Vital evidence: Change in the marine 14C reservoir around New Zealand (Aotearoa) and implications for the timing of Polynesian settlement

Precise and accurate radiocarbon chronologies are essential to achieve tight chronological control for the ~ 750-years since Polynesian settlement of New Zealand. This goal has, however, been elusive. While radiocarbon datasets in the region are typically dominated by marine and estuarine shell dates, such chronological information has been ignored by those interpreting the timing of key events because a detailed regional calibration methodology for marine shell, comparable to the highly precise Southern Hemisphere calibration curve, is lacking. In this paper, we present the first temporal 14C marine offset (ΔR) model for New Zealand based on paired estuarine/marine and terrestrial radiocarbon dates from 52 archaeological contexts. Our dataset displays significant offsets between the measured New Zealand data and the modelled global marine radiocarbon curve. These shifts are associated with oceanographic fluctuation at the onset of the Little Ice Age ~ AD 1350–1450 (650–500 BP). The application of a regional and temporal correction to archaeological shell dates provides complimentary information to terrestrial radiocarbon production and has the potential to add structure to the blurred chronology that has plagued archaeological theories about the colonization of New Zealand, and other Pacific islands, for decades.

www.nature.com/scientificreports/ shell dates for the NZ sequence. Regional chronological studies [2][3][4] have, however, concentrated on terrestrial materials that make up but a fraction of all available dates. In order to complement and enhance these terrestrial studies, we need greater clarity on what 14 C marine reservoir correction value to use. The standard approach for the last ~ 30 years has used the modelled global surface marine curve (e.g., Marine13 14 ) determined from a boxdiffusion model of global carbon exchange derived largely from 14 C atmospheric data. This curve corrects for the approximate 400-year difference between terrestrial and marine 14 C reservoirs. A local 'reservoir offset' (ΔR) is applied to the marine curve to account for regional variation 15 . The ΔR for a specific location '(s)' is calculated using the formula Rs(t) -Rg(t) = ΔR(s), where (ΔR(s)) is the difference between the global average (Rg(t)) and the actual 14 C activity of the surface ocean at a particular location (Rs(t); i.e., the archaeological marine date) at that particular time (based on the terrestrial 14 C date or another independent means by which to determine age, such as U-Th dates). This methodology assumes that the calibration curve accounts for temporal variation. Some marine ΔR work has been carried out around the NZ coastline 13,[16][17][18][19] using 'modern' (pre-AD 1950) shellfish of known collection date. Petchey et al. 19 identified regional variation between the different outlying islands-Norfolk, Kermadec and Chatham islands-because of the complex interplay of currents and water bodies (the Tasman Sea and the South Pacific Ocean). They settled on an average ΔR-value of − 7 ± 45 14 C years for NZ as a whole, with little discernible variation around the main coastline, but identified a much higher and variable ΔR for the Chatham Islands caused by upwelling of 14 C-depleted water along the Chatham Rise and Subtropical Front (which skirts the southeast coast of the South Island) (Fig. 1a). The identification of temporal change has been more difficult to evaluate because of the low precision and limited numbers of archaeological 14 C dates used, but the general consensus was that the regional ocean around NZ appears to have been stable over the recent past 17,18 .
This assumption of temporal stability in the marine reservoir offset is problematic. Recent 14 C values of independently U-Th dated black coral 14 C ages from Tasmania (Fig. 1a) 20 indicate significant reservoir shifts. These shifts are also evident in archaeological ΔR from central South Pacific islands 21 and have major implications for archaeological chronologies that rely on marine and estuarine shell 22 . It is likely that NZ waters are similarly affected, though this has not yet been quantified. In particular, the Tasmanian black coral data show a major positive ΔR shift, up to + 150 14 C years, between ca. 400-200 years BP (AD 1550-1750) and a negative shift of up to − 200 14 C years around 600 years BP (AD 1350). These ΔR offsets would introduce hundreds of years of offset into shell 14 C dates from NZ archaeological contexts, contributing to the blurring of calibrated shell ages. A complex interplay of climate and oceanographic conditions is likely responsible for these shifts, which will almost certainly have had an impact on human adaptation to NZ and may, in part, have been responsible for key economic changes evident in the excavated material remains. Recent revisions to the global marine 14 C curve 23,24 enable regional calibration options, but these options are based on climate models and not on measured data and do not come close to the resolution of measured terrestrial curves such as SHCal13 25 . Contained within the archaeological literature are a host of 'paired' (i.e., from the same context) shell and charcoal 14 C dates which can help evaluate the extent of this problem for NZ. www.nature.com/scientificreports/ In this paper, we present a temporal ΔR model for NZ based on published archaeological paired estuarine/ marine and terrestrial 14 C dates spanning the entire sequence since initial Polynesian settlement. This dataset is ten times larger than any previous regional chronological assessment, and is in close agreement with black coral 14 C data from Tasmania. Both archaeological and black coral datasets document significant changes in ΔR over the last 750 years which attest to changes in the ocean 14 C reservoir associated with major climatic events.

Methods
The marine reservoir age 'R' is the offset in 14 C age between the atmosphere and the global ocean. Regional offsets from R are termed the 'local marine reservoir age' , or 'ΔR' 15 . Calibration of marine 14 C dates involves application of a ΔR-value to the marine calibration curve (e.g., Marine13 14 ) to account for these regional offsets. A ΔR can be calculated from terrestrial and marine samples excavated from archaeological sites. A regional reservoir offset can also be calculated from known-age marine carbonates collected prior to atmospheric bomb testing 19 , or from samples where independently measured calendar ages can be obtained, such as coral dated by both U-Th and 14 C 20 . No matter what materials are used to determine ΔR, they must comply with a set of prerequisites 26 . For archaeological shell samples, the age is determined by dating short-lived (must be identified to species and/or element), 'paired' terrestrial materials from contemporaneous contexts.
A careful evaluation of each context and set of paired dates from each archaeological context was made prior to inclusion in our temporal model of changing marine ΔR. During this evaluation of legacy data, we found numerous issues that warrant mention, as indicated below. calibration issues. Any terrestrial sample with a calibrated age of less than 200 BP produces multiple calibrated ages because rapid fluctuations in 14 C have resulted in 'wiggles' in the calibration curve. In these regions it is near impossible to evaluate which of the multiple wiggles is the true age of the sample, and therefore what the exact marine offset is (Fig. S1). Consequently, any samples with a calibrated mean age of less than 200 BP is of limited use to our evaluation.
Context uncertainty. Typically, the most robust determinations come from well-defined features. We found the selection of charcoal and shell dates from single contexts was rare, with the majority of dates sourced from broader 'midden' layers. Radiocarbon dates from horticultural soils have not been included because these are typically mixed (plaggen soils) 27 . Where more than one terrestrial date was available from the same context we used the Chi square test to insure the 14 C results were indistinguishable (multiple terrestrial dates were only identified from thirteen contexts).
Differences between terrestrial and marine 14 C dates have previously led researchers to search for clues as to the cause of these offsets. Anomalies have been attributed to hardwater, taxa variation, upwelling, inbuilt age in charcoal and the incorporation of natural shell or reuse of previous midden material 12,28,29 . Often the integrity of the archaeological feature has been questioned on the basis of divergent terrestrial and marine 14 C results, even when there is limited evidence of disturbance. Certainly, disturbance remains a very strong candidate for apparent ΔR variation with both positive and negative ΔR values possible. In this study, we have ignored any pre-conceived assumptions of site disturbance, upwelling, taxa variation and/or hardwater impact, unless there is additional independent evidence to back up these claims.
Material suitability. The literature is full of recommendations regarding the suitability of material for 14 C dating and different pretreatment methodologies used. As the science progresses, many of these recommendations have been re-evaluated. With this in mind, we have carefully assessed preconceived ideas regarding the value of different sample types and different pretreatments.
All charcoal dates included in this evaluation have been identified to species that are considered suitable for 14 C dating. Unfortunately, it is rare for the original publications to specifically state if the material has been positively identified as small diameter twigs, and there are very few dates on seeds. Anderson 29 has speculated that even "short-lived" charcoal may have inbuilt age of 50-150 years, but this is difficult to evaluate for each pair of dates. Inbuilt age reduces the apparent age difference between the terrestrial and marine proxies and results in a more negative ΔR. Delamination of growth rings from larger branches during combustion is also possible, while differences have been noticed between the ages of twigs from shrubby plants and twigs from larger trees (W. Gumbley, pers. comm., Nov 2019). We have excluded all dates on bark because the rate of shredding and accumulation varies depending on taxa 30 . We have also excluded dates on tree fern (ponga) because they grow in a spiral pattern making it difficult to distinguish early from late growth (R. Wallace, pers. comm., Nov 2019).
Most shellfish precipitate their shells in equilibrium with the isotopic signature of dissolved inorganic carbon from the waters they live in 31 . Terrestrial organic material from rivers or rainwater runoff generally only have a small negative impact on the ΔR of filter-feeding estuarine bivalves because the uptake of this carbon is typically less than 10% of the 400-year difference between the marine and terrestrial 14 C reservoirs 32 . Areas with calcareous rocks can, however, result in very positive ΔR values (termed a hardwater effect) because the shellfish uptake ancient bicarbonate ions that percolate through the substrate. Typically, this is highly localised and the presence of limestone does not guarantee a hardwater effect in suspension feeding bivalves 32 . Deposit-feeding and herbivorous shellfish can ingest both young and old ('stored') carbon which may have a significant impact on 14 C ages in areas with limestone or old sediment 33 . In the NZ context, the deposit-feeder Amphibola crenata is the best studied example of this problem 34 , and dates of this taxa are excluded from our evaluation. All other shellfish taxa were evaluated on a individual basis. Although upwelling has been suggested as a cause for some anomalous shell ages there is no definitive evidence of any such influence on marine shell along the NZ coastline 19 www.nature.com/scientificreports/ Bone dating has a long and troubled history in the dating of archaeological contexts. Problems related to bone pretreatment and diet variability make this material incredibly complex to interpret. For this reason, the following constraints have been applied: 1. All 'collagen' dates are excluded (based on recommendations by Petchey 36  construction of the 'new Marine' regional calibration curve. We have created a regional curve for use in OxCal 42 using published U-Th and 14 C dates from coral sequences reported by Komugabe-Dixson et al. 20 An OxCal curve file (0.14c) was constructed from the calendar age, derived from the mean U-Th age, the 14 C age in BP and the standard uncertainty on 14 C age. This mock curve is based on limited regional inter-comparison and has not undergone rigorous statistical evaluation.

Results
We have identified 52 ΔR marine/terrestrial pairs recovered from contemporaneous deposits within 36 archaeological sites (Table S1). The majority of sites are located on the northwest coast of the North Island; the Coromandel and Auckland regions of NZ (Fig. 1b). There is an absence of material from the west coast of the South Island. Twelve contexts have a single terrestrial and a single marine date, while twenty-five contexts have a single terrestrial date for comparison with multiple marine dates. This limits assessment of deposit integrity, here determined by statistical agreement between terrestrial 14 C dates. Thirteen sites contain multiple terrestrial and multiple marine dates. These provide the most robust evaluation and the ability to spot anomalies caused by taphanomic, environmental, taxa specific and/or laboratory variables (Table S2). Black coral and archaeological ΔR values are plotted in Fig. 2. Both sets display the same trend whereby the ΔR before 600 BP (AD 1350) is slightly negative, followed by an extreme negative shift around 600 BP and a subsequent gradual return to more positive values around 500 BP (AD 1450). This positive ΔR is very brief and quickly returns to moderately negative values which remain relatively stable over the next 150 years. Between 0 and 350 BP (AD 1600 and 1950) ΔR values are more positive. The U-Th dated black corals produce tightly controlled calendar ages, but there are relatively few values between 750 and 500 cal BP and none between 670 and 554 cal BP. The terrestrial dates for each archaeological context have a wider spread in calendar age but this additional information enhances the extant coral data and may more closely represent the NZ situation than coral from Tasmania.
We suspect that the black coral and archaeological datasets only hint at the complexity of the marine 14 C environment. Additional research is required to map the temporal and geographic variability of the marine reservoir around NZ but, in the meantime, we have calculated the following ΔR offsets between 0 and 750 cal BP using a combination of the black coral dataset, modern pre-AD 1950 shellfish, and the archaeological pairs: 1. Between 0 and 100 cal BP we recommend the use of − 7 ± 45 14 Table S2 and Fig. 2).
There is, however, an obvious problem with the application of these ΔR values; we need to know fairly precisely the terrestrial age of the context to apply the correct ΔR to the marine calibration. Ideally, a regional curve should be developed from measured data from around NZ, but insufficient research has been carried out in these waters. Despite the absence of a detailed curve there is sufficient detail in the ΔR offset to identify a significant www.nature.com/scientificreports/ reservoir shift between 550 and 600 cal BP (AD 1350 and 1400) which has not previously been documented. A major wiggle in the terrestrial calibration curve is present at this time. The global marine curve displays a much attenuated wiggle, and the ~ 400 14 C year difference between these two curves remains (Fig. 3a,b). The extreme negative ΔR indicated in Fig. 2 suggests, however, that this difference is actually less than 100 years at this time in this region (see for example the terrestrial and marine curves in Fig. 3c). This has major implications for dating early sites. But, rather than being a reason not to use shell for dating, this offset could potentially enable greater refinement of site chronologies than presently possible. An example of this is given in Fig. 3, which shows paired terrestrial/marine dates from Layer 9 at Cross Creek, an early site in the Coromandel Peninsula, variously calibrated using a modern pre-AD 1950 ΔR of − 7 ± 45 14 C years (Fig. 3a), the average archaeological ΔR of − 15 ± 24 14 C years for the period between 650 and 600 cal BP (Fig. 3b), or the New Marine curve derived from the raw black coral data (Fig. 3c). Calibration using the New Marine curve (Fig. 3c) gives an age for Layer 9 of AD 1241-1329 (95.4% prob.), a result equivalent to the Kaharoa tephra-an important chronostratigraphic marker thought to immediately postdate human settlement of NZ. This tephra has been precisely dated to 1314 ± 6 AD (1310-1320 AD 68%) 44 . Table 1 presents the statistical probability for the deposition of the Kaharoa tephra at different times within the Cross Creek stratigraphy and the impact of different marine correction methods on the findings. Although the results are not conclusive, the new black coral and archaeological ΔR values (− 42 ± 44 and − 15 ± 24 14 C years (Table S2), and the New Marine curve, support deposition of the tephra between layers 5 and 7, or between layers 7 and 9, rather than before initial occupation at the site (i.e., before Layer 9) (Fig. S2 gives an extended model for layers 7 through to 9. The OxCal code is presented in Text S4).
These findings are of specific interest to a debate about the origins of a tephra sandwiched between layers 7 and 9. Furey et al. 45 have previously suggested the culturally sterile yellow "sand" layer was the Kaharoa Tephra. Jacomb et al. (p. 29) 3 , however has described this association as "unconvincing" and considers this tephra layer to be redeposited. Rat gnawed seeds found within the Kaharoa Tephra at Te Rerenga in the Coromandel are Figure 2. Temporal marine ΔR variation between 0 and 800 cal BP. Red diamonds = U-Th/ 14 C ages for Tasmania 20 (error bars not shown). Grey boxes = calibrated terrestrial age range for each site (horizontal) by ΔR error (vertical) (reported at 68.2% probability and calculated using https ://calib .org/delta r/) 41 . Black dots = median calibrated terrestrial age by mean ΔR-value for each archaeological site (values given in Table S1). Results from sites outlined by boxes are excluded from the temporal average ΔR-value reported in Table S2. Only archaeological sites with a median age > 200 cal BP shown. Dashed lines show period of 'Little Ice Age' transition 43 . Age of the Kaharoa Tephra (red bar) is based on Hogg et al. 44 (see Text S1 for information on the sites highlighted). www.nature.com/scientificreports/ a definitive indication that humans had made landfall in this region by this time 47 , so the inclusion of the ash between cultural layers at Cross Creek cannot be dismissed outright. In the absence of a renewed dating program for Cross Creek, improved calendar age resolution for shell and fishbone 14 C dates through the development of a regionally specific marine calibration curve will be crucial to solving this debate.

Discussion
Komugabe-Dixson et al. (p. 977-978) 20 attributed ΔR shifts along the Eastern Australian coastline to the variable influence of 14 C-depleted water from equatorial waters, possibly caused by El Niño, Southern Oscillation (ENSO) variability. They attributed older waters (i.e., more positive ΔR values) around Tasmania to an increased influence of older sub-Antarctic waters, but did not specifically discuss causes for the negative values observed in the black coral data that start around AD 1550 (400 BP). The extreme negative ΔR trend we have identified between 550 and 600 cal BP is an extension of this and, at its most negative, broadly matches the date of transition from the Medieval Climate Anomaly (∼AD 700-1350; 1250-650 BP) to the Little Ice Age (∼AD 1350-1450; 650-500 BP) (Fig. 2). Using a range of proxy climate records from subtropical and extratropical sources, Goodwin et al. (p.1212) 43 argue that there is a "relatively abrupt shift" in mean climate state after AD 1300 resulting in winder, wetter and colder conditions across NZ (associated with a movement north of the westerly wind belt that circles Antarctica combined with El Niño conditions). These climatic shifts have been linked to the apparent patterns of East Polynesian voyaging and migration 48 14 C years based on archaeological data collected in this paper; and (C) Age calibrated using the black coral dataset ('New Marine' calibration curve). The extended sequence analysis of layers 7 and 9 is presented in Fig. S2. Vertical dashed lines are the 68% prob. maximum age ranges of the marine dates (Wk-21363 and Wk-21355). Wk-21642 is a terrestrial date. The red line represents the calendar age of the Kaharoa tephra 44 . Green band = marine calibration curve (either Marine13 in Fig. A and B or 'New Marine' in Fig. C). Blue band = terrestrial calibration curve (SHCal13). Table 1. Statistical evaluation # of the placement of the Kaharoa Tepha at different points of the Cross Creek sequence. # P = probability that this determination occupies that position as determined using the Outlier () function in OxCal with no parameters set (this enables the program to calculate the probability that a sample is at a particular point in the sequence) 46 (OxCal code given in Text S4). ^The calendar age of the Kaharoa Tephra is modelled at 2σ resolution within these models (i.e., 1314 ± 12 AD). *Layers 8 and 6 are non-cultural deposits. Layer 8 is considered by Furey et al. 45 to be the Kaharoa Tephra. **Calibrated using Marine13 14 with specified ΔR correction applied.  20 suggest that weaker gyre circulation at this time may have resulted in a northward shift of cooler, older Sub-Antarctic waters from the Subtropical Front that skirts the southeastern coast of the South Island (Fig. 1). Unfortunately, the resolution of the archaeological data between AD 1650 and 1950 (300-0 BP) is limited because of the calibration wiggles in the paired terrestrial dates. Conditions around NZ are the result of a complex interplay of many climate systems, so it is not surprising that there are anomalies with this picture. Komugabe-Dixson et al. 20 recorded low marine 14 C reservoir ages for black coral from the Norfolk Ridge between AD 1700 and 1950 (250-0 BP). Petchey et al. 19 also report a negative ΔR (av. -49 ± 10 14 C years) for modern shells from Norfolk Island. They attributed this to increased absorption of atmospheric CO 2 caused by enhanced biological activity at the intersection of warm tropical waters and cooler waters of the Tasman Sea. There is no evidence of lower ΔR values in the NZ archaeological data at this time, but there are few examples from the far northern tip of the North Island where the Tasman Sea has greater influence.
The short chronology for Polynesian settlement and broad calibrated age ranges for 14 C samples have prevented the resolution of many debates in NZ archaeology. The research presented here indicates that a better understanding of the marine 14 C reservoir will improve this picture. In particular, it is evident that the marine 14 C signal does not always match atmospheric 14 C production because of the complex interrelationship of climatic and oceanic 14 C. This difference between the marine and terrestrial 14 C reservoirs provides a means by which we can refine chronologies that have been limited by plateaus and wiggles in the terrestrial calibration curve. Ultimately, it is important that a more precise and accurate regional marine calibration curve is produced from measured (not modelled) data. In the meantime, the ΔR trends identified here will enable renewed insight and investigation of the chronology of Polynesian settlement and development of Māori culture.