A Stratigraphical Basis for the Last Glacial Maximum (LGM)

A stratigraphical basis for the Last Glacial Maximum (LGM)

Philip D. Hughes a, Philip L. Gibbard b

aGeography, School of Environment, Education and Development, The University of Manchester, Oxford Road, Manchester M13 9PL, United Kingdom

bCambridge Quaternary, Department of Geography, University of Cambridge, Downing Place, Cambridge CB2 3EN, United Kingdom

ABSTRACT

The Last Glacial Maximum (LGM) is widely used to refer to the episodewhen global ice volume last reached its maximum and associated sea levels were at their lowest. However, the boundaries of the interval are ill-defined and the term and acronym have no formal stratigraphical basis. This is despite a previous proposalto define it as a chronozone in the marine records on the basis of oxygen isotopes and sea levels, spanning the interval 23-19 or 24-18 ka and centred on 21 ka. In terrestrial records the LGM is poorly represented since many sequences show a diachronous response to global climate changes during the last glacial cycle. For example, glaciers and ice sheets reached their maximum extents at widely differing times in different places. In fact, most terrestrial records display spatial variation in response to global climate fluctuations, and changes recorded on land are often diachronous, asynchronous or both, leading to difficulties in global correlation. However, variations in the global hydrological system during glacial cycles are recorded by atmospheric dust flux and this provides a signal of terrestrial changes. Whilst regional dust accumulation is recorded in loess deposits, global dust flux is best recorded in high-resolution polar ice-core records, providing an opportunity to define the LGM on land and establish a clear stratigraphical basis forits definition. On this basis, one option is to definethe global LGM as an event between the top (end) of Greenland Interstadial 3 and the base (onset) of Greenland Interstadial 2, spanning the interval 27.540 to 23.340 ka (Greenland Stadial 3). This corresponds closely with the peak dust concentration in both the Greenland and Antarctic ice cores and with records of the globalsea-level minima. This suggests that this definition includes not just the coldest and driest part of the last glacial cycle but also the peak in global ice volume. The end of the LGM event is marked by Heinrich Event 2, marking the onset of the collapse of the Laurentideat c. 24 ka, along with other ice sheets in the North Atlantic region. A longer and later span for the LGM may be desirable, although defining this in chronostratigraphical terms is problematic. Whichever formal definition is chosen, this requires the contribution of the wider Quaternary community.

1. Introduction

The term ‘Last Glacial Maximum’ has attained iconic status in Quaternary Science. The often capitalised letters form the acronym LGM and suggest a formal stratigraphical status.However,the search for a common definition for the LGM is not as obvious as the acronym implies. Hughes et al. (2013) argued that the geomorphological record of glaciations doesnot provide a clear definition for the LGM since the imprint of glaciers in the geological and geomorphological records is asynchronous around the world. This is not only for mid-latitude mountain glaciers (e.g. Gillespie and Molnar, 1994) but for some of the large continental ice sheets too.The current status of the LGM is defined not by the terrestrial record of global environmental change, but by changes recorded in the ocean floor sediments' oxygen isotope sequence.

In recognition of this paradox, Mix et al. (2001) used dated corals and marine oxygen isotope sequencesto suggestthe LGM as a chronozone. However, as noted by Walker et al. (2009) when defining the base of the Holocene Series, marine records do not offer the resolution to provide a formal stratigraphical basis for defining short-duration time divisions, exemplified by the LGM. Mix et al. (2001) argued that the LGM should be centred on the radiocarbon calibrated date of 21 cal. ka BP and should span the period 23-19 ka or 24-18 ka (all further ages in this paper are calibrated or in calendar kiloyears, expressed as ka). However, the boundaries for these chronozones are not defined in a specific “body of rocks” (cf. Salvador, 1994) but are arbitrary time intervals spanning events within in a number of different records. Thus the chronozone status advocated by Mix et al. (2001) and others such as the MARGO Project (2009) can only be considered an informal labelling of the LGM interval. In addition, there is now evidence that the timings of maximum global ice volume and the lowest eustatic fall are at the older end of the interval 24-18 ka (cf. Thompson and Goldstein, 2006) withPeltier and Fairbanks (2006) suggesting that the LGM occurred as early as 26 ka.

On land, the evidence for an LGM climate signal is even more transient, leading researchers often to lean heavily on the marine records for correlation. However, a truly global LGM time division requires definition in both land-based and marine-based environmental proxies in order to provide a stratigraphical unit that can be used effectively and meaningfully in correlation. This paper provides a critique of the various Quaternary records, both marine and terrestrial. In particular, this paper seeks to establish the most appropriate record(s) and a suitable stratigraphical definition for this iconic label in the Quaternary geological succession.

2. Current definitions of the LGM

The term ‘Last Glacial Maximum’ (abbreviated to LGM) refers to the maximum in global ice volume during the last glacial cycle. The LGM was originally described by CLIMAP Project Members (1976; 1981) as spanningthe interval 23,000 to 14,000 14C ka BP, with a mid-point at 18,00014C ka BP (Shackleton 1977). It is marked bytwo independent proxies: in the marine isotope record and changes in globalsea level; and, it is on this basis that the LGM was defined(Mix et al., 2001).

The δ18O signal in the marine record is known to lag global ice volume (Mix et al., 2001; Thompson and Goldstein, 2006) and, consequently, the globalsea-level minimum is likely to be closer to the true global last glacial maximum in terms of maximum ice volume. Based on evidence of globalsea-level change from the continental margin of northern Australia, Yokoyama et al. (2000) concluded that the global land-based ice volume was at its maximum from at least 22to 19 cal. ka BP. As noted earlier,the age of 21 ka is now widely used as a time marker for the acme of the global LGM (Mix et al., 2001; MARGO project Members, 2009).

The definition of the LGM in terrestrial records depends on the criteria applied. Shakun and Carlson (2010) used 56 records to recognise a climate-defined LGM. They suggested that a global average age of 22.2 ± 4.0 ka best defines the LGM. However, they recognised that “there is considerable variation in the timing of these extreme climate states in different records with the LGM…..spread over more than 10 kyr”(Shakun and Carlson, 2010, p. 1802). Shakun and Carlson (2010) found that the LGM within 56% of their records fell within the chronozone span of 23-19 ka, defined by Mix et al. (2001), and noted that this chronozone does not appear to capture the length or variability of the LGM. In their dataset the largest frequency of climate-defined ‘LGM’ events in the northern hemisphere are at 24 ka and 30 ka (Shakun and Carlson 2010, their Fig. 4), although younger ‘LGM’ events between 23-16 ka result in a global average close to 22 ka.

Whilst Mix et al. (2001) proposed that the LGM should be defined as a chronozone, the boundaries of such a unit, as determined from a particular type-section, remain elusive. Sea-level and ice-core evidence, which provided the basis for the LGM chronozone in Mix et al. (2001), provides broad indications of the glacial maximum event. However, the bracketing ages do not conform to the strict formal requirements of a chronozone (cf. Hedberg 1976; Salvador 1994). Furthermore, Mix et al. (2001) noted at that time that there were some inconsistencies between the different ice-core chronologies and that “some puzzles remain to be solved regarding ice-core chronologies near the LGM”. Since then, new ice-core records have been obtained and used to define events at the end of the last glacial cycle (e.g. Andersen et al. 2006; Rasmussen et al., 2006; Lowe et al., 2008; Walker et al., 2009) (Fig. 1). This paper provides a new examination of the LGM issue and considers both the chronostratigraphical and geochronological status of the interval. Further, it builds on suggestions proposed at the First International Conference on Stratigraphy (Hughes and Gibbard, 2014).

3.Defining the LGM in Marine records

3.1. Oxygen isotopes

Marine oxygen isotopes, determined from the tests of foraminifera from deep-sea floor sediments, have been viewed as a proxy for global icevolume since the 1960s (e.g. Shackleton 1967). Deep-water temperatures also play a role and this means that δ18O variability in benthic foraminifera is not entirely driven by ice volume (Shackleton 2000). Nevertheless, this effect can be accounted for (Shackleton 2000) and the basic tenet relating variations δ18O to shifts in global icevolume stills holds. Since benthic δ18O variability is driven by global icevolume, it also provides a globally synchronous record of glacioeustasy(Skinner and Shackleton, 2005). Whilst this is true when considering changes over longer timescales (>5 ka) and especially 100 ka glacial cycles (e.g. Shackleton 2000; Waelbroeck et al., 2002), the marine isotope record does not offer sufficient resolution to differentiate environmental events at millennial timescales. This is, in part,a consequence of the slow sedimentation rates in the deep oceans and especially bioturbation, which “is a virtually universal source of degradation for deep-sea records” (Shackleton, 1987, p. 183; McCave et al.,1995). However, in addition to this, there are other significant reasons why marine isotopes such as benthic δ18O cannot provide a globally correlative stratigraphical scheme for fine-resolution intervals such as the LGM (Gibbard 2014).

The oxygen isotope signal in the marine record is often assumed to be a proxy forglobal ice volume. However, it is not a straightforward as this and Shackleton (2000) highlighted that a substantial part of the 100 ka glacial climate cycle, recorded by δ18O in marine foraminiferal records, is a deep-water temperature signal and not an ice-volume signal. Skinner and Shackleton (2005) showed that fluctuations in benthic δ18O and MIS boundaries from different hydrological settings may be significantly diachronous. The use of benthic δ18O as a proxy for global ice volume as established by Shackleton (1967) begins to “break down at millennial time-scales and in particular across glacial–interglacial transitions”. Thus, for relatively short intervals such as the LGM, the marine isotope record is inappropriate for defining its span.

The timescale of the deep ocean δ18O signal has been constructed by assuming that solar forcing paces variations in δ18O variations (Hays et al. 1976) and this provides the basis for the SPECMAP timescale. Thompson and Goldstein (2006) calibrated this timescale using radiometric dating and found significant discrepancies in the orbital tuning with observed U-series ages from corals. They found that for Marine Isotope Stage (MIS) 2, SPECMAP ages were too young,up to 5.2katoo young at 17.9ka, the original date assigned to the trough in δ18O for thelast glacial cycle by Martinson et al. (1987, event 2.2. in their Fig. 18). The newly adjusted radiometric calibration of SPECMAP places the trough in δ18O at c. 23.1ka, the lowest sea levels bracketed between 23.1 and 24.6 ka, with the latter age corresponding to the greatest sea-level lowering of -132.1 m (Thompson and Goldstein, 2006).An offset also exists between the high-resolution coral-derived sea-level curve of Thompson and Goldstein (2006) and a global synthetic 'stack' of 57 marine oxygen isotope records compiled by Lisiecki and Raymo (2005) –(Fig. 2). In their 'stack', Lisiecki and Raymo (2005) correlated the interval before 22 ka with the radiocarbon-dated benthic δ18O record of Waelbroeck et al. (2001) and the interval from 22-130 ka with the high-resolution benthic δ18O record of Shackleton et al. (2000). In Lisiecki and Raymo’s (2005) LR04 'stack'the δ18O trough occursc. 5ka earlier (at 18 ka, as with SPECMAP) than the sea-level minimum of Thompson and Goldstein (2006). The lag of δ18O behind sea-level minima was noted by Mix et al. (2001, p. 637)who concluded that “the highest value of benthic foraminiferalδ18O is not precisely aligned with the lowest sea-level stand, and may even be offset in time by as much as a few thousand years”.

3.2.Corals

Whilst the marine oxygen isotope record can be used as a proxy for sea-level changes, it has low resolution. Corals can be used to date low globalsea-level stands and calibrate timescales based on marine isotope records (Thompson and Goldstein, 2006). Sites that are suited to the study of corals as a proxy for fluctuating globalsea levels are those which are located close to continental shelves, distant from areas of major global glaciation. Examples include the island of Barbados in the Caribbean Sea (Peltier and Fairbanks, 2006),the Sunda Shelf in Indonesia (Hanebuth et al., 2000), the Huon Peninsula in New Guinea (Chappell and Shackleton, 1986; Yokoyama et al., 2000), and the Bonaparte Gulf, north of Australia (Yokoyama et al., 2000). Based on evidence from Barbados, Peltier and Fairbanks (2006) argued that the global sea-level minimum occurred at 26 ka and remained low (within 5 m of the lowest stand) for 5 ka until 21 ka. This contrasts with the findings of Yokoyama et al. (2000) who concluded that the lowest sea-level stand in the Bonaparte Gulf occurred between 22 and 19 ka. Peltier (2002) argued that Yokoyama et al.’s (2000) mathematical analysis was flawed and this was reiterated byPeltier and Fairbanks (2006). As noted above, Thompson and Goldstein (2006) analysed U-series data from 11 papers and recalculated ages accounting for open-system behaviour (e.g. Thompson et al., 2003). Thompson and Goldstein (2006) found that the recalculated ages place the eustatic low stand between 23.1 and 24.6 ka. This appears to support the arguments of Peltier and Fairbanks (2006) and their idea of an LGM interval spanning the period 26-21 ka, rather than the earlier interval of 24-18 ka suggested by Mix et al. (2001).

4.Defining the LGM in terrestrial records

4.1. Glaciers

The term Last Glacial Maximum is often used to refer to the peak in global ice volume during the last glacial cycle. This is reflected in the marine oxygen isotope record and also globalsea levels recorded in corals (Mix et al., 2001). On land, Clark et al. (2009) reviewed 5704 14C, 10Be, and 3He ages that span the interval from 50 to 10 ka to constrain the timing of the LGM during MIS 2. They found that glaciers advanced between 33 and 26.5 ka and reached maximum positions between 26.5 and 20/19 ka, with rapid deglaciation occurring soon after. However, for the longer interval of the entire last glacial cycle the pattern of glacier advances, and in particular, the maximum extent of ice masses, was not synchronous in time and space. Hughes et al. (2013) noted that at high, mid- and lowlatitudes across the world, glaciers reached their maximum extent before MIS 2, in MIS 5, 4 and 3. It is well-established that mid-latitude mountain glaciers were asynchronous with the global record of ice volume (Gillespie and Molnar, 1994) but increasingly, new dating evidence has revealed that even some of the largest ice sheets were out-of-phase with the record of global ice volume determined from the marine isotope record (Hughes et al., 2013).

The East Antarctic Ice Sheet, which today represents the largest modern ice mass on Earth, retreated from its maximum position well before the global LGM during MIS 3 (Stolldorf et al., 2012) and in parts of the East Antarctic the ice-sheet thickness at the LGM was little different from that today (Mackintosh et al., 2007). In New Zealand, glaciers also retreated during MIS 3, with less extensive advances occurring in MIS 2, i.e. close to the global LGM (Putnam et al., 2013). In the Northern Hemisphere the Barents-Kara Ice Sheet, the third largest ice mass of this hemisphere, after the Laurentide and Fennoscandinavian, reached its maximum position early in the last glacial cycle, as early as 90 ka. In Asia, most glaciers reached their maximum before MIS 2, with the global LGM being recorded by a less extensive advance (e.g. Dortch et al., 2013; Owen and Dortch, 2014), or not recorded all at (Heyman et al., 2011; Stauch and Lehmkuhl, 2011).

The LaurentideIce Sheet was by far the largest ice mass on Earth during the last glacial cycle. Evidence from its eastern and southern margins support a maximum phase during MIS 2.At Martha’s Vineyard, in Massachusetts, boulders on a moraine marking the outer limit of the SE sector of the LaurentideIce Sheet yielded cosmogenic nuclide exposure ages that cluster between 22 and 25 kawith a mean age of 23.2 ± 0.5 ka(Balco et al., 2002). New production rates make these ages older andBalco and Schäfer(2006) suggestthat the ice front started to retreat from the Martha’s Vineyard moraine atc. 24 ka. These authors pointed out that this new age coincides with Heinrich Event 2, a major period of ice rafting in the North Atlantic. All later ice-front positions in New England and Connecticut indicateless extensiveglaciation (Balco et al., 2002; Balco and Schäfer2006). There is also evidence from other parts of the LaurentideIce Sheet margin which suggest an ice maximum before 23 ka, such as in Ohio (Lowell et al., 1999; Szabo et al, 2011), Illinois, (Curry et al., 2011) andNorth Dakota (Manz et al., 2005).In NW Canada, Zazula et al. (2004) used radiocarbon to date pro-glacial lake sediments and suggest that the maximum extent of the Laurentide Ice Sheet in this area occurred between 35 and 22 ka followed by a less extensivereadvance at 22-16 ka.These ages, from the outer margins of the last LaurentideIce Sheet,correspond to globalsea-level evidence compiled by Peltier and Fairbanks (2006) and Thompson and Goldstein (2006) who place the low-stands at 26 and 24.6 ka, respectively. Indeed, in their modelling of ice volume, Stokes et al. (2012) take 25 ka as the time of the LaurentideIce-Sheet maximum (Fig. 3).

The southeastern sector of the FennoscandinavianIce Sheet reached its maximum extent in MIS 2. Here, moraines have yielded a mean exposure age of 19.0 ±1.6 ka (Rinterknecht et al., 2006; using a production rate of 5.1 ± 0.3 atom/g/yr). These ages are likely to be slightly older (by up to 15%; Owen and Dortch et al., 2014) based on new production rates. Thes areas low as 3.77 atoms/g/yrfor terrestrial cosmogenic nuclides that been determined for northern Norway(e.g. Fenton et al., 2011).This is similar tothe rest of the world where production rates are now thought to be in the range of 3.7 to 4.5 atoms/g/yr(Balco et al., 2009; Putnam et al., 2010, Briner et al., 2012; Young et al., 2013). This means that the SE sector of the FennoscandinavianIce Sheet began retreating from its maximal position slightly earlier than 19 ka, possibly at c. 21 ka, though still slightly later than the LaurentideIce Sheet. The situation in the SW sector of the FennoscandinavianIce Sheet was different, and there is evidence that the southwestern sector reached its maximum extent earlier, during MIS 3 (Houmark-Nielson et al., 2011).