GEOCHEMISTRY GEOPHYSICS GEOSYSTEMS, VOL. 4, NO. 4, 1038, doi:10.1029/2002GC000441, 2003
Sedimentary record of disintegrating ice shelves in a warming climate, Antarctic Peninsula
Department of Geography, Queen's University,
Kingston, Ontario, Canada
Department of Geology, Hamilton College,
Clinton, New York, USA
[1] Seafloor sediments from beneath the former Larsen-A and Prince Gustav ice shelves document the recent breakup of the shelves and provide evidence for interpretation of previous events. At three of five sites, sediment texture coarsens upward with up to 40% sand at the surface. Radiometric 210Pb dating shows this to have occurred between 1985 and 1993, several years before breakup of the shelf, and that rates of accumulation of sediment on the seafloor doubled to quadrupled during this period. These events are related to the release of eolian sediment in periodic rapid draining of small lakes and crevasses on the ice shelf before breakup. X-ray radiographs of sediment cores also document the recent influx of coarse particles (gravel) related to ice rafting during ice shelf disintegration. Because sediment is released irregularly in time and space from well-separated point sources on the ice shelf during at least several years before final disintegration, only a portion of the seafloor is affected.
Components: 5959 words, 8 figures, 1 table.
Keywords: Larsen Ice Shelf; Antarctica; glacial marine sedimentation; lead dating; meltwater.
Index Terms: 3022 Marine Geology and Geophysics: Marine sediments—processes and transport; 1827 Hydrology: Glaciology (1863); 4558 Oceanography: Physical: Sediment transport; 4863 Oceanography: Biological and Chemical: Sedimentation; 9310 Information Related to Geographic Region: Antarctica.
Received: 11 September 2002;
Revised: 27 January 2003;
Accepted: 17 February 2003;
Published: 19 April 2003.
Citation: , Sedimentary record of disintegrating ice shelves in a warming climate, Antarctic Peninsula, Geochem. Geophys. Geosyst., 4(4), 1038, doi:10.1029/2002GC000441, 2003.
Copyright 2003 by the American Geophysical Union.
[2] The loss of over 9250 km2 of floating glacial ice of the Larsen and Prince Gustav ice shelves (as defined by Vaughn and Doake [1996]) along the eastern Antarctic Peninsula during the last decade is one of the most dramatic environmental changes observed anywhere on Earth [Rott et al., 1998; Skvarca et al., 1999; Domack et al., 2001b]. The demise of these ice shelves is related to decades of warming in the Antarctic Peninsula region [Vaughan et al., 2001] which has shifted the shelves into zones of ablation, exacerbated by meltwater ponding and hydrostatic propagation of crevasses on their surfaces [Scambos et al., 2001]. Yet it is probable that ice shelves along the Antarctic Peninsula have vanished and formed again during the Holocene [Domack et al., 2001b]. Recent studies have confirmed that even the Prince Gustav Ice Shelf is a feature limited to the last ~1.9 ka and probably the time prior to 6–7 ka [Pudsey and Evans, 2001]. Clearly, ice shelves are sensitive climatic indicators and the nature of their most recent demise needs to be scrutinized in light of the paleorecord of past fluctuations.
[3] In this paper we describe the textural characteristics of sediments from grab samples and cores dated by 210Pb and 14C analyses, taken from the seafloor beneath these ice shelves, and we infer the sedimentary signal of the breakup of the ice shelves.
[4] Access to the seafloor beneath the former Prince Gustav and Larsen-A ice shelves (Figure 1) during a winter cruise in May 2000 aboard U.S. Antarctic research vessel N.B. Palmer provided an opportunity to recover sediments for detailed analysis of deposition during the breakup of the ice shelves. The triple-tube multicorer and Bowers and Connelly multicorer (7.6 cm diameter) provided nearly undisturbed samples of the upper meter of the seafloor. Cores 5, 6, 7, 28 and 34 (Figure 1) were logged and sampled for grain size and dating by means of 210Pb activity. Subsamples of sediment finer than 2 mm were taken at 0.5-cm intervals from the surface to 10 cm depth in anticipation of defining the unsupported 210Pb activity, and at progressively greater intervals to 25 cm depth to determine the supported 210Pb activity [Appleby, 2001]. Samples submitted for 210Pb analysis were not separated by grain size. Grain-size analysis on other samples from these cores was performed with a Coulter LS-200® laser diffraction instrument after standard treatment to remove organic matter and disperse flocs [Last, 2001]. Particulate matter filtered from snow samples were measured using a Malvern Master Sizer E® laser diffraction instrument that allows very small samples to be analyzed.
|
|
[5] Twenty six sub cores were taken from Smith McIntyre sediment grabs by pushing a 6-cm-diameter tube vertically into the sediment immediately after the sampler came on deck. The Smith McIntyre preserves a 0.5 × 0.5 m intact sediment-water interface and undisturbed sediment to a depth of 15 cm, depending on the texture of the material. The cores were split and both halves were X-rayed using Industrex type M X-ray film. Individual grains greater than 2 mm diameter (gravel size) were counted in 2.5-cm increments following the method of Grobe [1987]. A similar procedure was followed using the replicate cores described above, thus providing a check on the grab data, and extending the results to greater depth. To facilitate comparison among our samples and with those collected by others, the results are presented below as counts per deciliter of sediment.
[6] Radiocarbon dates were obtained on selected surface materials in order to assess the potential of this method in establishing a chronology for the sedimentary records. We focused on encrusting bryozoans of two distinct relative ages recovered from a cobble dropstone on the sediment surface at station 4, and from surface organic matter at station 5, both in the Greenpeace Trough (Figure 1). The cobble had two generations of byozoan encrustation as determined by cross-cutting relationships of living and dead colonies. Calcium carbonate was removed from the stones with dental picks, ground with a mortar and pestle, cleaned in an ultrasonic bath for 10 min., treated with a dilute solution of H2O2 to remove surface organic material, and finally ultrasonically cleaned again. Acid insoluble organic matter was treated following standard procedures [cf. Domack et al., 1999]. Samples were sent to the University of Arizona Tandem Accelerator Mass Spectrometer Laboratory for 14C analysis. Ages are reported as uncorrected except for adjustment due to 
13C fractionation. No calibrations are reported, although we note a reservoir correction for the NW Weddell Sea of 1200–1300 yr [Björck et al., 1991].
[7] Textural and 210Pb analyses for the three cores from Greenpeace Trough are presented in Figure 2. In core 5 from the central region of the trough closest to the Antarctic Peninsula, there are distinct changes in texture at 12 cm depth and in the upper 3 cm. Below 12 cm the particle size distribution (PSD) is unimodal with mean grain size in fine silt between 5 and 7 µm. Above 12 cm a second mode appears in coarse silt which becomes progressively more dominant upwards. Mean grain size increases correspondingly to more than 10 µm. From 3 cm depth to the surface the fine mode disappears and the PSD shifts to coarse silt with up to 30% sand by volume; mean grain size rises to 24 µm. At the surface of the core the sand content and mean grain size decrease slightly from the maximum values at 0.5 cm depth.
|
|
[8] Cores 6 and 7 from the north-eastern part of the trough do not show the same pattern with respect to depth. PSD is more consistent throughout the upper 25 cm of the core; it is nearly unimodal in the fine silt range (mean 4 to 7 µm) with a much smaller second mode in the coarse silt range. Sand content is less than 2%. In these respects the PSD resembles the lower part of core 5.
[9] The 210Pb analyses (Figure 2c) show the characteristic exponential decay with depth except for lower values in the upper centimeter of core 5. Possible reasons for the abnormal pattern include: (1) the sediment may be bioturbated at the surface, mixing older sediment having lower concentration of 210Pb to the surface, or (2) the coarse grains have lower concentration of 210Pb than the finer-grained sediment and so dilute the activity levels. Bottom photographs reveal a benthic community dominated by epifaunal grazers and a depauperate infaunal community, indicating minimal mixing of sediment, especially in the deep Greenpeace Trough. We suggest that the second is more probable given the values for the three cores, the close correspondence between activity levels and sand content in the upper three samples, and the fact that the profiles show no mixing at greater depth. As well, X radiographs show little or now bioturbation of laminae in the upper portions of cores from these sites. Thus, our calculations of age based on a constant rate model ignore the irregularity in core 5. In each case, background activity was measured in a sufficient number of the lower samples to construct the dates with some confidence. We interpret that the dramatic change in PSD in the upper 1.5 cm of core 5 occurred since about 1989. The more subtle change about 12 cm depth is not precisely dated but appears to be about 300 years old. In this respect, core 5 is unique but it is possible that this shift may relate to growth of the Larsen-A Ice Shelf during the past several thousand years.
[10] The rates of sediment accumulation calculated from the dates shown in Figures 2c and 3c have a similar pattern with respect to time (depth). Core 6 has the lowest rates (<0.3 mm a-1 and <250 g m-2
|
|
[11] Cores from Prince Gustav Channel (Figure 3) reveal the same pattern. In core 34 there is very little change with depth in the core, except a slight amount more sand (up to 1%) in the upper 2 cm of the core. The PSD is nearly unimodal, with mean grain size of 6 to 7 µm and a weak second mode in the coarse silt and fine sand range. In 28 the pattern below 5 cm depth is similar, except that there is a coarser layer at 13 cm depth, and throughout the second mode is stronger, giving rise to sand contents up to 5%, and a correspondingly greater mean grain size of about 9 µm. Above 5 cm the PSD shows the same pattern as in core 5 with a coarser layer at 4 cm containing 35% sand, and from 2 cm to the surface which is inversely graded with 25% sand increasing upward to 37% at the surface.
[12] The pattern of 210Pb activity in these two cores shows similar irregularities near the surface as in core 5, and we ascribe the cause more likely to the coarser sediment content, although that argument is less convincing for core 34. For core 28 the background level may not have been reached in the samples available for our analysis, although the activity level in the lowest sample is similar to those in core 34 and the error in assigning dates is small. In core 28 the coarser sediment near the surface was deposited in the late 1970s (layer at 4 cm depth), and between 1993 and the present (upper 2 cm) with the lowest 210Pb activity level at the end of 1994. The reason for the sandy sediment at 13 cm (about 1930) is not known, although acoustic Doppler current profiles indicate significant lateral advection from under the present remaining ice shelf in the vicinity of the Seal Nunataks [Domack et al., 2001b]. A similar but weaker pattern appears in core 34 with a reduction in 210Pb activity after 1977 and especially during 1993.
[13] The rates of sediment accumulation in these cores at depth (0.4 to 1 mm a-1 and 300 to 500 g m-2 a-1) are similar to those from Greenpeace Trough. However, both have very much greater rates near the surface, with the greatest (2.2 to 2.5 mm a-1 and 1.4 to 1.5 kg m-2 a-1) occurring at 2 cm depth which dates as mid 1993 in both cores. Total accumulation of sediment associated with the Larsen Ice Shelf or open water in the Greenpeace Trough has been about 1.5 m since the early Holocene [Domack et al., 2001b], representing a mean accumulation of less than 1.5 mm/a during this period. This is somewhat greater than the early twentieth century rates but less than the rates associated with breakup at several sites as determined by lead dating (Figures 2 and 3).
[14] Of the 16 subcores collected in the Larsen-A region (Greenpeace Trough and Seal Nunataks: Figure 4), ten show the maximum of gravel abundance in the uppermost 2.5 cm [see also Evans and Pudsey, 2002]. Of the remaining six cores, four show a decrease, two have subsurface maxima, and one has only limited recovery. Two of the five cores from the Seal Nunataks region have less gravel toward the surface. This is a region that differs from the main axis of the Greenpeace Trough in the strength of bottom currents and the abundance of well-sorted sand.
|
|
[15] The nine subcores from the Prince Gustav Channel show similar results in that seven demonstrate a maximum in gravel abundance at the surface. Hence, out of the total of 26 surface cores examined, 65% show gravel maxima at the surface. This indicates that ice rafting is not a time-random process at most of the sub-ice-shelf sites. Rather, it is apparent that the sites beneath both the Larsen-A and Prince Gustav Channel experienced a recent peak in ice rafting. The evidence from dating presented above indicates the upper 2.5 cm of each grab sample accumulated in from a few years (at accumulation rates up to 4 mm a-1: Figure 3) to a few decades.
[16] Replicate cores from the lead dating sites (Figure 5) provide evidence of longer-term deposition of gravel at these sites. Core 5 shows a greater accumulation near the surface corresponding to the higher rates of accumulation documented by lead dating, while cores 7 and especially 6 have lower concentrations at their surfaces relating to the lower accumulation rates from lead dating. In the Prince Gustav Channel cores, the rates are higher at the surface corresponding with the greater accumulation determined by lead dating. However, in each case the accumulation of gravel has occurred throughout the record, especially at depths of about 30 to 60 cm which, given the rates of accumulation reported above, may correspond with the period before the development of an ice shelf during the past several thousand years. In any event, recent deposition of gravel associated with the disintegration of the Larsen-A Ice Shelf is not more dramatic than through a longer record as evidenced by these cores.
|
|
[17] The 14C dated bryzoans on the dropstone cobble (Table 1) at site 4 (Figure 1) and of dispersed organic material at site 5 also provide evidence of the complexity of the sedimentary environment beneath the Larsen-A Ice Shelf. Surface organic matter at site 5 is significantly older (9720 BP) than other surface ages on organic matter from the Antarctic continental shelf which are themselves much older than organic matter from other shelf regions of the world [Andrews et al., 1999; Domack et al., 1999; Pudsey and Evans, 2001]. This indicates that radiocarbon dating of sediments in the region of the Larsen Ice Shelf using bulk organic matter [Pudsey et al., 2001] will be extremely difficult to interpret without independent chronologic control.
[18] The pair of calcite dates are, however, more useful. The age of 1014 ± 30 BP is equal to, or younger than, the reservoir age [Björck et al., 1991] and indicates present growth in open water. In contrast the age of 1750 ± 30 BP, some 700 years older, suggests a period of epifaunal activity significantly older than the modern period of growth. We hypothesize that the old growth marks the penultimate period of open water within Greenpeace Trough prior to the growth and expansion of the Larsen Ice Shelf (which must be at least several hundred years old [Scambos et al., 2001]). The age of 700 yr BP is coincident with the onset of significant cooling determined from regional paleoclimatic data [Domack et al., 2001a] and just older than the growth an ice shelf in Lallemand Fjord on the western side of the Peninsula [Shevenell et al., 1996]. The test of this hypothesis awaits additional radiocarbon analyses and the application of alternate dating methods such as paleomagnetism and optically stimulated luminescence now underway by our colleagues.
[19] Site 4 is different from those reported above in the Greenpeace Trough where the evidence based on 210Pb analysis is that deposition of mud has occurred steadily for more than a century at least, and probably much longer, at rates of 0.25–1.6 mm/a. In the core from site 4, the upper unit 1A is missing; it is this unit in other cores that was proposed by Domack et al. [2001b] to represent the formation and presence of the Larsen Ice Shelf associated with cooling during the past several thousand years. Site 4 at 668 m depth is on an exposed ridge at the edge of the Greenpeace Trough. We suggest that currents dominantly from the south (beneath the Larsen-B Ice Shelf) slowed or prevented deposition of fine-grained sediment at this location.
[20] Both the gravel content and the texture of finer sediment provide a signal of the disintegration of the ice shelf in the Larsen-A and Prince Gustav Channel regions. Melt of Larsen-A based on the irradiance characteristics of the ice surface [Scambos et al., 2001] occurred in periods from 20 to 60 d/a from 1980 to 2000 except for (1) about 100 days of melt in 1993 when 234 km2 of the shelf in Prince Gustav Channel disintegrated, (2) 85 days in 1995 when most of the rest of the ice over Prince Gustav Channel (324 km2) and the Greenpeace Trough (1707 km2) disappeared, and (3) 85 days in 1998 by which time most of the ice north of Larsen-B had already vanished (Figure 1). The 210Pb dates (Figures 2 and 3) indicate that where accelerated deposition occurred, it began after about 1990 and in several cases reached its peak in 1993 and 1994, associated with the first of the exceptional warm years.
[21] The increase in sediment size and rates of accumulation, as well as the spatial irregularity of the signal, can be explained by the processes of disintegration of the ice shelf. A LANDSAT 7 image (Figure 6) recorded on 1 February 2000 shows a small portion of the Larsen-B Ice Shelf about 20 km south of the region shown in Figure 1. On the ice are many partially ice-covered to open-water ponds, each about 1 km2 in area, and each having a drainage basin on the ice of several square kilometers. The location of these ponds is at least partially determined by a series of crevasses parallel to the outer edge of the ice and spaced about 0.8 km intervals across the ice. As the ice melts, the ponds and crevasses collect water which facilitates more melting and enlargement due to the lower albedo of the water compared to the surrounding ice.
|
|
[22] We suggest that the sediment associated with breakup is from two sources: gravel represents material of glacial origin, transported within the ice shelf, while sand represents both glacial and eolian sediment, the latter originating from the subaerial weathering of nunataks and other exposed lands in the nearby region of the Antarctic Peninsula.
[23] Sediment in snow samples collected in December 2001 from six shallow pits on Robertson Island (Figure 1) averaged 16 mg of clastic sediment per liter of snow. Dusting of significant parts of the Larsen-B Ice Shelf was visible from a distance, suggesting that concentrations exceed this value in places. The grain size distributions were strongly bimodal (Figure 7) with an average of 45% sand; this indicates that transport both by suspension and traction occurred [Barrett et al., 1983]. The source of the sediment is especially the poorly indurated and heavily frost-weathered Cretaceous sedimentary bedrock exposed on the islands and a portion of the Antarctic Peninsula [Thomson and Harris, 1981].
|
|
[24] If our measurements of sediment concentration represent conditions on the ice shelves prior to disintegration, we estimate that a pond may sequester up to 0.2 tonnes of eolian fine sand derived from its drainage basin during the several years of its existence. At the height of the melt in February, streams were observed in several locations on the surface of the shelf (P. Skvarca, personal communication, Feb. 2002). These scavenge a significant amount of the eolian sediment from the snowpack and deliver it to the ponds. Sometime before the shelf disintegrates, some of the ponds and crevasses melt through the shelf and create a conduit for their water, allowing at least some of the sediment collected to pass into the sea below. Water depths as small as 15 m may be sufficient to drive the process [Scambos et al., 2001]. There is evidence in Figure 6 that this has happened; some of the ponds, especially in the southwest of the image and close to the outer edge, appear partially or completely drained, and abandoned shorelines are apparent around these former water bodies. Melting through of each pond almost certainly occurs at a different time, probably over a period of at least several summer seasons. Thus, if a plume of sediment is released from a single source at the bottom of the pond, one might expect the sediment to be deposited at various times during at least several years before final disintegration of the shelf.
[25] This is the pattern we see in the core records, with the earliest evidence of sand in core 28 about 1978, and in core 5 about 1992. It is interesting that the greatest rates of accumulation occur at the tops of the Prince Gustav Channel cores (Figure 3d), even though the rates lower in the cores are closer to those in the Larsen-A area (Figure 2d). We suggest that at the former site, the closer sources of unglaciated land on both James Ross Island (which has large ice-free areas of poorly consolidated Cretaceous sandstone) and the Antarctic Peninsula contributed more eolian sediment to the shelf surface and thus to the draining process described above. This may also be true of core 5 in which the rates are intermediate between background and those of Prince Gustav Channel; this core is the closest to the Nördenskjöld Coast and the Seal Nunataks of those studied in the region of the Greenpeace Trough.
[26] According to modeling results [Scambos et al., 2001], ice in the Larsen-B Ice Shelf advanced seaward from the grounding line about 500 m/a during the last several hundred years. If Larsen-A and the Prince Gustav ice shelves behaved similarly, during several years before disintegration, as ponds and crevasses contributed their water and sediment at different times, they did so at different places as the ice advanced. As well, current patterns below the ice may vary, so that the sediment is carried to different places on the seafloor at least several hundred meters below the bottom of the ice.
[27] Nomographs based on calculations of the dispersal of sediment falling through quiet water from a point source [Gilbert, 1990] show that a small volume of sediment (10 to 100 L) released instantaneously from a point source spreads in a cone-shaped pattern to a maximum radius of less than 10 m for medium and coarse sand, and to less than 100 m for fine sand. Water depths of less than 50 m (medium and coarse sand) or 500 m (fine sand) are required to achieve these radii. Water depths at all the study sites exceed these values even considering several hundred meters of shelf ice thickness. We suggest that the releases occur predominantly as relatively instantaneous events; melt progresses until the water in the pond or crevasse breaks through under an hydraulic head due to the small elevation above sea level of the water surface. This process quickly carries at least some of the sediment accumulated in the pond to the sea below.
[28] Given the spacing of the ponds and crevasses is of the order of 1 km (Figure 6), it is apparent that the clouds of sediment released by draining of the ponds cover only a portion of the seafloor, and it is possible that, due to the movement of the shelf, releases from two or more ponds could occur sequentially at or near the same place. Even if currents disperse the sediment, the footprint of the resulting deposit is not significantly greater than in quiet water, although it may occur in a different place. Thus, sites such as 6 and 7 do not record the release of sediment by this mechanism.
[29] Although some sites also experienced an increased flux of fine-grained sediment as documented above, the sudden increase in gravel implies that the breakup event was also associated with the release of glacial debris from within the ice itself. Photographs of the breakup of the Larsen-B ice shelf in 2002 revealed extensive bands of englacial material that were not visible across the surface of the ice shelf. Upon disintegration, the medial moraine and debris septums produced down-flow from the tributary junctions became exposed as the ice shelf fractured into thousands of icebergs (P. Skvarca, personal communication, 2002). It is this late stage process that we believe leads to the increase in coarse sediment on the seafloor.
[30] There is a recognizable signal of the breakup of the Larsen and Prince Gustav ice shelves in the sediments that were deposited beneath them. The processes are summarized in the cartoon presented as Figure 8. During the period when the ice shelf is stable or advancing (Figure 8a) both fine sediment and large clasts of glacial origin are released intermittently in small quantities by melting from the base of the ice shelf while eolian sand and dust accumulates in snow on the surface. During several years before final disintegration (Figure 8b) the release of basal debris increases as the ice shelf thins by accelerated melting, while continued eolian accumulation and stored eolian sediment is washed to melt ponds and crevasses due to increased melt on the surface of the shelf in warmer summer temperature. The catastrophic draining of melt ponds [Scambos et al., 2001] releases the sediment accumulated in them as clouds spreading the fine-grained material over the seafloor. Oceanic circulation beneath the ice shelf displaces laterally the more slowly settling sediment without greatly affecting the size of the footprint of the final deposit. The final, extremely rapid disintegration of the ice shelf (Figure 8c) releases large quantities of both surface and basal material as the shelf breaks into small fragments that quickly disperse seaward. The englacial medial moraines which are evident only during breakup as the ice at depth is exposed by the process of disintegration, are major sources of sediment at this time.
|
|
[31] However, because of the way the sediment was released during draining of ponds and crevasses on the ice shelves, and because of the uneven distribution of basal and englacial debris, the distribution of this record is spatially heterogeneous over the seafloor, with some sites receiving a large influx of both coarse and fine-grained sediments from basal melting and release of englacial debris, others receiving only the fine-grained material released from the ponds, and others receiving little or no sediment associated with breakup, and therefore not recording a signal of breakup. Nevertheless, along with biological and chemical evidence, the textural evidence and increased rates of accumulation provide an important signal that may be applied to the longer sedimentary record to assess the history of formation and decay of ice shelves.
[32] Grants from the U.S. National Science Foundation, Office of Polar Programs (OPP9814383), and the Natural Sciences and Engineering Research Council of Canada funded the work. We thank Captain J. Borkowski III, officers, crew, contractors and fellow scientists on NBP0103 for their extraordinary skill and effort during a challenging winter cruise. 210Pb analyses were performed by Jack Cornett (Mycore Scientific Co.). C. Pudsey and two anonymous reviewers provided thoughtful reviews of the paper.
) classes from 0.34 µm (11.5
