American sweet potato and Asia-Pacific crop experimentation during early colonisation of temperate-climate Aotearoa/New Zealand

Abstract

The American sweet potato (Ipomoea batatas) is a globally important comestible crop that features prominently in Polynesian lore; however, the timing and mode of its Oceanic transplantation remain obscure. New research from the Māori cultivation site M24/11 in Aotearoa/New Zealand, presented here, offers a re-evaluation of evidence for the early use and distribution of the sweet potato in southern Polynesia. Consideration of plant microparticles from fourteenth-century archaeological contexts at the site indicates local cultivation of sweet potato, taro and yam. Of these, only sweet potato persisted through a post-1650 climatic downturn it seems, underscoring the enduring southern-Polynesian appeal of this hardy crop.

Introduction

The historical distribution of the American sweet potato (Ipomoea batatas) crop across Oceania, an area where Asia-Pacific vegetables were otherwise dominant, is noteworthy (Yen 1974; Hather & Kirch 1991: 887; Ballard et al. 2005). Hardy, quick-growing I. batatas, now a globally significant comestible, was first transferred into the Pacific in separate events to open up dry, marginal environments for productive agriculture (Ballard et al. 2005; Iese et al. 2018). One early modern sweet potato journey into western Oceania revolutionised pre-nineteenth-century economics in cooler montane New Guinea (Yen 1974: 104–26; Ballard et al. 2005; Kirch 2017: 114–15). In Eastern Polynesia, I. batatas arrived directly from the Americas to be dispersed widely sometime after initial human settlement, c. AD 900, from a central island ‘ellipse’ (Figure 1A). Asia-Pacific crops remained prominent in this ellipse historically. However, transplanted I. batatas became the dominant food crop in southern Polynesia—as kumara on dry, sub-tropical Rapa Nui, and as kūmara for Māori in temperate-climate Aotearoa/New Zealand (Yen 1974; Ballard et al. 2005; Green 2005; Kirch 2017: 198–203, 208–10, 236, 238, 241–42, 265; Kirch et al. 2017: 172–73).

Figure 1. Location maps for places and regions of this study: A) Polynesia, including Eastern Polynesian homeland ‘ellipse’ region identified in grey fill (after Yen 1974; Green 2005); B) central Aotearoa (base data CC BY 4.0) (figure by Les O'Neill).

Māori grew annual single-season kūmara up to the crop's cool-climate limits on the leeward coast of Te Waipounamu/South Island (Figure 1B). From warmer localities across Aotearoa, semi-subterranean storage pits (rua kūmara) supplied dormant tuberous roots into winter months (Barber & Higham 2021). Māori also cultivated the Asia-Pacific geophytes Dioscorea yam/uwhi (probably D. alata) and semiaquatic taro (Colocasia esculenta), but less widely. Taro and uwhi required longer growing seasons and deeper, wetter and more fertile soils than kūmara (Yen 1974; Leach 1984: 17–32, 58–72; Barber 2012a; Kirch 2017: 200; Heider et al. 2021).

Debate persists around Polynesian sweet potato origins. Some scientists cite theoretical mechanics of ocean drift, seed resistance to salinity and modelled pre-human American-Polynesian genetic separation to argue that I. batatas became naturally transpacific (Muñoz-Rodríguez et al. 2018; Pereira et al. 2020; Temmen et al. 2022). Many anthropologists propose instead that South American or returning Polynesian voyagers introduced I. batatas, probably alongside the American gourd/hue (Lagenaria siceraria). They cite comparable Indigenous South American and Pacific names for sweet potato, which pivot around Proto-East Polynesian *kūmara (Figure 2; see also reviews in Scaglion 2005; Kirch 2017: 210; Anderson & Petchey 2020: 351, 370; Temmen et al. 2022). The absence of recorded native (beyond feral) I. batatas populations anywhere in the Indo-Pacific area, despite a well-documented natural Oceanic distribution of other Ipomoea spp. (Yen 1974: 223–30, 261–63; Green 2005: 48; Ladefoged 2005: 363; Muñoz-Rodríguez et al. 2018; 2022: 68–70), is also notable. This fits with a genetic model of pre-Columbian human I. batatas transfer into Polynesia (Roullier et al. 2013).

Figure 2. Location map of modelled and unmodelled chronometric ages for plausible reported pre-Columbian Polynesian I. batatas remains including related data and crop names. Report sources by location are: Anaho, Allen & Ussher (2013); Anakena, Berenguer et al. (2024); Kohala, Ladefoged et al. (2005); Pūrākaunui, Barber & Higham (2021, modelled); Tangatatau, Niespolo et al. (2019, modelled); Te Niu, Horrocks & Wozniak (2008); Whangamatā, Gumbley & Laumea (2019, modelled) (figure by Les O'Neill & Chris Jennings).

The timing of pre-Columbian I. batatas movements into and across Oceania is also debated. A once widely accepted c. AD 1000–1100 radiocarbon chronology on unidentified charcoal from Tangatatau rockshelter on Mangaia bracketed finds of I. batatas parenchyma at the site (Hather & Kirch 1991; Green 2005: 50). But more recent modelling of shorter-life radiocarbon plant ages for this context supports uranium-series dating of coral to AD 1361–1466 (Kirch et al. 2017: 172; Niespolo et al. 2019: 22–23, 31; Anderson & Petchey 2020: 352–53). From northern Polynesia, the earliest plausible radiocarbon report is AD 1290–1430 (95% probability) on a poorly preserved carbonised root at Kohala, Hawai‘i, identified “in favour” of I. batatas (Ladefoged et al. 2005: 367–68). From south-eastern Polynesia, macrobotanical I. batatas remains on Rapa Nui are no older then the thirteenth century AD, at most (Green 2005: 51; Anderson & Petchey 2020: 353–54; Muñoz-Rodríguez et al. 2022: 70–73, 78). And, in south-west Polynesia, the earliest reported radiocarbon age (AD 980–1280) on a carbonised I. batatas root from Pouerua, northern Aotearoa, represents an outlier from a post-1500 context (Anderson & Petchey 2020: 360).

Purported pre-Columbian palynomorphs cf. I. batatas have also been dated across Polynesia. These are primarily starch granules and occasionally pollens (see online supplementary material (OSM) Methods and Materials (OSM Methods hereafter). However, it is not always clear whether these palynomorphs have been, or even can be, differentiated from native Pacific Ipomoea spp., some of which are edible (Anderson & Petchey 2020: 353, 354, 360; Muñoz-Rodríguez et al. 2022: 69, 71–72, 74–76). Moreover, the disturbance or mobility of palynomorphs within sediments is indicated in core samples from Aotearoa and a Rapa Nui crater lake (Horrocks et al. 2012: 190, 192, figs. 2–4, tab. 1; Anderson & Petchey 2020: 360–62).

A few pre-Columbian I. batatas starch reports appear to have more reliable contexts. Mollusc tools dated generally to AD 1200–1400 from Anaho, Marquesas Islands, preserve starch granules attributed to I. batatas, discriminating I. pes-caprae (Allen & Ussher 2013: 2804, 2805, 2808, 2811, tabs. 2, 4–6). Intriguingly, South American human DNA may also arrive in the Marquesas group c. AD 1200 (Ioannidis et al. 2020). Coastal cultivation soils incorporating starch granules cf. Ipomoea from Te Niu, Rapa Nui, plausibly date to AD 1214–1436 (95% probability), although this radiocarbon estimate is on unidentified wood from “major forest burning” (Horrocks & Wozniak 2008: 130, tab. 1). Along this same coast at Anakena, starch granules on obsidian tools from a basal archaeological deposit are attributed to I. batatas with a modelled probability >90%. There are various radiocarbon determinations for this deposit, among which calibrated short-life palm endocarp dates are reported as AD 1326–1448 (Berenguer et al. 2024). In Aotearoa, a short-life Bayesian radiocarbon model dates a boundary transition for archaeological deposits with ‘sweet potato’ and taro starch to AD 1316–1399 (95% probability) at coastal Whangamatā, north-eastern Te-Ika-a-Māui/North Island (Gumbley & Laumea 2019: 103, 131–32, 206). Whangamatā also lies around the indistinct southern natural limit of Ipomoea in Oceania (Figure 2; OSM Methods). Well south of that limit, Polynesia's southernmost archaeological starch granules cf. Ipomoea are reported from midden-capped rua kūmara deposits in a coastal dune at Pūrākaunui, south-eastern Te Waipoumanu. These assumed I. batatas deposits are dated to AD 1430–1460 in a short-life Bayesian radiocarbon model, also at 95% probability (Barber & Higham 2021).

Seemingly, sweet potato invisibility in the earliest, unequivocally pre-1300 deposits of these island sequences supports persistent suggestions of delayed Polynesian I. batatas dispersal (Green 2005; Ladefoged et al. 2005; Anderson & Petchey 2020). In Aotearoa, the foundational food economy (c. AD 1300–1450) included native vegetables, shellfish and finfish, marine mammals and birds, especially the flightless moa megaherbivores (Aves: Dinornithiformes, extinct by c. 1450–1500). Polynesian foods introduced to the islands were domestic dogs (kurī, Canis familiaris), commensal rats (kiore, Rattus exulans) and crops (Anderson et al. 2014: 76–86). Among the latter, sweet potato has been seen as influential (Leach 1984: 54–63; Barber 2004, 2012a), but a recent study has called for a “critical review of nearly all” reported pre-1400 Aotearoa I. batatas dates (Anderson & Petchey 2020: 371). Pre-1400 cultivation becomes Asia-Pacific crop production primarily, with I. batatas deemed unimportant, if present (Anderson & Petchey 2020: 356–64, 371–72; see also Prebble et al. 2019, 2020).

These problems inform our study at Te Tau Ihu/northern Te Waipounamu, a colonisation-era distribution hub for meta-argillite adzes (toki). In this region, with no native Ipomoea, pre-contact kūmara evidence is diverse and potentially early (Barber 1996, 2010, 2012a, 2013, 2017). Our focus is a multi-century Māori cultivation complex (Barber 2013). Short-life radiocarbon ages are modelled and diagnostic palynomorphs targeted in light microscopy (see OSM Methods for details). We ask, when were I. batatas and other Polynesian crops transplanted to Te Tau Ihu and how might this transplantation inform our understanding of southern Polynesian colonisation?

Te Tau Ihu study region and site

Eastern Polynesian toki and moa and seal bones characterise Aotearoa's earliest papakāinga (home base) at Wairau Bar, eastern Te Tau Ihu (radiocarbon dated AD 1320–1360), as well as scattered sites further west (Barber 1996; Anderson et al. 2014: 77–79; Jacomb et al. 2014). Post-1500 earthworks (pā) across the region signal later Māori occupation. Māra (cultivations) are dated to the sixteenth–seventeenth centuries from floodplain soils modified by gravel additions around Appleby, southern Te Tai-o-Aorere/Tasman Bay. In neighbouring Mohua/Golden Bay, potentially earlier coastal māra are identified in eastern dunes and at the coastal Triangle Flat in the north-west (Barber 1996, 2010, 2012a & b, 2013, 2017).

Triangle Flat site M24/11 spreads across low sandy-shelly ridges behind Mohua's soft shore, approximately 1km north-east of a defended headland (pā). Site components include stratified māra soils and features, midden deposits, hāngi (earth ovens), burials and stone artefacts, consistent with a sustained papakāinga (Figures 3 & S1; Barber 2013, 2024). The stratigraphy at M24/11 is capped by black sand layer 1 (L1 etc. hereafter) that incorporates post-contact farm materials (Figures 4–6A, S1–S3; Barber 2013). In many places L1 overlies redeposited bivalve shells (mostly Austrovenus stutchburyi) from beach sediments. This shelly deposit which incorporates occasional Māori materiality (only) is differentially mounded and caps black sand māra soil. Together, the capping shell and māra soil become L2. Substrate depressions (labelled SD) from L2 soil extend roughly 200mm into natural, C-horizon shelly sand (L5). Small intrusions (<100mm) may be kūmara root moulds (e.g. Figure S3). L1 or L2 overlie discontinuous midden deposits of marine molluscs (primarily Paphies sp. and Myitilidae), finfish, cetaceans, smaller birds, kurī and kiore, designated L3. A lower, intermittent, anthropic black sand A-horizon base is designated L4. This fills a natural subsurface channel which is expanded in one area. As in L2 soil, substrate depressions are cut below beach shell caps from L4 soil into L5 substrate (Figures 4 & 5B; Barber 2013, 2024).

Figure 3. Triangle Flat site M24/11 from: A) western hill looking east over woolshed (cf. Figure S1); B) inland shelly ridge edge looking south over N21 E5 (by arrow) (photographs by Ian Barber; figure by Les O'Neill).

Figure 4. M24/11 excavation plan for units N19 E3–N21 E5 (from Figure S1), with lower section A'–B' (below, see also Figure S2) and inset photograph for N21 E3–E5 north (photo by Ian Barber; figure by Les O'Neill).

Figure 5. M24/11 sections presenting shelly mounded surfaces above planting pits filled with post-harvest beach shell (indicated within broken lines): A) lower L2 context between oblique baulk edges below discrete but disturbed beach mollusc cap (ii) and extensive mollusc deposit (i), N41 E15 (location Figure S1; also in Barber 2013: figs. 4, 6B–B'); B) L4 context at border of N21 E3–E4 (SD3, see Figure 2). Scale 100mm in each section (photographs by Ian Barber; figure by Les O'Neill).

Figure 6. Deep, proposed taro planting pit with surface depression in L4 context, N70 E31: A) stratigraphic section and inset photograph for texture; B–D) photomicrographs from surface depression fill including B) large (35μm) faceted starch granule with central fissures cf. I. batatas, scale, 10μm; C) bundles of needle-like structures 80μm long in brightfield (left) and polarised (right) light, cf. ‘short thick’ C. esculenta raphide bundles (see Figure S9A caption), scale 10μm; D) a granular aggregate (individually <7μm, e.g. by arrow) with dark tissue in brightfield (left) and polarisation (right) around projecting CaOx druses cf. C. esculenta, scale 20μm (photographs by Ian Barber & Rebecca Benham; figure by Les O'Neill).

The rounded to asymmetric substrate depressions at M24/11 resemble ‘planting pits’ on Rapa Nui that vary in size and crop use, and are so designated hereafter (Figures 4–6A cf. Stevenson et al. 2006: 934, figs. 9–12; Horrocks & Wozniak 2008; Barber 2010, 2012a, 2013). Durable lithic and beach mollusc sediments above archaeological Rapa Nui and M24/11 planting pits, respectively, preserved garden moisture and soils in place. This has facilitated cultivation reuse to the present on Rapa Nui (Barber 2010, 2013).

Previous palynomorph reports from M24/11 identified pollens, spores, starch granules, vascular materials, biogenic silica and calcium oxalate (CaOx) crystals in light microscopy, including micro-particles cf. I. batatas and C. esculenta (OSM Methods; Horrocks 2004: 328; Horrocks et al. 2004: 155; Barber 2024: app. 1). L2 pollens include intrusive post-contact pine (Pinus sp.) and ribwort plantain (Plantago lanceolata), a single hue (Lagenaria siceraria) grain, and edible, native sow thistle (Sonchus sp.) (the latter often considered an indicator of disturbance; cf. Prebble et al. 2019). Secure L4 sediments at N70 E31 incorporate native palynomorphs only (Figure 6A). These are dominated by forest pollens and considerably fewer grass phytoliths than L2 samples, consistent with a primary context (Barber 2024: app. 1).

Results

Light microscopy palynomorph analysis

The relatively well-preserved, semi-crystalline plant starch granules of our study were identified optically. In most cases this followed recognition of a birefringent (double refracted light) extinction cross in polarisation (Figures 7, 8, S4, S5). For species identification purposes we considered environment, context and known plant biogeography alongside specimen anatomy, allowing for limitations in Aotearoa starch granule reference work (Prebble et al. 2020; OSM Methods). Helpfully at M24/11, floral diversity is constrained in the predominantly young, dry, shallow, sandy topsoils, while starches of important local food plants have been documented in earlier light microscopy work (Barber & Higham 2021: 11–12; OSM Methods).

Figure 7. Photomicrographs (brightfield left, polarisation right) of polygonal, circular and cupule starch granules from M24/11 archaeological contexts cf. I. batatas. Diagnostic features resolved (cf. reference specimens from Figure S5, Table S1) include small, round or ovate cavities at the hila (round in A, N21 E4, L4 channel fill; B, N21 E4–5, L4, SD2, with visual distortion in brightfield through bubble; faint ovate in C, SD2); a prominent circular to oval cavity at the hilum enclosed by marked central lamellae (D, in channel fill), and fissures from cavities at the hila (E, winged from prominent cavity; F, faint, by arrow, partly obscured extinction cross. Scales 10μm in each panel (photographs by Ian Barber & Rebecca Benham; figure by Les O'Neill & Chris Jennings).

Figure 8. Photomicrographs (brightfield left, polarisation right) of grouped to semi-compound starch granules cf. I. batatas (e.g. Barber & Higham 2021: fig. 6D–F) from archaeological planting pits with arrows pointing to fissures from hila: A) N41 E15, lower L2; and B) N21 E4–5, SD2, L4, scales at slightly different magnification each 10μm (photographs by Ian Barber & Rebecca Benham; figure by Les O'Neill & Chris Jennings).

Hundreds of single, semi-compound and occasionally compound granules measuring 7–35μm in diameter, consistent with Ipomoea, presented in samples from L4 below dense mollusc sediments, and from L2 (e.g. Figures 6B, 7, 8, S4, S5). The granules are circular, subcircular and polygonal (multi-faceted) and present cavities or fissures at the hilum (granule origin) and on occasion, several lamellar growth rings (see OSM Methods). Granules of native, starchy Mohua plants lack cavities at hila, except in seeds from naturalised (pre-contact) karaka (Corynocarpus laevigatus) trees introduced from northern Aotearoa. However, karaka granules are <13.2μm and have no visible lamellae. Moreover, multiple lamellae are not diagnostic in laboratory starch contaminants that present cavities or fissures at hila (Type 1 in Crowther et al. 2014), nor reported in granules from L. siceraria rind. Large uwhi (Dioscoreaceae) granules and lab starch contaminants do present lamellae but are easily differentiated (see OSM Methods and discussion below). Accordingly, eight L4 granules with lamellae around vacuoles or fissures in circular, cupule or polygonal shapes, one semi-compound, are most likely I. batatas (Figures 7, 8 & Table S1). Individual and semi-compound granules with cavities or fissures cf. I. batatas from L2 suggest local persistence of cultivation (Figures 8 & S4).

In contrast, the starch granules of a deep, asymmetric, L4 planting pit in N70 E31 are predominantly smaller (<6μm), often aggregated, or enclosed in tiny (<4μm) granule masses. These forms resemble reference C. esculenta amyloplasts, and similar granules from northern Te-Ika-a-Māui taro soils (Figure 6; Horrocks et al. 2004: 155; Barber 2020; Prebble et al. 2020). There are comparable granular masses in L4 channel and SD2 fill, along with birefringent microcrystals that are bright and sometimes pleochroic (presenting variable colours) in polarisation (Figure S7; Barber 2024).

These birefringent L4 microcrystals resemble CaOx raphides and druses (needle-like and larger group formations of crystals, respectively) from C. esculenta corms and northern Te-Ika-a-Māui taro soils (Figures 6C–D & S8; OSM Methods). CaOx crystals do occur in other plant families and native Aotearoa Araceae, including duckweeds (Lemna sp.) with starch granules less than 10μm (albeit not in tiny-granule masses). Context and association therefore become more determinative at M24/11. Of note, a smaller granule aggregate encloses druses from fill of the deep N70 E31 pit, while microcrystals are incorporated within tiny-granule masses in channel and contiguous SD2 fills. Epidermal fragments cf. C. esculenta leaf in these L4 samples strengthen the interpretation of taro presence (Figures 6D, S7A–C & S8B–C; OSM Methods).

One granule group from SD2 includes large elongate, flattened ovoid shapes, individually >30μm, with no facets and highly eccentric extinction crosses. Collectively these features characterise larger starch granules of greater yam/uwhi (D. alata) that present individually and in aggregate (Figure 9, Table S2; cf. Allen & Ussher 2013: fig. 10c–d; see also OSM Methods). Large-granule lab contaminants with eccentric hila are uncommon and typically singular (Type III in Crowther et al. 2014), while Dioscoreaceae are not native to Aotearoa (see OSM Methods). Accordingly, this group is attributed as likely D. alata, representing the pre-modern world's southernmost credible Dioscorea cultivation datum (40°S, previously 38°S at Anaura Bay, eastern Te-Ika-a-Māui; Leach 1984: 64–66, 68). It is plausible that uwhi were composted (if not planted) in the contiguous channel and subsequently incorporated into the shallow SD2 as detritus, along with C. esculenta.

Figure 9. Photomicrograph (brightfield left, polarisation right) of grouped SD2 starch granules presenting flattened, ovate shapes with highly eccentric crosses cf. D. alata. Scale 10μm (photographs by Ian Barber; figure by Les O'Neill).

Radiocarbon analysis

Carbonised branches and twigs containing intact vascular structures and without fungal hyphae—indicating that they were fresh when burnt—were radiocarbon dated using accelerator mass spectrometry (AMS) (Figure S9). Three single dating entities were selected from a primary concentration of well-preserved twigs in the L4 channel fill. Wood was also dated from an intact surface of discarded charcoal and cooking stone materials capping the L4 channel. These capping samples were short-life outer sections of a carbonised puka (Griselinia lucida) branch and a southern rātā (Metrosideros umbellata) branchlet. These outer stem samples were processed at different radiocarbon laboratories as NZA 62661 and Wk-55215, respectively (Figures 4, S2 & S9C–D). Three angiosperm twigs were also dated from L2, where a previous AMS age had been derived (Figure S3).

These nine atmospheric AMS determinations are analysed around five standard L3 marine ages in a four-phase Bayesian model. Calibration curves Marine20 (with a local marine offset) and SHCal20 are applied with outlier analysis in OxCal v.4.4 (Bronk Ramsey 2009 for Heaton et al. 2020 and Hogg et al. 2020, respectively; also OSM Methods, Tables S3–S5). A historical terminus ante quem of 1836 mitigates the effect of multimodal L2 distributions across the wiggly post-1600 atmospheric curve (Figure 10, Tables S3 & S5).

Figure 10. Bayesian radiocarbon model for M24/11 phases calibrated by SHCal20 (Hogg et al. 2020) and Marine20 (Heaton et al. 2020) with local marine ΔR -166 ± 25 and outlier analysis (O: prior probability 0.05 or 5%) in OxCal v. 4.4 (Bronk Ramsey 2009 with data and references in Tables S3–S5). Posterior distributions in darker histogram fill are grey for atmospheric and green for marine above 68.3% and 95.4% ranges (figure formatted by Les O'Neill).

All prior and posterior L4 distributions including capping ages from different laboratories fall within the interval AD 1285–1405 at 95.4% probability (rounded to five, as below). No outliers are identified beyond 0.05, or 5% (Figure 10). Lower modelled L4 distributions for NZA 66620, 62797 and 62799 span 1290–1385 at 95.4% probability and 1300–1365 at 68.2% probability (Table 1). This secure fourteenth-century chronology covers granules cf. I. batatas from channel fill and the contiguous pit SD2 with its mixed crop remains (Figure 4). For radiocarbon ages above L4, highest posterior density intervals begin around the mid-fifteenth century from the L3 midden, and from the end of the seventeenth century for L2 soil (Figure 10, Table S3).

Table 1. Calibrations for atmospheric Conventional Radiocarbon Ages (CRA), L4, with Bayesian radiocarbon model outputs in OxCal v. 4.4.

* Gl = Griselinia lucida; Mu = Metrosideros umbellata.

† Rounded to 5 in OxCal v.4.4.

SHCal20 after Bronk Ramsey 2009; Hogg et al. 2020; see also Table S3.

Discussion

This study contributes the first credible evidence of pre-AD 1400 I. batatas on Te Waipounamu (cf. Anderson & Petchey 2020: 362–64) and the southernmost identification of D. alata in pre-contact Oceania. Moreover, the modelled L4 transition at M24/11 (AD 1310–1390; Figure 10) closely brackets the dates for the earliest, secure archaeological chronology in Aotearoa—the moa-hunting era Wairau Bar site (1320–1360; cf. Table 1 and Jacomb et al. 2014). The proximity of silcrete rock and regular local whale strandings help explain the appeal of early M24/11 settlement (Anderson et al. 2014: 81; Barber 2024). We infer that Te Tau Ihu moa-hunting and multi-cultigen (including I. batatas) cultivation were probably contemporary.

These results illustrate the more complete picture that combined archaeological and microbotanical analyses provide (cf. Barber & Higham 2021). Archaeological excavation and radiocarbon dating identified shallow, dry fourteenth-century planting pits that are indicative of kūmara cultivation. The deep pit in N70 E31, with its concave surface, is also suggestive of semiaquatic taro cultivation. The dominance of palynomorphs cf. I. batatas in the shallow pits and cf. C. esculenta in deeper N70 E31 is consistent. And microbotany alone has identified likely D. alata in SD2 from possible compost applications.

What should one make of this early multi-crop evidence? If the M24/11 channel is considered a likely crop detritus catchment, then there are fewer than 10 palynomorphs cf. C. esculenta compared with tens of granules of cf. I batatas in each analysed sample from channel fill. Moreover, fewer than 10 granules of cf. D. alata are present in any sample. This dominance of I. batatas from the outset of channel use resonates with the “relative frequency” of sweet potato starch on fourteenth-century tools from the homeland-region Marquesan Anaho site (Allen & Ussher 2013: 2811). Kūmara are also prominent in Māori accounts of crop introductions from Hawaiki, homeland of tradition and mythic abode of atua (deities) and the dead (cf. Figure 1; Anderson et al. 2014: 60–62, 67). In one stream of lore, original if not autochthonous non-agricultural inhabitants are so impressed by kūmara introduced from a Hawaiki visitor that they send waka (canoes) for new roots. The visitor's core name is frequently Rongo (southern var. Roko), the ancient Polynesian atua of agriculture, and of kūmara for Māori (Barber 2012b; Anderson & Petchey 2020: 354–56; Barber & Higham 2021: 13–15).

Given the advantages of the hardy, fast-growing I. batatas in areas of marginal agricultural productivity, contemporaneous attempts at uwhi and taro cultivation at M24/11 might seem curious. One possibility is that the site was a social production experiment to replicate high-value, multi-crop ‘Hawaiki’ settings (Leach 1984: 54–56). Here one might compare founding Wairau Bar settlers who assiduously reproduced high status homeland artefacts in local materials (Anderson et al. 2014: 37, 78). But with the American new outperforming the Asia-Pacific old, uwhi and perhaps even taro disappeared from pre-contact Te Waipounamu (cf. Leach 1984: 105–6). Māori kūmara planting rituals continued, imposing potent tapu (supernatural separation) on cultivation fields (Barber 2012b). Tapu-removing activities, people and cooked food were strictly prohibited, perhaps explaining why raw beach mollusc rather than cooked midden mulch was applied to māra at M24/11 (e.g. Figures 4 & 5; Barber 2013: 49–50). In our new chronology this ritual preference has fourteenth-century roots.

In examining change over time, the more frequent cold and wet conditions affecting late Holocene Aotearoa (c. AD 1500–1900) need to be considered. These conditions correspond approximately to the global ‘Little Ice Age’ climate convergence (Lorrey & Bostock 2017: 117). After AD 1600, Aotearoa was influenced further by “harsh” winds during one of the “strongest periods of westerly flow” for 4000 years (Lorrey et al. 2008: 71). Such adverse conditions may have affected marginal cultivation outputs and influenced the abandonment of extensive kūmara fields at Palliser Bay, south-eastern Te-Ika-a-Māui (Leach 1984: 35–43, 61–63; Barber & Higham 2021). But if so, microclimate M24/11 kūmara and, perhaps, limited hue (L. siceraria) cultivation persisted after 1650 (Barber 2024). Hundreds of starch granules cf. I. batatas are identified in a lower L2 planting pit below mounded beach shells. Extensive, stabilising eighteenth-century shelly surfaces may have been spread above pits to counter increasing wind effects (Figures 5A, 10, S4 & S5B–C; Barber 2013).

Coastal southern kūmara agronomy during the Little Ice Age likely relied on the crop's tolerance of impoverished sandy soils that enabled cultivation in warm, if less fertile, dune settings. Moreover, the low, spreading, wind-resistant canopy and fast-setting roots of I. batatas would facilitate greater storm resilience than plants with larger, exposed leaves such as C. esculenta (Yen 1974: 70–72; Iese et al. 2018: tab. 1; Gatto et al. 2021). Repeated re-introduction may have improved kūmara stock further. One nineteenth-century authority from the post-1600 Te Waipounamu iwi (tribe) Ngāi Tahu recalled that later waka introduced “better kinds” of kūmara (Barber & Higham 2021: 14). Original southern varieties have not survived, unfortunately (Leach 1984: 103–04), but archaeological evidence of leeward, eastern Te Waipounamu kūmara field systems and stores indicates some success (Barber 2004: 177–78, 185–86; Barber & Higham 2021). Ngāi Tahu traditions also acknowledge kūmara food and social production values, naming the crop in accounts of reciprocal ceremonial feasts (kaihaukai). In one history, kūmara were stored against an eighteenth-century east coast pā siege, just 12km north of Pūrākaunui's fifteenth-century rua kūmara (Barber & Higham 2021: 2, 13–15, 16, S1 Text 10–13). Kūmara persistence through changing southern seasons and climate underscores Kirch's (2017: 238) observation that sweet potato came “preadapted” to Aotearoa.

This research may even have applied uses. Today, sweet potato is the world's fifth most important comestible crop (Heider et al. 2021: 64). But adverse impacts of climate and other environmental changes, including cyclones, drought and desertification, have affected twenty-first century production. In response, new agronomies focused on natural I. batatas hardiness seek to maintain, if not improve on, the geophyte's naturally high nutritional value (Heider et al. 2021; Rosero et al. 2022). The ancient knowledge (ngā mātauranga o mua hei in Māori) and archaeology of marginal Oceanic I. batatas persistence might encourage if not inform these modern food security developments.

Conclusion

Dating of early cultivation and composting contexts containing proposed I. batatas palynomorphs in Te Tau Ihu points to a pre-1300 presence of sweet potato in the Polynesian homeland. From that Hawaiki, a pre-adapted I. batatas resilience, bequeathed by continental evolution, may have helped motivate early migrants to cross cooler waters for southern Polynesian isles (Barber 2012a). Certainly, the presence of shallow planting pits and large numbers of starch granules suggest that fast-growing I. batatas was the primary fourteenth-century crop at the temperate-climate M24/11 site. A colonising experimental strategy that prioritised hardy kūmara over taro and uwhi is indicated. In time, wider amplification of that strategy would support social and subsistence kūmara production in selected areas of Te Waipounamu (Barber 2010; Barber & Higham 2021). This ancient southern Oceanic agronomy underscores the modern food security value of sweet potato.