Abstract
This paper presents the results of an archaeological investigation into anthropogenic earth (oven) mounds located on the Murray River floodplain at Calperum Station in the Renmark region of South Australia. Six mounds were excavated and their contents examined. Sediment analyses were also conducted to assess magnetic susceptibility, grain size and loss on ignition. Radiocarbon age estimates were obtained on shell and charcoal. Mound contents primarily included anthropogenically burnt clay (heat retainers), charcoal, fragments of mussel shell as well as very minor quantities of other faunal material and stone artefacts (which were consistent with previous lithic assessments for the region). The radiocarbon age determinations from 15 samples indicate that mounds were formed by Aboriginal people on the Calperum floodplain from at least 3981–3723 cal BP and utilised up to the time of European invasion. The very minimal amount of faunal remains (other than mussel shell), artefacts and a general lack of other material evidence apart from clay heat retainers, confirms that these features were single purpose and not used as living areas. Sediment analyses and radiocarbon dates indicate a high degree of homogeneity within mounds but provide insights into an economic transition on the Calperum floodplain, at around 4000 cal BP involving a food-production procurement strategy based on heat retainer technology and the exploitation of emergent macrophytes.
Keywords
Introduction
Earth oven cooking has been reported from both Pleistocene and Holocene contexts in Australia (Balme and Beck, 1996; Berryman and Frankel, 1984; Brockwell, 2001a, 2001b, 2005, 2006a, 2006b, Brockwell et al., 2017; Coutts et al., 1976; Frankel, 1991, 2017; Godfrey et al., 1996; Johnson, 2004; Jones, 2016; Jones et al., 2017; Ó Foghlú, 2017; Ó Foghlú et al., 2016; Pardoe and Hutton, 2021; Sullivan, 1980; Whitau et al., 2018). Aboriginal oven mounds represent a specific category of earth oven that are believed to have developed through the repeated and formalised construction of earth ovens in the same location over a period of time, rather than a more casual use and placement of cooking/heating hearths (Balme and Beck, 1996; Berryman and Frankel, 1984; Brockwell, 1996a, 1996b, 1996c, 2001a, 2001b, 2005, 2006a, 2006b; Brockwell et al., 2017; Coutts et al., 1976; Godfrey et al., 1996; Johnson, 2004; Jones, 2016; Jones et al., 2017; Klaver, 1998; Lane, 1980; Martin, 2006; Westell and Wood, 2014). These mounds share similarities with sites found in Europe and North America, often termed ‘burnt mounds’ or ‘burnt rock middens’, which have been dated from approximately 30,000 and 10,000 cal BP respectively, and up to the present in some areas (Anthony et al., 2001; Beamish and Ripper, 2000; Black and Thorns, 2014; Buckley, 1991; Gillespie, 1991; Leach, 1981; Salazar et al., 2012; Sopade, 1997; Thoms, 2008, 2015; Wandsnider, 1997). The Australian mounds, however, date to no more than 5000 cal BP (Brockwell, 2006a, 2006b; Jones, 2016; Martin, 2006; Woodroffe et al., 1988: 97–99).
In Australia, earth mounds derived from earth oven cookery have been reported from discrete provinces in the northern Adelaide plains (South Australia), the Murray Darling Basin (MDB), Arnhem Land in the Northern Territory and in Queensland’s northern Cape York Peninsula (Figure 1) (Balme and Beck, 1996; Brockwell, 1996a, 1996b, 1996c, 2001a, 2001b, 2005; Berryman and Frankel, 1984; Brockwell et al., 2017; Coutts et al., 1976, 1979; Godfrey et al., 1996; Johnson, 2004; Jones, 2016; Jones et al., 2017; Klaver, 1998; Martin, 2006; Westell and Wood, 2014). The Riverland region of South Australia is one of several distinct mound provinces in the MDB (Figure 1). Large numbers of mounds have been reported in this province (Jones et al., 2017; Westell and Wood, 2014), however, a deeper analysis of their contents, chronology and spatial relationships to other mounds and site types, both within the Riverland and in the broader MDB, has not previously been undertaken. As such, this region provides an opportunity to fill a notable gap in mound studies in the MDB (and Australia) and to consider a range of important questions relating to the Holocene emergence of this distinctive site type. In this paper, we report on the contents analysis, radiocarbon dating and sediment analyses of six mounds from Calperum Station, located within the central River Murray region upstream of Renmark (Figure 2). This work was undertaken in collaboration with the Aboriginal traditional owners of the region who are represented by the River Murray and Mallee Aboriginal Corporation (RMMAC).

Location of major earth mound provinces in northern and south-eastern Australia, showing the number of previously published dates and those obtained in this study (in brackets) from each location.

The study area showing the location of mound sites and topography within the Calperum floodplain. The relief image is based on the Digital Elevation model (DEM) 5 m Grid of Australia derived from LiDAR© Commonwealth of Australia (Geoscience Australia) 2015.
Background
The MDB is a drainage catchment covering much of south-eastern Australia. Run-off is derived primarily from the western slopes of the Great Dividing Range (GDR) with the various rivers and tributaries ultimately converging into the River Murray. In South Australia the River Murray flows through an incised valley before emptying into the Southern Ocean (Figure 2). The River Murray had existed as a seasonally fast flowing and high discharge system across the Last Glacial Maximum (LGM) before assuming its contemporary hydrological form of lower peak flows and longer flooding cycles in the early Holocene (Fitzsimmons et al., 2013; Petherick et al., 2013). Pardoe (1995: 696–701) has argued that this hydrological transition promoted greater resource productivity in the riverine corridor and supported increased population densities compared to the faster, colder and less productive LGM systems.
Australian Aboriginal earth mounds (herein referred to as mounds) typically comprise circular or elliptical, mounded cultural deposits formed by heat retainer (stone, clay or termite mound depending on availability) and a matrix of ash, charcoal and sediment derived from the breakdown of heat retainer and/or underlying soils (Brockwell, 2006a; Jones et al., 2017; Martin, 2006; Pardoe, 2003; Westell and Wood, 2014). Within more erosive environments, they are occasionally recognisable only as dense lags of heat retainer where the finer grained matrix has been removed (Jones, 2016; Pardoe, 2003). Faunal material (predominately freshwater mussel shell) and stone artefacts are common in some mound assemblages (Pardoe and Martin, 2011). In some provinces mounds also occasionally contain human burials (Beveridge, 1889: 37; Littleton, 1999; Pardoe and Grist, 2001; Stone, 1911: 434).
In Australia, mounds are recognised in the archaeological record from the mid Holocene until the near present and are considered part of an emerging repertoire of socio-economic systems that supported and/or promoted a more intensive use of natural resources, and ultimately, higher Aboriginal populations (Klaver, 1998; Martin, 2006). Within the MDB, mounds appear after 5000 cal BP and begin to proliferate around 2000 cal BP (Figure 3). This period witnessed significant climatic variation brought about by increased cycling of the El Niño Southern Oscillation (ENSO) and Indian Ocean Dipole phenomena, which together are likely to have promoted a boom-bust scenario in various habitat zones (Moy et al., 2002; Petherick et al., 2013; Rodbell et al., 1999; Williams et al., 2015a: 104). Mounds appear to be tied to a raft of innovative resource procurement, settlement and landuse strategies in both northern and south-eastern Australian contexts that emerged in response to such conditions.

Histogram (a) and cumulative plot (b) of calibrated 14C available Australian earth mound ages (Balme and Beck, 1996; Berryman and Frankel, 1984; Brockwell, 2001b; Brockwell et al., 2017; Brockwell, 2005, 2006a, 2006b; Coutts et al., 1976; Godfrey et al., 1996; Johnson, 2004; Jones, 2016; Klaver, 1998; Littleton et al., 2013; Martin, 2006; Westell and Wood, 2014). 1
No ethnohistorical records pertaining to earth mound use exists for the Calperum floodplain, however, descriptions of earth oven cooking are reported from adjacent regions. These descriptions provide detail about their construction and use for cooking plant material (both for dietary carbohydrate and fibre manufacture) and meat. Berndt and Berndt (1993: 103–107), for example, described the use of the maramin (earth oven/steaming) method for the cooking of animal and vegetable foods amongst the Ngarrindjeri people of the lower Murray River, Lakes and Coorong:
A fire was made in a scooped-out depression and stones were thrown in so that they could be heated to the extent of retaining their heat. When the fire burnt down, the stones were moved about with a wooden poker to form a fairly level base. . . Cut kinyera grass and pilbala grass was strewn over the stones to keep the meat clean and unburnt. . .The oven was then covered with yalkura or other grass, followed by skins and finally sand so that the heat of the oven would be retained. . . Shortly after cooking, a pointed stick would be thrust through the sand. . . so that it reached the bottom stone base, and was then withdrawn; water was poured into the hole to increase the intensity of the steam.
Eyre (1845, 2: 289–291, 254), writing about his observations at Moorundie, near Blanchetown (Figure 2), also described earth oven cooking in some detail:
The native oven is made by digging a circular hole in the ground, of a size corresponding to the quantity of food to be cooked. It is then lined with stones in the bottom, and a strong fire made over them, so as to heat them thoroughly, and dry the hole. As soon as the stones are judged to be sufficiently hot, the fire is removed, and a few of the stones taken, and put inside the animal to be roasted if it be a large one. A few leaves, or a handful of grass, are then sprinkled over the stones in the bottom of the oven, on which the animal is deposited, generally whole, with hot stones, which had been kept for that purpose, laid upon the top of it. It is covered with grass, or leaves, and then thickly coated over with earth, which effectually prevents the heat from escaping.
[And for]:
vegetables and some kinds of fruits, the fire is in the same way removed from the heated stones, but instead of putting on dry grass or leaves, wet grass or water weeds are spread over them. The vegetables tied up in small bundles are piled over this in the central part of the oven, wet grass being placed above them again, dry grass or weeds upon the wet, and earth overall. In putting the earth over the heap, the natives commence around the base, gradually filling it upwards. When about two-thirds are covered up all round, they force a strong sharp-pointed stick in three or four different places through the whole mass of grass weeds and vegetables, to the bottom of the oven. Upon withdrawing the stick, water is poured through the holes thus made upon the hissing stones below, the top grass is hastily closed over the apertures and the whole pile as rapidly covered up as possible to keep in the steam.
In the broader MDB, ethnohistorical accounts suggest a strong association of mounds with the large-scale exploitation of bulrush roots (Typha spp.) (Beveridge, 1883, 1889: 32–34; Gott, 1999). Mitchell (1839, 2: 53, 60, 80–81, 134), for example, noted the significance of bulrush root as a staple food for Aboriginal populations along the Murrumbidgee and Lachlan Rivers, describing the use of earth ovens for the cooking and preparation of this resource and their association with ‘lofty ash-hills’ (see also Kirby, 1895: 27–28 for a similar account in the Swan Hill region of Victoria).
Mound chronology
The dating of mound sites in Australia has relied primarily on radiocarbon ages derived from charcoal or freshwater mussel shell (Balme and Beck, 1996; Berryman and Frankel, 1984; Brockwell, 2001a, 2006a; Brockwell et al., 2017; Coutts et al., 1976; Godfrey et al., 1996; Johnson, 2004; Jones, 2016; Jones et al., 2017; Klaver, 1998; Martin, 2006; Westell and Wood, 2014). Within the MDB, mounds have a maximum age of c. 5000 cal BP within the context of a longer timeline of Aboriginal occupation in this region extending to c. 45,000 cal BP (Allen and Holdaway, 2009; Bowler et al., 2003; Fitzsimmons et al., 2014; O’Connell and Allen, 2004).
The chronological distribution of previously published calibrated 14C ages, derived from mounds in both south-eastern and northern Australia, is presented in Figure 3. Of these, 86% date to within the last 3500 cal BP and 14% between 3500 and 5000 cal BP. In addition, 78% of all ages are less than 2000 cal BP. There is compelling evidence, therefore, that mound use in Australia emerged in the mid Holocene though proliferated in the last two millennia. The ages prior to 3000 cal BP were obtained from sites in the Hay Plain, Wakool River, Adelaide River, Murrumbidgee and Weipa regions (Figure 1).
Climatic influences and Mound Function
It is generally accepted that wetter than modern conditions prevailed across south-eastern Australia through the early–mid Holocene (Bowler and Hamada, 1971; Fitzsimmons and Barrows, 2010; Fitzsimmons et al., 2013; Gingele et al., 2004, 2007; Petherick et al., 2013: 69–70; Shulmeister and Lees, 1995). This period (commonly referred to as the mid Holocene climate optimum (MHCO)) has been posited as a time of increased biomass productivity and sustained growth in Aboriginal populations (Haberle and David, 2004: 177). Such population growth, however, was curtailed by a period of extreme climatic variability and increased aridity during the period from 6000 to 4000 cal BP (Moy et al., 2002; Petherick et al., 2013; Rodbell et al., 1999; Williams et al., 2015a: 104). The emergence of mounds corresponds to this shift from the MHCO to an ENSO-influenced climate, suggesting that mounds, and the intensified use of aquatic vegetation resources more generally, may have been an economic response to more challenging climate conditions. Although this is not to suggest that other socio-economic influences did not also play a role in this transition. Furthermore, the early dates obtained on mounds in regions separated by vast distances suggest that these sites developed independently though perhaps as a similar adaptive response to these conditions. Mounds have been dated, for instance, to 5054–4390 cal BP (ANU-3992) in the Northern Territory (Woodroffe et al., 1988: 99), 3291–2822 cal BP (WK-36516) in the Weipa region (Brockwell et al., 2017) and 5319–4425 cal BP (WK-4101) in the MDB (Martin, 2006).
Haberle and David (2004: 177) have similarly argued that the processing of toxic, though carbohydrate-rich, plants in northern Queensland by 3000 cal BP is evidence for a broadening of diet, positing a possible combination of climatic changes and social factors as causation (see also David, 2000 and McNiven, 1999: 158–159). Higher trending economic activity from 4000 cal BP – indicated in the data presented by Williams (2013: 6) may have been supported by the reorganisation of resource procurement strategies in relation to plant foods. The common association of mounds with seasonally productive wetland environments in disparate regions suggests that this strategy was specific to these conditions.
The development of mounds through intensive plant processing in earth ovens is, therefore, indicative of innovation within Aboriginal economic systems centred on floodplain environments. This was essentially a re-purposing, or at least a change, in the logistical scale of a traditional cooking technology, that probably arose in response to environmental and/or socio-cultural influences in the mid to late Holocene (Jones et al., 2017; Martin, 2006; Pardoe, 1995, 2003; Westell and Wood, 2014; Williams et al., 2015a).
Calperum study area
The Calperum floodplain forms one of a series of contiguous anabranch systems characterised by complex networks of seasonal flow paths, oxbows, intermittent lakes, backplain swamps and expansive floodplains contained within an incised valley up to 9 km wide. The geomorphology of the floodplain is based around a series of four alluvial terraces and related landforms that include abandoned channels, meander scrolls, source bordering dunes and lunettes. The floodplain is bounded by cliff lines and valley slopes that extend from a high mallee and dune-covered plain.
Mounds are a common feature of the archaeological landscape in this area (and the central western Murray more generally), and are most often located on the banks of anabranch creeks, backplain swamps, billabongs and oxbows (Jones, 2016; Jones et al., 2017; Westell and Wood, 2014). Riverland mounds (n > 207) display a relatively consistent morphology, being generally circular in plan with diameters of between 3 and 50 m (Jones, 2016; Jones et al., 2017; Westell and Wood, 2014). The mounds are between 0.2 and 0.7 m high, and surface observations have revealed that they invariably consist of a fine silt matrix containing burnt clay pellets, ash and charcoal (Jones, 2016; Jones et al., 2017; Ross et al., 2019; Westell and Wood, 2014: 46). Surface surveys have recorded mussel shell fragments in relation to approximately 40–50% of the mounds recorded in the Riverland floodplains of Calperum, Chowilla and Katarapko, though often in low quantities (Jones, 2016; Jones et al., 2017; Ross et al., 2019; Westell and Wood, 2014). Shellfish are known to have been a daily staple in this region (Balme and Hope, 1990; Garvey, 2017; Mulvaney, 1960; Weston et al., 2017), and indeed form extensive midden deposits, however, the low incidence of this material in mounds potentially indicates either the occasional consumption of mussel around these sites and/or the expedient use of mussel shell as tools for processing bulrush root and/or other material cooked in these large scale ovens (Angas, 1847 1: 55; Jones, 2016: 97; Roberts et al., 2021; Westell and Wood, 2014).
Mounds in this region are closely associated with anabranch channels and the margins of lagoons in the more regularly flooded (lower) parts of the floodplain. These environments represent prime habitat for emergent macrophytes such as Typha (bulrush) and Phragmites (common reed) spp. (see Figure 4 for two views of the Ral Ral Creek anabranch) (Jones, 2016; Jones et al., 2017; Westell and Wood, 2014: 48). Prior researchers in the region (as well as further afield in the broader MDB) have suggested that the mounds were primarily used to process Typha spp. for food and fibre and not for general habitation, evidence also exists for the exploitation of other aquatic genera such as Bolboschoenus and Triglochin (Jones, 2016; Jones et al., 2017; Martin, 2006, 2011; Westell and Wood, 2014: 48). The prominence of Typha is also supported by ethnohistorical evidence (Beveridge, 1869, 1883, 1889: 32–34; Eyre, 1845 2: 289–291, 254; Kirby, 1895: 27–28; Mitchell, 1839 2: 53, 60, 80–81, 134). It is noted, however, that some mounds were developed to provide refuge, as a secondary purpose, during periods of flooding in some locations within the MDB (Coutts et al., 1979).

Two views of Ral Ral Creek anabranch, Calperum. Photographs: R. Jones, May 2019.
An integrated study of earth mound sediments
The model outlined above, has informed this investigation into the potential broadening of regional diets from the mid Holocene which employs a range of techniques including an analysis of mound contents, 14C dating and sediment analyses, including magnetic susceptibility (MS), loss on ignition (LOI) and grain size. The latter techniques were selected for their potential to provide insights into mound formation processes as well as the changes in resource exploitation and landscape use – a similar approach was employed by Lowe et al. (2016) in their integrated study of sediments excavated from rock shelters in northern Australia. Variation in grain size profiles and LOI percentages in archaeological sediments, can inform insights into site formation processes and/or subsequent sediment disturbance (Lowe et al., 2016). MS analysis allows for the detection of magnetic minerals in sediments (Evans and Heller, 2003; Thompson and Oldfield, 1986), the identification of which can indicate both cultural and natural phenomena, including pedogenesis, chemical weathering and/or burning (Dalan and Banerjee, 1998).
Methods
Site selection
To date >60 mound features have been identified on the floodplain at Calperum through pedestrian surveys conducted over a series of field seasons from 2015 through 2019 (Jones et al., 2017; Westell et al., 2020). The excavation, contents analysis and sampling of six mounds for 14C dating and sediment analyses occurred during September and October 2019 and September 2020. Figure 5 shows mound RRWWS3/CAPRI_17_04 during excavation in October 2019.

Excavation of mound RRWWS3/CAPRI_17_04, with members of RMMAC in attendance. Photo: R. Jones, October 2019.
The project received Flinders University ethics approval (No. 6618) and a South Australian government permit (Ref: B423901). RMMAC members were involved in all field work and approved all excavation methods and locations. The sites selected for excavation were chosen to reflect a variety of landscape settings in order to consider nuances in mound establishment and preservation related to local flood regimes.
The contents analysis, radiocarbon dating and sediment analyses undertaken in this research relate to groups of mounds located in two distinct landscape contexts within the Calperum floodplain (Figure 2). The analysed mounds occur in three locations (Figure 2). The first group of two mounds (HCN20 and HCN21) are located on a thin silt levee on the northern side of Hunchee Lagoon in the higher (northern) part of the floodplain. The second group of three mounds (HIS_2_21, HIS_2_22, HIS_2_23) are located on the banks of a smaller billabong in the lower (southern) part of the floodplain. This close spatial grouping provided an opportunity to assess the chronological relationship, or otherwise, of a group of closely located mounds. The third location includes a single mound (RRWWS3/CAPRI_17_04) located on the northern bank of Ral Ral Wide Water, a large oxbow in the south-western part of the floodplain. A Leica GS16 real time kinetic Global Navigation Satellite System (GNSS) unit using a HxGN SmartNet correction was used to investigate individual mound height profiles as well as relative elevation in relation to the Australian height datum (AHD).
Excavation
Excavation at each of the mounds consisted of a single pit measuring 0.5 m × 0.5 m. A Leica TS16 robotic total station was used to provide spatial coordinates during excavation. Proforma context, trench and excavation unit documents were used to record all relevant details of the excavation and sampling, including soil characteristics and site context.
Pits were excavated in 5 cm excavation units (XUs) with all excavated material sieved on site through nested 1 and 3 mm sieves (mounds HIS_2_21, HIS_2_22, HIS_2_23, RRWWS3/CAPRI_17_04) or 3 and 7 mm sieves (HCN20, HCN21). All material captured in the sieves was retained for analysis together with a 1 kg sample of sieved sediment from each XU. Column samples of undisturbed sediment were also taken from one side of each excavation pit. Shell and charcoal samples for 14C dating were taken in situ using steel tweezers and/or a steel trowel and wrapped in aluminium foil prior to double bagging in zip-lock plastic bags. Radiocarbon samples were later weighed at the Flinders University Archaeological Research Laboratory using digital scales.
Analytical techniques
Contents and sediment analyses
The mounds on the Calperum floodplain are generally comprised of a mixture of fine sediments, nodules of clay heat retainers, charcoal, shell fragments, unburnt clay and root material from living and dead plants. The surfaces of all clay heat retainer nodules retained during field sieving were examined for any imprints and marks related to their manufacture. The comparative analysis of mound sediment grain size fractions was included in this study to assess variations in individual mound formation processes and landscape context. The sieved fractions from each mound were physically examined in the laboratory to identify any faunal, macrobotanical and artefactual materials. Bulk sediment samples collected from each excavation unit in the field were used for the sediment analyses described below.
Sediment magnetic properties were measured in the laboratory with a Bartington Instruments MS3 using a MS2B sensor. Sediment samples were weighed prior to testing, the instrument was zeroed between readings and measurement time was 0.1 s. Magnetic susceptibility (MS) profiles of low frequency, high frequency and frequency dependence were determined as a proxy for anthropogenic burning in order to determine the base of anthropogenically modified sediment in each mound. Low frequency profiles are shown in Figure 11.
Grain size profiles were determined in the laboratory by passing approximately 50 g of sediment through an Endecotts laboratory test sieve pack comprising 1, 0.5, 0.25, 0.125 and 0.063 mm aperture sieves which were vibrated for 5 min on an Endecotts Minor vibration device. Weights of each fraction were then calculated as a percentage of total retained weight of sediment. Grain size fractions higher than 1 mm were not included in the comparative analysis to exclude potential bias associated with the presence of large pieces of heat retainer.
Loss on ignition (LOI) was conducted in two stages on sediments previously dried overnight at 100°C in a muffle furnace at 550°C and 1000°C to determine LOI for organic and carbonate materials, respectively.
Samples of floodplain sediment were obtained from a related but separate study to provide non-site control data and a comparative analysis for the results obtained from the mound sediments outlined above. An auger was used to extract sediment samples from a location identified as a palaeo river channel 3 km south-west of the mound HCN21 (Figure 2), which were analysed for grain size, magnetic susceptibility and LOI. The auger samples were taken at depths of 16, 29, 42, 57 and 77 cm. Two age determinations of 4960–4846 cal BP (WK53028) at 16 cm and 2292–2004 cal BP (WK53029) at 57 cm indicated an inversion but was interpreted as confirming the sequence 0–57 cm from the surface as late Holocene.
14C age determinations
Thirteen mound samples were analysed by Accelerator Mass Spectrometry (AMS) at the Australian Nuclear Science and Technology Organisation (ANSTO) Centre for Accelerator Science at Lucas Heights. An additional two mound and two non-site control samples were processed at the University of Waikato Radiocarbon Dating Laboratory and the graphite analysed at the Keck AMS Radiocarbon Laboratory, University of California, Irvine, also using AMS. All ANSTO conventional ages were rounded according to the methods outlined in Stuiver and Polach (1977).
Conventional Radiocarbon Ages (CRAs) were calibrated using the OxCal 4.4 programme (Bronk-Ramsey, 2009) applying the SHCal20 atmospheric curve (Hogg et al., 2020). The calibrated age-ranges listed in Table 2 are reported at 95.4% probability. Gillespie et al. (2009) undertook sampling of freshwater mussel shell upstream form Calperum Station on Murray River tributaries (the Murrumbidgee and Darling Rivers) and concluded that no appreciable reservoir effect would occur in MDB samples. Westell et al. (2020: 165) concurred with their assessment and further noted that the ‘intervening geology dissected by the river is non-carbonate fluvial/alluvial sediments and underlying quartzose sands of the Pliocene Loxton-Parilla Sands’.
Results
Excavation
The stratigraphic sections obtained from the excavation of the selected mounds appear to indicate undifferentiated masses of mixed sediment (Figure 6), however, subtle variations in colour, texture, sediment characteristics and potentially moisture content, ultimately enabled stratigraphic differences to be discerned. The sediments contain a varying mixture of degraded clay heat retainers, charcoal, mussel shell fragments and river clay with the very occasional presence of other faunal and stone artefact material.

Stratigraphic diagrams for the six excavations showing identified sediment units. Elevation 0 marks the string line. Horizontal and vertical dimensions are in centimetres. The lower edge of each excavation unit (XU) is indicated by the letter L followed by the relevant XU number. MS, LOI and sediment grain size for all mounds. See Figure 10 for a direct comparison of grain size profile by depth level of each mound. (a) HCN20. (b) HCN21. (c) RRWW3S. (d) HIS_2_21. (e) HIS_2_22. (f) HIS_2_23.
Faunal material present included aquatic and terrestrial snail shell, crustacean gastroliths and fragments of small mammal and bird bones. Freshwater mussel shell was present in all mounds, represented either as a lens of large (1–4 cm) fragments (HCN20 and HCN21) and/or disseminated fragments of varying size (HCN20, HCN21, HIS_2_21, HIS_2_22, HIS_2_23 and RRWWS3/CAPRI_17_04). Six single intact valves were identified as Velesunio ambiguus, five of which were very small (<14 mm by 10 mm, 0.17 g) and the other larger (55 mm by 39 mm, 4.6 g), in all other cases the shell was too fragmented for identification. Terrestrial mollusc taxa found in the sediments included one specimen each of Cupedora and Cernuella, genera of the terrestrial Camaenidae and Geomitridae families respectively (the latter specimen, found in level 1 of mound HIS_2_23, was identified as an introduced species). Thirty-eight specimens of native freshwater taxa from the Planorbidae family were recorded, including the genera Isidorella (n = 35) and Gryaulas (n = 3) and two specimens of the genus Plotiopsis of the Thiaridae family. Planorbidae and Thiaridae members are found in association with aquatic plants in ponds, billabongs, swamps and slow moving streams (Ponder et al., 2020; Smith, 1992; Smith and Kershaw, 1979; Walker, 1988). A small number of gastroliths and claw fragments from the common yabby (Cherax destructor) were also present. Total faunal bone weight recorded was 14.5 g. Artefacts comprised minimal numbers of small fragments of chert and silcrete, and several irregular pieces of silcrete (n = 10, total weight 26.7 g) consistent with previous lithic assessments for the region (Thredgold and Roberts, 2017).
Mound HCN20 provided a good example of the possible range of heat retainer nodule sizes during its excavation revealing a number of large heat retainer nodules at the surface level of the excavation (Figure 7). The heat retainer component displays considerable variability in nodule size and colour. The latter can range from black through red, light orange, brown to light beige, interpreted as resulting from varying exposures to heat during repeated cycles of cooking (Figure 8) (Klaver, 1998; Martin, 2006; Pardoe and Martin, 2011: 54; Westell and Wood, 2014). The variation in grain size observed within the mounds likely relates, in part, to the breakdown of heat retainer nodules through burning, digging and raking of mound contents (Figure 9). The lower floodplain mounds (RRWWS3/CAPRI_17_04, HIS_2_21, HIS_2_22 and HIS_2_23) show the presence of heat retainer material in all excavated levels. HIS_2_21 also has evidence of bioturbation by ants (dry clay nodules with evidence of tunnelling) from level 4 to the base of the excavation (Stephen Hasiotis pers. comm. 2021). The higher floodplain mounds HCN20 and HCN21 show a reduction in the abundance of heat retainer material and increasing evidence of bioturbation by ants below levels 6 and 5 respectively.

View of HCN20 excavation showing large nodules of heat retainer at the surface. Photograph by R. Jones, September 2020.

RMMAC member Jennifer Grace holding examples of clay heat retainers. Photograph by A. Roberts, 13 April 2016.

Sediment grain size profiles for all mounds, level numbers refer to 5 cm spits.
The physical attributes of the mounds selected for excavation are summarised in Table 1.
Physical characteristics of selected mounds.
Sediment analyses
The granulometry results (Figure 9) indicate all mounds have relatively homogenous grain size distributions but some variations exist. For instance mounds HCN20 and HCN21 show an increase in the proportion of fine sediments below level 2, HIS_2_23 and HIS_2_21 show a regular grain size depth profile, and HIS_2_22 and RRWWS3/CAPRI_17_04 show a decrease in the proportion of finer sediments below level 2. The granulometry results from the non-site control data floodplain auger sediment samples (Figure 10) show significant differences to those obtained from the mounds. Grain size analysis shows higher percentages of coarser grained material at all levels than the profiles obtained for the mound sediments.

MS, LOI and grain size profiles for the Hunchee floodplain auger sediment samples.
The amount of organic material in all mounds ranges from 1 to 2.5%. Carbonate percentages range from 0.25% to 1.25% (Figure 11). Low frequency MS values of sediments were recorded between 0e+00 and 2e−06 SI units. The low frequency MS (Figure 11) show a gradual decrease with depth in all sites, except for mounds HIS_2_22, HCN20 and HCN21. HIS_2_22 shows an increase in low frequency MS, and LOI to level 4 then a subsequent gradual decrease. Mounds HCN20 and HCN21 exhibit a sharp reduction in all three parameters between levels 1 and 2 of each excavation except for carbonate in HCN20. LOI of organic material in mound RRWWS3/CAPRI_17_04 shows an increase in organic content between levels 4 and 5 and HCN21, which showed a slight increase of organic material at level 9. A discrete lens of sand at level 5 (Figure 11f) was noted during excavation of RRWWS3/CAPRI_17_04, which is potentially related to an increase in coarser grain fractions in the relatively low resolution grain size results (Figure 9) and is visible in the stratigraphic diagram (Figure 6). HIS_2_22 also demonstrates an increase in coarser material at level 5. LOI of carbonate show some variation, albeit over a limited range, with depth. Mound HCN20 increased from levels 1 to 3 then decreased below that level, mound HIS_2_21 increased with depth, HIS_2_22 increased from levels 1 to 4 then reduced, HIS_2_23 increased from levels 1 to 3 then reduced. The non-site control data floodplain auger sediment samples (Figure 10) show little variation in low frequency MS and LOI with depth, in contrast to the mound samples. Low frequency MS readings at the surface of the mounds are between 8 and 10 times higher magnitude than the equivalent floodplain results. However, this difference reduces with depth.

MS, LOI and sediment grain size for all mounds. See Figure 9 for a direct comparison of grain size profile by depth level of each mound. (a) HCN20. (b) HCN21. (c) HIS_2_21. (d) HIS_2_22. (e) HIS_2_23. (f) RRWW3S.
Thirteen ages (11 from mussel shell and two from charcoal) were obtained from the four lower floodplain mounds and two ages (one from shell and one from charcoal) were obtained from the higher floodplain mounds. Dates range from mound HIS_2_23 at 283 to 0 cal BP (OZZ572) through to mound HCN21 at 4812–4525 cal BP. The calibrated age estimations obtained for this research are plotted against depth in Figure 12.

Calibrated radiocarbon age ranges (with 2 sigma errors) plotted against depth for every mound for which more than one date is available.
Discussion
The close relationship to water features, the presence of large quantities of clay heat retainer and charcoal, and the minimal presence of lithic material and absence of general domestic debris supports the hypothesis that the Calperum floodplain mounds were used for a specific purpose unrelated to domestic occupation. A key observation to be made in relation to the mounds excavated at Calperum for this research relates to very minimal amounts of faunal material, other than shell derived from freshwater species of mussel and snails, and a minimal quantity of undiagnostic bone fragments (potentially rabbit, rodent and/or small marsupial). This aligns with the ethnographic and historical information which relates the formation of substantial earth mounds in the MDB to the exploitation of Typha spp. roots as a source of carbohydrate and fibre. Westell et al. (2020: 166–168) demonstrated the antiquity and ubiquity of river mussel exploitation in the South Australian Riverland through a radiocarbon dating programme (n = 31) conducted on shell middens located across both the Pike and Calperum floodplains. Their study indicated exploitation of freshwater mussel as a food source in the Riverland region, had begun by at least 30,000 cal BP, with an abrupt expansion of middens around 15,000 cal BP and again in the last millennia. This long history of resource use is evidenced by extensive shell midden deposits. The paucity and fragmented nature of mussel shell within mound sediments provides a striking contrast, and suggests that mussel shell found in earth mounds is the product of only casual consumption and/or possibly resulting from the use of shell as tools. Tools made by Aboriginal peoples from mussel shell have been reported ethnographically in the MDB (e.g. Berndt and Berndt, 1993; Eyre, 1845; Taplin, 1879) and in the archaeological record (Mulvaney, 1960, 1961; Roberts et al., 2021; Weston et al., 2017: 229, 236). Angas (1847 1: 55) observed the consumption of mussels as well as their use for the processing of bulrush-root fibre and other purposes.
The fresh-water mussels found in the muddy flats of the river are much sought after by the natives, who cook them by burying them in the ashes of their wooden fires. The shells are used to scrape the fibres of the bulrush-root, after it has been well chewed, for the purpose of making cord for their mats and baskets.
and:
The sharp edge of the mussel-shell is used as a knife, and the women crop their hair by this means. . .Another shell, found in the reeds, serves the purpose of a spoon (Angas 1847 1: 92).
The ethno-historical association of mussel shell with the processing of bulrush root suggests that this is a potential explanation for some of the mussel shell found in the Calperum mounds and indicates an opportunity for future research through use-wear on shell edges. The presence of the shells of other freshwater mollusc taxa, such as Isidorella, Gyraulus and Plotiopsis spp., and five very small complete half valves of V. ambiguus in mound sediments, may indicate their transference along with the mud adhering to the roots of freshwater plants which are their primary habitat (Smith and Kershaw, 1979: 76–89, 57–58). This would have occurred after harvesting from adjacent billabong and creek environments and prior to cooking in mounds established in close proximity to emergent macrophyte habitat. Given the relative absence of other faunal remains, this association provides circumstantial evidence of the relationship of mounds to this aquatic plant resource.
A comparison of grain size plots (Figures 9, 10 and 11) prepared from the sieve analyses of the Hunchee floodplain and mound sediments indicate coarser grain size fractions within the non-site floodplain sample than the mounds over comparable depth levels (0–45 cm, levels 1–9 respectively). The six mounds all exhibit a high level of sediment homogeneity (though not in respect of larger nodules of clay heat retainer, shell or charcoal) as a consequence of human and occasional bioturbation by insects and rabbits. Grain size, MS and LOI plots show anthropogenic agency through mound development processes. These include digging and raking, the setting of new fires, the regular introduction of fresh clay (as heat retainer) and the reuse of mound sediments for capping to complete the oven for cooking. The MS plots of mound sediments show a greater low frequency MS signal than the floodplain sediments except for mound HIS_2_21 which is likely related to its earlier age, early abandonment and later bioturbation (homogenisation) by ants and rabbits. MS and LOI readings in mound sediments generally reduce with depth which is potentially related to the region of intensive burning which shifts higher over time as the mound grows in size and lower levels become less disturbed, however, some local variation exists.
The sediment analysis results for mounds HCN20 and HCN21 indicate an increase in finer grain size fractions, a reduction in low frequency MS and a reduction in organic content below level 2 for both mounds, for which we also noted a reduction in heat retainer presence and a reduction in sediment mixing (Figures 6, 9 and 11). The higher coarser grained fraction percentage in the top excavation levels of these two mound deposits supports the interpretation that the original surfaces of these features have been removed by erosion, exposing base levels which show a coarser grained profile (Figure 9). The higher proportion of coarser grained material towards the bases of younger and more intact mounds, supports this conclusion (Figure 10). An increase in carbonate at level 3 in HCN20 is likely related to the concentrated shell lens found at this depth. The concentration of mussel shell in both HCN20 and HCN21 at levels 4–6 and 3–4 respectively, is in contrast to the nature of mussel shell found in the other excavated mounds. In the latter, shell is typically mixed more evenly through the deposit and is more highly fragmented. The results of the sediment analyses for HCN20 and HCN21 indicate a distinct change in sediment characteristics (with a predominance of silt and clay, and very fine sand fractions) at and below levels 2 and 3 respectively (Figures 6a, 6b and 9). This change most likely identifies the top of a transitional zone between the base of the mounds and the underlying levees (shown by the dotted lines in Figure 6a and b). The age estimate of 4812–4525 cal BP for charcoal found in level 9 (well below the shell lens at level 2) of the HCN21 excavation potentially represents the bottom of this transitional zone near or on a palaeo surface represented by the dense layer of clay which was evident at that depth. Additional funding has been approved for a more comprehensive radiocarbon dating of the lower levels of both the HCN20 and HCN21 mounds to explore this transition.
RRWWS3/CAPRI_17_04 contained a sandy layer, noted during excavation, at the base of level 4 extending into level 5. Both RRWWS3 and HIS_2_22 indicate a dominance of medium sized sand grains and a reduction in organic content prior to c. 450 cal BP, which was not observed in the older mound HIS_2_21 and the younger HIS_2_23 (Figures 6 and 9), which may reflect a discrete period of increased depositional energy related to a flood event.
A degree of homogenisation is present in mound sediments, however, we interpret the stratigraphic profiles and results of the analyses conducted as evidence of decreasing depositional disturbance with depth, over the period of active use. Clearly, low frequency MS and LOI results in mound sediments (other than HIS_2_21, which the radiocarbon results demonstrate is extensively disturbed) are not uniform, as found in non-site floodplain sediments. We would also argue that grainsize variations appear to reflect both anthropogenic and environmental influences. However, the sediment analyses must be interpreted carefully when considering changes in depositional conditions or to provide an accurate proxy for palaeo-burning intensity over time. Human agency, such as digging, firemaking, clay heat retainer addition and raking in the mounds, while they were actively being used, is clearly the major agent of bioturbation, complicated by natural depositional and erosional processes. The difference in grainsize profile between the two closely situated and contemporaneous mounds HIS_2_22 and HIS_2_23 and the similarity between HIS_2_22 and RRWWS3/CAPRI_17_04 potentially demonstrate the environmental nuances which exist in these landscapes. Similarly, the grainsize, MS and LOI profiles obtained for HCN20 and HCN21, together with discrete shell lenses found at levels 4–6 and 3–4 respectively, suggest anthropogenic agency of some complexity.
The oldest confirmed age determinations for the Calperum mounds were obtained from HCN20 and HIS_2_21 at 3723–3484 cal BP (WK52024) and 3981–3723 cal BP (OZZ566) respectively. As discussed above, the early age determination obtained for the lowest excavation level of mound HCN21 at 4812–4525 cal BP (WK52025) requires confirmation through additional research to determine its relevance to this discussion. However, the former dates demonstrate a relatively early establishment for earth mounds in the upper Riverland region, and are some of the oldest in the MDB. As indicated previously, the earliest earth mound radiocarbon ages in the MDB were reported by Martin (2006, 2011) from two mounds located on environmentally stable palaeo-lake lunettes near Balranald in south-western New South Wales (at c. 5000 cal BP). An explanation for the early age profile of the Balranald mounds likely relates to their locations on stable lake lunettes rather than in an active riverine environments (Jones et al., 2017; Martin, 2006). However, this research demonstrates that mounds approaching the antiquity of the Balranald examples can be found in parts of riverine corridors where topographic settings support their preservation. The research also highlights the value of thorough landscape surveys and site selection.
The sample of age determinations reported here provide clear evidence of the periods of mound operation in the Calperum region. The adjacent mounds HIS_2_22 and HIS_2_23 appear to be almost contemporaneous, whilst the later use of mound RRWWS3/CAPRI_17_04 overlaps with the oldest age for HIS_2_22 (Figure 12). The range of dates for mound HIS_2_21 are much older with age determinations of 1864–1729 cal BP (OZZ568), 2429–2150 cal BP (OZZ567) and 3981–3723 cal BP (OZZ566) (Table 2). The ages obtained for HIS_2_21 are inverted, suggesting a significant degree of internal mixing has occurred, as also indicated by the sediment results discussed above. The ages for RRWWS3/CAPRI_17_04 also exhibit a degree of inversion, though the bottom two samples are in temporal order with respect to one another. Whilst the inversions in HIS_2_21 and RRWWS3/CAPRI_17_04 are problematic for interpreting the internal stratigraphy of these sites, they are still valuable in terms of bracketing the age range of earth mound use at these locations. The dates from HIS_2_22 and HIS_2_23 (with the exception of OZZ571 from near the base of HIS_2_22) have overlapping age ranges which are statistically indistinguishable, suggesting a discrete period of mound activity over a few hundred years in each of these sites (Figure 12 and Table 2).
Summary of age determinations by feature.
Where mounds potentially overlay older occupation/activity sites, age determinations and sediment analyses at their base levels have provided insights into potential transitions in resource exploitation and technological innovation. The age range of 3723–3484 cal BP (WK52024) obtained from shell located in level 4 of the upper floodplain mound HCN20, confirmed our hypothesis that the eroded state of this mound, likely reflected a greater antiquity. Both HCN20 and HCN21 have lenses of mussel shell within their lower excavation levels which most likely represents the presence of an underlying shell midden in each case. The stratigraphy identified for HCN21 includes a zone of grey silty/fine clay over a layer of coarser clay (Figure 6b). The surface of HCN20 was 1 m higher than that of HCN21 which may indicate differences in inundation intensity prior to mound formation and potentially explains the difference in grain size profile between levels 2 and 5, for each mound, evident in Figure 9.
The lenses of mussel shell found in the transitional zone below the base levels of mounds HCN20 and HCN21 suggest that a midden had accumulated on the surface of the levee prior to the establishment of the mounds (Figure 6a and 6b). Thus, these sites potentially indicate a change in subsistence strategy and potentially other socio-economic behaviours along the Hunchee Lagoon around 3723–3484 cal BP. We argue that this involved a fundamental switch from a focus on the gathering of freshwater mussels, along the Hunchee lagoon, to the establishment of a food production system based on the innovative use of heat retainer technology for the seasonal exploitation of Typha root. This age, together with the early age of 3981–3723 cal BP (OZZ566) obtained for mound HIS_2_21 on the lower floodplain, potentially indicates the deployment of a major innovation across the landscape at Calperum. This was developed through the integration of a new or previously underused but seasonally abundant resource (Typha), a logistical change of scale in the use of an existing technology (heat retainer cookery), and active resource management through the use of fire, digging and harvesting, as discussed by Gott (1999: 42). Consequently, under this hypothesis, local Aboriginal people broadened their diet via the addition of a seasonal food production strategy to their subsistence repertoire. This hypothesis will be the basis for future investigations.
This study has aimed to expand on the fragmentary information derived from historical observation and ethnographic sources to clarify the relationship between the use of aquatic plant foods, the role of earth mounds and the influence of climate change in Aboriginal economic systems based on floodplain environments in the MDB. The period from the mid to late Holocene saw a trend towards aridity and lower discharge into MDB rivers (Bowler and Hamada, 1971; Fitzsimmons and Barrows, 2010; Fitzsimmons et al., 2013; Gingele et al., 2004, 2007; Petherick et al., 2013: 69–70; Shulmeister and Lees, 1995). A study of diatoms and macrophyte pollen through coring at Tareena Billabong, at the eastern edge of the Chowilla floodplain in the Riverland of South Australia (Figure 2), suggests there was a permanent fresh lagoon in this area which was isolated from the main channel at about 5000 cal BP. The lagoon reconnected after 3800 cal BP with the likely onset of ENSO related wet-dry cycles from that time (Gell et al., 2005: 450; see also Williams et al., 2008, 2015a). Gell et al. (2005: 450) concluded that ‘a regime of wetting and drying had become established’ around 3800 cal BP, with phases of higher inflow interspersed with shallow, brackish conditions. The correlation of this ecological change with that of the early use of mounds HIS_2_21 and HCN20 (Table 2) is evidence of the broadening of local diets in response to climate change and consequential adverse effects on traditional resources. The move to large scale seasonal exploitation of emergent macrophytes through the repurposing of an existing technology (heat retainer cookery) is an example of innovation by Aboriginal peoples in response to deteriorating environmental conditions. Our results indicate nuances in adaptive responses to environmental change, the introduction of broad-spectrum diets and the development of new socio-economic and cultural systems. This encompasses the adoption of new ideas and poses interesting questions about the process of implementation of technical innovation within Aboriginal societies. Thus providing a local perspective to the wider debate associated with environmental and social adaptation and socio-economic intensification.
Conclusions
A focus on content, sediment analyses and 14C dating associated with the intensive use of a common technology (heat retainer earth ovens) has been adopted in this paper to sample historical and environmental trajectories which have provided evidence of the implementation of technical innovation and changes in socio-economic practices associated with variable wetland environments during the mid to late Holocene. The analysis of earth mound sediments has proven to be a useful source of information for the investigation of plant based lifeways at Calperum. Contents and sediment analyses (using grain size, MS and LOI) have identified mound formation processes and purpose, and enabled the identification of a transition in subsistence procurement systems. Fifteen age determinations on shell and charcoal have framed this narrative and allowed a correlation with climate research, which has indicated a relationship between the early development of mounds at Calperum to a widespread deterioration in climate patterns which developed in the Australian region from the mid Holocene. The research detailed here has demonstrated a near continuous chronology for the use of earth mounds on the Calperum floodplain from at least 3981–3723 cal BP and possibly 4812–4525 cal BP, until the time of the European invasion. The age results indicate a continuity of mound use at Calperum that suggests earth mounds were an important element of subsistence procurement systems in this part of the MDB. Seasonal food production achieved through the management of a natural ecological niche, comprising emergent macrophytes in wetland environments, helped mitigate risk in an environment defined by variability.
The early ages obtained for mounds HCN20, HCN21 and HIS_2_21 highlight the potential for the preservation of archaeological sites (including mounds) over the long term despite the inherently dynamic nature of floodplain environments. Ultimately, while preservation is an important factor in mound chronology it is important to consider the landscape in a detailed way when interpreting the chronological distribution of mounds in riverine settings. Furthermore, the results confirm that mound use along the central Murray River corridor began soon after the initial uptake of mound use in the MDB, providing further insights into a regional pattern of economic change in this region from the mid Holocene, specifically around the use of food resources derived from aquatic plants.
Ulm (2013: 183–185) provided a general critique of the continental perspective of ‘intensification at the continental level’, emphasising that ‘regional cultural trajectories need to be disarticulated from the continental narrative to enable independent characterisation of local behavioural variability’. He emphasised the need to address chronological control, sampling and taphonomy and suggested that researchers refocus on open sites instead of the usual emphasis on rock shelter deposits in order to gain a greater picture of regional trends (Ulm, 2013: 187). Earth mounds are one expression of a typological continuum of open sites which have been largely ignored as a potential source of information on regional socio-economic trends during the mid to late Holocene. Particularly in respect of the exploitation of marginal plant foods. The broadening of diets by Australian Aboriginal populations from the mid Holocene, has been argued to be a response to environmental challenges and associated demographic and social factors which emerged at this time in support of higher population growth trajectories from the mid Holocene optimum (Edwards and O’Connell, 1995; Haberle and David, 2004; Williams et al., 2015a, 2015b). In this context we have focussed on specific wetland features and the introduction of innovative management and processing techniques which allowed the extraction of previously marginal and/or low trophic plant foods. These may have come to the fore as traditional resources declined, due to climate variation, ecological change and/or population growth.
Footnotes
Acknowledgements
This paper is based on PhD research being conducted by Robert Jones in collaboration with the River Murray and Mallee Aboriginal Corporation (RMMAC) and approved by Flinders University’s Social and Behavioural Research Ethics Committee (Project Number: 6618). The authors sincerely thank Assoc. Prof. Mick Morrison, all of the fieldwork volunteers and the RMMAC members who participated in the fieldwork for this project. The authors also thank the RMMAC Directors who have supported this work, Ms Diana Fusco of the Flinders University Palaeontology Department for assistance in the identification of faunal material, Mr Peter Hunt and Ms Shirley Sorokin of the South Australian Museum for guidance in the identification of freshwater molluscs, and the staff at Aboriginal Affairs and Reconciliation, Department of State Development, Government of South Australia, for their assistance with permits. The authors thank the Australian Landscape Trust for their ongoing support of this project as well as the reviewers of this article and the editors for their constructive guidance.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding was provided by an Australian Research Council Linkage Project (LP170100479). Dr Ian Moffat was the recipient of Australian Research Council DECRA Fellowship (DE160100703) and a Flinders University Research Investment Fund Grant. Radiocarbon dating at ANSTO was funded by an ANSTO research grant AP12533. The authors acknowledge the financial support from the Australian Government for the Centre for Accelerator Science at ANSTO through the National Collaborative Research Infrastructure Strategy (NCRIS). Additional funding was provided to Robert Jones by the Australian Archaeological Association Student Research Grant Scheme and a student grant from the Royal Society of South Australia.
Supplemental material
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
