A study of density and porosity is presented for the 1285-m-long AND-1B core recovered from a flexural moat in the McMurdo Sound (Antarctica) in order to interpret sediment consolidation in an ice-proximal location on the Antarctic shelf. Various lithologies imply environmental changes from open marine to subglacial, and are numerically expressed in high-resolution whole-core wet-bulk density (WBD). Grain density data interpolated from discrete samples range from 2.14 to 3.85 g/cm3 and are used to calculate porosity from WBD in order to avoid the 5%–15% overestimation and underestimation of porosities obtained by standard methods. The trend of porosity extends from 0.5 near the top (Pleistocene) to 0.2 at the bottom (Miocene). Porosity fluctuations in different lithologies are superimposed with 0.2–0.3 in sequences younger than ca. 1 Ma and 0.5–0.8 in Pliocene diatomites. The AND-1B porosities and void ratios of Pliocene diatomites and Pleistocene mudstones exhibit a large negative offset compared to modern lithological analogs and their consolidation trends. This offset cannot be explained in terms of the effective stress at the AND-1B site. The effective stress ranges from 0 to 4000 kPa in the upper 600 m, and reaches 13,000 kPa at the base of the AND-1B hole. We suggest an excess of effective overburden stress of ∼1700 and ∼6000 kPa to explain porosities in Pliocene diatomites and Pleistocene mudstones, respectively. This is interpreted as glacial preconsolidation by subsequently grounded ice sheets under subpolar to polar, followed by colder polar types of glaciations. Information on Miocene consolidation is sparse due to alteration by diagenesis.

In the austral summer season of 2006–2007, the Antarctic Drilling Project, ANDRILL, drilled a unique record of more than 1200 m of sedimentary rock, the AND-1B core, from a deep basin beneath the McMurdo Ice Shelf near Ross Island, Antarctica (Fig. 1). We followed a multiproxy approach in order to study the behavior of the Ross Ice Shelf and Ice Sheet under different climate conditions during the past 13 Ma (Naish et al., 2007). This approach includes a high-resolution data set of physical properties measured on the core during the drilling phase (Niessen et al., 2007) that also forms the major data source of this paper.

At the AND-1B drill site, the nature of the depositional environment varied drastically over time in response to global climate fluctuations (Naish et al., 2009; Levy et al., 2012). On orbital time scales, the system oscillated back and forth between two main states: the presence of a grounded Ross Ice Sheet as part of the Antarctic Ice Sheet during glacial times, and an open ocean environment representing the interglacial situation for much of the late Cenozoic. The respective deposits produced in such a highly dynamic environment are equally variable, and range from massive and stratified subglacial diamictites to open-marine diatomites and mudstones. Transitional facies reflecting intermediate climate states are found as claystones, mudstones, siltstones, and sandstones. In addition, the sediment core includes volcanic deposits of various types that differ geochemically as well as genetically and in particle size. Below ∼50 mbsf (meters below seafloor) most of the core is lithified, suggesting some postdepositional overprint by diagenesis (Krissek et al., 2007; Pompilio et al., 2007). The primary goal of this publication is to determine, describe, and interpret the porosity-depth relationships of the AND-1B core, taking into consideration the complex boundary conditions.

Whole-core physical properties of very high vertical resolution (Niessen et al., 2007) provide initial numerical information about variations in the depositional system and postdepositional alteration. Most of the AND-1B changes in lithologies and facies are somehow reflected in the downcore pattern of wet-bulk density (WBD), sonic velocity, and magnetic susceptibility, in particular in the upper 600 m of the core (Niessen et al., 2007). Although easy to measure on whole cores by gamma-ray absorption on multisensor tracks, WBD may be a complex parameter that is often not straightforward to understand and interpret. By definition, WBD is a function of grain density, water density, and porosity. As a simple approach omitting analysis of discrete samples, WBD is often used to calculate GRAPE porosities (gamma ray porosity evaluator, also referred to as fractional porosity). This is based on the assumption of constant densities of grains and pore water of 2.7 and 1 g/cm3, respectively (e.g., Geotek Limited, www.geotek.co.uk, manual accessed in 2000). As we demonstrate herein, fractional porosities can have significant errors if grain densities are not specifically analyzed, and, as in our case, the core consists of diatomaceous and volcanic materials. This is of importance because void ratios are calculated from fractional porosities, which, together with effective stress levels, provide fundamental parameters for assessing the consolidation history of the AND-1B core.

Sediment porosity is affected by a large number of different variables. At the sediment surface, initial porosity is a complex function of interrelated variables controlled by the depositional environment. Variables include grain size, shape, sorting, and mode of packing (e.g., Bryant et al., 1981, and references therein). With greater burial depth and time, porosities are further altered by postdepositional processes; these include squeezing out of initial pore water by overburden pressure, alteration of interparticle bonding, cementation, and tectonic stress (Chilingarian and Rieke, 1974). In situ effective stress is the primary control on the rate and magnitude of sediment consolidation (O’Regan et al., 2010a). In glacial environments effective stress due to temporal grounding of ice, in addition to that of the formation, may play an important role for understanding glacial history (Sættem et al., 1996; O’Regan et al., 2010a). This may also involve removal of overburden by glacial erosion and thus exhumation and elastic rebound (Bryant et al., 1981; O’Regan et al., 2010a). In order to understand and interpret the depth-porosity relationship of a long sedimentary record (or single core), it is important to differentiate between effects related to the depositional environments and the consolidation history.

Porosity-depth profiles and general consolidation behavior in shales and sandstones are studied extensively (e.g., Schmertmann, 1953; Magara, 1980; Bryant et al., 1981; Hamilton, 1976). Studies of the consolidation status of sediments with respect to processes in glacially dominated environments have mainly concentrated on the compaction of diamicton and diamictites, addressing questions about basal ice conditions and estimation of past ice dynamics and ice sheet thickness (e.g., Sættem, 1990; Solheim et al., 1991; Sauer et al., 1993; Christoffersen and Tulaczyk, 2003a; O’Regan et al., 2010a). Diatomaceous ooze or diatomites in longer core records exhibit abnormally high porosities and low consolidation, often associated with underconsolidation (Bryant and Rack, 1990; Bryant et al., 1981; Pittenger et al., 1989; Volpi et al., 2003). Studies of diatom deposits subsequently affected by grounding ice are very rare.

In this paper we present a porosity record of the AND-1B core based on WBD and grain density measurements that is more sophisticated than earlier preliminary records of AND-1B porosity (Morin et al., 2010; Konfirst and Scherer, 2012). It is demonstrated that a simple approach calculating porosity by assuming constant grain densities of 2.7 g/cm3 (e.g., Geotek Limited) is not sufficient for the AND-1B core, because significant parts of the core have grain densities well below and well above 2.7 g/cm3, such as diatomites and volcanic units, respectively. The physical character of the AND-1B core is discussed in the context of depositional effects and consolidation history. The void ratios as a function of effective stress are compared to general consolidation trends in order to identify additional reduction of void ratios in excess of those caused by stress levels due to field consolidation. This is important for the AND-1B core sediments because they were originally deposited in open water environments and subsequently were overridden by grounded ice. We present a brief assessment of glacial loading for a Pleistocene mudstone and a Pliocene diatomite. Further quantification of past ice-sheet thickness at the AND-1B site is beyond the scope of this study and will be addressed in the future.

Further relevance and a broader significance of this paper is given by the fact that the knowledge about consolidation based on porosity of long sediment records from the Antarctic continental margin is important to constrain or modify past glacial regimes reconstructed from sediment description and sample analysis (Naish et al., 2009; McKay et al., 2009; Scherer et al., 2007). This, in turn, can improve assessment of boundary conditions needed for ice-sheet modeling (Pollard and DeConto, 2009). More specifically, porosity-depth relationships presented here are essential for a variety of subsequent studies of the AND-1B core that are ongoing or being prepared. These include (1) reconstruction of ice thickness in the McMurdo Sound during the Last Glacial Maximum (LGM), (2) the quantification of preconsolidation stress based on ongoing consolidation tests using discrete samples, (3) basin analysis by means of back-stripping methods, (4) calculation of sediment accumulation rates and sediment budgets to reconstruct marine productivity, and (5) the investigation of porosity-velocity relationships and related acoustic properties of the AND-1B core for a better understanding of seismic records in the McMurdo Sound area.

Age Model and Glacial Regimes

The age model of the AND-1B core is based on a combination of 40Ar/39Ar ages determined on 14 volcanic samples (Fig. 2), microfossil biostratigraphy of diatomite units, magnetostratigraphy, and correlation of stratigraphic sequences with the marine isotope record (Wilson et al., 2007, 2012; Naish et al., 2009; Konfirst et al., 2011; Levy et al., 2012; McKay et al., 2012; Cody et al., 2012). The integrated chronology identifies several unconformities in the upper 600 m of the core that separate sections representing relatively continuous deposition (Wilson et al., 2007; Naish et al., 2009; McKay et al., 2009). Unconformities are interpreted as being the result of nondeposition and/or erosion due to long-term tectonic (Wilson et al., 2007; Levy et al., 2012) and shorter term glacial effects (McKay et al., 2009, 2012), making the consolidation-exhumation history of the core rather complex. The chronology is more robust in the upper 600 m of the core. The lower part is constrained only by 40Ar/39Ar ages of a lava flow at 648.37 mbsf and a cluster of volcanic clasts providing a maximum depositional age of 13.57 ± 0.18 Ma near the base of the core (Wilson et al., 2007; Fig. 2).

The core contains a succession of subglacial, glaciomarine, and marine sediments that comprise ∼58 depositional sequences (Krissek et al., 2007; Fig. 2). All sequences start with a glacial surface of erosion (GSE), some of which represent major unconformities (Wilson et al., 2007; Levy et al., 2012). McKay et al. (2009) identified three different glacial regimes (or motifs) that characterize ice-sheet extents associated with subglacial erosion and deposition across the Ross Embayment followed by ice retreats and open-marine conditions: motif 1, cold polar types of glaciations similar to the situation since the LGM; motif 2, subpolar to polar styles of glaciation conditions that were warmer than present, but that had limited amounts of subglacial meltwater; and motif 3, subpolar styles of glaciations that released significant volumes of meltwater. During the Middle Miocene a cold and fully polar glacial regime lacking evidence of subglacial meltwater was followed by a warmer glacial regime producing significant amounts of meltwater-induced sands and muds. The lack of diatomites indicative of open-marine deposition is explained by glacially induced turbidity inhibiting extensive plankton blooms. From the early Pliocene until the early Pleistocene the glacial regime changed from subpolar to polar conditions, as documented by a decrease of subglacial meltwater sediments, leading to fluctuations between colder subglacial diamictites and open-marine diatomites. Diamictite-diatomite alterations dominate most of the upper 600 m of the core. During the middle Pleistocene transition, the glacial regime returned to full cold polar glaciations, documented by subglacial diamictites intercalated with clay and mudstones that originated from deposition under an ice shelf during interglacial times (McKay et al., 2012). This style of deposition is consistent with the situation beneath the present Ross Ice Shelf, with a lack of significant subglacial meltwater sediments and diatom ooze.

From sediment descriptions there is a great deal of evidence for past grounding events of ice sheets at the AND-1B core location (Krissek et al., 2007). Sediments directly above and below unconformities interpreted as a GSE (Fig. 2; McKay et al., 2009, 2012) often display evidence of soft-sediment deformation, clastic dikes, and fractures and brecciation, including horizontal alignment of clasts indicative of shearing by glacial overriding (McKay et al., 2009). The diatomite below the GSE typically is deformed, sheared, and intermixed with subglacial till derived from the GSE, which is attributed to ice pushing during glacial advance (McKay et al., 2009).


A custom-built drilling system comprising a UDR-1200 rig, jack-up platform, hot-water drill, sea riser, and diamond-bit wireline coring string was set up by the ANDRILL Project on the McMurdo Ice Shelf at 77.89°S, 167.09°E (sediment surface at 917 m below sea level; Falconer et al., 2007). Soft-sediment coring barrel assemblies were used during the spud-in process to a depth of 41.88 mbsf. These core sections were recovered in liners. Below 41.88 mbsf the strata was mostly lithified and recovered without liners to the final depth of 1284.87 mbsf (Falconer et al., 2007). Core was recovered in 3 m and 6 m runs and was delivered from the drill rig to the drill-site laboratory complex, where it was cleaned and cut into 1 m lengths prior to logging. Lithified core was placed into one-half of a split PVC pipe to enhance core handling and maintain core integrity.


Whole-Core WBD Measurements and Data Correction

High-resolution measurements of whole-core gamma-ray attenuation (GRAPE) were performed at the drill-site laboratory by means of a nondestructive multisensor core logger (MSCL; Geotek Limited) a few hours after the cores were retrieved. Core was logged in liners (to 41.88 mbsf) and on split pipe (41.89–1284.87 mbsf) in 1–4 cm depth intervals (Niessen et al., 2007). Core quality in liners was relatively poor but generally good on split pipe. Data logged on disturbed sections of core were removed during processing (Niessen et al., 2007). The conversion of gamma counts into density values was achieved after calibration using aluminum standards following the approach of Best and Gunn (1999; described in detail elsewhere; see Niessen et al., 2007). Standard material density of industrial aluminum (2.71 g/cm3) was used for WBD data processing at the drill site (Niessen et al., 2007). Within the framework of this study, a check has shown that the actual density of the aluminum calibration material used as standards for AND-1B is 2.8 g/cm3. Therefore, a correction was applied to the initial WBD data set by means of a regression function:
where y is WBD corrected and x is WBD initial from Niessen et al. (2007).

Determination of Grain Densities through Discrete Sample Measurements

Bulk sediment samples (n = 1210) with a volume of ∼20 cm3 each were taken over the entire length of the sediment core, providing a resolution of ∼1 sample/1 m of core. Sample bags were sealed immediately after sampling to prevent drying of samples. All samples were freeze-dried and the mass was determined from the wet and dry sample, respectively, to obtain the gravimetric water content. In preparation for further geochemical analyses, the sample blocks were crushed to remove visible clast fragments (>2 mm) before undergoing pre-grinding by mortar and final treatment in a planetary mill to produce analytically fine matrix powder. An automatic gas displacement pycnometer (AccuPyc 1330 Pycnometer; for details, see Micromeritics Operator’s Manual, v. 3.03, http://www.micromeritics.com) was used to retrieve the pycnometer density of matrix samples. Based on the previously determined gravimetric water content, a correction as presented in Weber et al. (1997) (after Hamilton, 1971; Gealy, 1971) was applied to account for the effect of salt initially dissolved in the pore water and dried with the sample, resulting in grain densities for only the matrix.

Calculation of Fractional Porosities

In order to calculate porosities for all depth intervals where reliable WBD data are available, grain density data for discrete samples were interpolated to the level of resolution of the MSCL data. The interpolation takes into account different lithologies as documented on 1 m logs (http://coreref.org/projects/and1-1b/viewer/455#0; Krissek et al., 2007), which vary on a much higher vertical resolution than expressed in the simplified lithological overview log (Fig. 2). These high-resolution lithology data were then grouped into 17 lithological classes (Table 1), which allowed sorting and separating the data according to lithology and depth.

The interpolation of grain density was carried out using KaleidaGraph software (www.synergy.com/; Interpolate Curve Fit; mathematics described in Stineman, 1980) separately for each lithology class from top to bottom of the core corrected by linear interpolation between lithological boundaries not represented by discrete samples. Subsequently, the interpolated grain densities from different lithologies was reassembled for the entire data set and added to the MSCL data. The high-resolution data of interpolated grain densities were smoothed over 10 points using a Gaussian filter (KaleidaGraph software) in order to avoid unrealistically sharp shifts of grain densities (GD) at threshold boundaries of different lithological composition, as most lithological transitions of the core are gradual except for GSE (http://coreref.org/projects/and1-1b/viewer/455#0).

Fractional porosities (FP) were then calculated according to the Geotek Limited manual (www.geotek.co.uk, accessed 2000):
where GDi is the lithology-specific interpolated grain density as described above and ρw is pore-water density. We have assumed a constant ρw of 1.03 g/cm3, because pore-water density for in situ conditions differs insignificantly (0.99–1.055 g/cm3) as calculated from conductivity of pore-water samples (after Fofonoff and Millard, 1983; data from Pompilio et al., 2007).
The void ratio (e), defined as the volume of voids divided by the volume of solids, is calculated from FP (Hamilton, 1976):

In some intervals the interpolated grain density of matrix was found to be lower than the respective WBD value. In this case, a negative porosity would have resulted, which is not valid. These intervals are associated with the occurrence of pebble-size clasts and strongly cemented horizons, which were avoided for grain density sampling. These data points illustrate the errors inherent in the method of grain density interpolation. A threshold was set at FP = 0.005, below which porosities are considered as incorrect and were discarded (147 of 66,630 data points including negative FP).

Intergranular or effective stress (σ′), also called effective overburden pressure, was calculated according to Holtz and Kovacs (1981):
where u is the pore-water or neutral pressure and σ is the total vertical stress:
where ρi and ρw are the subbottom WBD and water density of a given depth interval zi or zw, respectively, g is the gravitational constant, and n is the number of layers for which density data exist. Because no in situ density data could be evaluated for the AND-1B borehole, in situ stress calculations were carried out using the WBD of the core.
In order to assess the effect of glacial loading, the overburden pressure caused by a possible ice sheet was quantified for normal vertical stresses (Mulder and Moran, 1995; O’Regan et al., 2010a):
where ρice is the ice density (0.917 g/cm3; Denton and Hughes, 2002), hi is the height of the ice sheet, hw is the water depth, and Hi is the part of the ice sheet not supported by seawater. Similar to previous publications (O’Regan et al., 2010a, and references therein), it is assumed that the total overburden associated with the ice above sea level is transferred to the sediments the moment the ice becomes grounded. This is given when the height of the ice sheet (hi) is equal to the product of water depth (hw) multiplied by the ratio of seawater density (ρw = 1.03 g/cm3) over ice density (ρice). When the ice is in contact with the seafloor, the added force acting on the sea bed is only due to that part of the ice sheet (Hi) not supported by seawater (Mulder and Moran, 1995). For simplification, isostatic loading and rebound are not considered (Fig. 3).

Whole-Core WBD Record

WBD values range from ∼1.25 g/cm3 to 3.11 g/cm3, with one exceptional maximum value of 3.85 g/cm3 measured through a pyrite nodule at 997.80 mbsf (Fig. 2A). Over the entire length of the core WBD values increase by ∼0.6 g/cm3 from top to bottom. Superimposed on the depth gradient are strong fluctuations in WBD related to different lithologies. In general, mean values for the different lithologies are mainly distributed between two end-member types that relate to the high-amplitude oscillations in WBD, particularly from ∼100–600 mbsf: diatomites with a mean WBD of 1.52 g/cm3 and diamictites with a mean of 2.13 g/cm3. At ∼648 mbsf, locally high WBD values of ∼2.63 g/cm3 are attributed to a submarine lava flow. A 40-m-thick interval of volcanic sand at the bottom of the core shows exceptionally low density values for that depth (∼2.0 g/cm3), with an offset of ∼0.3 g/cm3 from the general downcore trend (Fig. 2A).

The high scatter in the WBD data with distinct peaks is, in most cases, caused by the occurrence of clasts. A total of 7454 WBD spikes were identified; this is ∼11% of the total WBD measurements, compared to a total of 86,333 counted clasts (Pompilio et al., 2007; Talarico et al., 2011). Regression plots (Fig. 2B) reflect the generally weak relationship of clast occurrence and the actual WBD value. The correlation coefficient is very low when all lithologies are considered undifferentiated. The same is true for diamictites. For the diatomites the correlation is relatively higher.

Matrix Grain Densities from Bulk Samples

Showing a general depth trend, which strongly resembles that of the WBD data, most grain density values range between 2.45 and 2.75 g/cm3; there is great variability even within samples of similar lithology, reflecting the heterogeneity of the material (Fig. 2A; Table 1). Two lithologies, however, deviate significantly: diatomites with grain densities between 2.14 and 2.4 g/cm3, and volcanic sandstones having significantly higher grain densities, to 2.93 g/cm3, in the top part of the core and lower grain densities of ∼2.55 g/cm3 in the bottom part. An exceptional feature is the submarine lava flow with a local density maximum of 2.8 g/cm3 at 647.25 mbsf. A general increase of 0.1 g/cm3 in grain densities from top to bottom is found within nonvolcanic lithologies.

A major change in the distribution of grain densities is apparent between the upper and lower 600 m of the core. Grain density values in the upper part show a strongly bimodal distribution related to lithology (Fig. 4). Diatomite samples with low densities can clearly be distinguished from the rest, which are less defined and reflect the influence of the various remaining lithologies with a dominance of diamictites (Fig. 4). A distinct cluster of only a few samples of ∼2.9 g/cm3 marks the outstanding character of the volcanics in the upper part of the core. For the lower half, from 600 to 1285 mbsf, the count maximum has shifted from 2.6 g/cm3 to 2.7 g/cm3 and overall, data are more confined (Fig. 4). The opposite trend accounts for volcanic samples, as the ones from the upper half of the core show comparably higher grain densities, whereas volcanic samples from lower parts of the core are characterized by lower grain densities compared to nonvolcanic ones (Fig. 4).

Porosities in the AND-1B Core and Effective Stress

Fractional porosities in the AND-1B core show values mainly ranging from 0.2 to 0.7 depending on depth interval and dominant lithology (Fig. 5). As observed in the record of WBD, the occurrence of clasts is reflected by punctual offsets from the general pattern, in this case toward lower porosities. The highest porosities are reached in diatomites with more or less constant downcore values between 0.5 and 0.7. A reduction of porosities from ∼0.5 to 0.2 describes the general top-to-bottom trend, which shows prominent local deviations, i.e., low-porosity diamictites in particular in the upper 100 m of the core, high-amplitude variability between 100 and 600 mbsf, and a high-porosity volcanic sandstone at the bottom part. The range of porosities within each diatomite is relatively large and offsets of single data points relate to the occurrence of clasts. The mean content of biogenic opal and diatomite porosities covary with depth (Fig. 6).

The effect of grain densities on the AND-1B porosities becomes obvious when the difference is calculated of data assuming a constant grain density of 2.7 g/cm3 minus data using interpolated grain densities (Fig. 5). In particular between 180 and 1000 mbsf, porosity would be overestimated between 5% and 10% with the strongest deviation in diatomites of as much as 20%. In a few intervals of volcanic material the porosity would be underestimated between 5% and 10%.

The trend of effective stress due to overburden versus depth below seafloor (Fig. 7) exhibits an interval of relatively rapid increase down to 150 mbsf followed by a very low increase between 150 and 600 mbsf. The rapid increase is mainly caused by the relatively high WBD of Pleistocene diamictites, and the low increase is mainly caused by low WBD of diatomites. The trend in volcanic material does not differ notably from other lithologies (except for diatomites). The plot of effective stress versus void ratios exhibits mainly three different groups of data (Fig. 8). (1) Pleistocene diamictites cluster at relatively low void ratios and stress levels between 110 and 800 kPa. (2) Diatomites retain relative high void ratios to as high as 4.5 at stress levels between 1000 and 10,000 kPa. (3) For most of the remaining data (except for some volcanic material) void ratios decrease from 1–2 to 0–1 at stress levels between 800 and 12,000 kPa.

Grain Densities and the Effect on Porosity

The grain densities of AND-1B bulk samples (Fig. 2A) are primarily a function of the varying petrological and geochemical compositions of the matrix material (Pompilio et al., 2007). Because the compositions of the matrix material are linked to the occurrence of two lithologic end-member types, diatomites representing open-marine conditions and diamictites indicative of subglacial or proglacial environments (Naish et al., 2009; McKay et al., 2009), a strong first-order environmental control on the actual grain density value is apparent for both fluctuations and the general downcore trend.

For diatomites the negative correlation between the biogenic opal content and grain density (lower regression plot in Fig. 6) suggests that the range of grain density values for the diatomites relates strongly to the degree of purity of the samples with respect to the biogenic opal content. Diatom skeletons are composed of silica, but incorporate water molecules in the amorphous structure of biogenic opal to 2–15 wt% (Müller and Schneider, 1993), which causes the low grain densities of diatomites. Recycling of diatomaceous material into overlying diamictites due to glacial erosion is commonly described for the depth interval between 180 and 600 mbsf of the AND-1B core (motif 2; McKay et al., 2009; Fig. 2A), but not observed in other diamictites of the AND-1B core. This may explain the slightly lower grain densities of the diamictites of motif 2 (Fig. 2A). Changes in diamictite composition to explain these lower grain densities can be ruled out. Provenance analyses indicate similar regional cluster compositions in diamictites and mudstones from both the upper and the lower parts of the core (Monien et al., 2010; Pompilio et al., 2007; Talarico and Sandroni, 2009). A major change in diatom abundance and preservation characteristics is observed at ∼580 mbsf, from diatoms being abundant and relatively unaltered above to being rare and diagenetically altered below this depth (Scherer et al., 2007). The variable content of diatom frustules as primary deposits in diatomites or reworked in diamictites appears to be one major reason for the observed fluctuations and depth trend in grain density.

Variations in grain densities of volcanic samples are the products of the initial petrologic composition such as mafic, intermediate, and felsic being attributed to the different sources of the volcanic material (Pompilio et al., 2007). However, chemical alteration of the volcanic glass is an important contributor, for example, accompanying hydrothermal activity and burial-related diagenetic processes. Grain density values from volcanic samples from the upper 600 m can clearly be related to the initial chemical composition of relatively unaltered material. For example, high density values for 3 samples at ∼113 mbsf initiate from mafic volcanic sands, whereas the interval from 133 to 144 mbsf shows an intermediate composition for the volcanic sands (Pompilio et al., 2007). The situation is more complex for the lower 600 m, where grain densities of some volcanic material are <2.7 g/cm3. Most of the deposits from this interval have undergone modifications of initial textures and compositions, especially through palagonitization (Pompilio et al., 2007). This seawater-induced alteration process of volcanic glass results in a decrease of grain density through Si, Al, Mg, Ca, Na, and K losses and H2O gain (Gerard and Stoops, 2005). The occurrence of palagonitized deposits is mainly observed for the intervals of 660–705 mbsf and 1230–1270 mbsf, where volcanic siltstones and sandstones are similarly affected (Pompilio et al., 2007) and where grain densities are <2.7 g/cm3 (Fig. 2A).

In addition, the combination of biogenic, volcanic, and nonvolcanic terrigenous deposits in the sediment core gives a complex picture of alteration and burial diagenesis. This involves pyritization through the replacement of the biogenic silica and, in a few cases below 580 mbsf, recrystallization of opal-A into opal-CT (Scherer et al., 2007). In addition, aspects of fluid migration and cementation have to be taken into account. The occurrence of both carbonate and pyrite cement as major diagenetic features is described for the AND-1B core (Krissek et al., 2007). Dolomite was detected in the form of several discrete horizons in the upper part of the core (Scopelliti et al., 2011), whereas extensive occurrence of pyrite only occurs below 400 mbsf (Krissek et al., 2007).

The increase in grain density with increasing depth is thus a likely combination of the two aspects of a successively increasing diagenetic overprint as well as a decreasing content of biogenic opal. In particular opal, content and palagonitization are responsible for grain densities <<2.7 g/cm3 in many depth intervals of the AND-1B core. Without considering the actual grain densities, the calculation of porosities would, in places, overestimate pore volume significantly (Fig. 5).

Effect of Cementation on Porosity

Carbonate cements occur in a variety of forms, but the distribution of cement is highly heterogeneous in all lithologies (Krissek et al., 2007; Pompilio et al., 2007; Scopelliti et al., 2011). Disseminated carbonate cement was noted in the matrix of the uppermost Pleistocene diamicts of the core as shallow as 30 mbsf. The increased effect of carbonate cement downcore caused the shift from soft-sediment names (e.g., diamict) to rock names (e.g., diamictite) by ∼50 mbsf (Krissek et al., 2007). At this transition there is no notable difference between the porosities measured in soft sediments as compared to those measured in the lithified core (Fig. 5). This suggests that disseminated cement has an effect on core integrity, but its effect on pore volume is small. Depth intervals described by Krissek et al. (2007), in which carbonate cement is common to abundant compared to those where cement is low to absent, are not marked by notable shifts in porosity (Fig. 5). The same is true for common pyrite cement below ∼400 mbsf, where the overprint locally obscures sediment texture and stratification (Krissek et al., 2007). For such an interval (∼780–920 mbsf; Fig. 2A), there is no notable difference in the depth trend of porosity (Fig. 5). This indicates that pyrite cementation, as with carbonate, does not reduce porosity to a level relevant for the discussion herein.

There are exceptions where the cement has reduced the pore volume of the AND-1B core significantly. This is the case when carbonate cement is concentrated locally into thin micritic beds, most commonly within intervals of diatomite but also within silty claystone (Krissek et al., 2007). For the porosity data of diatomites, a strong effect of cementation can be observed near the base of DU-IX (diatomite unit) where carbonate cement is abundant (Krissek et al., 2007), and the porosity is reduced from >0.4 elsewhere in the unit to 0.1–0.4 (Fig. 6). This enhanced cementation is associated with a higher number of clasts observed in the same interval (Konfirst and Scherer, 2012), which contributes to the reduction of porosity in diatomites. In other intervals with micritic beds the effect on porosity is less pronounced or hardly visible (Fig. 6). Niessen et al. (2007) provided a typical example how micritic beds in the AND-1B core are characterized by physical properties. Between 130 and 136 mbsf within an alternation of mudstone, siltstone, and sandstone, p-wave velocities increase significantly from a base level of ∼2 km/s to 3–4 km/s in 3 thin intervals, each as much as 0.5 m thick (Fig. S1 in the Supplemental File1). This is associated with a distinct drop of electrical conduction of the core and an increase in WBD from 1.8 to 2.5 g/cm3. In these intervals the porosity quantified in this study drops from 0.45–0.6 to 0.4–0.05. It has been suggested that carbonate cement builds up a framework between particles, which blocks pore connections (and reduces permeability), increases elastic moduli, but does not fill the voids completely (Niessen et al., 2007). Thus, the effect of cementation on velocity and conduction is large while it is moderate to small on porosity. This also implies that core intervals not affected by intense cementation retain interconnected pore space to larger depth. This means that pore water is subjected to squeezing during long-term consolidation.

Although not yet quantitatively analyzed for the entire AND-1B core, cementation has notably reduced porosity in places. In general, strongly cemented intervals are thin (mostly <0.5 m; Fig. S1 [see footnote 1]), minor in terms of quantity, and by no means dominating the AND-1B porosity record. Thus, a diagenetic influence on consolidation has to be considered for the AND-1B core, but it does not explain most of the variability and depth trend of porosity. However, it is possible that thin cemented layers conserve the porosity of the lower strata because they reduce the permeability and limit vertical pore-water migration (Pompilio et al., 2007). Because cementation may have affected and altered the lower part of the core more than the upper part, which is difficult to quantify in this study, we interpret the porosity record below the lowermost diatomite (DU-XIII, ∼ 600 mbsf; Figs. 6 and 7) only briefly, and discuss the upper part in more detail.

Core Rebound

Core rebound must be taken into account if porosity and effective stress data from the AND-1B core are compared to data of laboratory consolidation tests. The porosity of all sediments of the AND-1B core must increase when they are removed from in situ stress conditions. Rebound in porosity after removal from overburden pressure at depth was quantified for different lithologies of Ocean Drilling Program (ODP) cores (Hamilton, 1976). From a depth of 500 mbsf rebound porosity is smallest in diatom ooze (3%) and largest in terrestrial sediments (8.5%). These expansions were quantified in soft sediments (Hamilton, 1976). Since most of the AND-1B core is lithified, it is expected that the cemented framework reduces core expansion significantly. Correlation of borehole to core data is good for the AND-1B site (Williams et al., 2012) and yielded small differences in the two depth scales. A constant correction of 1.5% depth expansion had to be applied to the AND-1B borehole data to match the core physical properties (Morin et al., 2007). This includes effects of wire expansion and pipe stretch (Morin et al., 2007) and may also attest to core expansion and increase in porosity. Although the precise effect of core rebound on porosity is not known, we assume that the effect is small. Therefore, the core porosity data discussed here are not corrected for rebound when compared to laboratory consolidation data. This means that AND-1B porosities under in situ stress conditions are slightly lower and inferred levels of overconsolidation are slightly higher than quantified in this study.

Initial Porosities

In order to understand the consolidation history of the AND-1B core, it is important to consider initial porosities of the different lithologies as they are (or should be) observed at the sediment surface shortly after deposition. This information is not available from the AND-1B core because rotary drilling did not recover surface sediments (Falconer et al., 2007). However, initial porosities near the sediment surface of the AND-1B site were determined during ANDRILL pre-site survey studies or were determined elsewhere (Table 2).

In Windless Bight, only a few kilometers away from the AND-1B drill site (Figs. 1 and 3), gravity cores recovered Holocene muds <1 m thick from a grounding-line distal marine environment under the McMurdo Ice Shelf (Fig. 1; Barrett et al., 2005). These muds contain some diatom frustules and have porosities of 0.7 to >0.8 (Table 2). They provide a modern analog for interstratified mudstones and sandstones found at the top of Pleistocene glacial-interglacial sequences of motif 1 in the AND-1 core (Fig. 2) that can contain variable amounts of diatoms ranging from traces to 50% (McKay et al., 2009, 2012). As part of the same type of sequence and glacial regime, an early Holocene diamicton is described from the same site-survey location (Table 2), and is interpreted as a melt-out deposit from the basal debris zone shortly after the retreat of grounded ice as part of the regional deglaciation of the Ross Ice Sheet (or Ross Ice Shelf) since the LGM (McKay et al., 2008). This diamicton has a porosity of 0.5 and provides a modern analog for stratified (Ds) and granulated (Dg) diamictites of the Pleistocene AND-1B sequences. Both Ds and Dg diamictites are interpreted as glaciomarine deposits proximal to the grounding line (Fig. 9; McKay et al., 2012).

The initial porosity of subglacial diamicton deposited under grounded ice, the most prominent type of glacial deposits found in the AND-1B core, can only be assessed. At the present transition from the West Antarctic Ice Sheet to the Ross Ice Shelf, boreholes drilled to the bottom of ice stream B (Figs. 1 and 3) allowed sampling of modern subglacial sediments from the basal zone of the 1050-m-thick ice (Tulaczyk et al., 2001). Porosities of tills cored to a depth of 3 m below the ice cluster at ∼0.4 near the sediment surface (Table 2); very low preconsolidation stress levels were calculated, ranging from 2 to 25 kPa (Tulaczyk et al., 2001). The basal water pressure of Stream B is close to the total isostatic ice pressure, or floatation level of the ice, so that the effective stress is very low (Engelhardt et al., 1990). The situation under Stream B may provide a modern analog for AND-1B diamictites, which are massive (Dm) and interpreted as subglacial deposits (McKay et al., 2012).

There are no porosity data available for Holocene diatom ooze from the McMurdo Sound area (McKay et al., 2008) to serve as modern analog for the other lithological end member of the AND-1B core. A diatom ooze example from the top of Maud Rise exhibits high porosities (Table 2) and may be taken as an analogue of initial porosities of open-marine diatomites in the AND-1 core. The hollow configuration of diatom frustules allows the development of high-porosity sediments. Thus, high initial void ratios of sediments and low degrees of consolidation are associated with high biogenic silica content (Bryant and Rack, 1990). We have not found a suitable modern analog for the volcanic deposits of the AND-1B core.

Consolidation of the AND-1B Sediments

In the upper 600 m of the AND-1B core the intercalation of diatomites with diamictites plus the massive diamictites of motif 1 near the top makes the overburden heavier and the field consolidation stronger than, for example, the diatomaceous sediments of ODP Leg 178 sites from near the Antarctic Pennisula (Fig. 7; Volpi et al., 2003) and from several other sites around Antarctica and elsewhere, where thick subglacial diamicton units are missing (compiled as reference data; Fig. S2 [see footnote 1]). A similar effect is observed below 800 mbsf, where thick and heavy diamictite units of the AND-1B core force the field consolidation toward higher effective stress levels compared to the laboratory consolidation of mud (Fig. 7). This explains why grounding-line distal and open-marine sediments of the AND-1B core have lower porosities than most of the reference data at similar depth in the formation (Fig. S2). Here the question arises whether the low porosities determined in the AND-1B core are due to the specific field consolidation or a reflection of overconsolidation in addition to that of rock overburden. Overconsolidation occurs when the maximum past overburden stress has been larger than the modern in situ stress (Bryant et al., 1981).

Overconsolidation may be identified if void ratios are plotted as a function of effective stress. Together with the initial porosities under the Ross Ice Shelf (Table 2), the porosity-stress data of the AND-1B sediments are compared to consolidation trends of different marine sediments (Fig. 8). The data from volcanic sediments (Fig. 7) are not included in this comparison, because no initial porosities are available, no suitable reference data exist about consolidation of volcanic deposits comparable to the volcanic debris of the Victoria Land Basin, and the volcanic material has been altered by postdepositional processes.

For normal consolidation and void ratios as a logarithmic function of effective stress different sediments compact with increasing overburden pressure along characteristic lines (virgin consolidation; e.g., O’Regan et al., 2010a, 2012b; Tulaczyk et al., 2001). Virgin consolidation is referred to as the state of loosest possible compaction at a given level of effective stress (Holtz and Kovacs, 1981). From consolidation experiments originally carried out by Hamilton (1976), it was demonstrated that regardless of the original composition of the sediments and the initial porosity, these compaction lines or curves appear to be converging to a common cluster of void ratios between 0.3 and 0.5 at a pressure of ∼10,000 kPa (Bryant et al., 1981). Exceptions are diatomites and plastic clays, which compact from their initial porosities toward higher void ratios of ∼1 at ∼10,000 kPa (Hamilton, 1976). We have used three different data sets to compare the normal consolidation trends of different marine sediment types to the observed consolidation of AND-1B (Fig. 8).

  1. From laboratory consolidation of various Deep Sea Drilling Project data, Hamilton (1976) established sediment-specific profiles of the reduction of porosity and increase of density under overburden pressure in the seafloor. For pelagic clay, terrigenous sediments and diatom ooze depth gradients of density and porosity were calculated (Hamilton, 1976; see Supplemental File [footnote 1] for more details: Algorithms used to express and calculate normal consolidation for void ratios as a function of effective stress—effective stress). These data were converted into effective stress and void ratios using Equations 4–6 and 3, respectively (Fig. 8; see Supplemental File [footnote 1] for more details: Algorithms used to express and calculate normal consolidation for void ratios as a function of effective stress—effective stress, void ratios). The data for pelagic clay are only valid to 300 mbsf where the clay merges the trend of terrigenous sediments valid to 1000 mbsf (Fig. 8). The specific data for diatom ooze are valid to 500 mbsf (Hamilton, 1976).

  2. Consolidation experiments of diatomaceous sediments from the Weddell Sea (ODP Leg 113) cover the stress range from 1 to 3000 kPa (Bryant and Rack, 1990) and can be extrapolated along the straight line of virgin consolidation from 1300 to 4500 kPa in order to be comparable to the AND-1B stress data (Fig. 8). Although having different initial porosities, both consolidation curves of diatom ooze (A and B, Fig. 8) exhibit similar trends below 200 kPa (Fig. 8). This suggests a general validity for normal consolidation of sediments with very high diatom content (>90%).

  3. Due to the broad range of different types and depositional conditions of diamictites (and different initial void ratios), it is difficult to reproduce a general trend for normal consolidation of diamictites. As a simple approach we use the initial porosities of Holocene diamicton deposited under the floating Ross Ice Shelf near the McMurdo Ice Shelf site (Table 2; Fig. 8), and assume virgin consolidation along a straight line approaching a void ratio of ∼0.2 at 105 kPa (Fig. 8).

Plots of the different initial porosities of surface sediment near the AND-1B site suggest that normal consolidation with increasing normal overburden pressure would follow consolidation lines close to A–D (Fig. 8). Thus, for the AND-1B core not affected by additional stress under grounded ice, stress and void ratios of diatomites, mudstones, and diamictites should plot close to the lines of diatom ooze, clay and/or terrigenous sediments, and diamicton, respectively (Fig. 8). At stress levels to 3000 kPa most of the void ratios of diatomites, mudstone, and other sediments (mostly sandstones) clearly exhibit void ratios well below the trends of normal consolidation. This suggests that significant proportions of the AND-1B strata are overconsolidated. If sediments were subject to preconsolidation by an additional temporary load of overburden, this would result in lower void ratios than expected from normal consolidation, because sediments will not fully re-expand and void ratios will not return to their original level once the additional overburden pressure is reduced (Holtz and Kovacs, 1981).

Using algorithms of specific normal consolidation trends (B, C, D, E; Fig. 8), hypothetical void ratios were calculated from effective stress data of AND-1B diatomites, mudstone, other lithologies (mostly sandstones), and diamicton (see Supplemental File [footnote 1] for algorithms and more details: Algorithms used to express and calculate normal consolidation for void ratios as a function of effective stress—hypothetical void ratios of AND-1B for normal consideration). These hypothetical void ratios were subtracted from void ratios measured in the AND-1B core, expressed as residual void ratios and plotted versus core depth (Fig. 10). Negative residuals (Fig. 10) represent porosities, which are lower than would be expected for normal consolidation of different lithologies and vice versa (Fig. 8). The depth trend of residual void ratios exhibits two main features (Fig. 10).

1. In the upper 600 m individual lithologies show relatively large scatter in the data, more pronounced in shales and/or siltstone and most pronounced in diatomites. For diatomites this may be caused partly by different opal and clast contents (Fig. 6), which should have an influence on consolidation. In this respect the approach used here is too simple, calculating void ratios for all diatomites along one single line of consolidation. The same may be true for shale and siltstones, the consolidation of which is possibly affected by differences in initial grain size (Bryant et al., 1981). Thus, the scatter may largely represent the general deviation from a single consolidation line due to differences in composition. This deviation becomes smaller below 700 mbsf (∼6000 kPa, Figs. 10 and 9), because consolidation is converging to a small range of void ratios at 105 kPa regardless of composition and initial porosities (Fig. 8; Bryant et al., 1981).

2. Above 400 mbsf, and within the glacial motif 2 (McKay et al., 2012), there is a clear increase to negative residuals in diatomites, mudstone, and other lithologies (mostly sandstones) superimposed on the scatter in the data. More pronounced negative residuals characterize the top 80 m of the record interpreted as glacial motif 1. This suggests that the diatomites and mudstone interpreted as interglacial deposits (McKay et al., 2009, 2012) were affected by significant preconsolidation. In the uppermost 80 m of the core some preconsolidation is also indicated for diamictites. The weak trend toward positive residuals in diamictites from 500 to 100 mbsf is probably due to an overestimation of the normal consolidation trend caused by the rather low initial porosity (E in Fig. 8). Below 400 mbsf and at effective stress levels higher than 3000 kPa, preconsolidation cannot be interpreted from the data presented in this paper (Figs. 8 and 10). Above this stress level the void ratios of AND-1B agree with normal consolidation trends and preconsolidation may be overprinted by consolidation due to actual overburden stress.

Overconsolidation in sediments affected by ice grounding can arise from a number of different processes, including subglacial freezing (Christoffersen and Tulaczyk, 2003a, 2003b), shear-induced failure under drained conditions, one-dimensional time-dependent compaction, and glacial erosion (O’Regan et al., 2010a). The effects of these different processes are difficult to quantify for the AND-1B core because many variables involved are unknown (such as shear strength) and data from rock consolidation tests are needed. In particular the question of the thickness of eroded sediments can only be assessed because the large unconformities found in the core could represent times of nondeposition and/or erosion (Wilson et al., 2007; Naish et al., 2007). In the upper 700 m of the core, major unconformities are at 96, 151.3, 438.6, and 596.35 mbsf and comprise ∼22% of the total time corresponding to the depth of 700 mbsf (6.8 Ma; Levy et al., 2012). None of these unconformities exhibit large shifts in porosity trends (Fig. 5). This implies that preconsolidation and erosion of overburden may have occurred, but the imprint on porosity is not seen because the eroded material represented a smaller overburden than what exists today. The same effect on porosity can result from nondeposition and normal consolidation across the unconformities. For example, in the upper 700 m of the core the largest unconformity at 438.61 mbsf represents ∼820 k.y. (Cody et al., 2012). An overburden of 300–400 m of sediments would have been eroded at or before 3.596 Ma, assuming that sedimentation was continuous with rates similar to strata above and below the GSE (Levy et al., 2012). Because this is less than the current overburden, it is not possible in this case to distinguish erosion from nondeposition in the porosity record. However, the fact that even at shallow depth in the core the GSEs are not characterized by significant downward increases in porosity suggests relatively small rates of erosion, consistent with the conclusions drawn from sediment descriptions (McKay et al., 2009, 2012). In any case, removal of overburden significantly thicker than modern would be needed to explain the observed low porosities by erosion of older strata. The AND-1B age model does not indicate such rates of sediment removal, so erosion cannot explain the overconsolidated character of the sediments.

At the base of polar ice sheets subglacial water and debris can be accreted by basal freeze-on (Christoffersen and Tulaczyk, 2003a, 2003b). The process induces changes in physical properties of subglacial deposits. In clay-rich till the pore space may be too small for in situ growth of ice crystals, so pore water is extracted out of the till by a freezing-induced thermo-osmotic process (Christoffersen and Tulaczyk, 2003a, and references therein). According to results of numerical modeling, this leads to depression of pore-water pressure, which causes an increase of subglacial effective stress and consolidation. As a result, till strength is increased and porosity can be reduced from 40% (void ratio 0.66) to <25% (void ratio 0.33) in cases of prolonged periods of subglacial freezing (>200 yr; Christoffersen and Tulaczyk, 2003b). An actualistic example of basal freeze-on was discussed for the West Antarctic Ice Stream C, where a debris-bearing basal zone ∼20 m thick was observed via borehole video camera (Christoffersen and Tulaczyk, 2003b).

Because the intensity of AND-1B preconsolidation and the processes involved appear to be different for the Pleistocene and Pliocene sequences (motifs 1 and 2, respectively; Fig. 10), specific effects are discussed separately. This discussion is restricted to the upper 600 m of the core because that is the only glacial-derived deviation from normal consolidation indicated by the data presented here.

Consolidation of the Pleistocene Sequences (Motif 1)

For the Pleistocene sequences of motif 1 (Fig. 2A), it has been argued that removal of sediments at the GSE was minimal (McKay et al., 2009). McKay et al. (2012, p. 106) interpreted the degree of erosion from sedimentary evidence: “Sequences 2 to 6 (Fig. 9) are all underlain by stratified or granulated diamictites (facies Ds and Dg) that pass upward into massive diamictites (Dm), and overlie mudstones (Facies M, Mc) of the underlying sequences. That shearing by glacial overriding was not significant enough to homogenize these proglacial ice sheet advance facies suggests that overriding of the ice sheet was relatively nonerosive in these sequences, and peak interglacial conditions appear to be preserved in the upper 60 m of the AND-1B core.” From 40 to 60 mbsf several interglacial mudstones have very low void ratios, from 0.2 to 1.4, and distinctly negative residual void ratios, suggesting significant overconsolidation (Fig. 9). These mudstones are interpreted as grounding-line distal sub–ice-shelf deposits correlated to Marine Isotope Stages 9–15 (McKay et al., 2012). Therefore, grounding of ice and overconsolidation in the depositional environment can be excluded. Overconsolidation as a result of postdepositional removal of overburden sediments seems to be negligible for these interglacial deposits. Thus, the low void ratios of the Pleistocene mudstones are interpreted as a result of succeeding glaciations when ice grounded at the AND-1B location associated with glacial processes reducing pore fluids in subglacial deposits. In this context it is interesting to note that the different diamictite facies (Ds, Dg, Dm; McKay et al. 2012) intercalating with the mudstones are not reflected by porosity or void ratios (Fig. 9). This suggests that the different proglacial and subglacial sedimentary environments did not leave an imprint on diamicton porosity or, more likely, the original signature is lost by overcompaction under grounded ice of subsequent glaciations. Processes leading to overcompaction can include sediment sheering, basal freezing, and vertical loading by ice sheets during glacial times.

Sedimentary evidence of shearing as a result of motion of grounded ice (McKay et al., 2012) implies that large shear stresses were transmitted to the underlying sediments of AND-1B. Under these conditions, porosity loss can arise from contractive failure of the sediments and cannot be attributed to simple vertical loading (O’Regan et al., 2010a). However, glacial deformation of sediments as the dominant control of mechanical failure has not affected the entire thickness of an individual motif 1 sequence in AND-1B. The glacial deformation associated with each sequence boundary usually occurs over a zone of 1–6 m, which in most cases represents a continuum from undeformed glaciomarine and/or marine sediments, up into a glaciomarine and/or marine facies that has increasing degrees of postdepositional disturbance upsection, before passing into the highly sheared and homogenized subglacial till (McKay et al., 2012). This suggests that low porosity/void ratios caused by subglacial sediment deformation are largely restricted to the diamictite units and cannot explain the negative void ratio residuals and consolidation of undeformed glaciomarine mudstones (Fig. 9).

It has been argued that freeze-on driven mechanism of till consolidation may be responsible for strongly consolidated till layers present at the bottom of the Ross Sea, over which the West Antarctic Ice Sheet advanced during the LGM (Anderson, 1999, p. 72; Christoffersen and Tulaczyk, 2003b). The same ice sheet has covered the location of the AND-1 core from where LGM till has not been recovered (Dunbar et al., 2007). However, strongly consolidated AND-1B diamictites of previous Pleistocene glacial ice-sheet advances were deposited in a subglacial environment similar to the LGM.

A “cold ice” scenario, i.e., a vertical temperature profile entirely below freezing with negligible surface melting and subglacial outwash, is consistent with the motif 1 glacial regime interpreted from sedimentological evidence, and is suggested to explain the lack of meltwater deposits (McKay et al., 2009). Thus, it is possible to interpret the overall very low porosities and void ratios of the Pleistocene sequences 2–9 (motif 1; McKay et al., 2012) by very strong compaction under a cold ice sheet with conditions comparable to the LGM. This may include some reduction of pore water due to basal freezing.

Interpretations of enhanced consolidation and stiffness of sediments due to subglacial freeze-on are restricted to tills (Christoffersen and Tulaczyk, 2003a, 2003b). This includes arctic examples from the last glacial advance of the Fennoscandian Ice Sheet in Denmark (Christoffersen and Tulaczyk, 2003a) and offshore Norway ∼300 m below sea level (Sættem et al., 1996). The Fennoscandian examples from Denmark and Norway have in common that the entire till sequence is not affected by basal freeze-on (Sættem et al., 1996; Christoffersen and Tulaczyk, 2003a); in both cases the upper strongly consolidated till is overlying a weaker and softer till. This is interpreted as a result of basal freezing associated with late Pleistocene deglaciation. A switch from basal melting to basal freezing can be triggered by several mechanisms, one of which is thinning of the ice cover during deglaciation. At increased rates of ablation and thinning only the top part of the till, which is facing the freezing ice base, undergoes high levels of effective stress. This leads to uneven levels of effective stress (Christoffersen and Tulaczyk, 2003a). Till profiles are characterized by an upward increase in hardness associated with a decrease in porosity toward the end of a glacial cycle. In the AND-1B core, glacial cycles 5 and 6 may be such examples, as porosity/void ratios decrease from 0.42/0.8 to ∼0.2/0.4 near the tops of these cycles (Fig. 9). This magnitude of porosity change can be caused by basal freezing (Christoffersen and Tulaczyk, 2003b). If so, basal freezing would not have affected the entire diamictite units of the AND-1B core. It remains questionable whether the porosity of the interglacial mudstones has been reduced significantly by basal freeze-on as well; there is no known sedimentary evidence. Alternatively, the overconsolidation of mudstones was caused by increased levels of effective stress due to vertical loading of ice during successive glaciations, including the LGM.

During the LGM at the AND-1B site, the Ross Ice Sheet reached a thickness of ∼700 m above present sea level (Denton and Hughes, 2002; Fig. 3). With a lower LGM sea level of –120 m (Lea et al., 2002) and according to Equations 7 and 8, an additional effective stress of ∼6500 kPa would have been applied to the surface sediments under the ice. This is the effective overburden pressure caused by grounded ice having a thickness above sea level in excess of that needed for the ice to be grounded (∼720 m; Fig. 3). Compaction would work under the assumption of free pore-water drainage to the ice margin (Sættem et al., 1996). Compared to the lines of normal consolidation and depending on lithology (Fig. 8), this additional stress is capable to reduce the initial void ratio of mud from 4 to ∼0.5. This level of preconsolidation stress is sufficient to explain most of the low void ratios (negative residuals) of both interglacial mudstones and glacial diamictites of motif 1 sequences.

Changes in Consolidation at the Mid-Pleistocene Transition

For both mudstones and diamictites of ANT-1B, the inverse upward porosity gradient between 80 and 90 mbsf (transition between motifs 1 and 2; Fig. 2) can largely be explained by a change from wet-based to dry-based ice sheets. The change in the glacial regime (McKay et al., 2012) is associated with a significant rise in preconsolidation stress levels during the middle Pleistocene transition (1–0.8 Ma). This rise may be related to differences in ice-sheet thickness above sea level and basal water pressure under warmer than present subpolar (motif 2) to polar style of glaciations with limited amounts of subglacial meltwater (motif 1), as concluded from the sedimentological evidence of the AND-1B core (McKay et al., 2009, 2012). We interpret the increase in preconsolidation stress during the middle Pleistocene transition as the result of larger overburden pressure caused by thicker Pleistocene ice sheets, possibly enhanced by effects of basal freeze-on.

The preceding interpretation also implies that the basal stress of Pleistocene ice sheets covering the drill site during the past 1 m.y. was not (or at least not fully) transmitted into older strata deeper in the formation, as indicated by older strata retaining higher levels of porosity than the top of the formation younger than 1 Ma. It is possible that the time of peak glaciations and maximum basal stress affecting subglacial deposits was relatively short, and not all excess pore pressure dissipated. This might have been encouraged by early diagenetic micritic beds reducing formation permeability, as discussed herein. Thus, it is suggested that most of the upper 600 m of AND-1B strata affected by subglacial preconsolidation still carry some evidence of the individual glaciations that caused the GSE, rather than having a dominant final overprint by subglacial stress of the last glaciation, in this case the LGM. This idea is supported by the fact that we find a good correlation of different porosity levels in the AND-1B core to different glacial regimes superimposed on the general consolidation trend due to overburden pressure (Fig. 7).

Consolidation of the Pliocene Diatomite Intervals

In the consolidation of the Pliocene sequences (motif 2), the diatomites are important because the units DU-II to DU-XI have negative residual void ratios (Fig. 10), suggesting some levels of preconsolidation. In contrast, motif 2 diamictites do not deviate much from normal consolidation trends in the data presented here. The fact that diamictite residual void ratios are mostly positive from 400 to 80 mbsf (Fig. 10) should not be interpreted as an indication of underconsolidation, because it is more likely that the consolidation of diamicts is underestimated by our approach and any preconsolidation of these diamictites cannot be resolved without consolidation tests. Therefore, the physical properties and consolidation of diamictites of motif 2 are not interpreted further herein. Nonetheless, for a better understanding of the consolidation of motif 2 deposits in general, the physical character of the Pliocene interglacial diatomites of the upper 600 m of the AND-1B core can be viewed in more detail, as they were subjected to glacial loading at each GSE (Fig. 2). Conceptually this is similar to the situation of the Pleistocene mudstones as discussed herein.

One aspect all diatomite records have in common is the fact that they retain porosity levels with increasing burial depth below seafloor better than other sediments in deep-sea records. This also implies that the diatom strata can resist increases of effective stress during their consolidation better than all other lithologies (Bryant et al., 1981). For the AND-1B diatomite intervals, this is evident from the plot of void ratios versus depth below seafloor (Fig. 7). This specific physical signature of diatom ooze and sediment enriched in biogenic silica has often led to the conclusion that the deposits are underconsolidated. Bryant et al. (1981), for example, reported on seemingly underconsolidated diatomaceous oozes from the Japan Trench that had structural strength to resist the process of consolidation. Pittenger et al. (1989) described sediments from the Voring Plateau in the Norwegian Sea with high porosities and low degrees of consolidation that they associated with high biogenic silica contents. Volpi et al. (2003) found anomalous consolidation trends in sediment cores from the Antarctic Peninsula, i.e., undercompacted sediments at a depth of 600 mbsf (Fig. 7), that they related to the presence of biogenic silica. Here we use the comparison of void ratios as a function of effective stress for normal consolidated diatom ooze to argue that the AND-1B diatom units are significantly overconsolidated. As discussed herein, this can be caused by glacial loading and/or sediment shearing. Subglacial freeze-on processes are not considered because the sediment description is indicative of the presence of meltwater at the base of motif 2 ice sheets (McKay et al., 2009).

In order to quantitatively assess patterns of diatom comminution resulting from compaction and progressive shear stress, Scherer et al. (2004) subjected diatomaceous sediment to laboratory stresses to 85 kPa, comparable to those beneath the modern Ross ice streams (Engelhardt et al., 1990). Progressive but unequal changes in absolute and relative diatom abundance occurred with compaction and shearing, with distinctive patterns of diatom valve comminution. The results can be used to distinguish glacially sheared sediments from undisturbed hemipelagic sediments, and from sediments fragmented by normal stress compaction (Scherer et al., 2004, 2005). The study also demonstrated that subglacial diatom fragmentation can occur at relatively low levels of effective vertical stress compared to effective stress levels calculated for diatomite units in the AND-1B core (Fig. 7). Analysis of AND-1B core reveals that all diatomite units are characterized by a rich abundance of very highly fragmented diatoms (Scherer et al., 2007). Fragmentation is likely due to compaction by sedimentary and glacial overburden. After a detailed study of DU-XI, Konfirst and Scherer (2012) concluded that physical interaction of frustules with sand grains during consolidation is responsible for diatom fragmentation; they found no evidence for fragmentation as the result of glacially induced shearing. Field-consolidation data from the AND-1B core exhibit effective stresses ranging from ∼1300 to 4500 kPa for DU-IV to DU-XIII, respectively (diatomites with >35% opal; Figs. 6 and 7). Thus, dominated by overlying diamictite units, effective overburden pressure alone (Fig. 7) could probably explain sufficiently the fragmentation of AND-1B diatoms; however, it cannot explain the negative residual void ratios of DU-II to DU-XI (Fig. 10).

According to consolidation tests on near-surface marine diatom ooze (Bryant and Rack, 1990, consolidation curve B in Fig. 8) the initial porosity/void ratio of 0.85/5.4 is reduced to 0.73/2.7 after being subjected to effective stress of 1300 kPa (ODP Leg 113, 690C, 7.5 mbsf, 88% diatom content). The ratio of diatom content to biogenic opal is ∼2:1 (Volpi et al., 2003), which makes the sample from ODP Site 690C comparable to, for example, DU-IV (43% opal). The range of porosity/void ratios in DU-IV clusters at ∼0.6/1.5 (Figs. 6 and 7); this suggests further compaction of DU-IV in excess of that caused by effective overburden pressure (1300 kPa) to a level of ∼3000 kPa. At this level the consolidation line B corresponds to a porosity/void ratio of 0.64/1.8 (Fig. 8). According to the consolidation trend B, a porosity/void ratio of 0.56/1.3 is to be expected at the base of DU-XII (4500 kPa; Figs. 6 and 7). This is about the same level as measured in the core (∼0.55/1.22). Therefore, the consolidation of DU-XII can be explained by effective overburden pressure alone. However, we cannot exclude that the same order of stress (or some proportion) has been the result of preconsolidation. As a working hypothesis we suggest that the opal-rich diatomites of AND-1B (DU-IV–DU-XIII) have been subjected to preconsolidation by grounded ice on the order of 3000 kPa vertical stress, probably every time after a GSE formed. This would explain the negative residual void ratios of DU-IV–DU-IX (Fig. 10). It would also explain why the porosities from DU-IV–DU-XI are quite similar regardless of the increase of field consolidation stress from 1300 to 3000 kPa in the formation. The slight decrease of porosities in DU-XII and DU-XIII may be due to field consolidation in excess of 3000 kPa (Figs. 6 and 8). In order to transform effective stress of 3000 kPa to the seafloor, a grounded marine ice sheet must have had a thickness above sea level of 330 m in excess of that needed for the ice to be grounded (Equations 7 and 8). This thickness is significantly less than reconstructed for the LGM and probably other cold stages after the middle Pleistocene transition.

  1. Only a few volcanic sections of the AND-1B core exhibit grain densities >2.7 g/cm3. Significant parts of the core have grain densities <<2.7 g/cm3, as low as 2.14 g/cm3, mainly for two reasons: accumulation of marine biogenic silica, which occurs as primary deposits in diatomites or has been reworked into diamictites by subglacial processes, and seawater-induced alteration of volcanic glass (palagonitization). The accumulation of marine biogenic silica is only significant above 600 mbsf, and palagonitization is only significant below 600 mbsf. For the calculation of porosity from multisensor-track density, grain densities were interpolated from discrete samples to avoid errors of –5% to +15% that arise when grain density is assumed to be 2.7 g/cm3 (a common practice for quantification of GRAPE porosity).

  2. Cementation by disseminated carbonate and pyrite (below 400 mbsf) is mainly responsible for lithification of the core below 50 mbsf. Cement has only subtle effects on porosity and consolidation trends, but its framework is interpreted to suppress porosity rebound after removal from in situ stress conditions. Cement concentrated locally in micritic beds can reduce fractural porosity by as much as 0.4. Cemented micritic beds are not significant in number and total thickness compared to the rest of the core. There is no evidence that diagenesis plays a dominant role in the consolidation history of the AND-1B core, certainly not in the upper 600 mbsf, where no porosity offset is obvious between soft sediments and lithified core.

  3. When compared to normal consolidation trends of different marine sediments, interglacial deposits (mudstones, diamictites) of the upper 400 m of the AND-1B core exhibit significantly lower void ratios than suggested by the effective stress level they were recovered from (hypothetical void ratios). The inferred overconsolidation can be explained by grounding of ice sheets during subsequent glaciations. Although some diamictites exhibit abnormal low porosities, the indication of preconsolidation is limited using void ratios as a function of effective stress. Subglacial processes that induce preconsolidation include glacial loading (vertical stress), basal shearing, and probably basal freeze-on.

  4. The downcore variation of porosity and consolidation correlates with different glacial regimes. In particular, the significantly lower porosities in the upper 80 mbsf (Pleistocene, motif 1) can only be explained if the stresses imparted by Pleistocene glacial ice sheets below the drill site were not transmitted into underlying strata of higher porosities (Pliocene–late Pleistocene, motif 2). The same is suggested for the motif 2 glaciations that imparted subglacial stresses to underlying diatomite units but not to the base of the sedimentary column. Thus, it is suggested that the record of void ratios compared to hypothetical void ratios calculated from effective stress of compressed (but otherwise undisturbed) interglacial deposits can be used to assess preconsolidation during the subsequent glaciations (ice thickness), at least semiquantitatively.

  5. Ice sheets of the same order of thickness as that of the LGM (∼700–800 m above present sea level) can explain the consolidation of sediments of the AND-1B sequences 2–9 deposited after the middle Pleistocene transition (1–0.8 Ma; McKay et al., 2012). In order to explain the low porosity/void ratio of Pliocene AND-1B diatomite units by preconsolidation, the ice thickness might have been 300–350 m above present sea level.

  6. For Miocene glacial regimes there is only a small difference notable in the porosity record between sediments deposited during cold polar types of glaciations compared to warmer subpolar types of glaciations due to larger depth and advanced alteration of the core (>600 mbsf). In general, the ability to make more solid conclusions on preconsolidation in glacial environments requires dedicated sampling and laboratory-based geotechnical analysis of sediments in future coring efforts.

The ANDRILL (Antarctic Drilling) Project is a multinational collaboration between the Antarctic programs of Germany, Italy, New Zealand, and the United States. We acknowledge the ANDRILL MIS (McMurdo Ice Shelf) Science Team as well as cochiefs Tim Naish and Ross Powell for their contribution to the successful drilling project, and all technicians and research assistants for their laboratory help. This study was supported by the budget of the Alfred Wegener Institute (AWI) and partly supported by the Bremen International Graduate School for Marine Sciences (GLOMAR) funded by the German Research Foundation (DFG) within the frame of the Excellence Initiative by the German federal and state governments and by DFG grant KU 683/8. The bathymetric map base for Figure 1 is courtesy of N. Ott and D. Graffe (AWI). Johann Philipp Klages compiled data for, and prepared an earlier version of, Figure 3, and Olivia Thomas kindly improved the phrasing of the original manuscript. We thank Matt O’Regan and two anonymous reviewers for fruitful comments and suggestions that improved the manuscript significantly.

1Supplemental File. PDF file of 2 supplemental tables and 2 supplemental figures that include information on cemented layers, details on methodology of grain-density interpolation, algorithms used to express and calculate normal consolidation for void ratios as a function of effective stress, and a comparison of AND-1B diatomite fractional porosities with data from diatomaceous intervals in other marine sediment cores. If you are viewing the PDF of this paper or reading it offline, please visit http://dx.doi.org/10.1130/GES00704.S1 or the full-text article on www.gsapubs.org to view the Supplemental File.