ash content (Fig. 6). With one exception, the tephra events with N25,000 counts/gram contained either a bimodal mix of dominantly brown/black ash with lesser amounts of colorless ash or were comprised completely of brown/black ash. The tephra event at 6.5 ka, however, was a bimodal mix of dominantly colorless ash and lesser amounts of brown/black ash. 4. Discussion 4.1. Evolution of SST variability in the Irminger Sea The abundances of the polar species N. pachyderma (s) and of nonpolar foraminifera groups in the Hole 919A sediment counts can be used as proxies for relative sea surface temperatures (SST) estimates in the Irminger Sea at the time of deposition. These proxies appear to delineate three intervals during the past 630 kyr with distinct SST characteristics (Fig. 3). First, between 630 and 486 ka, polar and nonpolar taxa experienced moderate variations in their abundances, although these variations were not always in tune with global climate trends as defined by marine isotope stages. For example, prior to 514 ka, the percentages of N. pachyderma (s) generally were higher in glacial stages and lower in interglacials, however, from 514 to 482 ka, N. pachyderma (s) percentages consistently were at or above 88%, indicating that the surface ocean in the Irminger Sea was cool during this time, in contrast to warm global climate trends. The second distinct SST interval began at 485 ka and ended between 211 and 190 ka. This interval was characterized by high-amplitude and short-duration SST changes, as indicated by percentages of N. pachyderma (s) that fluctuated between N90% and 20% within 1.5 to 7 kyr. These extreme and rapid abundance fluctuations in relative abundance primarily occurred during interglacials, but these also occurred throughout glacial stage 12. Interval 2 (485 190 ka) is also characterized by a two-step decrease in the abundance and diversity of warm water taxa. The first step occurred by the end of MIS 9 when decreases in SST appear to have been extreme enough to negatively impact many of the nonpolar foraminifera groups. Between 338 and 309 ka, T. quinqueloba, G. bulloides, G. scitula, and Orbulina species all decreased in abundance or temporarily

K. St. John et al. / Marine Geology 212 (2004) 133 152 143 disappeared from the record. Based on a diatom extinction event at ~315 ka, Koç et al. (2001) also argued for a significant cooling in sea surface conditions in the Irminger Sea during MIS 9. Warm water foraminifera made a temporary recovery at the transition to interglacial MIS 7 but experienced a second drop in abundance and diversity between 211 and 190 ka. During this time, all nonpolar foraminifera groups, diatoms, and radiolarians virtually disappeared from the Hole 919A record (Figs. 3 and 4). This change reflects the local effect of global cooling at the MIS 7 6 transition. Irminger Sea cooling is consistent with the interpretation by Funder et al. (1998) that ice sheets in the Scoresby Sund area reached their maximum extent during MIS 6. An additional observation of both intervals 1 (630 486 ka) and 2 (485 190 ka) is that light y 18 O excursions tend to precede both N. pachyderma (s) abundance decreases and total foraminifera abundance increases in interglacial MIS 15, 13, 11, 9, and 7 (Fig. 3). This may mean that deglaciations during this time are marked by early meltwater input followed by later warming of SSTs and bioproductivity increases. A similar observation was made by Elliot et al. (1998) for the last 60 kyr in the Irminger basin. The third SST interval was established since 190 ka and continued through the remainder of the Pleistocene. This interval is characterized by persistently cold SSTs in the Irminger Sea during both glacials and interglacials. This is documented by consistently very high (N95%) percentages of N. pachyderma (s) from 190 ka to glacial MIS 4 at 69 ka and high but slightly more variable percentages of N. pachyderma (s) since 69 ka. This modest change in SST conditions interpreted at 69 ka is consistent with the results of Funder et al. (1998) who recognized a shift to milder climates at 70 ka in the Scoresby Sund region. 4.2. Evolution of Arctic front variability in the Irminger Sea The modern-day Arctic front transects the Irminger Sea from the southwest to the northeast and defines the boundary between cold, ice-laden Arctic waters, and relatively warmer and more saline North Atlantic waters (Fig. 1; Malmberg, 1985; Kuijpers et al., 2003). During late Pleistocene glacial times, the position of the Arctic front sometimes shifted far to the southeast, resulting in increased IRD deposition in the North Atlantic (~508 N) and reduced heat and moisture transport to the polar region (Ruddiman and McIntyre, 1981; Smythe et al., 1985). Qualitative estimates of diatom abundance and preservation have already demonstrated that Site 919 lies in a position to monitor fluctuations in the Arctic front (Koç and Flower, 1998). Those results indicated that the Arctic front was east of Site 919 during glacial stages 16, 12, 10, 6, and 2 and west of Site 919 during glacial stages 20, 18, 8, and 4. To further investigate the history of Arctic front, the percentages of N. pachyderma (s) and T. quinqueloba for the 0 630-ka interval of the Hole 919A record are compared here (Fig. 7). This approach is based on observations made by Johannessen et al. (1994) that the Arctic front is characterized by a faunal transition, with N. pachyderma (s) dominant in Arctic waters and T. quinqueloba more abundant near the sea ice edge in North Atlantic water. Wright and Flower (2002) used relative percentages of these two taxa in records from Feni Drift and Bjorn Drift, south of Iceland to track the position of the Arctic front between 500 1000 ka; we have applied their approach to the Hole 919A record. In general, prior to 190 ka, there were highamplitude variations in the abundances of both N. pachyderma (s) and T. quinqueloba during interglacials. This suggests that the Arctic front and its associated sea ice edge repeatedly migrated to a position near but slightly to the northwest of Site 919 during most interglacials. The abundances of both taxa were generally less variable during glacials. N. pachyderma (s) abundances were generally high, and T. quinqueloba abundances were generally low, suggesting that the Arctic front migrated to a more stable position far to the southeast of Site 919 during most glacial periods. The primary exception is glacial stage 12; the repeated high-amplitude variations of the abundances of both taxa during MIS 12 imply a very dynamic Arctic front, more similar in character to the typical interglacial Arctic front. The general abundance patterns of N. pachyderma (s) and T. quinqueloba changes near the end of MIS 7. T. quinqueloba disappeared almost completely from the Hole 919A record in sediments younger than 205 ka, whereas abundances of N. pachyderma (s)

144 K. St. John et al. / Marine Geology 212 (2004) 133 152 Fig. 7. Arctic front (AF) indicator % N. pachyderma (s) and sea ice edge indicator % T. quinqueloba at Hole 919A for 0 630 ka. Arrows on % N. pachyderma (s) indicate that, when percentages were high, the AF was southeast of Site 919, and when low, the AF was northwest of Site 919. Arrows on % T. quinqueloba indicate that, when percentages were high, the AF was near Site 919, and when low, the AF was far from Site 919. Marine isotope interglacial stages are shaded and numbered. generally remained quite high since 190 ka. This suggests that the Arctic front primarily has been located far to the southeast of Site 919 since ~205 190 ka, migrating somewhat to the northwest only during MIS 4 and in the current interglacial. Overall, our foraminifera-based interpretation of Arctic front variability generally agrees with Koç and Flower s (1998) diatom-based interpretations. The results of this study are also consistent with Wright and Flower s (2002) interpretation that the Arctic front usually shifted to the northwest Atlantic during interglacials of the past 620 ka. However, unlike previous interpretations, this study indicates that the Arctic front was quite variable and frequently northwest of Site 919 during glacial stage 12. In addition, the foraminifera data indicate that the Arctic front was atypically positioned to the southeast of Site 919 during interglacial stages 5 and 13 (from 515 486 ka). 4.3. Evolution of nutrient supply to the Irminger Sea Interglacial stages in the Hole 919A record were generally characterized by relatively high abundances of planktic and benthic taxa and high y 13 C signals (Fig. 4). The glacial interglacial patterns of taxa abundances are interpreted to represent the local impact of global climate changes on nutrient inventories in the Irminger Sea. The y 13 C pattern is in agreement with the interpretation of Flower (1998) that the Hole 919A y 13 C signal since 960 ka was largely controlled by global inventory changes. In addition to the glacial interglacial pattern of productivity, the Hole 919A record also includes a distinct change in the abundance of siliceous planktic taxa (diatoms and radiolarians), benthic foraminifera, and pyritized burrows after MIS 11. Prior to ~380 ka, these groups all experienced extended episodes of high abundance. However, beginning within MIS 11, benthic foraminifera abundances decreased by a factor of 2 to 5, pyritized burrows disappeared until a brief reappearance around 200 ka, and diatoms and radiolarians decreased by a factor of 10 (Fig. 4). There are at least two possible reasons for the change in the planktic and benthic abundances after MIS 11. One possibility is that, after ~350 ka, the Arctic front moved farther away from Site 919 and did not return to a nearby position until a period from ~262 to ~220 ka (Fig. 7). The migration of the Arctic front presumably would also mean the migration of the high primary productivity zone in the surface waters associated with the Arctic front (Knudsen and Eiriksson, 2002) and the sea ice edge (Sathyendranath

K. St. John et al. / Marine Geology 212 (2004) 133 152 145 et al., 1995; Heinrich et al., 1996). Benthic productivity would also decrease because the main food source for benthic communities in the deep ocean is production in the water column (Graf, 1992; Knudsen and Eiriksson, 2002). However, there are two problems with this explanation: (1) the decreases in taxa abundances appear to precede the shift in the Arctic front by at least 30 kyr, and (2) when the Arctic front briefly returned to a position near Site 919 at 262 ka, the expected increases in taxa abundances did not occur. An alternate explanation for the pattern of benthic productivity before and after MIS 11 may be related to changes in the chemistry and circulation of the Denmark Strait Overflow Water (DSOW), which bathes Site 919. This subsurface water mass influences thermohaline circulation and global climate because it contributes to the formation of North Atlantic Deep Water (NADW; Broecker and Denton, 1989; Raymo et al., 1990; Kase and Oschlies, 2000). The benthic abundances prior to MIS 11 may be evidence of a strong influx of well-oxygenated DSOW. As discussed above, increased productivity near the Arctic front (Sathyendranath et al., 1995; Heinrich et al., 1996; Knudsen and Eiriksson, 2002) and/or increased nutrient supply from thermocline depths may explain the concurrent changes in siliceous planktic abundances. A decrease in oxygen to Hole 919A, resulting from decreased flow of DSOW, could account for the decrease in benthic productivity and diversity during and after MIS 11. Decreased DSOW during MIS 11 is consistent with interpretations of concurrent events in the North Atlantic and the Greenland Iceland Sea. Thunell et al. (2002) described a decrease in NADW production during MIS 11 based on a benthic y 13 C record from the western North Atlantic; diminished DSOW presumably could have contributed to the decrease in NADW production. In addition, the Icelandic basin experienced a period of extreme carbonate dissolution and high IRD input at ~400 ka, which Huber et al. (2000) related to a meltwater pulse during the interglacial. The largest IRD input to Hole 919A also occurred in MIS 11 (419 ka; Fig. 5), just before the decreases in benthic and siliceous planktic groups. This set of observations raises the possibility that broad scale iceberg discharges from along the east Greenland coast may have created a freshwater pool in the Greenland Sea during MIS 11 that negatively impacted the production of DSOW, and consequently the Irminger Sea benthic community, as well as the production of NADW. 4.4. Evolution of iceberg discharges to the Irminger Sea 4.4.1. The temporal pattern of IRD accumulation IRD events in the Irminger basin since MIS 16 were deposited under a range of climatic conditions, including glacial stages, interglacial stages, and climatic transitions (Fig. 5). IRD events since 190 ka generally occurred when SSTs in the Irminger Sea were cold, as indicated by high abundances (z80%) of N. pachyderma concurrent with increased IRD (Fig. 8). In contrast, from 630 to 190 ka, IRD events generally coincided with warm water pulses, as marked by decreased N. pachyderma (s) abundances and increased abundances of nonpolar planktic foraminifera (Figs. 3 and 8). Inasmuch as the Arctic front was positioned near (e.g., MIS 11 and 12) or to the northwest (e.g., during MIS 9) of Site 919 during much of this time (Fig. 7), the associated SST and nutrient gradients (Sathyendranath et al., 1995; Heinrich et al., 1996) were probably responsible for the increases in iceberg melt and the changes in polar and nonpolar planktic productivity. 4.4.2. IRD composition, iceberg sources, and dispersal The presence of hematite-stained IRD in the Irminger basin reflects the discharge of debris-laden icebergs from the east central coast of Greenland and possibly Svalbard (Bond and Lotti, 1995; Bond et al., 1997; van Kreveld et al., 2000). In this study, typically 20 40% of all the lithic (nonvolcanic) grains deposited at Hole 919A were hematite-stained (Fig. 5). Modern values for this area do not exceed 15% hematite-stained grains, and nearly all modern values N15% are in the Greenland Iceland sea (Bond et al., 1997). In addition, the abundance of hematite-stained grains at nearby site VM 28-14 was also low (10% or less) during the last glaciation (10 30 ka; Bond and Lotti, 1995). This indicates that the modern and late Pleistocene discharges of icebergs from northern red bed regions to the Irminger Sea lie in the low end of the range observed over the last 630 kyr.

146 K. St. John et al. / Marine Geology 212 (2004) 133 152 Fig. 8. Comparison of N. pachyderma (s) abundances, N. pachyderma (s) oxygen isotope variations, colorless ash input, and total IRD input versus age. Marine isotope interglacial stages are shaded and numbered. Dashed lines linking the oxygen isotope and colorless ash curves mark times when explosive eruptions coincided with light y 18 O excursions. When these dash lies also are linked with the IRD curve, it indicates IRD events also occurred at these times. The values of 3000 and 20,000 grains/g were used as the respective lowest abundance levels marking significant colorless ash and IRD peaks in this comparison; both levels are indicated by horizontal dotted lines. Marine isotope interglacial stages are shaded and numbered.

K. St. John et al. / Marine Geology 212 (2004) 133 152 147 In light of these conclusions, some IRD events elsewhere in the North Atlantic may require reinterpretation. For example, Bond and Lotti (1995) concluded that lithic peaks in the eastern North Atlantic (DSDP 609 and VM 23-81), which had high percentages (z15%) of hematite-stained grains and correlated to D O events in the GIS during MIS 2 3, must have resulted from discharges of ice in the Gulf of St. Lawrence. This conclusion is supported by good evidence for a Gulf of St. Lawrence source for lithic events in the western North Atlantic (e.g., Keigwin and Jones, 1995; Piper and Sleene, 1998) and by the judgment that the abundances of hematite-stained grains transported through the Denmark Strait are too low to explain the high abundances of those grains in the eastern North Atlantic. However, this study shows that Pleistocene icebergs commonly carried significant abundances of hematite-stained grains southward, out of the Greenland Iceland Sea and at least as far south as Site 919 in the Irminger basin. Thus, discharges of ice from eastern Greenland may have played a greater role in North Atlantic lithic (IRD) events than previously thought. While hematite-stained grains commonly formed 20 40% of the IRD input to Hole 919A, multiple episodes also were recorded when hematite-stained grains increased to 65 100% of the IRD input. At each of the times when the proportion of hematitestained grains was high (65 100%), total IRD input was low (Fig. 5). Most of these high hematite low total IRD events occurred during interglacials, but one also occurred during glacial stage 14. This inverse relationship between total IRD and the importance of hematite-stained IRD suggests that the Irminger Sea received a constant rainout of icerafted debris derived from central east Greenland and possibly Svalbard, but this input was often diluted by IRD input from other glaciated regions closer to the site of deposition. Other major IRD components identified in this study were felsic and mafic crystalline rock fragments and sedimentary rock fragments (Fig. 5). Previous compositional work identified the dominant respective source areas of such IRD in the Irminger basin to be the Precambrian igneous and metaigneous crystalline basement of SE Greenland, the Tertiary flood basalts near Scoresby Sund, and exposures of Paleozoic sedimentary units along the central east Greenland coast (Bond and Lotti, 1995; Krissek and St. John, 2002; St. John and Krissek, 2002; Fig. 1). The results here show that ice discharges from glaciated regions north of the Irminger basin contributed a greater proportion of IRD to Hole 919A than did ice discharges directly to the west inasmuch as the inputs of sedimentary rock fragments, mafic crystalline rock fragments, and hematite-stained grains are generally greater than the input of felsic crystalline rock fragments (Fig. 5). This interpretation is consistent with previous findings that identified a greater input of basaltic and sedimentary dropstones than gneissic dropstones at Hole 919A (Shipboard Scientific Party, 1994). A significant change in the input of provenance-specific IRD to Hole 919A occurred at ~255 ka. In sediments younger than 255 ka, IRD derived from glaciers draining mafic crystalline outcrops and sedimentary outcrops rarely was deposited at Hole 919A, while input from glaciers draining felsic crystalline outcrops remained variably low, and input from glaciers in red bed regions increased slightly. To account for this set of observations, we suggest that semipermanent sea ice may have been established since 255 ka in the fjords along the Scoresby Sund coast, blocking or redirecting the discharge of icebergs from areas draining basaltic and some sedimentary outcrops. An increased presence of sea ice in the Scoresby Sund region since 255 ka is consistent with evidence that ice sheets in the Scoresby Sund area reached their maximum size during glacial MIS 6 and that long-lasting sea ice cover was present over the continental slope off Scoresby Sund during that time (Funder et al., 1998). Differences in the thermal regimes of the glaciers draining the basalts and non-red bed sedimentary units and those draining the red beds could also explain the major decrease in the input of mafic crystalline and sedimentary rock fragments to Hole 919A since 255 ka, while the supply of hematitestained grains continued. The discharge of debrisladen icebergs containing mafic and sedimentary rock fragments would have been reduced if the glaciers draining those source areas shifted from being warm-based to cold-based at ~255 ka. More information from land-based glaciological studies is needed, however, to evaluate this possibility.

148 K. St. John et al. / Marine Geology 212 (2004) 133 152 4.5. Evolution of Icelandic explosive volcanism as recorded in the Irminger basin 4.5.1. Source and composition Tephra events at Hole 919A commonly were dominated by brown/black ash with lesser amounts of colorless ash (Fig. 6). Geochemical analysis (Lacasse et al., 1998) of Hole 919A tephra layers, zones, or pods indicated that each tephra event had an affinity with either a basaltic or a rhyolitic source in Iceland. We assume that the dispersed ash events identified in this study have similar Icelandic affiliations, with brown/black ash indicating input from a basaltic eruption, and colorless ash indicating input from a rhyolitic eruption. 4.5.2. Stratigraphic distribution Shipboard scientists visually identified six volcanic ash beds in sediments from Hole 919A (Fig. 6; Shipboard Scientific Party, 1994). Lacasse et al. (1998) suggested that the apparent lack of ash layers between ~14 and 60 mcd in Hole 919A either reflected a decrease in the volcanic/rifting activity as no ash fall events were recorded in the Icelandic basin (site 907) sediments between MIS 8 and 15 (245 620 ka) or may have resulted from an increase in sea ice during that time, which prevented the deposition of ash layers on the sea floor. However, the data presented here show multiple tephra (N25,000 counts/gram) events within the 14 to 60 mcd interval, including the six largest tephra events in our record. This suggests that ash input to the western Irminger basin was significant throughout the mid- to late Pleistocene. These events are dispersed and cannot be identified by visual core descriptions primarily because of dilution by glaciomarine and biogenic sediment components as well as mixing by bioturbation. These tephra events can be identified, however, by point count analysis of the total sediment population. 4.5.3. Transport modes Long-distant transport of ash from an explosive volcanic eruptions to a deep marine site of deposition can occur primarily in three ways: (1) ash can be transported directly through the atmosphere to open surface water, or (2) ash can be transported through the atmosphere and fallout onto sea ice and be transported as IRD, or (3) ash can be transported through the atmosphere and fallout onto glacial surfaces, subsequently being transported to a marine site as IRD. Additionally, in the special case of subglacial volcanic eruptions, ash may be carried to the adjacent sea via jokulhlaups (glacial floods) and related turbidity currents (Geirsdottir et al., 2000; Maria et al., 2000) but probably would not be dispersed into the atmosphere for any long distance transport. For at least the last 300 kyr, the primary dispersal direction for Icelandic tephra has been to the north and northwest under the influence of a prevailing wind and ocean current system similar to that of today (Haflidason et al., 2000; Wallrabe-Adams and Lackschewitz, 2003). Therefore, it is likely that most ash transported to the Irminger Sea arrived under the influence of the southward-flowing East Greenland Current (Fig. 1). Most of the tephra events identified at Hole 919A were diluted by other marine sediment components, and most of these dispersed tephra events cooccurred with increased IRD (Fig. 5). Assuming past wind regimes similar to those of today, this suggests that the Icelandic ash initially was transported through the atmosphere to glaciated coastal regions of Greenland or to sea ice along the Greenland coast and subsequently was ice-rafted to the western Irminger basin under the influence of the East Greenland Current. Direct atmospheric transport was probably a secondary mode of transport for this ash. Such transport must have occurred at least six times to explain the texture and bedding characteristics of the discrete ash beds (including chronostratigraphic marker beds, Ash Zones 1 and 2) in Hole 919A sediments (Fig. 6; Shipboard Scientific Party, 1994; Lacasse et al., 1998). Any subglacially erupted ash transported to the Irminger Sea via jokulhlaups and associated turbidity currents presumably would be concentrated in the northeastern part of the basin, adjacent to the Icelandic margin, and less so at Site 919 in the western part of the basin. Explaining the input of volcanic ash by surface currents (iceberg or sea ice-rafting) brings into question the time that could have elapsed between eruption and ultimate deposition at Hole 919A. Lacasse and van den Bogaard (2002) considered this question for Plio Pleistocene ice-rafted tephras recovered from the marine basins surrounding Iceland

K. St. John et al. / Marine Geology 212 (2004) 133 152 149 (including the Irminger basin). They concluded that the time involved was short, ranging from a few years to several hundred years, and was below the resolution of current y 18 O stratigraphy. Larsen et al. (1998) have shown that the residence time of volcanic material in some Icelandic glaciers to be as much as 800 years or more. Thus, a reasonable upper time limit for ash transported to Site 919 via ice-rafting is several hundred years, and perhaps 1000 years, whereas the time elapse was probably only on the order of days to weeks for ash transported to the seafloor site via atmospheric fallout. 4.5.4. Timing of tephra events A final topic to consider is the temporal relationship between tephra events and evidence of climatic change. We focus this discussion on the rhyolitic tephra events because rhyolitic ash is more likely to be erupted explosively, and explosive eruptions are more likely to be either a driver of climate change or a consequence of climate change. Airborne ash from explosive eruptions may drive climate change by increasing the reflectivity of the atmosphere, causing cooling. For example, increased explosive eruptions along the North Pacific rim may have helped rapidly propel that region into large-scale glaciation at 2.67 Ma (Pruher and Rea, 1998, 2001). An increase in explosive eruptions may also be a consequence of climate change, particularly for icecovered Icelandic volcanoes. Ice unloading from Icelandic volcanoes during deglaciations that has been shown to have influenced both eruption rates (Hall, 1982; Sigvaldason et al., 1992; Jull and McKenzie, 1996) and compositions of Icelandic magma (Maclennan et al., 2002). Lacasse and Garbe-Schönberg (2001) established that rhyolitic tephra in the northern North Atlantic (including Site 919) and Arctic oceans records the sources and timing of explosive volcanism in Iceland and the Jan Mayen area. Therefore, the record of rhyolitic tephra at Hole 919A may provide some insight on the relationship between Icelandic explosive volcanism and climate change over the last 630 kyr. The abundance pattern of colorless ash documents numerous explosive eruptions since 630 ka. The majority of the explosive eruptions coincide with light y 18 O excursions during both glacial and interglacial marine isotope stages (Fig. 8). Many of these colorless ash events (N3000 counts/gram) also coincide with IRD events (N20,000 counts/gram), but approximately 50% do not, eliminating the possibility that the concurrence between colorless ash events and light y 18 O excursions simply reflects the positive relationship between increased icerafting and y 18 O meltwater signals. Instead, the Hole 919A data supports the hypothesis that episodic deglaciations in Iceland and the associated glacioisostatic changes resulted in an increase of explosive volcanism in Iceland (Hall, 1982; Sigvaldason et al., 1992; Maclennan et al., 2002). Unlike previous datasets (Sejrup et al., 1989; Haflidason et al., 2000), however, the data presented here shows that this causal relationship between deglaciations and explosive eruptions was not restricted to interglacial marine isotope stages but also occurred within glacial stages. The difference in conclusions may reflect the difference in sampling strategy between this study and previous studies; examining dispersed ash input using point count data produced a more continuous record of the relationship between explosive volcanism and y 18 O than could be developed from a study restricted to discrete ash beds. 5. Conclusions The paleoceanographic evolution of the Irminger basin since 630 ka contains both long-term patterns and significant developmental steps, which are interpreted from Hole 919A sediment counts and isotopic data. The evolutionary development of oceanic and climatic conditions in the Irminger basin during this time span provides a history and therefore a context in which the millennial-scale cycles and events of the last glacial cycle can be considered. The primary long-term depositional patterns that persisted throughout all or most of the last 630 kyr in the Irminger basin include the following: (1) Persistent deposition of ice-rafted debris derived from red beds in central east Greenland and possibly Svalbard, with abundances more typical of sites in the Greenland Sea and more abundant than modern Irminger basin values,

150 K. St. John et al. / Marine Geology 212 (2004) 133 152 suggesting that the modern and late Pleistocene discharges of icebergs from northern red bed regions to the Irminger Sea lie in the low end of the range observed over the last 630 kyr. (2) IRD deposition between 630 and 190 ka that appears to have been largely controlled by the position of the Arctic front. (3) An extensive record of Icelandic volcanism based on dispersed ash, which links explosive eruptions to deglaciations (i.e., light y 18 O excursions) in both interglacial and glacial stages during the last 630 kyr. (4) A pattern of deglaciations between 630 and 190 ka marked by early meltwater input followed by later SST warming and bioproductivity increases in interglacial stages, similar to what was concluded for the last 60 kyr in Irminger basin (Elliot et al., 1998). Significant developmental steps in the paleoceanographic evolution of the Irminger basin during the past 630 kyr include the following: (1) Between 630 380 ka, increased supply of DSOW (relative to today) based on benthic assemblage changes. (2) Between 485 and 190 ka, extreme and rapidly fluctuating interglacial (and glacial stage 12) SSTs, related to a proximal but migrating Arctic front. (3) At ~380 ka, diminished supply of DSOW, possibly initiated by the influx of meltwater from broad-scale iceberg discharges along east Greenland. This negatively impacted benthic communities in the Irminger basin and may have contributed to the diminished supply of NADW. (4) First at ~338 309 ka and again at ~211 190 ka, a two-step cooling of sea surface conditions in the Irminger basin, after which both glacials and interglacials were colder as the Arctic front migrated to the southeast. (5) After ~225 ka, significantly reduced iceberg discharge to the Irminger basin from glaciers draining mafic crystalline outcrops and sedimentary outcrops in east central Greenland possibly due to semipermanent sea ice along the Scoresby Sund coast. Acknowledgments We wish to thank Terri King for assistance in generating the mcd scale and students Callie Rowe, Jennifer Milliken, and Trey Kendrick for assistance in sample processing. Jon Eiriksson and Bill Ruddiman provided insightful and thorough reviews for which we are most appreciative. References Bond, G., Lotti, R., 1995. Iceberg discharges into the North Atlantic on millennial time scales during the last glaciation. Science 267, 1005 1010. Bond, G., Showers, W., Cheeseby, M., Lotti, R., Almasi, P., demenocal, P., Priore, P., Cullen, H., Hajdas, I., Bonani, G., 1997. A pervasive millennial-scale cycle in North Atlantic Holocene and glacial climates. Science 278, 1257 1266. Broecker, W.S., Denton, G.H., 1989. The role of ocean-atmosphere reorganizations in glacial cycles. Geochimica et Cosmochimica Acta 53, 2465 2501. Elliot, M., Labeyrie, L., Bond, G., Cortijo, E., Turon, J.L., Tisnerat, N., Duplessy, J.C., 1998. Millennial-scale iceberg discharges in the Irminger basin during the last glacial period: relations with the Heinrich events and environmental setting. Paleoceanography 13, 433 446. Elliot, M., Labeyrie, L., Dokken, T., Manthe, S., 2001. Coherent patterns of ice-rafted debris deposited in the Nordic regions during the last glacial (10 60 ka). Earth and Planetary Science Letters 194, 151 163. Flower, B.P., 1998. Mid- to Late Quaternary stable isotopic stratigraphy and paleoceanography at Site 919 in the Irminger basin. In: Saunders, A.D., Larsen, H.C., Wise, S.W. (Eds,), Proceedings of the Ocean Drilling Program. Scientific Results vol. 152. pp. 243 248. Funder, S., Hjort, C., Landvik, J.Y., Nam, S., Reeh, N., Stein, R., 1998. History of a stable ice margin east Greenland during the middle and upper Pleistocene. Quaternary Science Reviews 17, 77 123. Geirsdottir, A., Hardardottir, J., Sveinbjornsdottir, A., 2000. Glacial extent and catastrophic meltwater events during deglaciation of southern Iceland. Quaternary Science Reviews 19, 1749 1761. Graf, G., 1992. Benthic pelagic coupling: a benthic view. Oceanography and Marine Biology Annual Reviews 30, 149 190. Haflidason, H., Eiriksson, J., Van Kreveld, S., 2000. The tephrachronology of Iceland and the North Atlantic region during the Middle and Late Quaternary: a review. Journal of Quaternary Science 15, 3 22. Hall, K., 1982. Rapid deglaciation as an initiator of volcanic activity a hypothesis. Earth Surface Processes and Landforms 7, 45 51. Heinrich, R., Freiwald, A., Wehrmann, A., Schaefer, P., Samtleben, C., Zankl, H., 1996. Nordic cold-water carbonates; occurrences and controls. In: Reitner, L., Neuweiler, F.,