Minimal Holocene retreat of large tidewater glaciers in Køge Bugt, southeast Greenland

Køge Bugt, in southeast Greenland, hosts three of the largest glaciers of the Greenland Ice Sheet; these have been major contributors to ice loss in the last two decades. Despite its importance, the Holocene history of this area has not been investigated. We present a 9100 year sediment core record of glaciological and oceanographic changes from analysis of foraminiferal assemblages, the abundance of ice-rafted debris, and sortable silt grain size data. Results show that ice-rafted debris accumulated constantly throughout the core; this demonstrates that glaciers in Køge Bugt remained in tidewater settings throughout the last 9100 years. This observation constrains maximum Holocene glacier retreat here to less than 6 km from present-day positions. Retreat was minimal despite oceanic and climatic conditions during the early-Holocene that were at least as warm as the present-day. The limited Holocene retreat of glaciers in Køge Bugt was controlled by the subglacial topography of the area; the steeply sloping bed allowed glaciers here to stabilise during retreat. These findings underscore the need to account for individual glacier geometry when predicting future behaviour. We anticipate that glaciers in Køge Bugt will remain in stable configurations in the near-future, despite the predicted continuation of atmospheric and oceanic warming.

. The oceanography of the Køge Bugt area. (a) 1:900,000 map showing the bathymetry of the continental shelf 13 and the locations of CTD 14,15 and seal dive 12 profiles. Symbols mark the locations of the CTD profiles (circles) and seal dive profiles (triangles). The location of core ER1116 is marked by a star. The inset map shows the location and extent of the main figure. Background imagery are Landsat 8 scenes 16 . (b) Temperature profiles from deep (>300 m) seal dives in Køge Bugt 12 . (c) Temperature profiles from CTD stations; these were produced with data from the World Oceanographic Database 2013 (WOD13) and the International Council for the Exploration of the Sea (ICES) 14,15 . This figure was created using ArcMap 10.1 and Adobe Illustrator CS6. Figure S2. Distinguishing between turbidites and layers with high ice-rafted debris content. (a) and (b) Line scan and X-ray imagery showing an example of a stacked turbidite sequence (core POR13-05, Upernavik, northwest Greenland). The turbidite is characterised by a sharp, undulating erosional basal surface, a general fining-upward trend, and stacked layers of sandy/gravelly sediment. (c) and (d) Line scan and X-ray imagery of a layer of high ice-rafted debris content (core POR13-16, Upernavik, northwest Greenland). This layer is characterised by diffuse upper-and lower-boundaries, numerous coarse sand and pebble-sized clasts, and a lack of internal layering structures. This figure was created using Adobe Illustrator CS6.
ER1116 was immediately sealed and stored vertically for transport to the core facility at The Geological Survey of Denmark and Greenland (Copenhagen, DK).

Examination of sediment for disturbance events
Marine sediments in high-latitude environments are frequently disturbed by iceberg ploughing and by mass-wasting events, such as debris flows and turbidity currents 18,19 . It is important to identify these events when establishing the stratigraphy of a sediment core. In this study the effect of iceberg ploughing was mitigated by coring from the centre of the deep trough in Køge Bugt, this is thought to be well-below the keel-depth of even the largest icebergs here. We used line scan and X-ray imagery to examine core ER1116 for the presence of turbidites and to distinguish these from layers rich in ice-rafted debris on the basis of morphological and sedimentological characteristics (Fig. S2). Core ER1116 shows no evidence of large turbidites, although we consider it possible that minor, millimetre-scale turbidites may be present.

3/13
Dating constraints and age model Dating sediments from glacier proximal environments is extremely challenging; harsh conditions restrict marine life and 14 C dating is not always possible. The scarcity of organic material is compounded by high sedimentation rates in these settings 20 . Sediments in ER1116 were age-constrained by five 14 C dates and five 210 Pb age determinations. Radiocarbon dates were obtained from benthic calcareous foraminifera, in the upper-half of the core these are sparse and it was necessary to combine planktonic and benthic foraminifera with bivalve shell fragments to obtain sufficient material for 14 C dating. The base of ER1116 (172-174 cm) was dated to 9120 ±150 cal. years BP ( Table 2). The remaining 14 C age determinations exhibit a simple age-depth relationship, with the exception of the date from 88-89 cm (Fig. S3). Sediment from this interval was very sparse in calcareous benthic foraminifera tests. It was necessary to supplement the benthic tests with planktonic foraminifera and shell fragments from an unidentified species of bivalve to provide enough organic material to measure the 12 C/ 14 C ratio. Detailed inspection of the supplementary bivalve shell during preparation for 14 C analysis revealed a dull lustre and 'chalky' texture; this suggests it may have been composed of replaced carbon (C. Patrick, personal communication, 2014). This interpretation is supported by the anomalously old 14 C age; this date is subsequently excluded from the age model (Fig. S3).
Results from lead-210 dating (Table 1) demonstrate that the core top is composed of modern material; this provides a tie-point for the age model here (Fig. S3). Intervals below this contained no unsupported (excess) 210 Pb ( Table 1). The lack of unsupported 210 Pb in these samples is consistent with the modelled age of these sediments (Fig. S3).

Age model
The age-depth relationship is established using simple linear interpolation between chronostratigraphic tie-points (Fig. S3). Sedimentation rates were highest in the early-Holocene (∼46 cm kyr −1 ) and dropped to ∼15 cm kyr −1 by 8000 years BP. A single sedimentation rate of ∼16 cm kyr −1 is calculated for the entire period from 6300 years BP to the present. It should be noted that this may not accurately reflect sedimentation rates over shorter timescales within this interval.
The age-model is relatively coarse; the absence of organic material in late-Holocene sediments means chronological uncertainties in this interval are large (Fig. S3). Nonetheless, the 14 C and 210 Pb results indicate that the sediments cover the last 9100 years of sedimentation in Køge Bugt. This is supported by linescan and X-ray imagery which shows no evidence for  23 , and a marine reservoir correction of 400 years, ∆R = 0 24,25 . Calibrated ages show the median of the 2σ range with 2σ uncertainties. The material from 88-89 cm was composed of benthic and planktonic foraminifera tests in addition to shell fragments from an unidentified species of bivalve.

Foraminifera assemblage analysis
Detailed results from foraminifera assemblage analysis are presented here (Figs. S4, S6, S5, and S7). Foraminiferal assemblage analysis was undertaken every 8 cm in ER1116. Additional subsamples were collected at 4 cm intervals in areas of the core where rapid shifts in species assemblage occur. Foraminifera tests were collected from 1 cm slices of core material, consequently, the species assemblage in an individual sample represents an average of conditions during the period of sediment accumulation. Subsamples were wet-sieved, the 100 µm to 1 mm size fraction was selected for analysis 26 . Foraminifera tests were concentrated by flotation using heavy liquid (CCl 4 , 1.66 g cm −3 ) and then picked from a graticuled tray using a stereo microscope. A total of 30 calcareous and 13 agglutinated benthic species were identified within ER1116. Three species of planktonic species were identified. Robust identification of individual species within a genus was problematic in some instances, especially where unique identifying features are extremely subtle (e.g. Buccella, Elphidium, and Islandiella species). Results are shown from calcareous species that individually accounted for more than 1% of the total assemblage from at least two sampling intervals 27 ; these species constitute >98% of calcareous benthic tests.
Foraminifera count, species type, and concentration data are presented in Fig. S4. The data show a dramatic decrease in foraminifera concentrations from ∼8000 to 4500 years BP (Fig. S4a). This is followed by a shift at ∼4500 years BP from assemblages dominated by calcareous foraminifera to those predominantly composed of agglutinated specimens (Fig. S4b). This shift is also accompanied by a drop in the concentration of planktonic foraminifera. The late-Holocene, from 4000 years BP onwards, is characterised by very low foraminifera concentrations, the dominance of agglutinated species, and the near-absence of planktonic species.

Foraminiferal assemblage zones
Three broad foraminiferal assemblage zones are identified from the data; these are based on the occurrence of key indicator species (Figs. S5, S6, and S4), the dominance of calcareous or agglutinated foraminifera (Fig. S4b), and the concentration of 5/13 Figure S4. Foraminifera counts, relative percentages, and concentrations. (a) Foraminifera test counts, benthic foraminifera (magenta), and planktonic foraminifera (blue). (b) Relative percentage of calcareous (blue) and agglutinated (magenta) foraminifera. (c) Concentration of foraminifera in core ER1116 (per gram of wet sediment). Concentration data are not available from the base of the core (9120 ±150 cal. years BP). This figure was created using Microsoft Excel 2013 and Adobe Illustrator CS6.

6/13
benthic foraminifera (Fig. S4c). The boundaries between zones are arbitrarily drawn at the mid-point between the sampling intervals where a change is observed; the shift in assemblage may occur anywhere within this interval. The palaeo-environmental significance of each foraminifera assemblage zone is discussed below.

Assemblage zone I: 9100 to 4800 years BP
Assemblage zone I was dominated by calcareous species; agglutinated foraminifera were virtually absent from this interval. The total concentration (n g −1 ) of foraminifera tests was high at the start of this interval and drops progressively (from ∼300 to ∼50 g −1 ). A substantial dip in foraminifera concentration occurred at ∼8500 years BP (Fig. S4c), but this was not accompanied by an attendant change in the species assemblage; this remained similar to the rest of zone I (Figs. S5 and S6).
The foraminiferal assemblage during this interval was dominated by Cibicides lobatulus, Elphidium spp., and Cassidulina neoteretis. Islandiella spp., Melonis barleeanus, and Pullenia bulloides were notable accessory species (Fig. S5). C. lobatulus is widely associated with coarse sediments and high-energy environments [28][29][30][31][32][33] . Less is known about the environmental preferences of Elphidium species. Elphidium excavatum f. clavata forms a large proportion of this genus grouping and is generally associated with cooler water masses and glacier proximal conditions 28,30,34,35 . Similarly, C. reniforme is considered an Arctic species, and is related to cooler water masses near glacier termini [28][29][30]35 . C. neoteretis has a well-established association with Atlantic water masses and is often viewed as diagnostic of warm, saline oceanographic conditions 29,30,36,37 . This is consistent with the presence of P. bulloides which is also linked with warm Atlantic waters 25 . M. barleeanus is a detrital feeder that dwells within the sediment matrix (infaunal); it is commonly associated with an abundance of buried organic matter 30,38 . The Islandiella genus grouping is composed of Islandiella helenae and Islandiella norcrossi; both are typical of cool, Arctic shelf conditions and may also be associated with enhanced productivity linked to seasonal sea-ice 25,29,30 .
The foraminiferal assemblage data suggest that oceanographic conditions in Køge Bugt from 9100 to 4800 years BP were characterised by elevated hydrodynamic energy levels, the incursion of warm, saline Atlantic water, and possibly also intervals of cooler and fresh Arctic water masses. The foraminifera assemblage data also indicate that Køge Bugt was a glacier proximal environment during this interval and was subject to seasonal sea-ice cover. Finally, the presence of M. barleeanus is consistent with a period of enhanced sedimentation in Køge Bugt.

Assemblage zone II: 4800 to 1500 years BP
Assemblage zone II was characterised by a significant reduction in the concentration of benthic foraminifera and an increased, but variable, abundance of agglutinated species (Figs. S4 and S6).
The foraminiferal assemblage is primarily composed of Elphidium spp., Buccella spp., M. barleeanus, and Islandiella spp. (Fig. S5). The Buccella genus grouping is composed of Buccella tenerrima and Buccella hannai arctica; these species are associated with sea-ice and enhanced productivity 25 . The accessory species Nonionella labradorica and Trifarina fluens are also connected to high productivity environments and are often related to the occurrence of seasonal sea-ice or oceanic frontal zones 25,31,37 . A reduction in C. neoteretis and C. lobatulus abundance through this interval suggests the establishment of cooler, less-energetic hydrographic conditions. Agglutinated species are abundant; this is consistent with a period of cool, Polar oceanic conditions 29,39 .
The foraminiferal assemblage data suggest that the oceanography of Køge Bugt during this interval was characterised by cool and fresh Polar water masses, frequent sea-ice cover, periods of relatively rapid sedimentation, and a reduction in hydrodynamic energy levels at the seabed. The data also indicate an oceanographic cooling trend through zone II, with the establishment of cold Arctic conditions by the end of this interval.

Assemblage zone III: 1500 years BP to the present
Assemblage zone III was characterised by the dominance of agglutinated foraminifera (>70%, Fig. S4b) and very low total foraminifera concentrations (<20 g −1 , Fig. S4c). This assemblage zone was characterised by the presence of Ammoglobigerina globigeriniformis, Cribrostomoides spp., Silicosigmoilina groenlandica, and a range of other agglutinated species (Fig. S6). There is relatively little detailed information about the environmental preferences of agglutinated Arctic foraminifera. However, agglutinated foraminiferal assemblages are known to occur in areas of the modern Arctic where oceanographic conditions are dominated by cold, low-salinity, Polar water masses 29,39 . Textularia earlandii and Spiroplectammina biformis are associated with Polar water masses and glacier proximal conditions 29,39 . Adercotyma glomerata may be associated with cooled Atlantic water 28,33 . Although very limited, the assemblage of calcareous species is consistent with cold, low-salinity oceanographic conditions (Fig. S5). Foraminifera data generally indicate that the late-Holocene in Køge Bugt was characterised by cold, low-salinity oceanographic conditions.

Sortable silt analysis
Data from laser diffraction analysis are shown in Fig. S8. The raw data (0.3-63 µm) from all intervals exhibited bimodal distributions (Fig. S8 inset). The secondary peak in particle sizes smaller than 10 µm is a common feature in laser diffraction  data; this is an artefact caused by the deflocculation of clay mineral aggregates during sample pre-treatment 40,41 . Excluding sediment fractions smaller than 10 µm mitigates this; we focus on the sortable silt fraction (10-63 µm 42 ), these data have normal distributions. Fig. S8 shows both the mean grain size and the standard deviation of the sortable silt fraction. The standard deviation (σ ) provides a measure of sorting; small σ values are indicative of a well-sorted sediment. The sortable silt σ can also be used to assess the robustness of laser grain size data as a proxy for palaeocurrent strength. The sortable silt mean grain size and the degree of sorting are weakly correlated (R 2 =0.239). However, the longer-term pattern illustrates that periods of high sortable silt mean values (i.e. stronger current) are generally also characterised by a high degree of sorting (Fig. S8). This is consistent with finer material being winnowed out of the sediment and suggests that sortable silt mean data are a viable proxy for benthic palaeocurrent vigour in this setting. The only interval where this relationship fails is from 9000 to 8000 years BP; during this period large sortable silt grain sizes are accompanied by high σ values (Fig. S8). The cause of this remains unclear. We speculate that it may result from the large influx of ice-rafted debris during this interval (Fig. 4c); this may have overwhelmed the current-sorting signal.