1. Glaciological Factors

Thickness gradients (dH/dx) and the pattern of flow lines displayed by the present-day Ross Ice Shelf have been used in the 18 ka reconstruction (Fig. 1). Our model argues for maintenance of an ice-shelf regime over the major part of the Ross Sea with forward migration of the ice-shelf grounding line by 300–350 km and commensurate movement of the ice front 150–450 km in order to sustain volume continuity with creep thinning. The mode of ice discharge out of Marie Byrd Land would have been via an arrangement of ice streams. Some variations in their location may well have occurred according to grounding-line position, bottom topography, catchment area, and mass-balance conditions. Since, however, the seabed troughs occupied by the ice streams are structurally controlled (Rose, in press; Robertson and others, in press) it is unlikely that their positions would have been drastically modified.

Substantial grounding would have characterized the coastal zone adjacent to Victoria Land assisted here by off-shore islands and shallow submarine banks (Pennell, Crary, and Mawson Banks, Franklin Shoal, Ross, Beaufort, and Franklin Islands) and thick ice contributed from Byrd Glacier. We consider that the enlarged grounded areas especially in western Marie Byrd Land would have been dominated by active, low-gradient ice streams extending several hundred kilometres inland of the grounding line and separated by slow-moving ridges and domes. The pattern would correspond closely with that observed in Marie Byrd Land today Reference Rose(Rose, 1979). Inland of the ice streams, surface slopes may well have been similar to those of other grounded ice sheets of comparable depth. This pattern implies only nominal thickening of the ice sheet and does not seek the substantial change to the overall surface profile demanded by the Maine reconstruction (Fig. 2).

Fig. 2.

Inferred profile through expanded ice sheet-ice shelf in the Ross Sea along ice stream C. Comparison with current profile and that proposed by Hughes and others ( 1979), Surface profile and ice thickness taken principally from radio echo-sounding with trianglesfrom seismic shooting, Sea bed in Ross Sea shows maximum and minimum elevations in a 50 km band of the flow line. Location of DSDP, RISP, and Byrd Station cores are also shown. Current Byrd–ice stream D surface profile shown with 300 m thickening at Byrd Station. Inset shows location of ice-stream flow lines and cores, present-day suiface contours in 10² m.

This conclusion is supported by several glaciological investigations in West Antarctica. Along the Byrd Station strain network (BSSN), Reference WhillansWhillans (1976) has deduced that the West Antarctic ice sheet has maintained near steady-state conditions (in terms of ice-sheet shape, mass balance, and velocity field) during a minimum of 20 ka. He considers that his conclusions on stability may relate to an area well beyond the BSSN (and even to most of Marie Byrd Land) and is consistent with our postulated thickening.

Tentative interpretation of isotopic and total-gas content of core from Byrd Station Reference Robin(Robin, 1977) suggests a net climatic warming of 5–8 deg since the coldest period of the Wisconsin, and an elevation decrease of the ice sheet by 600 m of which about 300 m could be due to down-slope flow. The remaining 300 m may represent an upper limit to the real decrease in ice-sheet thickness, but again compatible with the magnitude of the retreat of the grounding line and the subsequent thinning indicated by our model. This result is clearly divergent from the 1 000 m surface lowering and 1 760 m thinning required by the Maine analysis at Byrd Station (Figs 1 and 2).

2. Ross Sea Bathymetry

Sea-bed topography in the Ross Sea embayment provides an important base-line against which sea-level changes may trigger ice-sheet-ice-shelf advances and groundings.

Improved knowledge of the configuration of the Ross Sea continental shelf has been embodied into two charts by Reference Hayes and DaveyHayes and Davey (1975) and Reference Vanney and JohnsonVanney and Johnson (1976). Beneath the Ross Ice Shelf seismic reflection shooting has enabled reconnaissance mapping of the sea bed (Robertson and others, in press). Inland of the Ross Ice Shelf grounding line radio echo-sounding has provided the first detailed study of the ice-rock interface in Marie Byrd Land (Rose, in press).

The important bathymetric elements within the embayment are linear depressions and sub-parallel submarine banks (Fig. 3). The former are currently occupied by ice streams in the grounded (Marie Byrd Land) area Reference Rose(Rose, 1979) but extend out across the shelf at a depth of 500–600 m below sea-level (Robertson and others, in press), rising slightly towards the continental edge. It has been demonstrated that these troughs are structurally controlled basement features (Rose, in press; Robertson and others, in press). A number of shallow banks are present separating depressions. In the western Ross Sea these are clearly linear features parallel to the troughs, but to the east there is some disagreement between the chart of Hayes and Davey and that of Vanney and Johnson (allowing for differences in their contour intervals). In the former case the alternating pattern of shoals and swales is maintained, while the latter depicts a zone of shoals striking across the grain of the suspected troughs.

Fig. 3.

Location of Eltanin piston and other rotary cores in the Ross Sea ]and their relation to bathymetry (in metres below sea-level ). Units I and II ref er to material exposed on the sea floor; their ages are given in Table I and discussed in the text.

Important in evaluating bathymetric and sea-level interactions is the isostatic state of the Ross Sea shelf. Early gravity measurements over the Ross Ice Shelf suggested that a —12 mgal average free-air anomaly was related to incomplete isostatic adjustment from former ice loading (Reference BennettBennett, 1964). Additional data are now available on the Ross Ice Shelf (Bentley and others, in press) and for the Ross Sea (Reference Hayes and DaveyHayes and Davey, 1975). Some of the gravity gradients over the eastern Ross Ice Shelf (15 mgal in 500 km) may indeed be explained by former ice loading, a conclusion consistent with both the Maine model and mine, since both require some further grounding over this part of the Ross Ice Shelf. In the Ross Sea, however, marine free-air gravity data demonstrate little correlation with bathymetry (i.e. none similar to that displayed in North America and Scandinavia where the gravity field often closely mirrors topography (Reference WalcottWalcott, 1970)). Such observations would suggest possible influence of other controlling factors such as tectonic activity. Indeed analysis of RIGGS data (Bentley and others, in press) points to such geological influences on the gravity field of the Ross Ice Shelf and indicates that the gravity results cannot be unambiguously interpreted in terms of recent large-scale ice loading. Substantial vertical movements during the mid to late Cenozoic are already documented (Reference BarrettBarrett, 1975).

3. Sea-level

Crucial factors in all reconstructions of the Ross Sea region are the magnitude and duration of major sea-level changes. A major eustatic sea-level fall (S _f ) of > 150 m would fully ground and thicken the present Ross Ice Shelf where much of the water column, especially in the eastern sector, is less than 200 m thick (Reference Clough and RobertsonClough and Robertson, 1975)- Since the floor of the now open Ross Sea lies at similar, and over considerable areas at higher, elevations than beneath the shelf, the grounding line would continue to advance towards the continental shelf edge as predicted by Reference WeertmanWeertman (1974). If S_f < 50 m the result would be only a small forward movement of the grounding line (i.e. proportional to bed slope) as the sea floor is sufficiently deep to maintain buoyancy. There is, therefore, a critical value for S _f which will trigger full-scale grounding in the Ross Sea. Based on current ice-shelf thickness gradients and detailed bathymetry we suggest a number between —120 and —130 m. This agrees closely with Reference Thomas and BentleyThomas and Bentley (1978), whose calculations of thickness for an equilibrium grounding line take into account various values of the sliding-law parameters.

If such a eustatic change were sustained for only a short period, however, an expanding ice sheet might not have sufficient time to achieve equilibrium, and subsequent rising sea-level would trigger premature retreat. The value taken by the duration S_t of sea-level lowering, which has not been stressed in previous studies, is thus of considerable importance, although difficult to estimate. A marine ice sheet buttressed by an ice shelf, may have a complex response depending upon creep and sliding laws, friction between ice shelf and its margins, accumulation-rates within interior catchments, the presence and dynamics of ice streams, etc. In consideration of these factors and the work of Reference Thomas and BentleyThomas and Bentley (1978), who examine the reverse case of ice-sheet retreat, we suggest that for grounding-line advance across the Ross Sea (a distance of up to 1 000 km) to reach equilibrium, the response time may lie between 10³ and 10⁴ years. Advance-rate is not likely to be linear, but with slow initial forward motion as ice is thin, cold, and possibly frozen to the bed. As it thickens, basal ice will warm and “soften” speeding the advance, with possible development of fast basal sliding and strong preferred fabrics. The process may later slow as interior catchments become starved of accumulation by their increasing separation from moisture sources. The net result of such response times is that the amplitude of short-term sea-level events (e.g. maximum lowering during the late Wisconsin) will be filtered out and hence not experienced by the ice shelf.

It is at present difficult to evaluate precisely the magnitude of sea-level changes in the Ross Sea during Wisconsin times. This is due to the complex interaction between isostatic adjustment in the ocean basins and transient sea-level tilting induced by gravitational attraction of ice masses (Reference Farrell and ClarkFarrell and Clark, 1976; Reference ClarkClark, 1976). Several estimates of the volume of water abstracted by ice during Stage 2 have been given; the most reliable result from stable-isotope analyses of marine cores with high sedimentation rates. Reference Ninkovich and ShackletonNinkovich and Shackleton (1975) report a change in δ¹⁸O of — 1.6‰ although earlier studies by Shackleton and Opdyke (1973) suggested a figure of — 1.2‰. These values represent lowering of sea-level by 160 m and 120 m respectively if we apply the rule-of-thumb: S _f = 100δ (Shackleton and Opdyke, 1973) although S _f = 110δ is also used. These figures include the effects of both north and south polar areas, although we expect the “northern” component to dominate.

Calculation of the resulting depression of sea-level transmitted to the Antarctic coastal zone has to take account of global ocean-floor recovery from unloading and gravitational effects induced by first Northern Hemisphere ice-sheet growth and later Antarctic ice thickening. Some estimation of S _f in the Ross Sea embayment with regard to these factors has recently been made. Reference ClarkClark and others (1978), Reference Lingle and ClarkLingle and Clark (1979), and Clark and Lingle (in press) have computed sea-level lowering around Antarctica caused by Northern Hemisphere ice-sheet growth to be between 75 and 100 m. J. A. Clark (in a personal communication) suggests that a sea-level stand of — 120 m in the Ross Sea would represent an extreme lower limit, and in our modelling we have taken this as a maximum case. Furthermore we believe that independent marine geological data provide indirect evidence for S _f < 130 to 120 m by giving limited support for ice-shelf conditions but not grounded ice.

1. Lithostratigraphy and Depositional Environment

The upper sedimentary sequence flooring the Ross Sea, as sampled by piston and rotary cores, comprises three primary units (Table I). Unit I is a soft, soupy, diatomaceous, silty clay, apparently little reworked and containing only a few per cent fraction of sand. The absence of large clasts or granules and abundant diatoms (40% by volume) are particularly significant and argue for little ice rafting in the Ross Sea with prevailing ice-shelf, pack-ice, and/or open-water conditions.

Table I.Litho- and chrono-stratigraphy of Ross Sea cores Dates favoured in this paper are boxed

Immediately beneath Unit I there is a clear change to a much stiffer silty claystone containing scattered pebbles and granules (up to 60 mm in size), but with only 2–5% diatoms. Reference Kellogg, Kellogg, Truesdale and OstermanKellogg and others (1979) recognize a thin (0.1–0.5 m) transition zone between Units I and II (Consisting of a well-sorted sand. The bulk of Unit II is typically a marine till of between 18 and 40 m thickness. There appears to be little or no evidence of reworking: only at Deep Sea Drilling Project (DSDP) Site 273 in the western Ross Sea, do the Project team (Initial Reports, 1975) consider that Unit II may have slumped down the edge of Pennell Bank into the flanking sea-floor trough. This is suggested by seismic profiles, bottom topography, a diachronous, mixed diatom fauna, and lack of current activity. The source of materials would be Unit III. A different opinion is held by Reference Kellogg, Kellogg, Truesdale and OstermanKellogg and others (1979), who believe that Unit II has been extensively reworked, a view they support by the apparent recycling of faunal and floral constituents which exhibit fragmentation and silicification.

Between Units II and III there is evidence of a major unconformity, considered to be equivalent to the angular contact observed on seismic-profiler records from the Ross Sea (Reference Houtz and DaveyHoutz and Davey, 1973; Reference Hayes and DaveyHayes and Davey, 1975). A pronounced increase in lithification occurs across the unconformity to a thick sequence of hard, pebbly silty-claystone which we interpret as a typical marine till. J-9 cores appear to penetrate similar strata to Unit III. Reference Brady and MartinBrady and Martin (1979) find evidence from the palaeontological content of the cores for open water conditions with abundant ice rafting.

2. Chronology

According to Reference FillonFillon (1975) the age of Unit 1, given by foraminiferal and palaeomagnetic polarity studies, is Brunhes (0–0.69 Ma b.p.) but it is probable that the early Brunhes is absent and the sediments are much younger. Sedimentation rates typical of glacio-marine deposits, and as determined for old sequences in the Ross Sea from recognition of palaeomagnetic events, are 10–20 m Ma^–1 (Reference FillonFillon, 1975, Reference Fillon1977). On the basis of these rates the extrapolated age for the observed thickness of Unit I is between 20 and 60 ka b.p. The work of Reference Truesdale and KelloggTruesdale and Kellogg (1979) suggests that Unit I could be at least Holocene in part: they present a ¹⁴C date of between 5 and 11 ka b.p. on a composite core sample.

The dating of Unit II is not precise. On the grounds of foraminiferal studies and palaeomagnetic polarity stratigraphy, Reference FillonFillon (1975) has ascribed the sequence to Gauss epoch (2.4–3.4 Ma b.p.). Reference Kellogg, Kellogg, Truesdale and OstermanKellogg and others (1979), however, assert that Fillon’s dating is incorrect, and on the basis of three diatom species infer a late Pleistocene age for the uppermost part of Unit II.

Palaeontological investigation of both foraminiferae and bivalves, combined with a well-defined palaeomagnetic polarity stratigraphy, indicate that Unit III spans the period between Oligocene and upper Miocene (25–7 Ma b.p.) (Initial Reports, 1975). Reference Brady and MartinBrady and Martin (1979) found abundant microfossils (diatoms, silicoflagellates, and benthonic Foraminifera) in J-9 core which they interpret as suggesting a mid-Miocene age for a succession that appears equivalent lo Unit III.

3. Sediment Distribution Pattern

Figure 3 indicates the relative age distribution of sediments in the Ross Sea. The youngest materials (Unit I) are concentrated in the linear topographic lows, while older exposed detritus (principally Unit II) occupies, almost without exclusion, the uppermost stratigraphic position on neighbouring swells or highs. Unit I would appear to be thin close to the front of the Ross Ice Shelf, but thickens towards the continental shelf edge.

4. Discussion

There is obvious disagreement on the antiquity of Unit II between Reference FillonFillon (1975) and Reference Kellogg, Kellogg, Truesdale and OstermanKellogg and others (1979). Fillon’s Gauss age is most probably too old. The normal magnetic polarity could well be Brunhes not Gauss, and criticism by Kellogg and others of Fillon’s foraminiferal interpretations (including the possibility of reworking) appear justifiable. Nevertheless it is also difficult to substantiate the Late Pleistocene age that Kellogg and others base on diatoms alone. Substantial problems of chronostratigraphic resolution are associated with micro-palaeontological dating in high-latitude environments where species diversity is low (and Unit II is demonstrably diatom-poor). Considerable care is required in establishing the age range of a complete assemblage—it is rarely sufficient to use one or two “key” species by themselves. Although at this time supplementary, independent, datable evidence, is not available to refine Unit II age beyond Brunhes, several other factors lead this writer to believe that Unit II is older than Wisconsin.

Gravity cores taken at J-9 from the sea floor beneath the Ross Ice Shelf (Figs 1 and 2) disclose an extremely thin (few mm) lag deposit overlying compact marine mud containing small erratic pebbles and a high percentage of glacial flour (Reference Webb, Webb, Ronan, Lipps and DeLacaWebb and others, 1979; Reference Brady and MartinBrady and Martin, 1979). The latter unit has been dated on the basis of diatom assemblages and comparison with DSDP and DVDP core materials to the mid-Miocene (c. 14 Ma b.p. or Unit III correlative). There is a pronounced absence of a more recent, less-consolidated, reworked layer above Miocene strata. According to the Maine model (Fig. 1 inset) J-9 must have witnessed substantial thickening and grounding of the ice shelf in Wisconsin times.

In common with areas to the north in the now open Ross Sea, Reference Kellogg, Kellogg, Truesdale and OstermanKellogg and others (1979) suggest that such events would have produced a distinctive till-like layer (Unit II).

The findings at J-9 are better explained by assuming Unit II is older than Wisconsin and adopting the expanded-ice-shelf model presented in this paper. Figure 1 indicates that J-9 lies very close to the inferred Late Wisconsin grounding line, but that since it occupies one of the sea-floor troughs it may have been the site of an ice stream. In the brief period when the grounding line was at its maximum seaward extent, this ice stream was probably responsible for active erosion of sea-bed sediments close to J-9. Subsequent grounding-line retreat would have heralded a much longer period with floating ice-shelf conditions over the J-9 site. Debris-poor basal shelf ice, “cleaned” by melting, combined with bottom-current scour, would have suppressed heavy sedimentation, giving rise to the observed thin pebble pavement. Our reconstruction does not, therefore, anticipate a thick Unit II till at J-9. Furthermore the distribution of sediments over the whole Ross Sea area with youngest materials in the topographic lows provides little support for the idea of ice universally grounded in the embayment, let alone the contention that a “Ross Ice Sheet” was responsible for the erosion of distinctive sea-floor channels out to the edge of the continental shelf during the last glaciation. It should also be noted that where the youngest deposits appear to be located on shoals (Fig. 3) these areas lie close to the edge of the continental shelf where local grounding of an expanded ice shelf was not predicted (Fig. 1).

A more reasonable age for Unit II is derived by attributing it to one or more of the several pre-Wisconsin glacial episodes during Brunhes times. Reference Shackleton and OpdykeShackleton and Opdyke (1973), for instance, document seven or eight glacials during this period. Furthermore their stable-isotope studies reveal that some of these glacial events were longer and more severe than the Wisconsin. Stages 6, 12(?), and 16–18 appear the most favourable in terms of duration and inferred ice volumes and sea-level lowering for the grounding of ice in the Ross Sea and reworking of sea-floor sediments into Unit II till. They date as approximately 128–195 ka b.p., 440–472 ka b.p., and 592–688 ka b.p. respectively (Shackelton and Opdyke, 1973).

We conclude that sedimentary evidence currently available does not support extensive grounded ice over the Ross Sea but gives tacit confirmation of enlarged ice-shelf conditions in Wisconsin times. It is likely that a concentrated investigation of Ross Sea marine deposits especially beneath the ice shelf may eventually provide the best independent test of Wisconsin glacial models of the region, and programmes addressing this problem should be actively encouraged.