Aminostratigraphy and sea-level history of the Pleistocene Bridgewater Formation, Mount Gambier region, southern Australia

A geochronological framework based on amino acid racemisation (AAR) and constrained by previously reported optically stimulated luminescence (OSL) ages is presented for the evolution and paleosea-level record of the Pleistocene Bridgewater Formation of the Mount Gambier region, of southern Australia. Within the study area, the Bridgewater Formation is represented by late early Pleistocene [Marine Isotope Stage (MIS) 23 at 933 ka] to Holocene barrier shoreline successions deposited during sea-level highstands. Regional monotonic uplift (0.13 mm yr–1) and pervasive calcrete development during the Pleistocene have preserved the sequence of calcarenite (mixed quartz-skeletal carbonate sand) shoreline complexes from denudation. AAR analyses confirm that the barriers generally increase in age landwards and correlate with sea-level highstands associated with interglacials as defined by the marine oxygen isotope record. AAR analyses on the benthic foraminifer Elphidium crispum have proved more reliable than the whole-rock method in extending the age range of AAR dating of these relict shoreline successions. Paleosea-levels from the coastal plain are as follows: MIS 7, –9 ± 2 m; MIS 9, 4 ± 1 m; and a minimum sea-level of 2 ± 2 m is derived for MIS 11. Paleosea-level could not be determined for MIS 15, 19 or 23 as diagnostic sea-level indicators were not identified within these sedimentary successions. Dismal Range, dated at 933 ± 145 ka (MIS 23), represents a correlative feature to the East Naracoorte Range but is some 25 km seaward of the Kanawinka Fault compared with the same barrier at Naracoorte. Mingbool Range (788 ± 18 ka) is of similar age to the West Naracoorte Range (MIS 19) and formed as an arcuate shoreline complex that became attached to the higher relief of the area represented by the Mount Burr Volcanic Province. The higher topographical relief resulted from crustal doming of the Oligo-Miocene Gambier Limestone caused by the intrusion of magma associated with the volcanic province. The AAR age of 788 ± 118 ka for Mingbool Range indicates that the Mount Burr volcanics predate the deposition of this shoreline complex.


INTRODUCTION
Amino acid racemisation (AAR) has been widely applied in geochronological studies of Quaternary coastal sedimentary successions where other methods cannot readily be used (e.g. the absence of coral suitable for U/Th dating). AAR has been commonly applied to marine molluscs and to carbonate sediments using the whole-rock technique (Hearty & O'Leary 2008). Advances in the analysis of amino acid isomers (Kaufman & Manley 1998) have enabled single grains such as individual tests of foraminifers to be analysed (Hearty et al. 2004;Kaufman et al. 2008Kaufman et al. , 2013. Foraminifera are well suited to AAR techniques as their tests approach a closed system, with the long-term retention of indigenous amino acids during diagenesis, and they can be confidently identified at the genus level, reducing further complexities in AAR dating.
This study evaluates critically the utility of AAR for the dating of individual tests of the benthic foraminifer Elphidium crispum and compares these results with the dating of sediments using the whole-rock method across the coastal barrier successions of the Mount Gambier coastal plain. These results provide a geochronological framework to constrain the evolutionary development of the Pleistocene to Holocene coastal barrier landform succession of the Bridgewater Formation and its paleosea-level record for this region.
sedimentary carbonates that were originally defined at Cape Bridgewater in western Victoria (Boutakoff 1963). The successive paleoshoreline features trend subparallel to the modern coastline ( Figure 1). The Holocene Younghusband Peninsula and associated back-barrier Coorong Lagoon (directly northwest of Robe) provide a modern analogue for the Pleistocene barriers of the Robe and Mount Gambier regions. The Mount Gambier coastal plain, mapped in detail by Hossfeld (1950) and Sprigg (1952), correlated the barrier landforms with those on the Robe coastal plain, 100 km northwest of Mount Gambier, based on the lateral continuity of these relict shoreline features. Sprigg (1952) suggested that the barriers formed during successive interglacial sea-level highstands related to Milankovitch forcing. Paleomagnetic investigation revealed the most landward barrier within the study area, the East Naracoorte Range, to be of reversed magnetic polarity (Idnurm & Cook 1980). More recent geochronological investigations have shown that the barriers on the Robe coastal plain predominantly increase in age landwards (Huntley et al. 1993(Huntley et al. , 1994Huntley & Prescott 2001;Murray-Wallace et al. 2001). Owing to the combined effects of slow (monotonic) regional uplift (0.07 mm yr À1 in the Robe region and 0.13 mm yr À1 in the Mount Gambier region; Murray-Wallace et al. 1998 and the return of interglacial sea-levels to a broadly common datum ( §6 m), some of the barriers across the Robe and Mount Gambier coastal plains are composite structures having formed in more than one interglacial. The Woakwine Range near Robe, for example, has been shown, based on morphostratigraphical evidence, to comprise three major depositional sequences (Woakwine I, II and III) (Sprigg 1952;Schwebel 1984). The ages of Woakwine I and II correlated with MIS 5e and 7, respectively, based on thermoluminescence (TL) dating (Huntley et al. 1993(Huntley et al. , 1994Murray-Wallace et al. 1999).
The barrier shorelines of the Mount Gambier coastal plain are well preserved; regional uplift removed the landforms from the influence of more recent coastal erosion and a temperateÀMediterranean climate fostered the development of regionally extensive calcrete profiles that preserved much of the original barrier landform morphology. In a global context, Australia is relatively tectonically stable at the continental scale owing to its intraplate setting and its geographically widespread stable cratonic regions. However, gentle regional uplift has occurred across the Mount Gambier coastal plain associated with Quaternary volcanism that influenced the spacing of successive interglacial barriers (Murray-Wallace et al. 1998). The genetic association of volcanism and localised regional uplift is shown by differential elevations of the upper surface of the Oligo-Miocene Gambier Limestone, which defines a broad dome near the volcanic centres and rises from present sea-level at Port MacDonnell to 60 m above present sealevel (APSL) to the east of Mount Gambier (Cook et al. 1977;Murray-Wallace et al. 1998; Figure 1).
Rates of uplift vary across the coastal plain in relation to proximity to the centre of uplift, the Mount BurrÀMount Gambier volcanic province ( Figure 1). In the Mount Gambier region the rate of uplift is 0.13 mm yr À1 , 100 km farther northwest between Robe and Naracoorte the uplift rate is approximately 0.07 mm yr À1 (Murray-Wallace et al. 1996); in the northwestern limit of the coastal plain near the River Murray mouth region, a history of subsidence is evident. These differing uplift rates result in differential spacing of the barriers across the coastal plain and their contrasting elevation above sea-level along the length of individual relict shoreline complexes. In the Mount Gambier region, barrier shorelines are approximately 7 km apart; in the Robe region they are~5 km apart; and farther north, at the mouth of the River Murray, the barrier shorelines of the Bridgewater Formation commonly coalesce owing to localised subsidence (Murray-Wallace et al. 2010).
The Oligo-Miocene Gambier Limestone underlies the Bridgewater Formation at shallow depths in the Mount Gambier region, and is also commonly expressed as a subaerially exposed, karstified marine abrasion surface in which dolines and depressions are common. Farther inland, the Bridgewater Formation unconformably overlies the Neogene Coomandook/Werrikoo Limestone. The volumetric size of the barriers varies slightly from barrier to barrier but on average extend up to 30 m above the surrounding land surface and are commonly 2 km wide in shore-normal cross-section (Murray-Wallace & Cann 2007). Barriers are primarily composed of mediumto coarse-grained skeletal carbonate sands deposited by eolian processes. Many of the cutting exposures reveal that several of the barriers have formed during a single sea-level highstand that is defined by the absence of unconformities within the barrier successions, apart from the basal unconformity separating the Gambier Limestone from the overlying Bridgewater Formation. The MacDonnell, Burleigh and Caveton ranges are examples of barriers formed during single highstand events. In contrast, Robe Range and Compton Range are composite structures that formed in more than one sea-level highstand.
The lack of surface drainage within the region accentuates the high carbonate productivity offshore on the Bonney Shelf, a temperate carbonate setting (James & Bone 2011; Murray-Wallace 2014). The resultant high calcium carbonate content of the Bridgewater Formation sediments ranges from 66 to 99 vol%. Common skeletal carbonate constituents within the sediment include echinoids, bryozoans, mollusc fragments and abundant foraminifers. Dominant foraminiferal species include: Ammonia beccarii, Discorbis dimidiatus and Elphidium crispum, the latter the focus of this study. E. crispum was selected for AAR analysis, as its multiple chambers and septal partitions render the species less prone to mechanical abrasion.
Small quarries and road cuttings within the calcarenite successions of the Bridgewater Formation in the study area revealed a range of bedforms reflecting different sedimentary facies. High-angle trough-cross beds signifying eolian dune deposition, low-angle laminar bedding of lagoonal facies and flint cobble paleobeaches, are well preserved (Figure 2). Identification of these bedforms aids recognition of former depositional environments and the definition of paleosea-level.

METHODS
In the absence of corals suitable for U-series dating, AAR was used to constrain the age of the barrier sequences. The high calcium carbonate content of the calcarenites and the abundance of fossil molluscs and foraminifers, combined with moderate current mean annual temperatures (CMAT; 13.3 C at Mount Gambier) and hence moderate diagenetic temperatures, render AAR an appropriate dating technique for the Pleistocene successions across the coastal plain. Optically stimulated luminescence (OSL) was also used to provide an independent age estimate for calcarenite (eolian facies) from Compton Range.
In a global context, few coastal regions preserve such a lengthy record of successive Quaternary interglacial highstands owing to the low preservation potential of relict shoreline successions in the geological record (Murray-Wallace & Woodroffe 2014). The location of the southern Australian coastline, within the far field of former Quaternary ice sheets, deems the sea-level record derived from this region globally significant, as inferred paleosea-levels closely approximate ice-equivalent sealevel. The elevation of the coastal barrier successions and relevant paleosea-level indicators were determined based on surveys tied into local benchmarks. Measurements were conducted using a Dumpy Level and checked using a handheld GPS device. The uncertainty in quantifying heights is approximately § 10 cm.

Sample site locations
Sediment and fossil shell samples were collected from representative exposures within each of the eight barriers across the coastal plain and the modern beach ( Figure 1). The depositional environment at each location was determined through the description of sedimentary facies architecture and paleoenvironmental reconstructions from molluscs and foraminifera present within the successions (see Appendix for stratigraphical descriptions of each sample site). Samples were collected from at least 2 m below the ground surface of each barrier to minimise the effects of diurnal and seasonal temperature changes on racemisation rates; the extent of racemisation measured in the fossil samples reflects longer-term changes in diagenetic temperature.

AAR procedures
Sediment samples analysed using the whole-rock method were sieved to the 250À500 mm fraction, sonicated in purified water for 5 min and rinsed to remove any foreign surface constituents. On average, 10 subsamples of~150 mg were analysed from each sample site and processed separately. Each subsample was etched with 2 M HCl to remove a third of its mass and any non-indigenous amino acids or degraded and mineralogically altered surface materials. Samples were rinsed in distilled water, bathed in 3% H 2 O 2 and rinsed again. Subsamples were hydrolysed in 7 M HCl for 22 h at 110 C to recover the total hydrolysable amino acids. Before hydrolysis, 50 mL aliquots of the unhydrolysed solutions One of three relict flint cobble beach facies (x) exposed within Rabbitors Road cutting through the Burleigh Range. The flint cobble units indicate a prograding coastline and similar modern features occur at the present coastline at Racecourse Bay (5 km east of Port MacDonnell). Flint concretionary lenses, exposed within the Gambier Limestone, which crops out within the region, are abraded by coastal processes to form shingle (cobble) beach deposits. Flint is not identified in coastal successions farther north on the Robe coastal plain as the Gambier Limestone does not crop out in this region. (c) Well-laminated beds of fine to very fine-grained sand within a lagoonal facies exposed in the lee of Caveton Range. Marine molluscs are common within this exposure and while well-preserved, have been transported post mortem. (d) A range of bedding patterns are identified within calcarenite at Baxter's Quarry, Mount Gambier within the Compton Range. A laterally persistent unconformity separates an upper unit of steeply bedded eolianite with dune foreset bedding, from a lower unit of subaqueously deposited calcarenite where bedding is finer and at a low angle. (e) A densely packed bed of articulated, well-preserved, mature Ostrea angasi within calcarenite at Fort O'Hare Quarry, Dismal Range, western Victoria. (f) Shallow-dipping laminar beds within calcarenite of the Gambier Range identified within Myora Forest, Mount Gambier. (g) Stratigraphical sketch of facies architecture within Baxter's Quarry, Compton Range identifying contrasting bedding patterns above (Compton I) and below (Compton II) the unconformity and paleosol, illustrating that this coastal barrier is a composite structure.
were collected for analysis of the free amino acids. Samples were then filtered to remove any undigested sediment and dried in vacuo. L-Homoargenine stock solution was used to rehydrate the samples. L-Homoargenine is used as an internal standard to quantify amino acid concentrations during chromatography. Samples were analysed using an Agilent 1100 reverse-phase high-performance liquid chromatograph (RP-HPLC) with a Hypersil BDS C18 column and auto-injector, following the methods of Kaufman & Manley (1998). Specifically, the amino acid residues underwent precolumn derivitisation using o-phthaldialdehyde (OPA) together with the chiral thiol, N-isobutyryl-L-cysteine (IBLC) to yield fluorescent diastereomeric derivatives of the chiral primary amino acids.
Analytical procedures for shell were similar to the whole-rock method. Where preserved, the umbo of bivalve shell species was preferentially analysed to avoid the effects of intrashell amino acid D/L variation on racemisation (Brigham 1983;Sejrup & Haugen 1994). Up to three subsamples from each umbo were analysed, and where possible at least three individuals from the same species were analysed at each site. After cleaning shells with a dental drill, subsamples were sonciated for 5 min and rinsed in purified water. Subsamples were weighed and followed the analytical procedure outlined for the whole-rock method.
Tests of individual Elphidium crispum were handpicked from the sediment. Mono-specific samples of Elphidium crispum were analysed to avoid potential genus effects on racemisation (King & Neville 1977;Kaufman et al. 2013). Foraminifers that appeared best preserved (i.e. had retained their overall shape and ornamentation; see Figure 5a) and characteristic peripheral keel, colour and lacked cementation or recrystallisation, were preferentially selected, as they are more likely to have retained amino acids. Tests were sonicated for 2 min and then rinsed, dried, bathed in 3% H 2 O 2 for 2 h and rinsed again. Once dry, tests were placed in sterilised, conical bottomed vials and dissolved in highpurity 6 M HCl. When an individual foraminifer was analysed, 7 mL of HCl was added to the vial, but when multiple (up to 10) foraminifers were analysed per vial (which were found to yield higher concentrations of amino acids and accordingly more reliable Gaussian D/L peaks), 10 mL of HCl was used. Samples were then capped with N 2 , hydrolysed at 110 C for 22 h, dried under a vacuum and rehydrated with a L-homoargenine solution (0.01 mM L-homoargenine C 0.01 M HCl C 0.77 mM sodium azide) at the same volume used to hydrolyse samples. Foraminifers were analysed using the chromatographic procedures previously outlined.
Kosnik & Kaufman (2008) list a range of data-screening criteria, and several of these were employed in this study to reject potentially contaminated or reworked samples. Serine (SER) racemises relatively quickly compared with other amino acids, is present in low concentrations in Pleistocene fossils and is commonly used as an indicator of sample contamination by younger amino acids. In this study, data were rejected where SER D/L values were less than 0.1 in materials of Pleistocene age, following the methods of Lachlan (2011). D/L values for the amino acids of interest that were more than 2s outside the mean value were rejected as anomalies. The covariance of aspartic acid (ASP) and glutamic acid (GLU) was also used to identify significant outliers, although with increased age (e.g. middle Pleistocene samples), aspartic acid loses the power to discriminate, and ASP D/L values may start to decrease with time, particularly for diagenetically modified materials (Kimber & Griffin 1987). In these older shells and sediments, valine (VAL) was used instead of ASP to discriminate the relative age of the deposits.
To monitor machine performance interlaboratory comparisons (ILC) samples (Wehmiller 1984) were analysed (see footer of Table 1). The standard deviation of the derived D/L values for each ILC was then derived and the coefficient of variation (CV) determined. The average of these three values yielded CV D 1.4% and represented a value of instrument reproducibility that was used as part of the final uncertainty estimates of AAR numerical ages (after Lachlan 2011).

Independently dated samples and age calibration
AAR numeric ages were derived from D/L values of Elphidium crispum by calibrating samples using OSL techniques. Specifically, AAR was calibrated with OSL data previously reported by Blakemore et al. (2014). OSL determines the time that sediment was last exposed to sunlight and thus the time of burial. The AAR method determines the time of cessation of protein formation within carbonate secreting organisms, which, for shortlived organisms, generally corresponds with death. The age derived from AAR analyses will therefore predate the sediment burial event as defined by OSL. This age difference, however, is accommodated within the uncertainties of the derived AAR numeric ages. The residence time is the time lag between carbonate formation and the production of skeletal carbonate grains and their final incorporation within sedimentary deposits. The D/L values of modern beach sediment were thus subtracted from all derived D/L values in the older wholerock sediments before numeric ages were determined. In a similar manner, D/L values of modern foraminifers were subtracted from their fossil equivalents before determination of numeric ages.
Numeric ages were derived using a parabolic racemisation kinetic model that may be applied without the knowledge of the thermal history of fossils, as long as the fossils have been subjected to a similar thermal history and remained deeply buried during diagenesis (Clarke & Murray-Wallace 2006). This kinetic model was also adopted by Murray-Wallace et al. (2001) to derive AAR numeric ages from whole-rock analyses of barriers of the Bridgewater Formation on the Robe coastal plain. Numeric ages were derived using the equation: where t represents the derived AAR numeric age, (D/L) s is the extent of racemisation at one sample site derived from the average of multiple subsamples, (D/L) t0 represents the time zero (or effectively the age of An OSL age of 124 § 10 ka from MacDonnell Range, the last interglacial barrier of the Mount Gambier coastal plain (Blakemore et al. 2014), was used to calibrate racemisation rates. In the present study, an OSL sample was also collected from the upper eolianite unit within Baxter's Quarry, Compton Range ( Figure 2) to further constrain the age of this deposit but was not used in calibrating rates of racemisation. Quartz grains of 180À212 mm were analysed from the matrix using standard purification procedures, which include an etch in 40% hydrofluoric acid for 45 min to remove external alpha-dosed rinds (Aitken 1988). OSL techniques used UV emissions from single grains of quartz and a modified SAR protocol (Murray & Wintle 2000) where 500À1000 single grains were loaded onto aluminium single-grain discs and measured in a TL-DA-20 Risø unit with a single-grain attachment (Bøtter-Jensen et al. 2000, 2003. Quartz grains were stimulated for 2 s using a 10 mW 532 nm ND:YV0 4 solid-state diode pumped green laser. Resultant UV emissions were detected by an Electron Tubes Ltd photomultiplier tube fitted with 7.5 mm of Hoya U-340 filter. Concentrations of 238 U, 235 U, 232 Th and 40 K were determined using high-resolution gamma spectrometry on dried and powdered sediment samples. Measurements were then converted to dose rates using the conversion factors of Mejdahl (1979) and the beta-dose attenuation factors of Stokes et al. (2003), allowing for estimates of long-term water contents, cosmic-ray dose rate (Prescott & Hutton 1994) and an internal dose rate of 0.03 Gy kyr À1 (Feathers & Migliorini 2001).

RESULTS
The extent of aspartic acid, glutamic acid and valine racemisation for fossil molluscs, sediments (whole-rock method) and foraminifers is presented in Tables 1À3, respectively. The result of an OSL analysis on eolianite from the upper portion of the Compton Range (Compton I) exposed in Baxter's Quarry is presented in Table 4.

Shells
Owing to successive changes in molluscan biofacies across the coastal plain, no single species of mollusc was identified as occurring within all of the barrier successions. Accordingly, it is not possible to empirically define the extent of racemisation in a single species for a complete time series across the coastal plain. The oyster Ostrea angasi was the most commonly identified mollusc species in the more inland barriers yet D/L values derived from analysed shell are lower than expected when compared with foraminifer D/L values from the same deposit. This reflects the platy carbonate structure of Ostrea shells that potentially promotes the leaching of more highly racemised amino acids bound in lower    molecular weight peptide residues from these fossils.
Ostrea has previously been identified as a poor matrix for retaining indigenous amino acids (Kimber & Griffin 1987). Several mollusc species were analysed from the lagoonal facies at Lake Hawdon South and Caveton Range (Figure 2), and reveal a significant spread of D/L values consistent with a genus effect on racemisation rate (Miller & Brigham-Grette 1989).

Whole rock
Generally, D/L values for sediment analysed using the whole-rock method increase with distance inland (and inferred barrier age). The large gastropod Turbo undulatus from the raised Holocene beach at Port MacDonnell has previously been radiocarbon dated and yielded an uncalibrated age of 2473 § 25 yr BP (Wk-34195; Blakemore et al. 2014). However, carbonate sand from this deposit yielded a significantly higher D/L value, comparable with those derived from last interglacial sensu lato eolianites of Robe Range (Figure 3). This indicates that skeletal carbonate sand from the stacked Robe Range eolianites at Port MacDonnell has been reworked into the raised beach during the Holocene sea-level highstand that Lewis et al. (2013) proposed had been up to 1 m APSL. Whole-rock D/L values do not increase for the older barrier successions inland from Caveton Range, which was correlated with MIS 9 based on a thermoluminescence (TL) age of 320 § 22 ka (Murray-Wallace et al. 1996; Figure 3). This most likely represents in situ leaching of the skeletal carbonate grains and the preferential loss of lower-molecular-weight, but more highly racemised, amino acids including free amino acids, resulting in lower D/L values for the total hydrolysable amino acids in older sediments. Figure 3 compares D/L values of sediment with foraminifers from the same deposits in the form of a simple series of presumed increasing age, in this case successive barrier features. There is a significantly stronger trend of increasing D/L values with proposed barrier age in foraminifers than sediment analysed through the whole-rock method. The AAR numeric ages for these successions are presented later and provide a more nuanced interpretation of barrier ages.

Foraminifers
Analyses of multiple foraminifers per vial yielded good concentrations of amino acids (commonly at least 50 picomole/mg). In several samples single foraminiferal tests, amino acid concentrations were not sufficient (i.e. D and L peaks could not be confidently identified and separated from the baseline, and concentrations were less than 20 pmol/mg) and results were rejected. Importantly, where results of single foraminiferal tests were not rejected, no significant difference was identified between D/L values derived from analyses of multiple foraminifers (within the same vial) and D/L values of single foraminifers from the same sample site. Overall, glutamic acid D/L values for single and multiple foraminifers from the same site generally overlap within 1s of the mean (Figure 4). However, in two instances (Robe III and Baxter's Quarry lower unit, Compton Range II), D/L values do not overlap, and D/L values derived from multiple foraminifers are lower than those for individual tests. Where up to 10 tests were analysed in a single vial, it may be that one or several of these tests are contaminated (i.e. recrystallised or cemented; see Figure 5) or are genuinely of different ages and may have been reworked into the sediment. This may lower the overall D/L value slightly but not significantly enough that the value is rejected completely.
AAR numeric ages (Table 5) were determined from D/L values of Elphidium crispum that were deemed to be more accurate and precise than derived D/L values from fossil shell and sediment samples, because of the possible diagenetic alteration, particularly in the older barrier successions. Generally, numeric ages increase with distance of the barrier from the modern shoreline.

AAR applications
An improved correlation in terms of progressively increasing D/L values and the proposed age of the barriers shoreline successions (based on their geographical position relative to the modern coastline) is observed in fossil foraminifers than in sediment analysed using the whole-rock technique. This may be due to the a Concentrations determined from beta counter measurements of dried and powdered sediment samples. b Determined from U, Th and K concentrations measured using a portable gamma-ray spectrometer at field water content. c Time-averaged cosmic-ray dose rates (for dry samples), each assigned an uncertainty of §10%. d Field/time-averaged water contents, expressed as: (mass of water / mass of dry sample) £ 100. The latter values were used to calculate the total dose rates and OSL/TL ages. e Mean § total (1s) uncertainty, calculated as the quadratic sum of the random and systematic uncertainties. An initial dose rate of 0.03 Gy ka À1 is also included. f Paleodoses include a §2% systematic uncertainty associated with laboratory beta-source calibrations. g UV SG OSL signal measured using single grains of quartz-700 grains proceeded with 21% of the grains emitting an acceptable luminescence signal, with the De derived from a FMM. h Uncertainties at 68% confidence interval.
preferential selection of better-preserved foraminiferal tests from sediment samples, and the rejection of results where serine (SER) D/L values are low and thus indicating contamination from a younger source. Skeletal carbonate sands of the Bridgewater Formation comprise a variety of sedimentary constituents, which may racemise at different rates. The whole-rock method will accordingly average these contrasting extents of racemisation resulting in a lower precision in derived ages. Thus, the analysis of mono-specific foraminifers provides a more reliable assessment of age. Previous AAR investigations on foraminifers have concentrated on tests from marine cores where temperature variations are suggested to be minimal (Wehmiller & Hare 1971;Hearty et al. 2004;Kaufman et al. 2013). This study highlights the potential use of AAR on foraminifera from marginal marine and eolianite barrier sequences.
Barrier ages, coastal evolution and sediment reworking  Schwebel (1978Schwebel ( , 1984 but compares well with the OSL age (61 § 3.6 ka) for this unit at Robe (Banerjee et al. 2003). AAR numeric ages for Elphidium crispum within Robe II indicate protein synthesis within the foraminifers ceased at approximately 79 § 18 ka (Table 5), which does not overlap with the OSL age from this deposit (Blakemore et al. 2014). While OSL is of a higher resolution than AAR, it determines when sediment was last exposed to sunlight rather than when the skeletal carbonate grains were formed. Therefore, it is possible that the barrier was originally deposited during MIS 5a but has since been reworked by eolian processes, perhaps during a windier and more arid climate associated with MIS 3. This may account for differences between AAR and OSL ages for this deposit. No significant difference is identified in D/L values or AAR numeric ages between the two eolianite units (Robe II and III) of Robe Range at Port MacDonnell (Table 5). Schwebel (1978Schwebel ( , 1984 suggested the eolianite units were of MIS 5a and MIS 5c age, respectively, at the type locality of this barrier complex at Robe. AAR does not have the resolution to distinguish age differences at the level of interstadials within interglacials. Burleigh Range, the next barrier inland from the last interglacial (MIS 5e) MacDonnell Range, has been correlated with MIS 7 based on a TL age of 237 § 16 ka  Tables 2 and 3. (Murray-Wallace et al. 1996). D/L values derived from whole-rock samples and fossil shell fragments, from a flint shingle beach facies within this barrier, are significantly higher than MIS 7 age and indicate that sediments have been reworked from an older source. However, D/L values derived from foraminifers in the coeval lagoonal facies immediately landward of Burleigh Range exposed at Laslett Road, yield an AAR numeric age of 218 § 35 ka corresponding with MIS 7. It is therefore suggested that Burleigh Range and associated back-barrier lagoon formed during the penultimate interglacial (MIS 7) but that a significant component of the skeletal carbonate sediment was reworked from an older source. This example highlights the potential difficulty of using the wholerock method to determine the age of coastal successions. The term 'Burleigh Range' was introduced by Crocker & Cotton (1946). The geochronological results reported here confirm that the range is equivalent to Reedy Creek Range to the north, as originally suggested by Hossfeld (1950) based on regional geomorphological mapping.
An AAR numeric age of 334 § 83 ka was obtained on Elphidium crispum from back-barrier lagoon facies associated with Caveton Range. This result is in accord with a previously derived TL age of 320 § 22 ka from the same facies (Murray-Wallace et al. 1996) and confirms a correlation with MIS 9 and the West Avenue Range to the north of the Mount Burr Volcanic Province.
At Baxter's Quarry, two distinct depositional successions of calcarenite occur within Compton Range and are separated by a laterally persistent unconformity. A 15 cm thick paleosol occurs beneath the unconformity surface and has been selectively weathered creating a recess within the exposure (Figure 2). The partially eroded paleosol occurs throughout the entire 75 m long exposure. Below the unconformity, bedding is of a low angle with herringbone and swaley bedforms evident, signifying a subaqueous depositional environment. Above the unconformity surface, bedding is thicker and of a much higher angle. Eolian foreset beds are identified at several localities within this upper unit suggestive of a laterally migrating coastal dune at the time of deposition. In this study, an OSL age of 390 § 35 ka (Table 4; Figure 7) was derived from the upper, eolian-deposited unit indicating deposition during MIS 11 using a Finite Mixture Model (FMM) that implies the succession beneath the unconformity and the paleosol are significantly older. This statistical model for single-grain paleodose determination was selected, as it requires the Figure 4 Extent of racemisation (total hydrolysable amino acids) within single and multiple tests of the foraminifer Elphidium crispum within Bridgewater Formation successions on the Mount Gambier coastal plain. D/L values are plotted against the sampled barrier, with barriers plotted in order of distance from the present shoreline. Uncertainties are at the 1s level. Robe II and Robe III represent samples from eolianite at Port MacDonnell and Shelly Beach, respectively. The MacDonnell Range sample is from the sand quarry at Swarts Road. The Burleigh Range sample is from the lagoonal facies at Laslett Road. Caveton Range is represented by a sample collected from lagoonal facies along Rabbitors Road. The Gambier Range sample was collected from subaqueous facies at Gooch Road. Compton II was sampled from the lower calcarenite unit exposed at Baxter's Quarry. The Mingbool Range sample is from Don's Quarry and the Dismal Range sample was collected from subaqueous facies exposed near Mingbool village. majority of grains to be bleached prior to deposition. The eolian nature of this unit supports this assumption, but the three dose populations reveal that there has been limited sediment mixing of older and younger grains. The largest dose population was chosen to represent the paleodose for this sample and the time since burial. This result most likely represents a minimum age owing to the high paleodose of the sample and the proximity to the age limit of the method. The geochronological results for Compton Range are consistent with the morphostratigraphical evidence that the barrier is a composite structure. The lower, subaqueously deposited succession exposed within Baxter's Quarry is of MIS 15 age (AAR numeric age for Elphidium crispum; Table 5) and is overlain by a younger, MIS 11 succession as revealed by the OSL age. D/L values for Elphidium crispum Compton I and II are not significantly different and thus reworking of sediment from Compton II may have occurred as Compton I was formed.
The conclusion that Gambier Range is of MIS 15 age while the upper portion of Compton Range (Compton I) is of MIS 11 age, implies that the MIS 11 transgression flooded the lee of the Gambier Range, two interglacials later, after its original formation and established a younger barrier 5 km landward of the structure. This depositional scenario contrasts with the general picture of coastal progradation of successively younger barriers across the coastal plain, but is consistent with the relative sea-level histories for the two interglacials. MIS 11 was a significantly warmer interglacial than MIS 15 (Masson-Delmotte et al. 2010) and was characterised globally by higher sea-levels possibly as much as 6À13 m APSL (Bowen 2003;Raymo & Mitrovica 2012).
An AAR numeric age of 788 § 118 ka for Mingbool Range suggests that it correlates with MIS 19 and is of a similar age to the West Naracoorte Range. Dismal Range yielded an age of 933 § 145 ka (MIS 23) and correlated  with the East Naracoorte Range at Naracoorte, previously dated at 935 § 178 ka (Murray-Wallace et al. 2001). The series of barriers across the Mount Gambier coastal plain are more closely spaced than in the transect from Beachport to Naracoorte (Figure 1). In the Naracoorte area, the strike of the East and West Naracoorte Ranges is broadly consistent with the trend of the Kanawinka Fault (Figure 1). However, the correlative features, the Dismal and Mingbool ranges, are located some 25 km seaward of the southeast extension of the Kanawinka Fault, and relate to the combined effects of regional crustal doming of the local basement (Oligo-Miocene Gambier Limestone) associated with the Pleistocene Mount Burr Volcanic Province, and the location of an inferred former estuary of the Glenelg River immediately landward of Dismal Range (Hossfeld 1950). This latter region is characterised by a higher-level plateau surface of subdued relief and barrier ranges are not discernible. In plan view, the Mingbool Range represents an arcuate-shaped barrier that at the time of formation became attached to the higher relief of the Mount Burr Volcanic Province, a feature termed the Mount Burr Peninsula by Sprigg (1952). Combined with the arcuate-shaped barrier to the northeast (Stewart Range), these barriers formed a large cuspate foreland. Importantly, the barriers indicate episodes of volcanism associated with the Mount Burr Volcanic Province, which, although difficult to date in view of the highly weathered basalt, predated the formation of Mingbool Range (i.e. > 788 § 118 ka).
In summary, while barrier successions are shown to generally increase in age landwards and can be correlated with the marine oxygen isotope record, reworking of sediment by wind and wave action is common. In view of the slow rate of uplift of the Mount Gambier coastal plain (0.13 mm yr À1 ) and return of interglacial sea-levels to a broadly common datum, not all of the barriers were uplifted beyond the influence of coastal erosion by the onset of the following interglacial and return of high sealevels. Evidence for the reworking of skeletal carbonate sand is identified in the raised Holocene beach deposit at Port MacDonnell, Robe Range, Burleigh Range and Compton Range, providing an insight into the complex coastal evolution of this region. In a sense, the coastal barriers remain dynamic features until they ultimately become 'fossilised' as relict landforms through coastal emergence and pervasive calcrete development on the surface of the landforms.
Uplift-corrected paleosea-levels AAR numeric ages, based on the apparent parabolic kinetic model applied to D/L values of Elphidium crispum from the barrier shorelines, reveal a correlation with sea-level highstands associated with Pleistocene interglacials (Table 5; Figure 6). The age of each barrier generally increases with distance inland, although some complexities are identified. The AAR numeric ages also support the suggestion by Sprigg (1952) that barriers correlate with successive interglacials, particularly for the past three interglacials, and with the barriers farther northwest on the Robe coastal plain that have also been correlated with marine oxygen isotope stages (Huntley et al. 1993(Huntley et al. , 1994Murray-Wallace et al. 2001).
Several paleosea-level indicators were identified across the Mount Gambier coastal plain, ranging from fixed sea-level indicators such as back-barrier lagoonal facies and beach facies to relational indicators such as subtidal deposits. The paleosea-level indicators are of a lower resolution than fossil corals ), but provide a first-order approximation of interglacial sea-levels. The uplift rate of the coastal plain has been previously defined at 0.13 mm yr À1 based on the elevation of last interglacial (125 ka) lagoonal facies in the lee of Woakwine Range, in the Mount Gambier region, at 18 m APSL and a value of 2 m APSL for the last interglacial sea-level (Murray-Wallace et al. 1996). The latter value was derived from several locations over a 500 km sector of western Eyre Peninsula, a geotectonically highly stable area represented by the Gawler Craton (Murray-Wallace & Belperio 1991; Murray-Wallace 2002) and, given its geographical proximity, was applied to the Mount Gambier region. Accordingly, it was estimated that the Mount Gambier coastal plain had uplifted 16 m within 125 ka, and this was used as the basis for deriving the long-term uplift rate. Paleosea-levels for the Mount Gambier region were estimated using this previously derived uplift rate and the AAR numeric ages reported in this paper (Table 5). We note that if a higher sea-level value for the last interglacial maximum were adopted, the inferred rate of uplift would be correspondingly lower.
Murray-Wallace et al. (2001) identified lagoonal facies in the lee of each successive barrier between Robe and Naracoorte, and suggested that these represent a more accurate paleosea-level indicator than beach facies, in view of higher sea-levels owing to wave set-up during storms along this high-energy coastline. The surface of the lagoonal facies more closely approximates mean sealevel and at the time of deposition was covered by about 1 § 0.5 m of lagoonal waters. Beach facies may be deposited within a wider range of tidal datum, which at Port MacDonnell is 2 m in fair conditions and significantly more during storm events, and are less well preserved in the geological record. In the Mount Gambier region lagoonal facies were not identified for each barrier. The outcropping Gambier Limestone in the Mount Gambier region represents a subaerially exposed marine abrasion surface modified by karst weathering processes. Enclosed depressions are common and the identification of lagoonal facies is more challenging particularly for the older successions where the relief is inverted by karst weathering so that small, isolated examples of lagoonal facies may occur perched on higher ground. Lagoonal facies associated with Burleigh and Caveton ranges occur at 36 and 38 m APSL, respectively. Upliftcorrected paleosea-level for MIS 7 (218 ka) is À9 § 2 m and C4 § 1 m for MIS 9 (Table 5). These results correlate well with the oxygen isotope record from marine and ice cores and other global morphological sea-level markers (e.g. Gallup et al. 1994;Vezina et al. 1999;Schellmann & Radtke 2004;EPICA Community Members 2004;Muhs et al. 2012).
Bedforms within Gambier, Compton II, Mingbool and Dismal ranges indicate deposition in a subaqueous setting. However, these sedimentary structures are not diagnostic paleosea-level indicators, and thus paleosea-level for these barriers (MIS 15, 19 and 23, respectively) cannot be determined. Compton I is an eolian unit and would have been deposited at least 1 m above mean sealevel. A minimum paleosea-level for MIS 11 is thus calculated to be C2 m § 2 m. It was not possible to define paleosea-level for Mingbool and Dismal ranges in western Victoria as uplift rates in this region, 50 km east of Mount Gambier, could not been determined. Uplift rates decrease with distance from the Mount Burr Volcanic Province but the lack of a last interglacial shoreline deposit in the Nelson area (directly seawards of Dartmoor) prevents the calculation of uplift rates for this region.

Mount Gambier coastal plain paleosea-level record within a global context
Proxy evidence from the EPICA Dome C Antarctic ice core indicates that MIS 5e and MIS 11 were the warmest interglacials during the past 800 000 years (Masson- Figure 6 Correlation of barrier successions on the Mount Gambier coastal plain, from Port MacDonnell to Dismal Range, with sea-level highstands as identified through the marine oxygen isotope record (Lisecki & Raymo 2005). The location of the transect is highlighted in Figure 1. Delmotte et al. 2010). In this investigation, a 2 m APSL value was adopted for the height attained by sea-level during the last interglacial maximum (MIS 5e), derived from studies of the Glanville Formation, a succession of shelly coastal facies from western Eyre Peninsula, South Australia (Belperio et al. 1995;Murray-Wallace 2002). Here a consistent shoreline datum of 2 m APSL based on the height of the transgressive landward feather edge back-barrier lagoon facies is noted for several sites over 500 km of coastline. This represents the most consistent value for MIS 5e successions in southern Australia. Elsewhere in southern Australia, significant differences in the transgressive limit of MIS 5e coastal successions have been noted and are dominantly an expression of differential neotectonism relating to contrasting geotectonic settings (Murray-Wallace & Belperio 1991;Murray-Wallace 2002).
We note that the value adopted for MIS 5e sea-level is lower than that described from many field areas around the world. Considerable controversy has arisen concerning the maximum height of sea-level during the last interglacial maximum since the initial compilation of sea-level observations by Veeh (1966). Veeh (1966) concluded that sea-level was approximately 6 m APSL during MIS 5e based on a range of values between 2 and 9 m APSL.
Before the Mid-Brunhes Event (MBE) ca 430 ka (Jansen et al. 1986), the transitions from glacials and interglacials are suggested to have been of reduced amplitude, as evidenced by marine oxygen isotope records (Lisieki & Raymo 2005) and ice-core records (EPICA Community Members 2004), and thus interglacials before the MBE are characterised by lower-magnitude sea-level highstands (Masson-Delmotte et al. 2010). The elevation of MIS 15 and MIS 23 coastal deposits on the Mount Gambier coastal plain indicate that sea-level may have been significantly lower than at present during these intervals, but the paucity of fixed sea-level indicators prevents a more definitive assessment for paleosea-level for these isotope stages. The significantly smaller size of barriers correlating with MIS 15 and older conforms with these less intense interglacials where cooler waters may have resulted in reduced carbonate production on the Bonney Shelf. These older successions may also have experienced some limited denudation, and thus their present sizes may not relate solely to the paleoproductivity of the Bonney Shelf during previous interglacials.
On the Mount Gambier coastal plain, the paleosealevel for MIS 9 is suggested to have been C4 § 1 m. This compares well with other coastal successions of the same age globally (e.g. Hearty & Kindler 1995;Vezina et al. 1999) and the EPICA Dome C ice core, which suggests that ice volume was similar to that of present (EPICA Community Members 2004).
The marine oxygen isotope record presented by Shackleton & Opdyke (1973) from the equatorial Pacific core V28-238 suggests that MIS 7 was significantly cooler with higher ice volumes than the Holocene interglacial, MIS 5e and MIS 9, and accordingly sea-level during the penultimate interglacial (MIS 7) may have been significantly lower than at present. Several studies of MIS 7 sea-level support this suggestion and propose that MIS 7 sea-level may have been as low as À18 m (e.g. Camoin et al. 2001;Bard et al. 2002). The EPICA Dome C ice core record derived a similar record indicating that MIS 7 was a cooler interglacial with greater ice volume (Masson-Delmotte et al. 2010). However, several studies have suggested that MIS 7 sea-level was closer to present sea-level (e.g. Vezina et al. 1999;Muhs et al. 2011Muhs et al. , 2012. Paleosea-level for MIS 7 on the Mount Gambier coastal plain was determined at À9 § 2 m APSL and is broadly in accord with relative sea-levels inferred from oxygen isotope records. The global variability of inferred paleosea-levels for MIS 7 is likely to vary systematically at a global scale and show similar spatial variation to the Holocene shoreline, but of a different magnitude. Factors contributing to this variation include proximity of some field sites to the location of Pleistocene ice sheets and the ongoing crustÀmantle adjustments following the Last Glacial Maximum, uncertainties in uplift rates that relate to the assumed value of ice-equivalent sealevel for MIS 5e, and the effect of local salinity and temperature contrasts in some marine-based oxygen isotope records for the estimation of ice volumes. The Mount Gambier coastal plain provides a globally significant record of Quaternary sea-level highstand Figure 7 Single-grain OSL equivalent dose (D e ) distribution (as a radial plot) for Baxter's Quarry, centred on CAM 255 Gy. The radial plot indicates three dose populations as highlighted by the blue, green and red lines. The paleodose for the sample of 340 Gy is the upper dose population defined by a Finite Mixture Model and was used to provide an age estimation of this deposit. events, as the region lies in the far field of former Quaternary ice sheets, and the slow uplift and favourable climate of the region has preserved the interglacial coastal successions. While oxygen isotope records provide continuous, global proxies of former ice volumes and hence indirectly paleosea-level, the Mount Gambier coastal plain preserves a punctuated but eustatically dominated record of interglacial sea-levels that may supplement these records.

CONCLUSIONS
(1) AAR analyses on the fossil benthic foraminifer Elphidium crispum yield more reliable D/L values than by the whole-rock method for the Quaternary interglacial barrier shoreline complexes from the Mount Gambier coastal plain in southern Australia. This is particularly evident for the older, more inland barrier sequences of Compton, Mingbool and Dismal Ranges. AAR numeric ages for Elphidium crispum correlate by proxy with interglacial sealevel highstand events of the marine oxygen isotope record and extend back to MIS 23. The barriers generally increase with age landwards; MacDonnell (MIS 5e) Burleigh (MIS 7), Caveton (MIS 9), Gambier (MIS 15), Compton I (MIS 11), Compton II (MIS 15), Mingbool (MIS 19) and Dismal Range (MIS 23).
(2) Sedimentary facies analysis and AAR numeric ages reveal that the evolutionary development of the coastal barrier successions within the region was complex with reworking of older sediments into younger barriers owing to the slow rate of uplift of the coastal plain and coastal erosion during a subsequent sea-level highstand for some of the barriers. Carbonate sediments and fragmented shell within Burleigh Range are significantly older than the coevally deposited back-barrier lagoon facies, suggesting that a significant component of the barrier is derived from reworked sediments. Several barriers across the coastal plain (such as Robe Range and Compton Range) are composite structures, having formed during more than one interglacial. (3) The arcuate-shaped barrier of Mingbool Range (MIS 19) signifies that during its formation the barrier became attached to the higher relief of the Mount Burr Peninsula. Accordingly, the volcanism associated with the Mount Burr Volcanic Province must predate the formation of Mingbool Range (788 § 118 ka).