INTRODUCTION
Radiocarbon dating of archaeological sites in Oceania has been a contentious issue, as significant modifications have been made to long-accepted cultural-historical models based on redating and reevaluation of key sites across the region (e.g. Anderson Reference Anderson1991; Spriggs and Anderson Reference Spriggs and Anderson1993; Mulrooney et al. Reference Mulrooney, Bickler, Allen and Ladefoged2011; Weisler and Green Reference Weisler and Green2011; Wilmshurst et al. Reference Wilmshurst, Hunt, Lipo and Anderson2011). Multiple interpretations of individual 14C dates or groups of dates have resulted from reexaminations of the appropriateness of sample material (e.g. long- versus short-lived wood charcoal; McFadgen et al. Reference McFadgen, Knox and Cole1994; Allen and Wallace Reference Allen and Wallace2007; Allen and Huebert Reference Allen and Huebert2014), questions of association and cultural context (Dean Reference Dean1978), and the increasing trend for the more precise accelerator mass spectrometry (AMS) dating (Weisler Reference Weisler1989: 121; Rieth and Hunt Reference Rieth and Hunt2008; Rieth et al. Reference Rieth, Morrison and Addison2008).
The careful selection and critical dating of marine shellfish offer an opportunity to date in situ cultural deposits, especially when no other sample material is available. Shells are abundant at many coastal sites throughout the Pacific, and using shells for dating archaeological deposits has a long history in archaeology. Shell taxa are usually identifiable to genus and often species, and problems of inbuilt age can be minimized by selecting short-lived, filter-feeding taxa or young individuals of long-living shellfish (e.g. Tridacna).
14C dating of marine-derived materials, however, yields an apparent age averaging 400 yr older than contemporaneous terrestrial organisms. This global average is built into constructed marine 14C calibration curves (e.g. Marine13 curve of Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Buck, Cheng, Edwards, Friedrich, Grootes, Guilderson, Haflidason, Hajdas, Hatté, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013), but regional and local variation from this global average must be accounted for by local/regional marine correction (ΔR) values (Dye Reference Dye1994; Reimer and Reimer Reference Reimer and Reimer2003; Petchey et al. Reference Petchey, Phelan and White2004, Reference Petchey, Anderson, Hogg and Zondervan2008, Reference Petchey, Allen, Addison and Anderson2009; Weisler et al. Reference Weisler, Hua and Zhao2009; Taylor and Bar-Yosef Reference Taylor and Bar-Yosef2014; Hua et al. Reference Hua, Webb, Zhao, Nothdurft, Lybolt, Price and Opdyke2015). This situation is further complicated by temporal and spatial variation in ΔR, which can vary at a magnitude of hundreds of years over decades, centuries, or millennia (Hua et al. Reference Hua, Webb, Zhao, Nothdurft, Lybolt, Price and Opdyke2015). While ΔR values are available for some Polynesian islands, they are not known, are uncertain, or reflect limited representation in space and time for many other archipelagos. The latter is true for the Samoan Archipelago.
For early efforts to date shell in the Samoan Archipelago, an established ΔR value did not exist. Consequently, dates on shell in Samoa normally were not used unless appropriate charcoal samples were unavailable. For example, Green and Davidson (Reference Green and Davidson1974) used only charcoal for the 45 dates from their extensive investigations, while Jennings and Holmer (Reference Jennings and Holmer1980) used shell for 17 of 34 dates, with all of their early dates (up to about AD 500) from shell. Kirch (Reference Kirch1993a: 87) recognized the need for a ΔR value for dates on marine shell from the To‘aga site on Ofu Island, so in the absence of a ΔR value for Samoa, or anywhere in West Polynesia, he used ΔR values from the North Pacific at Eniwetok in Micronesia, and Hawaii, Tahiti, and Mo’orea in East Polynesia (data taken from Stuiver et al. Reference Stuiver, Pearson and Braziunas1986). These values were pooled to calculate a weighted average of 100±24 yr (Kirch Reference Kirch1993a: 87).
Since Kirch’s work, there have been attempts to determine a more accurate ΔR for Samoa (Table 1). Phelan (Reference Phelan1999: 100–1) proposed a working value of 57±23 yr based on the average of known-age pre-bomb shells from the island of ‘Upolu, although he also calculated a value of –230 (no uncertainty reported) from proximally situated charcoal and shell dates from an archaeological site on Ofu Island (Figure 1). He noted the disparity in the two values and suggested reasons for the variance, favoring the possibility that the association between the two samples was problematic. In comparison, Cleghorn and Shapiro (Reference Cleghorn and Shapiro2000) argued that no ΔR value was needed based on their analysis of paired archaeological samples of charcoal and shell from Ta‘u Island (Figure 1), but later reevaluation by Petchey and Addison (Reference Petchey and Addison2008: 81) on this same material reported a value of –75±55 yr. Petchey and Addison (Reference Petchey and Addison2008) also recalculated Phelan’s (Reference Phelan1999) ‘Upolu values based on a then-current calibration data set (Marine04 curve of Hughen et al. Reference Hughen, Baillie, Bard, Beck, Bertrand, Blackwell, Buck, Burr, Cutler, Damon, Edwards, Fairbanks, Friedrich, Guilderson, Kromer, McCormac, Manning, Bronk Ramsey, Reimer, Reimer, Remmele, Southon, Stuiver, Talamo, Taylor, van der Plicht and Weyhenmeyer2004) and pooled those values with two new pre-bomb samples from Tutuila to propose a weighted average ΔR of 25±28 yr for the archipelago. Petchey et al. (Reference Petchey, Anderson, Hogg and Zondervan2008, Reference Petchey, Allen, Addison and Anderson2009) subsequently modified this value slightly to 28±26 yr based on an additional date on a known-age, pre-1950 shell from ‘Upolu, a new recalculation of the Phelan dates, and rejection of the archaeologically derived paired charcoal and shell dates from Ta‘u and Ofu. Their concluding recommendation to use 28±26 yr for the Samoan Archipelago implies that there is no significant spatial or temporal variation in the ΔR (Petchey et al. Reference Petchey, Allen, Addison and Anderson2009: 2242).
The Samoan Archipelago and its position in the Pacific Ocean
Summary of previous ΔR estimates for the Samoan Archipelago.
* Weighted average based on previous research from several different archipelagos in the mid-Pacific.
In this paper, we address the questions of spatial and temporal variability in ΔR for Samoa by dating individual Acropora spp. coral fingers from cultural contexts using AMS 14C analysis and uranium-thorium (U-Th) series—a procedure that worked successfully for establishing a ΔR for the Hawaiian Islands for the past ~1300 yr (Weisler et al. Reference Weisler, Hua and Zhao2009). Employing new data from Ofu Island, Manu‘a Group, American Samoa, we provide a robust ΔR for the 1st millennium BC and evaluate previous dates of marine samples from archaeological contexts on the island.
STUDY AREA
Located in West Polynesia some 4000 km southwest of Hawaii, the Samoan Archipelago extends in a roughly east-west line stretching across 370 km from 13° to 14° south latitude (Figure 1). Although the Samoans constitute a single ethnic group, the islands today are divided into two political entities: Savai‘i, Manono, Apolima, and ‘Upolu comprise the Independent State of Samoa (formerly Western Samoa), while Tutuila, Aunu’u, Ofu, Olosega, and Ta‘u compose the US territory of American Samoa, with the latter three, at the eastern end of the archipelago, collectively known as Manu‘a.
The islands of the Samoan chain were formed of basaltic lavas and pyroclastics from a series of shield volcanoes (Wright Reference Wright1986). The Manu‘a Islands are the youngest, with Ofu and Olosega dating to within the last 400,000 yr, although McDougall (Reference McDougall2010: 709) estimates that most of the shield-building took place about 300,000 yr ago. Ta‘u was formed less than 100,000 yr ago and possibly quite a bit later (McDougall Reference McDougall2010: 208). The islands vary in size based on differences in age of flows and erosional impacts. Savai‘i and ‘Upolu are among the largest islands in the central Pacific (1695 and 1125 km2, respectively). Tutuila, ~70 km east of ‘Upolu, is considerably smaller but is the largest island in American Samoa (140 km2). Lying ~95 km farther to the east are the small islands of Manu‘a: Ofu (7 km2), Olosega (5 km2), and Ta‘u (39 km2). Ofu and Olosega are separated by less than 100 m, while Ta‘u is 14.5 km to the southeast.
The focus of this study is Ofu Island where a relatively narrow coastal flat is found along much of the south and west coasts. The three widest areas of this flat, which are the most conducive to early settlement, are (1) Va‘oto Plain, at the south point of the island; (2) To‘aga, on a long stretch of the south coast; and (3) Ofu, on the west coast, where the modern village is situated (Figure 2). Coastal soils are largely calcareous (Nakamura Reference Nakamura1984), although terrigenous sediments have eroded from the interior and deposited in the back beach areas of the coastline, particularly over the last 2 millennia (Kirch Reference Kirch1993b; Quintus et al. Reference Quintus, Clark, Day and Schwert2015).
Ofu Island with the location of sites mentioned in the text
Not surprisingly, each of these three coastal flats has evidence of early and long-term human occupation, which has been documented at four sites: Va‘oto (AS-13-13), Coconut Grove (AS-13-37), To‘aga (AS-13-1), and Ofu Village (AS-13-42) (Figure 2) (Kirch and Hunt Reference Kirch and Hunt1993a; Quintus Reference Quintus2015; Clark et al. Reference Clark, Quintus, Weisler, Pierre, Nothdurft and Feng2016). All four sites are ceramic-bearing, with a range of broadly comparable lithic and shell artifacts. Faunal remains are common, with shellfish particularly abundant. While multiple 14C dates on charcoal are available from all four sites, most are from unidentified wood charcoal. To provide a more refined and accurate chronology, coral samples recovered from in situ cultural contexts and from immediately below cultural deposits have been dated by the U-series technique at the Coconut Grove and Va‘oto sites. This dating technique, which typically produces (2σ) age uncertainties of <15 yr (Weisler et al. Reference Weisler, Collerson, Feng, Zhao and Yu2006), improved the accuracy of the chronologies. Moreover, by pairing the U-series dates of the coral samples with AMS 14C dates from three of the same coral fingers, we provide a determination of ΔR for Ofu Island specifically and potentially the Manu‘a Group for the colonization period of 1st millennium BC.
The Va‘oto and Coconut Grove sites, where coral was obtained for this study, are on the Va‘oto Plain. While the sites are 425 m apart, their deposits are stratigraphically distinct (Figure 2). A central feature of the plain is the Va‘oto wetland, which is a small freshwater marsh (2.4 ha) likely fed by a freshwater spring. The marsh is backed (north) by a near-vertical basalt cliff face that rises over 250 m and is bounded on the east and west by talus spurs that extend out from the cliff. Immediately seaside (south) of the marsh is a concrete runway that extends east-west across the plain, with each end terminating near the arcing coastline. Seaside of the marsh and runway is Coconut Grove, an arcing stretch of prograded sand beach with three low paleobeach ridges, progressively rising to the modern shoreline dune crest. Today, the wetland is at or below sea level, but sea level nearly 3000 yr ago was 1–2 m higher than at present (Kirch Reference Kirch1993b; Quintus et al. Reference Quintus, Clark, Day and Schwert2015). Beneath the shallow marsh (about 30 cm deep) is a thin band of silt overlying coralline beach sand, indicating that the marsh area was at one time a shallow sea floor or beach deposit.
Nearly 2 km southwest along the coast from Va‘oto is the To‘aga site (AS-13-1) on the coastal plain of the same name. Kirch and Hunt (Reference Kirch and Hunt1993b) conducted test pit transects traversing the plain but concentrated their excavations where they found early ceramic deposits. Fourteen 14C determinations from the site (nine on shell) were reported by Kirch (Reference Kirch1993a), who concluded, based on what he considered to be the oldest reliable charcoal date from a cultural context (Beta-35601, 2900±110 BP), that To‘aga was settled “by the end of the second millennium B.C.” (Kirch1993a: 91). As noted above, Kirch used a ΔR of 100±24 yr derived from widely separated sites in the mid Pacific for calibrating the more numerous shell dates. He observed that the resulting calibrations “are entirely consistent with the calibrated charcoal dates,” even though there is some variance in the two sets of dates (Kirch Reference Kirch1993a: 87). It is worth noting that the early charcoal dates were derived from samples composed of dispersed flecks of unidentified charcoal, so the old-wood effect may be at play. In addition, the dating was done by the conventional 14C technique rather than AMS, resulting in relatively large 14C age uncertainties. Consequently, we reassess the shell dates below using the new ΔR reported here.
METHODS
Sample Material and Selection
For this study, three coral samples were selected for U-series and AMS 14C analysis of each coral finger, one from Coconut Grove and two from Va‘oto. All the samples are of the Acropora genus, which are the most abundant and widely distributed corals of the Indo-Pacific (Veron Reference Veron2000: 177). The single sample from Coconut Grove comes from the top of a culturally sterile calcareous sand layer below the basal cultural deposit. The two samples from Va‘oto are from cultural deposits, though from different units and stratigraphic contexts. The dates of these coral samples, discussed below, place them in the 1st millennium BC, which has not been considered in prior discussions of ΔR variation in the archipelago. These samples, therefore, are essential for calibrating marine derived materials from the colonization period of Samoa.
Sample Screening
The Acropora spp. branch samples were first examined to determine the general state of preservation. To exclude samples with diagenesis, coral branches with obvious water rounding were not considered further for U-series dating. Only branches that exhibited sharp and well-preserved corallites, except where abrader use is evident, were selected. Figure 3 illustrates the coral branches used for U-series dating and 14C analysis. These pristine-appearing branches were subsampled for analysis of diagenetic alteration from deleterious products including marine aragonite and calcite cements, meteoric cements, and dissolution and extensive bioerosion using scanning electron microscopy (SEM) (Nothdurft and Webb Reference Nothdurft and Webb2009; Sadler et al. Reference Sadler, Webb, Nothdurft and Dechnik2014; Hua et al. Reference Hua, Webb, Zhao, Nothdurft, Lybolt, Price and Opdyke2015). Small representative pieces were cut with a diamond saw and analyzed with SEM for identifying pore-filling cements. Then, samples were embedded in resin, polished, and etched with 1% formic acid solution for 15 s to observe the preservation of skeletal microstructure. Screening was performed at the Queensland University of Technology using a Hitachi TM3000 SEM with energy dispersive spectroscopy (EDS). For each coral sample, the screening was conducted on immediately adjacent samples as to those used for U-series dating and 14C analysis.
Acropora spp. coral branches used for U-series dating in this study: (a) Sample 2014-16; (b) Sample 2014-15; (c) Sample 2014-22. These samples are pristine and have undergone minimal postdepositional erosion as suggested by the well-preserved corallites (lumpy skeletal growths). See Table 1 for provenance information.
The method described here has been used previously to date unmodified Pocillopora spp. branch coral (Weisler et al. Reference Weisler, Hua and Zhao2009) and abrading tools fashioned from Acropora sp. branches (Burley et al. Reference Burley, Weisler and Zhao2012). In the latter example, Acropora spp. branches exhibited pristine as well as worn surfaces indicative of abrader use (e.g. Burley et al. Reference Burley, Weisler and Zhao2012: Figure 3). In these examples, the pristine surfaces suggested that coral branches were collected live; therefore, U-series dates reflected the time of death. However, in examples where corallites are completely removed from abrading use, our SEM methods can be used to check for diagenesis and selection of appropriate tools for dating purposes.
U-Series Dating
A subsample of material from each of the coral specimens was cut and the exterior corallites removed with a diamond-edged circular saw. Material was crushed with bone cutters and an agate mortar and pestle to approximately 1-mm grain size. Cleaning procedures follow those described in Clark et al. (Reference Clark, Roff, Zhao, Feng, Done and Pandolfi2014a, Reference Clark, Zhao, Roff, Feng, Done, Nothdurft and Pandolfi2014b) and were performed in ultraclean labs. In brief, coral fragments were soaked overnight in ~15% H2O2 and rinsed with Milli-Q™ H2O in an ultrasonic bath multiple times until the water was clear. Samples were then dried on a hotplate at 40°C and fragments examined under a microscope to select for analysis the cleanest coral pieces free from signs of alteration and clay or infilled cement contamination.
Approximately 150 mg of the clean unaltered coral pieces were weighed and transferred into precleaned Teflon® beakers with 30 µL of a mixed 229Th-233U spike. Samples were digested with the stepwise addition of concentrated HNO3 until completely dissolved. Several drops of H2O2 were then added to each sample beaker to remove any final organics before being tightly capped and left on a hotplate overnight at 120°C to allow for complete homogenization of the sample-spike solution and removal of organic material. Samples were subsequently dried down and redissolved in 0.7 mL of 7N HNO3. Uranium (U) and thorium (Th) elements were then purified using a modified anion resin ion-exchange method after Chen et al. (Reference Chen, Edwards and Wasserburg1986).
U and Th isotope ratios were measured on a Nu Plasma multicollector inductively coupled plasma mass-spectrometer (MC-ICP-MS) with a DSN-100 nebulizing system and a modified CETAC ASX-110FR autosampler, at the Radiogenic Isotope Facility, University of Queensland following procedures described in Clark et al. (Reference Clark, Roff, Zhao, Feng, Done and Pandolfi2014a, Reference Clark, Zhao, Roff, Feng, Done, Nothdurft and Pandolfi2014b). Samples were measured against NBS-960 U-metal, NBL CRM 6-A pitchblende ore, Harwell uraninite (HU-1), and YB-1 speleothem standards. All U-Th ages were calculated using the Isoplot/Ex version 3.0 program following 230Th procedural blank extraction (Ludwig Reference Ludwig2003). The bulk-earth activity value of 0.82 (atomic ratio ~4.4 × 106), with a designated 50% arbitrary uncertainty (Richards and Dorale Reference Richards and Dorale2003), was used to correct for initial 230Th as the detrital 232Th component likely derived from terrestrial sources postburial (Shen et al. Reference Shen, Li, Sieh, Natawidjaja, Cheng, Wang, Edwards, Lam, Hsieh, Fan, Meltzner, Taylor, Quinn, Chiang and Kilbourne2008; Clark et al. Reference Clark, Roff, Zhao, Feng, Done and Pandolfi2014a). The lab numbers and provenance information for the U-series dated coral samples are presented in Table 2.
Lab numbers and provenance for U-series-dated corals.
AMS 14C Analysis
U-series-dated coral samples containing >1 yr of growth were used for 14C analysis. This approach aimed to avoid possible large seasonal/annual variability in surface ocean 14C (Druffel et al. Reference Druffel, Griffin, Guilderson, Kasahgarian, Southon and Schrag2001; Guilderson et al. Reference Guilderson, Schrag and Cane2004; Hua et al. Reference Hua, Woodroffe, Smithers, Barbetti and Fink2005), which could influence the results of our study. A subsection of each coral was cut along the growth axis using a diamond saw. The outer portion of the coral was removed and only the internal core of the coral skeleton was used for 14C analysis. The samples were then cleaned with Milli-Q water in an ultrasonic bath several times for 20 min each to remove any surface contamination. They were dried in an oven at 60°C overnight. Dried coral samples were mechanically homogenized using a mortar and pestle before hydrolysis. The cleaned samples were hydrolyzed to CO2 using 85% phosphoric acid and then converted to graphite using the H2/Fe method (Hua et al. Reference Hua, Jacobsen, Zoppi, Lawson, Williams, Smith and McGann2001). A small portion of graphite from each sample was employed for the determination of δ13C using the Micromass IsoPrime elemental analyzer/isotope ratio mass spectrometer (EA/IRMS) at ANSTO. The remaining graphite of each sample was loaded into an aluminum cathode by rear pressing for 14C analysis. AMS measurements were carried out using the STAR facility at ANSTO (Fink et al. Reference Fink, Hotchkis, Hua, Jacobsen, Smith, Zoppi, Child, Mifsud, van der Gaast, Williams and Williams2004) with a typical analytical precision of ~0.3% (1σ).
RESULTS
SEM Analysis of Whole Coral Sections
SEM indicates that the skeletal components of the majority of samples are unaltered with largely pristine skeletal aragonite (Figure 4). Samples are generally pristine and the internal core of the coral skeletons considered unaltered. In those samples that were affected by alteration, the diagenetic effects were minimal and primarily confined to the exterior portions of the coral skeleton. The removal of the external skeleton before crushing and microscopic vetting of the crushed coral fragments after undertaking the H2O2 cleaning procedure eliminated any sample fragments that may have contained altered material. For this reason, all samples were considered suitable for U-series dating and 14C analysis.
Scanning electron microscope images in backscatter mode and polished and etched (left column) of rough cut (right column) coral skeleton surfaces. Images illustrate that samples analyzed in this study are pristine skeletons with no evidence of dissolution or replacement and with inner pores of the skeletons free of marine or meteoric diagenetic cement: (a–b) Sample 15; (c–d) Sample 16; (e–f) Sample 22.
Precision and Accuracy of U-Series Dates
U-series data in Table 3 show 232Th concentrations similar to values of other Pacific island corals of a similar age (e.g. Cobb et al. Reference Cobb, Charles, Cheng, Kastner and Edwards2003; Weisler et al. Reference Weisler, Collerson, Feng, Zhao and Yu2006, Reference Weisler, Hua and Zhao2009; Burley et al. Reference Burley, Weisler and Zhao2012). 232Th values range between 0.019 and 1.39 ppb, with an average concentration of 0.44 ppb. These values are relatively low and indicate that initial 230Th component from detrital 232Th is minimal or negligible, resulting in excellent age precision. U-Th age uncertainty averages less than 0.4% (2σ) of the U-Th age, with the errors of U-series dates used in this analysis ranging between 8 and 11 yr for age determinations between 2375 and 2796 cal yr BP.
U-Th isotope data and ages for Acropora spp. branch coral analyzed in 2014.
Samples have an average initial δ234U of 145±1.1 per mil (‰), which is within the range of modern seawater and pristine or modern coral values of 147±5‰ (Stirling et al. Reference Stirling, Esat, McCulloch and Lambeck1995; Robinson et al. Reference Robinson, Belshaw and Henderson2004; Shen et al. Reference Shen, Li, Sieh, Natawidjaja, Cheng, Wang, Edwards, Lam, Hsieh, Fan, Meltzner, Taylor, Quinn, Chiang and Kilbourne2008; Anderson et al. Reference Anderson, Stirling, Zimmermann and Halliday2010; Clark et al. Reference Clark, Roff, Zhao, Feng, Done and Pandolfi2014a, Reference Clark, Zhao, Roff, Feng, Done, Nothdurft and Pandolfi2014b). Modern corals contain average U concentrations between 2.5 and 3.5 ppm (Yokoyama and Esat Reference Yokoyama and Esat2004) and modern Pacific Acropora spp. are known to range from 2.9 to 3.3 ppm (Clark et al. Reference Clark, Roff, Zhao, Feng, Done and Pandolfi2014a). The Samoan samples all lie within the modern range of U values, falling between 2.7–3.2 ppm with an average U concentration of 2.9 ppm. Samples that fall within modern U concentration and δ234U ranges indicate that U uptake or loss associated with diagenetic alteration of coral material was unlikely.
All the Samoan samples fulfill the criteria, outlined in Scholz and Mangini (Reference Scholz and Mangini2007), to identify diagenetic factors that affect both age precision and accuracy. These include 232Th concentrations less than 2 ppb, U concentrations that fall within modern coral values (i.e. 2.5–3.5 ppm), and δ234U that fall within modern seawater and coral values (i.e. 147±5‰). Thus, the Samoan samples are considered reliable.
AMS 14C
The conventional 14C ages of the coral samples, after correction for backgrounds (accelerator and chemistry) and isotopic fractionation using measured δ13C values, are reported in Table 4.
Coral 230Th and AMS 14C age results and calculated ΔR values for Ofu.
* Model marine 14C age was reversely calculated so that the calibrated age (1σ) range of this marine 14C age was equivalent to the U/Th age (2σ) range using Marine13 data (Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Buck, Cheng, Edwards, Friedrich, Grootes, Guilderson, Haflidason, Hajdas, Hatté, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013), ΔR=0 yr and CALIB v 7.0.2 program.
DISCUSSION
The results of our U-series and AMS dating of each of three coral samples yield a weighted mean ΔR of –101±72 yr for the period ~2350–2800 cal BP on Ofu Island. We used a weighted average with the weight in all calculations of weighted mean values being errors associated with their individual ΔR values. The error associated with the proposed ΔR average is large because of the spread in the single values, ranging from –51 to –194 yr (see Table 4). The considerable range of ΔR represented by the three different samples may represent a natural range of ΔR variations at a given region (i.e. leeward Ofu Island). These data are presented in Table 5 along with previously published Samoan ΔR values. Our Ofu ΔR does not overlap at 1σ with any previous values, singular or averaged, derived from the analysis of known-age pre-bomb shells from Tutuila or ‘Upolu (Tables 1 and 5).
ΔR values for Samoa from this study and previously published research.
a From Petchey et al. (Reference Petchey, Allen, Addison and Anderson2009).
b From Petchey et al. (Reference Petchey, Allen, Addison and Anderson2009), recalculated from Phelan (Reference Phelan1999).
c From Petchey and Addison (Reference Petchey and Addison2008).
d Reported here based on U-series and AMS dating.
e Reported here based on calculations from previously published data.
Our value does overlap at 1σ with a ΔR of –155±190 yr for Ofu (Petchey and Addison Reference Petchey and Addison2008: 81) based on paired charcoal (2600±170 BP, Beta-35603) and shell (2330±80 BP, Beta-35604) samples reported by Kirch (Reference Kirch1993a: 89) from the 1st millennium BC deposit at To‘aga. There are, however, potential problems with that estimation. The samples were not collected with the intent of finding a suitable association of charcoal and shell for ΔR determination, but, instead, they, like Phelan (Reference Phelan1999) before them, used a date on charcoal and a date on shell reported by Kirch (Reference Kirch1993a: 89) as being from “the same depositional context” in a 2 × 1-m excavation unit. The large uncertainty for the charcoal date is perhaps explained by the fact that the sample was a collection of charcoal flecks dispersed through a 16-cm-thick layer (Kirch Reference Kirch1993a: 89; Kirch and Hunt Reference Kirch and Hunt1993b: 77). The charcoal was not identified, so some degree of inbuilt age due to old wood cannot be ruled out, a possibility that was acknowledged by Kirch (Reference Kirch1993a: 87) to account for the apparent older age of the charcoal than the shell, and by Petchey and Addison (Reference Petchey and Addison2008: 84) to explain their somewhat divergent ΔR values between Manu‘a and the rest of the archipelago. Based on these samples, Phelan (Reference Phelan1999: 100) had arrived at a very different ΔR value (–230 yr with no associated uncertainty reported) than that subsequently calculated by Petchey and Addison (Reference Petchey and Addison2008: 81) (–155±190 yr) based on a different methodological approach. Neither Phelan nor Petchey and Addison used the To‘aga value in the calculations that produced the weighted ΔR values for Samoa that they reported (Table 5). Given these considerations, despite the temporal and geographic proximity of To‘aga with our Va‘oto Plain sites, we do not include the To‘aga values in our weighted average.
Our value also overlaps at 1σ with two 2nd millennium AD ΔR values from Ta‘u Island also derived from paired shell-charcoal dates. One of these values is –75±55 yr (unidentified charcoal, date of 700±50 BP; Beta-109584) based on reassessment by Petchey and Addison (Reference Petchey and Addison2008: 81) of material recovered by Cleghorn and Shapiro (Reference Cleghorn and Shapiro2000) from Faga, on the north coast. The second value is from Ta‘u Village, on the west coast, of –14±59 yr [coconut (Cocos nucifera) endocarp charcoal, date of 611±30 BP; Wk-21993] (Petchey et al. Reference Petchey, Allen, Addison and Anderson2009: 2240).
These differences in ΔR values between the western (‘Upolu and Tutuila) and eastern islands (Ofu and Ta‘u) of the archipelago may reflect spatial variation similar to that reported from several locations in the Pacific (Dye Reference Dye1994; Petchey et al. Reference Petchey, Phelan and White2004). While spatial ΔR variation in Samoa has been downplayed by Petchey and Addison (Reference Petchey and Addison2008: 83), who cite the general overlap between calculated values from Ta‘u, Tutuila, and ‘Upolu, the new ΔR presented here suggests that a lower value is more accurate for the eastern islands (–101±72 yr for Ofu and –47±43 yr for Ta‘u; see Table 5).
Alternatively, or in conjunction with spatial variation, the differences between previously published values for the archipelago and those reported in this study may reflect temporal differences (Table 5). Temporal variation in ΔR has been reported for several regions (e.g. Yu et al. Reference Yu, Hua, Zhao, Hodge, Fink and Barbetti2010; Taylor and Bar-Yosef Reference Taylor and Bar-Yosef2014; Hua et al. Reference Hua, Webb, Zhao, Nothdurft, Lybolt, Price and Opdyke2015), although few studies have explicitly explored temporal ΔR variation for the Samoan Archipelago. Petchey et al. (Reference Petchey, Allen, Addison and Anderson2009: 2241) considered a possible effect of time but concluded that “in the South Pacific there is greater ΔR variability between the various island groups than that evident over time.” Taylor and Bar-Yosef (Reference Taylor and Bar-Yosef2014: 150), however, have pointed out that “using 14C measurements on modern pre-bomb marine shell to obtain data on reservoir effects is valid only for monitoring contemporary marine reservoir effects” (emphasis in original). The implication is that the use of archaeological samples is needed in order to establish time-based marine reservoir effects.
As mentioned, there is no overlap at 1σ between our 1st millennium BC ΔR of –101±72 yr and those based on known-age pre-bomb shell samples (Phelan Reference Phelan1999; Petchey and Addison Reference Petchey and Addison2008; Petchey et al. Reference Petchey, Allen, Addison and Anderson2009). The –101±72 yr value does overlap with the –75±55 yr value derived from 2nd millennium AD material at Faga on Ta‘u (Cleghorn and Shapiro Reference Cleghorn and Shapiro2000; Petchey and Addison Reference Petchey and Addison2008), but the charcoal sample on which the Faga date was based was unidentified and may have had an inbuilt age effect. If so, the associated ΔR value would be too low, as the difference between shell age and charcoal age is smaller than it should be. The value of –14±59 yr at Ta‘u Village was also based on paired charcoal and shell dates, with the charcoal identified as likely to have been from (short-lived) coconut endocarp (Petchey et al. Reference Petchey, Allen, Addison and Anderson2009) and thus free of inbuilt age. That value overlaps at 1σ with the ΔR values derived from known-age pre-bomb shells on Tutuila and ‘Upolu. However, this value is subject to some degree of additional uncertainty based on the vagaries of the associations of the shell and charcoal.
The movement of materials within seemingly intact archaeological deposits in the Pacific has been empirically documented by Khaweerat et al. (Reference Khaweerat, Weisler, Zhao, Feng and Yu2011), and the proximity of materials in an archaeological deposit may be the result of postdepositional processes rather than temporal association (Wood and Johnson Reference Wood and Johnson1978). We note that the shell and charcoal samples used in the Ta‘u Village calculation are vertically separated by 5 cm, and the shell sample appears to be in layer 5 but at the interface with the older layer 6, below (Addison Reference Addison2008: 13). Another dated charcoal sample from the same unit is from layer 6 but only 13 cm below the shell sample (Wk-21992) (Addison Reference Addison2008: 16). Petchey et al. (Reference Petchey, Allen, Addison and Anderson2009: 2240) state that the boundary between layers 5 and 6 was difficult to discern. The lower charcoal sample dates 100–400 yr earlier than the charcoal sample from layer 5 used to calculate the ΔR of –14±59 yr (Wk-21993; Addison Reference Addison2008). Based on the range of these dates, it is likely that these two layers represent a considerable amount of time. Whether the shell sample is, in fact, fully contemporaneous with the charcoal sample used to calculate the ΔR value is therefore uncertain.
Since our dates are derived from coral and the other ΔR values were derived from shells, the question may arise as to whether coral may differ from shellfish species, especially from different marine environments. In general, 14C data from both corals and shellfish are indicative of dissolved inorganic carbon (DIC) of surrounding seawaters. However, the 14C content in some shellfish species may not reflect that of DIC of surrounding sea waters at the time of accretion of their calcium carbonate skeletons, which could lead to problems in 14C dating of those species. Petchey (Reference Petchey2009; see also Hua Reference Hua2013) attributed such possible differences between specimens to the dietary preferences and habitats of the shellfish. There are insufficient data at present to take the possibility of such differences into consideration in our calculation of ΔR, but it is an issue deserving of further investigation and discussion.
In light of the above discussion, while variation in reported ΔR values is present in Samoa, we cannot determine whether differences are due to spatial or temporal issues. This situation stems from a highly limited array of archaeologically derived samples used in calculating ΔR. Moreover, the previously calculated ΔRs from archaeological materials are subject to some level of uncertainty due to the use of paired charcoal and shell dates, as this methodology relies on the assumption that spatial proximity in an archaeological deposit equates with temporal proximity of the samples, which may not be a valid in all instances. Additionally, using unidentified wood charcoal samples for constructing a ΔR introduces another unknown level of uncertainty (e.g. the possible old-wood problem). These issues are particularly relevant for the calculated ΔR values from Ta‘u Island and To‘aga.
To assess the impact of our proposed ΔR for 1st millennium BC sites in Manu‘a, we recalibrated four shell 14C dates from the To‘aga site on Ofu Island reported by Kirch (Reference Kirch1993a). Of the other shell dates from To‘aga, two are clearly outliers (clearly too old for the cultural deposit) and three are post-2000 cal BP, so those dates are not recalibrated here. The recalibrations were done in OxCal v 4.2 (Bronk Ramsey Reference Bronk Ramsey2009) using the Marine13 curve (Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Buck, Cheng, Edwards, Friedrich, Grootes, Guilderson, Haflidason, Hajdas, Hatté, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013) with a ΔR of –101±72 yr. We compared these results to calibrations using previously proposed ΔR values (Table 6).
Results of recalibration of shell dates from To‘aga reported by Kirch (Reference Kirch1993a) using different ΔR values. All age calibration was performed using OxCal v 4.2 (Bronk Ramsey Reference Bronk Ramsey2009) and the Marine13 calibration curve (Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Buck, Cheng, Edwards, Friedrich, Grootes, Guilderson, Haflidason, Hajdas, Hatté, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013).
a Calibrated using a ΔR of 100±24 yr (Kirch Reference Kirch1993a).
b Calibrated using a ΔR of 57±23 yr (Phelan Reference Phelan1999).
c Calibrated using a ΔR of 28±26 yr (Petchey et al. Reference Petchey, Allen, Addison and Anderson2009).
d Calibrated using the new ΔR of –101±72 yr reported in this study.
Given the lower ΔR value presented here, the recalibrated dates are earlier than previously reported. Unfortunately, all of the recalibrated shell dates exhibit relatively large age spans, which reduce the utility of these dates for precisely defining island colonization. Nevertheless, the earliest of these dates (Beta-35604) now overlaps with a recently modeled date of island colonization at ~2650–2700 yr ago (Clark et al. Reference Clark, Quintus, Weisler, Pierre, Nothdurft and Feng2016). Interestingly, this date features more overlap with the charcoal date (Beta-35603, 2600±170 BP) from the same context with which it was paired by Phelan (Reference Phelan1999) and Petchey and Addison (Reference Petchey and Addison2008) as described above. This suggests the possibility that the charcoal sample was not adversely affected by inbuilt age as has been suggested. Two other newly calibrated age ranges are of interest (Beta-25033 and Beta-25034) as they come from the same unit but different stratigraphic contexts. These two determinations (conventional 14C ages) overlap at 1σ, although the stratigraphically lower sample has a younger range. Given the range size, however, it is unclear whether there has been postdepositional movement in the sand deposit or whether this pattern is evidence of rapid deposition toward the end of the 1st millennium BC. The latter interpretation seems most likely since no other signs of disturbance were noted by the original researchers (Kirch and Hunt Reference Kirch and Hunt1993b: 51–2). In short, we find that the recalibrated dates using the ΔR of –101±72 yr provide a better match with other dates from the island (Clark et al. Reference Clark, Quintus, Weisler, Pierre, Nothdurft and Feng2016) than the previously proposed ΔR values of Phelan (Reference Phelan1999) and Petchey et al. (Reference Petchey, Allen, Addison and Anderson2009).
CONCLUSIONS
Marine 14C reservoir effects have long been known potentially to display spatial and temporal variations that need to be taken into account when calibrating 14C dates of marine samples. The existence of such variability has not been fully documented across the Samoan Archipelago, but the data reported here (previously published ΔR values along with those of the current study) suggest that one or both factors are at play in these islands. In previous studies based on pre-AD 1950 (pre-bomb) marine shells from ‘Upolu and Tutuila, Petchey et al. (Reference Petchey, Allen, Addison and Anderson2009) recommended a ΔR of 28±26 yr, but the paired AMS and U-series dates on archaeological coral samples from the 1st millennium BC on Ofu that are presented here indicate a weighted average ΔR of –101±72 yr. With these data in mind, we recommend using the ΔR of –101±72 yr for the 1st millennium BC in Manu‘a, and the 28±26 yr ΔR for calibrations for dates within the 2nd millennium AD in the western islands (Savai‘i to ‘Upolu and Tutuila). Until new data allow for additional locally determined ΔR values, we recommend that for all other dates (1st millennium BC in the western islands, 2nd millennium AD in Manu‘a, and 1st millennium AD in all islands) researchers use both ΔR values to see how the results of each compare with AMS dates on identified, short-lived wood charcoal.
14C dates based on marine-derived material from Samoa can help build chronologies, but until such time as the spatial and temporal variability of ΔR is documented from across the archipelago, culture-historical sequences should be anchored to AMS 14C dates of short-lived charcoal or U-series dates of culturally associated fresh coral fingers. Dates on marine shells can assume greater importance as the ΔR value, or values, in Samoa are increasingly refined. The refinement of the ΔR values in Samoa can be accomplished through an expanding array of samples from sites at various locations on all islands and from time periods covering the entire span of human occupation of the islands. Those samples must come from (1) paired charcoal and shell AMS dates derived from identified short-lived plant species and filter-feeding shellfish from secure associations (e.g. from a fireplace) and/or (2) compared AMS and U-series dating of corals. The development of such a refined series will have to take place through the contributions of multiple researchers. Such efforts will contribute to more effective use of marine reservoir corrections for the Samoan Archipelago, and in so doing, contribute to our understanding of marine reservoir corrections in general and Samoan prehistory specifically.
ACKNOWLEDGMENTS
We are deeply grateful to the people of Ofu Island, American Samoa for allowing us to conduct our research, and for their hospitality, cooperation, and assistance during our time on their beautiful island. This research was supported in part by the US National Science Foundation (Grant No. 1229417). Additional support was provided by North Dakota State University. Emma St Pierre’s postdoctoral fellowship was supported by strategic funding from the Deputy Vice Chancellor, University of Queensland to Marshall Weisler. We thank Tara Clark (University of Queensland) for input and assistance while processing coral samples for U-series dating.
