The Marshall Islands, Eastern Micronesia

20 Pages·2006·0.59 MB·English

Analysis of Plant Microfossils in Archaeological Deposits from
Two Remote Archipelagos: The Marshall Islands, Eastern Micronesia,
and the Pitcairn Group, Southeast Polynesia1
Mark Horrocks2,3,5 and MarshallI.Weisler4
Abstract: Pollen and starch residue analyses were conducted on 24 sediment
samples from archaeological sites on Maloelap and Ebon Atolls in the Marshall
Islands, eastern Micronesia, and Henderson and Pitcairn Islands in the Pitcairn
Group, Southeast Polynesia. The sampled islands, two of which are ‘‘mystery
islands’’ (Henderson and Pitcairn), previously occupied and abandoned before
Europeancontact,comprisethreetypesof Pacificislands:lowcoralatolls,raised
atolls, and volcanic islands. Pollen, starch grains, calcium oxylate crystals, and
xylem cells of introduced non-Colocasia Araceae (aroids) were identified in the
MarshallsandHenderson(ca.1,900yrB.P.and1,200yrB.P.attheearliest,re-
spectively).Thedataprovidedirectevidenceof prehistorichorticultureinthose
islands and initial fossil pollen sequences from Pitcairn Island. Combined with
previous studies, the data also indicate a horticultural system on Henderson
comprising both field and tree crops, with seven different cultigens, including
at least two species of the Araceae. Starch grains and xylem cells of Ipomoea sp.,
possiblyintroducedI.batatas,wereidentifiedinPitcairnIslanddepositsdatedto
the last few centuriesbeforeEuropean contact in 1790.
During the late Holocene settlement of This involved starch staples produced
theeasternPacific,agriculturalpracticescom- through a variety of ‘‘wet’’ and ‘‘dry’’ field-
bining tropical Asian-Pacific and American cropping methods (Kirch 1994) and tree
crops were adapted to a diversity of island cropping (Yen 1973). The extension of these
environments(Yen1973,Spriggs1982,Ayres practicestoremoteislandsof oftenverysmall
andHaun1990,Kirch1994,Bellwood2005). area with barren soil, unreliable groundwater
supply, unpredictable rainfall, and frequent
marine inundation was one of the most chal-
1TheMarsdenFundadministeredbytheRoyalSoci- lenging accomplishments of this diaspora.
etyofNewZealandprovidedsupportforM.H.Funding
The starch staples include four species of
fromtheWennerGrenFoundationforAnthropological
Research, the United Nations, the U.S. National Park the Araceae, Colocasia esculenta (taro), Alocasia
Service,theAustralianSouthPacificCulturesFund,and macrorrhiza (elephant ear taro), Cyrtosperma
the Historic Preservation Office (Republic of the Mar- merkusii (syn. chamissonis, giant swamp taro),
shall Islands) provided support for M.I.W. Manuscript
and Amorphophallus paeoniifolius (elephant
accepted2June2005.
2MicrofossilResearchLtd.,31MontLeGrandRoad, yam); and Dioscorea spp. (yams), Musa spp.
Mount Eden, Auckland 1003, New Zealand (e-mail: (bananas), Artocarpus altilis (breadfruit), and
www.microfossilresearch.com). Ipomoea batatas (sweet potato). In total, there
3School of Geography and Environmental Science, are approximately 30 species of subsistence
University of Auckland, Private Bag 92-019, Auckland,
plants listed (see, e.g., Kirch 1989, Loy et al.
NewZealand.
4SchoolofSocialScience,UniversityofQueensland, 1992, Matthews 1996). Many of these origi-
St.Lucia4072,Queensland,Australia. nated in Asia and, recent research suggests,
5Correspondingauthor. also in the Sahul continental region (Haberle
1995, Lebot 1999, Denham et al. 2003, Mat-
thews 2003, Bellwood 2005). They were in-
PacificScience(2006),vol.60,no.2:261–280
:2006byUniversityof Hawai‘iPress troduced throughout the Pacific by early
Allrightsreserved human colonizers. Ipomoea batatas (sweet
261
.
262 PACIFIC SCIENCE April2006
potato) and Lagenaria siceraria (bottle gourd) this growing body of evidence, however, ma-
originated in South America, their introduc- jorgapsinthePacificrecordremain.Inaddi-
tions a result of Polynesian contact (Hather tion, human settlement of atolls is extremely
andKirch 1991, Green 2000). underresearched (Weisler 1999).
Researchontheage,development,anddi- Although ethnographic accounts are valu-
versity of early agriculture in the Pacific has able sources of information, we would like to
been hindered by transformation of tradi- know which species were introduced where
tional cropping systems after European con- and when. The so-called ‘‘mystery islands’’
tact and archaeological scarcity of crop are particularly problematic in this respect
fossils. However, recent advances have been because they were uninhabited at the time
made with regard to the latter at sites across of Western contact. There are more than a
thePacificbytheidentificationof plantmac- score of these, typically small targets at some
rofossilsandespeciallybytheapplicationof a distance from their nearest neighbors (Irwin
range of microfossil techniques, namely pol- 1992). Given the isolation, small size, and
len,starchresidues,andphytoliths.InMicro- especially resource-poor nature of these hab-
nesia, prehistoric pollen (and macrofossils) of itats (including low-nutrient soils and unpre-
the Araceae and Artocarpus altilis has been re- dictable water resources), it is uncertain if
ported in the Mariana, Caroline, and Mar- horticulture waspracticed at all.
shall Islands (Beardsley 1994, Athens et al. Duringanarchaeologicalexpeditiontothe
1996, Athens and Ward 2002, 2004). In Mel- Pitcairn Group in Southeast Polynesia in
anesia, Araceae starch residues have been 1991–1992, sediment samples from deposits
identified at Lapita settlements in the Bis- thought to be prehistoric gardening areas
marck Archipelago (Crowther 2005) and were collected (Weisler 1995). The project
Vanuatu(HorrocksandBedford2005).Simi- was investigating the survival of small human
larly, Lentfer and Green (2004) identified groups on isolated, marginal landfalls, and an
Musa phytoliths at a Lapita site in the Bis- island-wide archaeological survey and exten-
marcks. In Polynesia, starch residues, phyto- sive excavations in habitation sites were con-
liths, and pollen of a suite of introduced ducted (Weisler 1995, 1997). In another
cultivated plants have been identified in New project samples were collected in 1993–1997
Zealand, comprising Ipomoea batatas, Colocasia from the Marshall Islands in eastern Micro-
esculenta, Lagenaria siceraria, and Broussonetia nesia (Weisler 1999, 2001a). That project
papyrifera (paper mulberry) (Horrocks was investigating the timing and nature of
2004a). In the Society Islands, Parkes (1997) Araceae horticulture across the rainfall gradi-
reported pollen of Ipomoea batatas, Colocasia ent in the Marshalls. Extreme remoteness is
esculenta, and Cordyline fruticosa (syn. termina- a common feature of these two distant archi-
lis, Pacific Island cabbage tree). Athens pelagoes.Thesitespresentedtheopportunity
and Ward (1997) reported pollen of Colocasia to test different tropical soil types in a range
esculenta, Cordyline fruticosa, and Aleurites mo- of environmental settings for preservation
luccana(candlenuttree)inHawai‘i,andCum- of plant microfossils and to look for direct
mings (1998) reported pollen of Ipomoea evidence of prehistoric horticulture near the
batatas and starch granules of Colocasia on outer margins of both Polynesia and Micro-
EasterIsland.MacrofossilsidentifiedinPoly- nesia. The outer Pacific archipelagoes were
nesia include Ipomoea batatas root (commonly among the last habitable places on Earth to
referredtoastuber)fromHawai‘i(Rosendahl be colonized by people. These islands are
andYen1971),theCookIslands(Hatherand ecologicallyfragileandoftensmall,providing
Kirch 1991), and New Zealand (e.g., Yen microcosms for the study of human adapta-
andHead1993);Aleuritesmoluccanaendocarp tion to diverse, extreme environments. Pre-
fromtheCooksandHendersonIsland(Kirch sented here are results of recent analyses of
et al. 1992, Weisler 1995); and Araceae and pollen and starch residues in the sediment
Musa leaf also from Henderson (Weisler samples from archaeological sites in these ar-
1997, Hather and Weisler 2000). Despite chipelagoes.
.
PlantMicrofossilsfromtheMarshallIslandsandPitcairnGroup HorrocksandWeisler 263
The Study Areas, Sites, and Sample Context stratigraphically below a 1,580G80 yr B.P.
date (Beta-69516 [Weisler 2001a]). The pit,
marshall islands. Situated in eastern 73 cm in diameter, had subparallel sides to a
Micronesia, the Marshall Islands comprise depth of 54 cm. Sample 3 is from MLMl-3,
29 atolls and five coral islands without a TP19, 21, and 24, buried A horizon dating
centrallagoonalignedintworoughlyparallel to 1,910G70 yr B.P. (Beta-79576 [Weisler
alignments. The climate is tropical, with the 1999]). Because the buried A horizon was
archipelago traversing a rainfall gradient thin, sediment was taken from the sidewall
from @1,000 mm in the north to @4,000 of three contiguous test pits. This sample is
mm in the south. Mean annual temperature associated with the construction of an adja-
is@27 (cid:2)C. Most of the atolls and islands are cent pit currently used for cultivation of
currently inhabited; the two northernmost Cyrtosperma merkusi.
and driest atolls, Taongi (Pokak) and Bikar, The Ebon and Enilok islet (Ebon Atoll)
are uninhabited but are visited occasionally samples comprise samples 4–8. Samples 4–7
to collect birds and turtles. Samples from are from MLEb-2I, TP2 layers I, II, III, and
twoatollswereanalyzedinthisstudy:Maloe- IV, whichcorrespond to samples 4–7(Figure
lapinthecenterof thearchipelagoandEbon 1A). A 4-m-long trench was excavated
inthefarsouth.The71isletscomposingMa- through the rim of a suspected horticultural
loelap Atollhave a total terrestrial area of 9.8 pit, and stratigraphic samples were taken
km2. Samples from one of these islets, Kaven from all four layers; layer III was the buried
(2.3 km2), were analyzed. The 22 islets of A horizon dated to 1,930G40 yr B.P. (Beta-
Ebon Atoll total 5.6 km2. Samples from two 92123 [Weisler 2002a]), layer I the near-
of these, the largest, Ebon (2.7 km2), and surface A horizon, layer II a prehistoric habi-
Enilok (0.4 km2), were analyzed. Low coral tation layer, and layer IV the sterile subsoil.
atolls such as the Marshalls, usually only ca. Sample 8 is from MLEb-38, TP2, 65 cmbs
2–3 m above sea level, are prone to marine within a large pit feature measuring at least
inundation during storms. Soils are typically 132 cm wide and@33 cm deep, below a date
humus-poorandporous,withpotablesubter- of 1,560G70 yr B.P. (Beta-92134 [Weisler
ranean water restricted to the largest islets of 2002a]).
anatoll.Radiocarbonevidencesuggestsoccu- pitcairn group. The Pitcairn Group,
pation from ca. 2,000 yr B.P. (Weisler 1999). located in Southeast Polynesia, comprises the
Detailed site descriptions and stratigraphic raised coral island of Henderson, two coral
contextare given in Weisler (2001a, 2002a). atolls (Oeno and Ducie), and the Pleistocene
Allsamplesarefromprehistorichabitation volcanic island of Pitcairn. Only Pitcairn
or suspected horticultural areas (see Appen- Island is currently inhabited. The climate
dix). The Kaven islet (Maloelap Atoll) sam- is subtropical, with meteorological records
ples comprise samples 1–3. Samples 1 and 2 from Pitcairn showing a mean annual rainfall
are from MLMl-3, TP9, feature 1, a sus- of 1,716 mm but with considerable inter-
pectedstoragepitforArtocarpusaltilis(bread- annualvariability,andameanannualtemper-
fruit) paste. Small pits, <1 m wide and ature of @21 (cid:2)C (Spencer 1995). Samples
@0.5 m deep (Weisler 2001b), were often from Henderson and Pitcairn Islands, both
used to store Artocarpus paste to extend the ‘‘mystery islands’’ and therefore uninhabited
availability of this seasonal crop. Especially at European contact, were analyzed in this
on drought-prone islands, food storage was study. Soils throughout Henderson (37.2
an important buffer against famine, a strategy km2), as on low atolls, are humus-poor, po-
well documented in the Marquesas Islands rous, and undeveloped, although at 30 m
(Handy1923,Linton1925).OurMarshallIs- abovesealevel,theislandisnotsubjecttoin-
lands sample 1 from Kaven islet (Maloelap undation.Potablewaterisfoundascavedrips
Atoll) was taken from inside the pit and following ample rains. Radiocarbon dating of
sample 2 outside the pit, both at 53 cmbs all major habitation sites suggests a 600-yr
(centimeters below surface). The samples are Polynesianoccupationsequencecommencing
.
264 PACIFIC SCIENCE April2006
Figure1.SampledprofilesfromtheMarshallIslandsandHendersonIslandinthePitcairnGroup.Samplenumbers
andpositionsareshowninboxes(seeAppendixfordescriptions).A,Northprofileof MLEb-2Itrench,EbonIslet,
EbonAtoll,MarshallIslands(fromWeisler2002a).ThedarkenedlayerIIIisaburiedAhorizon.B,Atypicalprofile
(TP1Northprofile)showingthesuspectedgardeninglayeratsiteHEN-23,HendersonIsland.
1,100 yr B.P. (Weisler 1995). Suspected The sample areas consist of soil pockets, up
horticultural soils in the northern and north- to 3 m across, between bedrock outcrops and
eastern part of the island are associated with stone mounds. The archaeology of Hender-
relic stands of the Polynesian introduced son Island is discussed by Weisler (1995,
shrub Cordyline fruticosa and were sampled. 1997, 1998).
.
PlantMicrofossilsfromtheMarshallIslandsandPitcairnGroup HorrocksandWeisler 265
Pitcairn Island (4.5 km2) is much smaller layersII–IVarecolluvialwithminoramounts
than Henderson, although it is higher (ca. of dispersed macroscopic charcoal, layer V is
330 m) with well-developed soils and a a cultural deposit, layer VI has minor
greater variety of habitats. There are few amounts of dispersed charcoal, and layer VII
springs on Pitcairn, and water flows down hasno macroscopic charcoal.
some of the valleys after rain. Radiocarbon
evidence from Pitcairn is minimal, but occu- materials and methods
pationshouldbesimilartothatatHenderson
at ca. 1,100 yr B.P.; the earliest date associ- A total of 24 sediment samples, eight from
ated with human activity is 615G105 yr the Marshall Islands (samples 1–8), five from
B.P. (Weisler 1995). Two soil profiles were Henderson Island (9–13), and the remainder
sampled on Pitcairn. The first comprised (14–24) from Pitcairn Island, were taken
stratigraphic samples taken from the east side directly from sections and examined for evi-
of asmalldrainageatDownFence(8mfrom denceof plantmicrofossils(pollenandstarch
awalkingtrackandbridge),thoughttobethe residues).Descriptionsandprovenanceof the
result of landscape disturbance upslope on samples are given in the Appendix. Starch
the plateau, the main gardening area today. residue analysis was carried out on all sam-
The second profile comprised stratigraphic ples. Pollen analysis was carried out on all
samples taken at The Landing, situated at samples except 2,7, and13.
the base of a slope, ca. 3.5 m above sea level Pollen (and spores) was recovered using
and <10 m from the water. This sequence the standard acetylation and hydrofluoric
recordsuplandlandscapechange.Furtherde- acid method (Moore et al. 1991). The sum
tails of Pitcairn Island archaeology are dis- comprised 200 pollen and spores, and slides
cussed by Gathercole (1964) and Weisler were scanned for types not found in the
(1995). count. Relative concentrations of fragments
The Henderson Island samples comprise of microscopic charcoal, recovered along
samples 9–13. Samples 9 (layer I) and 10 with pollen, were noted subjectively.
(layer II) are from a suspected prehistoric Starch residues were recovered by density
gardening area (site HEN-20 [Weisler separation using the method given in the
1997]) on the north end of the plateau, just work of Horrocks (2004b). Presence or ab-
inland from site HEN-3. Samples 11–13 are senceof starchresidueswasnoted.Identifica-
from a suspected prehistoric gardening area tion of starch residues was carried out with
(site HEN-23) near the northwestern edge the aid of a modern reference collection,
of the plateau between rock shelter sites which includes the starch staples (see begin-
HEN-11and12intheclifffacebelow(Weis- ning of article), minor starch crops, and
ler 1997) (Figure 1B). Layer I is recent hu- many of the remaining plants known to have
mus, II a suspected gardening layer, and III beencultivatedintheprehistoricPacific.The
is sterile subsoil. many studies available on starch grain mor-
ThePitcairnIslandsamplescomprisesam- phology were also consulted (Reichert 1913,
ples14–24.Samples14–17(layersI–IIIB)are Seidemann 1966,Hoover 2001).
from Down Fence, just inland from the The radiocarbon age determinations were
northeastern coast of Pitcairn Island in a carried out by Beta Analytic Inc. using stan-
small gulch that drains from the plateau at dard procedures for conventional sample
Flatlands (Weisler 1997) (Figure 2). Layer I analysis (Weisler 1995).
is anhistoricA horizonandIIispossiblyhis-
torically deposited upland clay. IIIA and IIIB
results
are very late prehistoric or very early historic
layers with dispersed macroscopic charcoal.
Palynology
Samples 18–24 (layers I–VII) are from The
Landing, located on the west side of Bounty All except one of the samples analyzed for
Bay (Figure 3). Layer I is the A horizon, pollen has high concentrations of micro-
Figure2.SamplefromtheEastprofileatDownFence,PitcairnIsland.Samplenumbersandpositionsareshownin
boxes(seeAppendixfordescriptions).ThelateradiocarbonagedeterminationforlayerIIIBsuggestsrapidaccumula-
tionof overburden.
.
PlantMicrofossilsfromtheMarshallIslandsandPitcairnGroup HorrocksandWeisler 267
Figure3.SamplefromTheLanding,PitcairnIsland.Samplenumbersandpositionsareshowninboxes(seeAppen-
dixfordescriptions).Muchof theprofilematerialiscolluviumfromupslope.
scopic charcoal. The exception, the lower- All of the Pitcairn Island samples con-
most sample of The Landing profile from tained sufficient pollen for analysis (Figure
Pitcairn(sample24,Figure3),hasarelatively 4). Like the Marshall Island samples, mono-
verylowcharcoalconcentration.Preservation lete fern spores (55–80%) dominated the
of pollen ranged from moderately good to palynomorph assemblages in the shorter Pit-
poor.Noneof thesamplesfromtheMarshall cairn profile (Down Fence), although Myrta-
Islands had sufficient palynomorphs for ceae shrub (up to 25%) and Pandanus tree
counting, although a few types were noted (10–15%) pollen and trilete fern spores (up
during scanning (Figure 4). Samples 9, 10, to 15%) also had substantial values. In the
and12 from Henderson Island contain abun- longer Pitcairn profile (The Landing), as-
dant pollen, the assemblages of which are semblages were dominated by both monolete
overwhelmingly dominated by monolete fern fern spores (30–60%) and Pandanus (20–
spores (both 95% psilate and 5% patterned). 55%) pollen. Small amounts of Asteraceae
The monolete spores in sample 12 were too shrub pollen (long spine) were noted in the
degraded to differentiate. Traces of a few uppermost two samples. Pollen of several
other types palynomorphs were found in the disturbance-related herbaceous taxa, notably
Hendersonsamples,notablypollenconsistent Chenopodiaceae, Poaceae (grasses), and Ci-
with Alocasia/Amorphophallus in sample 9 chorieae (a tribe of Asteraceae with distinc-
(Figure 5) and pollen of Pandanus in samples tive pollen), occurs in the upper part of the
9 and10. Because the former is first reported profile. Cyperaceae (sedges) pollen was noted
here, modern examples are shown in Figure in the lowermost sample (sample 24 [Figure
6. Pollen of woody trees was not found in 3]). In addition, pollen slides of this sample
anyof the Henderson samples. show a high concentration of other plant
-e
nh
et
percates
nar
ep
ollse
pe
esentedlin
reprdott
ns).Barsofslides,
criptionning
dessca
eg
mplurin
sad
ored
fot
ndixesn
eu
pd
Apesi
eehr
(src
upsta
Groor
cairncount
ePitafter
thed
andnot
dsen
anoll
MarshallIslepresentsp
heþrng.
mfromtounting,heLandi
acT
andstarchresiduediagrwithsufficientpollenfornceprofilefromthatof
n e
PollempleswnF
4.saDo
re iner
Figu tagesshort
.
PlantMicrofossilsfromtheMarshallIslandsandPitcairnGroup HorrocksandWeisler 269
material,whichalthoughdegradedappearsto rocks (2004a), and Horrocks et al. (2004b)].
comprise mainly sclerenchyma cells (not Thisisbasedonstarchgrainandxylemvessel
found in any of the other pollen samples in morphology, and raphide concentration.
this study). Pollen of woody trees was not (Starch grains are highly concentrated in the
foundin any of the Pitcairnsamples. tubers, whereas calcium oxylate crystals and
xylem are found throughout the plant, with
raphides in particular often in very high con-
Starch Residue Analysis
centrations.)Althoughxylemwithhelicalwall
In 13 of the samples examined for starch res- thickening may be found in other taxa, this
idues we identified a range of starch residue was virtually the only type of xylem found
types (Figure 4). Two of the samples not in the samples. However, xylem vessels (and
yielding starch residues, samples 2 (outside druses) of the four cultivated Pacific Araceae
pit, Marshalls) and 13 (sterile subsoil, Hen- are difficult to distinguish betweenspecies.
derson [Figure 1B]), were virtually devoid of The following rationale for Araceae mi-
any organic material. In the Marshalls and crofossil identification is from the work of
Henderson samples, starch residues comprise Horrocks and Bedford (2005). Of the four
spherical to subspherical starch grains, cal- Araceae, Colocasia esculenta can be differenti-
cium oxylate crystals (raphides and druses), ated from the other three on the basis of size
and xylem vessels with helical walls (Figures of starchgrains(Loyetal.1992).Theformer
5 and 7). Residues in the Pitcairn samples has a mean grain size of <5 mm (most grains
are different, comprising ovate to subtriangu- are <3 mm) whereas Alocasia macrorrhiza,
lar, often bell-shaped starch grains, with a Cyrtospermamerkusii,andAmorphophalluspaeo-
vacuoleatthecentralhilum,fissuredinlarger niifolius have mean grain sizes of 10–14 mm.
grains, and xylem vessels with slitlike pits However, overlap in standard deviations
(Figure 8). Calcium oxylate crystals were not within this ‘‘non-Colocasia’’ group, together
foundwiththestarchgrainsandxylem inthe withdegradationandpossibleswellingof fos-
Pitcairnsamples. sil grains due to gelatinization, makes differ-
Preservation of the starch residues ranged entiation to species on this basis in our case
from good to poor. The starch grains were unreliable.
present individually or attached to cellulose ThefourPacificAraceaeeachhaveseveral
tissue and many showed signs of gelatiniza- types of raphide. However, the types pro-
tion: discoloration, swelling, and loss of bire- duced in the most abundance, ‘‘whisker ra-
fringence.The color change isfrom colorless phides’’ (Loy et al. 1992), are difficult to
toshadesof amberorbrown(Horrocksetal. distinguish between species. Fragmentation
2004a). Most xylem vessels were also dis- and degradation exacerbate the problem of
colored.Xylemwaspresentasbundlesof ves- raphide differentiation. Types distinctive to
selsorfragmentsof vesselelements.Raphides species are produced in relatively far lower
were present mostly embedded in cellulose amounts, and none of these was identified
tissueandoccasionallyinbundlesandasindi- with certainty in either the Marshall Islands
vidualfragments. or Henderson Island samples. Thus, on the
The starch residues in the Marshall Is- basis of morphology of microfossils found
lands–HendersonIslandsampleshaveacom- in the samples, we cannot confidently distin-
bination of attributes that occur in the guish between the four Araceae species.
Araceae,afamilyof mainlyherbaceousplants However, notwithstandingswelling of grains,
with aerial or underground stems or rhi- Colocasia esculenta can almost certainly be
zomes, often swollen as an adaptation for ruled out for the starch grains shown in Fig-
starchstorage(e.g.,corms,rhizomes[allcom- ures 5 and 7, typical of those found in the
monly referred to as tubers]). [Photomicro- samples, on the basis of size. Although some
graphs and descriptions of microparts of of the very large, discoloured grains we clas-
modern reference samples of Araceae are sifyasswollen(Figure7D)arewithinthesize
given in the works of Loy et al. (1992), Hor- range of several Dioscorea species(mean grain