holocene volcanic activity, vegetation succession, and ancient human land use: unraveling the...

nology 143 (2007) 83–105www.elsevier.com/locate/revpalbo

Review of Palaeobotany and Paly

Holocene volcanic activity, vegetation succession, and ancienthuman land use: Unraveling the interactions on

Garua Island, Papua New Guinea

C. Lentfer a,b,⁎, R. Torrence c

a Archaeology Program, School of Social Science, University of Queensland, St. Lucia, Queensland 4072, Australiab Centre for Geoarchaeology and Palaeoenvironmental Research, School of Environmental Science and Management,

Southern Cross University, Military Road, Lismore NSW 2480, Australiac Anthropology, Australian Museum, 6 College Street, Sydney NSW 2010, Australia

Received 30 June 2005; received in revised form 2 June 2006; accepted 8 June 2006Available online 1 September 2006

Abstract

An integrated approach to the reconstruction of vegetation history and human land use during the Holocene on Garua Island,Papua New Guinea analysed sediments and plant microfossils (phytoliths and starch granules) together with archaeological data.The long-term record is punctuated by a series of volcanic disasters, where repeated cycles of massive destruction were followed bydiffering cycles of forest regeneration. The plant microfossil record shows that instead of long-term forest recovery, the overallpattern of regeneration was progressively more disrupted. Through time regeneration was halted earlier in the sequence and thenreverted to increasingly open plant communities dominated by grasses. The temporal patterns of burning, stone artefact discard,and plant introductions demonstrate that the increased impact of human systems of land management was primarily responsible forthe temporal patterning. Most notably, the study shows that human interference begins much earlier than expected given previousarchaeological research and relatively intensive burning and landscape modification, possibly indicating cultivation, predates theintroduction of Lapita pottery.© 2006 Elsevier B.V. All rights reserved.

Keywords: volcanic impact; vegetation recovery; land use intensification; Lapita; phytoliths and starch; Papua New Guinea

1. Introduction

Aseries ofmajorHolocene eruptions from theDakatauaandWitori volcanoes deposited thick layers of airfall tephraover vast areas of the island of New Britain in Papua NewGuinea (Machida et al., 1996; Boyd et al., 1999; Torrence

⁎ Corresponding author. Archaeology Program, School of SocialScience, University of Queensland, St. Lucia, Queensland 4072,Australia. Tel.: +61 2 66854210; fax: +61 2 66854210.

E-mail address: [email protected] (C. Lentfer).

0034-6667/$ - see front matter © 2006 Elsevier B.V. All rights reserved.doi:10.1016/j.revpalbo.2006.06.007

et al., 2000). Not surprisingly, these environmentaldisasters had major impacts on the human history ofsettlement in this region. Archaeological research hasrevealed that periods of abandonment in the order ofseveral hundred to one thousand years followed the largestevents (Torrence et al., 2000; Torrence, 2002a,b). Theeffects of these disasters on human cultures, however,particularly in terms of changes in land use, aresurprisingly difficult to discern when the volcanichistory is examined over the long term, i.e. the Holoceneas a whole. Instead of a punctuated pattern tracking the

mailto:[email protected]

http://dx.doi.org/10.1016/j.revpalbo.2006.06.007

Fig. 1. Location map of Garua Island showing the FAO site and Dakataua and Witori volcanoes.

84 C. Lentfer, R. Torrence / Review of Palaeobotany and Palynology 143 (2007) 83–105

volcanic history, changes in the nature and distribution ofstone artefact assemblages – almost the only archaeo-logical data – are slow and directional. This pattern has

been interpreted as the consequence of an increasinglyintensified system of land use (Torrence et al., 2000;Torrence, 2002a,b). To date, the reconstruction of

Table 1Holocene eruptions identified on Garua Island (Machida et al., 1996;Torrence et al., 2000; Petrie and Torrence, in preparation)

Eruption Cal BP Volume(km3)

Type

Dakataua Dk 1410–1240 10 Phreatomagmatic, plinian,ignimbrite forming

Witori W-K3 1880–1550 6 Phreatomagmatic, plinian,ignimbrite forming

Witori W-K2 3640–3360 30 Phreatomagmatic, plinian,ignimbrite forming

Witori W-K1 6200–5740 10 Plinian, ignimbrite forming

85C. Lentfer, R. Torrence / Review of Palaeobotany and Palynology 143 (2007) 83–105

human land use has been based solely on archaeo-logical evidence, but in this paper we examine theseissues from the very different perspective of vegeta-tion history as reconstructed primarily through a studyof phytolith assemblages. We begin by analyzing thesedimentary sequence through particle size, claymineralogy and, especially, plant microfossils. Sec-ondly, we attempt to understand the effects of volcanicactivity on the pattern of plant regeneration at severaltemporal scales. Thirdly, we consider the impacts ofhuman disturbance on vegetation history. The studyprovides a rare picture of the complex interactionsamong environmental disasters, vegetation history andhuman disturbance. The results provide a new pictureof the history of human land use in this region whichchallenges some previous reconstructions.

2. Setting

Garua Island is located in a sheltered bay adjacent tothe Willaumez Peninsula in the province of West NewBritain in Papua New Guinea (Fig. 1). It is in closeproximity to the Dakataua and Witori volcanoes on thenearby mainland of West New Britain, and has beensubject to volcanic ash fallout from each of them. Con-sequently, the whole island is overlain with layers ofvolcanic tephra. Several of these, including W-K1, W-K2, W-K3 and Dk, form distinctive marker beds and arewell dated (Table 1).

The island is predominantly comprised of two vol-canic cones, partly fringed by a narrow coastal plain.The studied archaeological site, coded FAO, overlooksthe sea from its position on the broad top of a distinct hillon the flanks of the eastern volcanic cone. It has steepslopes on all but the south side where there is a longridge running up from the stream formed between thetwo volcanic halves of the island. A series of 1 m2 testpits were excavated at this locality during the 1990s byan archaeological research team and several were sam-

pled for geochemical and microfossil analyses (e.g.Therin et al., 1999; Torrence et al., 2000; Torrence andStevenson, 2000; Parr et al., 2001).

The focus of this paper is a 1 m2 test pit labeled 1000/1000 (Fig. 2). Archaeological excavation to a depth of310 cm followed natural strata which were arbitrarilysubdivided into 10 cm thick spits, with the exception ofthe unconsolidated airfall tephras (Layers 7 and 10)which do not contain artefacts and were thereforeremoved in bulk. Only one AMS radiocarbon date wasobtained from 1000/1000 because of poor preservationof organic material but AMS dates from nearby test pitswhere the stratigraphy was duplicated provide addition-al information (Table 2) and these have been supple-mented by a series of obsidian hydration determinations(Torrence and Stevenson, 2000). The identification ofthe upper units, in particular Layers 7 and 10, by grossphysical and geochemical means as derived from well-dated eruptions of the Dakataua and Witori volcanoes(see Table 1) tightens the chronological framework forthe stratigraphic sequence.

3. Sedimentary history

The first objective of the study was to interpret thesedimentary sequence represented by the stratigraphicprofile of the FAO 1000/1000 test pit (Fig. 2). To someextent, the upper section of the test pit, Layers 7 to 11,matches the physical appearance of the well-dated Ho-locene sequence for the area as summarised in Table 1.At FAO, unconsolidated airfall tephras (W-K2, Layer 7;Dk, Layer 10), characterised by relatively coarse lapilliand pumice, are overlain with tephra accretion (contin-uous build up of fine ash layers, as described by Neall,1977) and variable palaeosols. During excavation,Layer 9 was recognised as a volcanic tephra but it didnot have the distinctive physical characteristics of any ofthe known Holocene tephras. In contrast to the up-per part of the section, Layers 6 to 1 consist of heavilyweathered clay sediments. The light colour, coarsertexture, and the presence of pumice fragments enabledLayers 5, 3 and 2 to be tentatively identified as airfalltephras, but none matched the physical appearance ofW-K1. Finally, the depositional history of the heavy claylayers, 6, 4 and 1, was unclear.

To further our understanding of the stratigraphy andto enable a more definitive reconstruction of sedimen-tary processes at FAO, a number of analyses of theparticle size, geochemistry, clay mineralogy, andweathering were undertaken. The distributions ofphytoliths, burned phytoliths and starch granules inthe sediments were also examined.

Fig.2.

Sectio

ndraw

ingof

FAO

1000/1000andsummaryof

theHoloceneprehistory.


Table 2Radiocarbon dates from the FAO archaeological site

Unit Layer Material Lab. No. 14C Calibrated BPa

1001/999 11 Nutshell Beta-72139, CAMS-13072 1100±60 1180–9201000/1010 8 Nutshell NZA 3738 2439±64 2720–2350990/990 8 Nutshell NZA 3729 2452±67 2720–23501000/1000 6 Nutshell NZA 2901 3532±66 3990–3630a OxCal v. 3.9 (Bronk Ramsey, 2003).


3.1. Particle size

Particle size analysis of the coarse-grained gravel andsand, fine sand, silt and clay component of sedimentswas used to confirm the presence of the various tephralayers, previously identified, and determine levels ofsoil development. Following the excavation of the FAO1000/1000 test pit, 43 sediment samples were collectedfrom the face of the north wall (see Fig. 2) down acontinuous column from the top to the base of thedeposit — samples 1 to 24 at 5 cm intervals, andsamples 25 to 43 at 10 cm intervals. Particle size wasdetermined for a selection of these samples, represen-tative of each of the layers with the exception of Layer 3(Table 3; Fig. 3) [see Therin (1994) and Therin et al.(1999)]. From this analysis, distinct peaks in coarse sandand gravel in samples 6 and 18 (Layers 10 and 7respectively) mark the Dk and W-K2 tephras. Likewise,peaks in samples 31 and 41 show the tephra horizons ofLayers 5 and 2. In contrast, Layer 6 is distinguished by

Table 3Layers from FAO 1000/1000 showing sample locations for phytolithsand starch (see Fig. 2)

Layer Phytolith samples Starch samples a

1 43, 42 431–2 41 413 40, 393–4 384 37, 36, 35, 34, 33 37, 344–5 32 325–6 31 316 29, 27, 26, 25, 24, 23, 22, 21 30, 28, 26, 247 20, 19, 18, 15 18, 157–8 168 15, 14, 13, 12 138–9 119 10, 9 109–10 810 7, 6, 5 610–11 411 3, 2, 1 2

Sample 30 was not available for phytolith analysis.a These samples were used for particle size analysis.

its comparatively high clay component, indicativeperhaps of extensive weathering and soil development.There are no coarse-grained horizons indicative of theW-K1 tephra in this layer. Layer 8 is more coarse-grained than Layer 6 and might be thought a tephra, butLayer 8 has much higher levels of silt and clay relativeto the adjacent unconsolidated W-K2 tephra Layer 7.Similarly, the uppermost layer of topsoil, Layer 11, isconsiderably finer-grained than the unconsolidated Dktephra layer, Layer 10, and Layers 4 and 1 are fineroverall than the adjacent tephra Layers 5 and 2. Theparticle size analysis, therefore, clearly identifies uncon-solidated and coarser-grained tephra layers. From thesilt and clay content, Layers 1, 4, 6, 8 and 11 appear tohave various levels of soil development and are all likelyto have been derived from finer-grained tephra accre-tion. Notably, Layer 9, representing the W-K3? tephra,stands out as being different to the W-K2 and Dk un-consolidated tephra layers, by having a higher fine sandand silt component. It is more similar to Layer 8, and islikely to have resulted from similar depositional pro-cesses. Finally, there is no evidence for the 6200–5740 cal BP W-K1 eruption in the deposit.

3.2. Geochemistry of glass shards

Geochemical analysis was undertaken to characterizeeach of the tephras and to see if those in the upper leveldeposits were derived from the same set of sources astephras in the lower level (Machida et al., 1996;Torrence et al., 2000). Concentration of calcium oxidevs. iron oxide was determined for each of the tephralayers (Layer 10, 9, 7, 5, 3 and 2) using SEMmicroprobe analysis (Fig. 4). In the upper section, theW-K2, W-K3? and Dk tephras are each characterized bya different distribution of points on the scatter plots andare clearly derived from different eruptions emanatingfrom magmas with different geochemistries. Notably,the geochemistry of the tephras in Layers 5, 3 and 2 inthe lower part of the sedimentary sequence is clearlydifferent to those above. They have very similar patternsof distribution with all points forming tight clusters

Fig. 3. Results of particle size analysis for 16 samples (from Therin, 1994).


towards the bottom left of the scatter plots. It is verylikely that they originated from the same source al-though to date this volcano has not been identified.

3.3. Clay mineralogy, and patterns of weathering

Mineralogy of clays and levels of weathering areanother useful means for determining rates of tephradeposition, levels of surface exposure and soil devel-opment. Sub-samples of 42 of the 43 sediment samplestaken from FAO were fine-ground for X-ray diffractionanalysis as well as analysis of total carbon content andmetals. Phytoliths extracted from the same sedimentsamples using the standard heavy liquid extraction pro-cedure described in Lentfer and Boyd (1998, 1999) wereexamined to further determine degree and possiblecauses of weathering. At least 150 phytoliths were ex-amined from each sample. The frequencies of weathered(pitted) and burnt phytoliths, as defined by Piperno andBecker (1996), were recorded from each phytolith resi-due sample and converted to percentage value (Fig. 5).

The results of the XRD analysis show clear shifts inmineralogy in relation to 1) depth, 2) weathering pat-terns, and 3) rates of tephra deposition (Fig. 5). Feldspar,marked by peaks at 28 at 2θ, is the dominant mineral inthe upper sediments representing the soil/tephra accre-tion layer (Layer 11), the unconsolidated Dk tephra(Layer 10) and the unconsolidated tephra, Layer 9.However, a new peak also begins in Layer 9 at 8.5 at 2θ.This appears to be related to weathering and soil devel-opment and may represent a hydrated kaolinitic claytype, halloysite. Further tests are necessary to confirmthis. This new peak gradually grows stronger, showingmaximum amplitude in sample 13 in the accretion layer(Layer 8) above the unconsolidated W-K2 tephra. Fur-ther down the section, feldspar becomes the dominant

mineral again as weathering lessens in the unconsoli-dated W-K2 tephra layer (Layer 7) and weathering lev-els decline. Phytolith weathering levels are relativelylow from the surface to the base of Layer 7.

Beginning at the top of Layer 6, carbon levels dropdramatically and there is a sharp increase in phytolithweathering levels and in iron and manganese concentra-tions. A new peak corresponding to kaolinite at 12.2 at 2θbegins, growing stronger at sample 22 where high levelsof phytolith weathering and burning also occur. Furtherdown the profile (Layers 5 to 1) the mineralogy reverts toa dominance of ?halloysite, beginning at sample 23, pea-king in samples 25 and 26, and gradually declining in thesamples below. Kaolinite dominance appears again at thelevel of sample 28 and from there it continues to the baseof the profile, although some ?halloysite is still present.Phytolith weathering levels are mostly high throughoutthe lower section of the profile. A marked drop off inphytolith weathering occurs in samples 40 and 41 at thebase of Layer 3 as evidenced by excellent preservation ofepidermal silica skeletons.

3.4. Distribution of plant microfossils

Distribution and abundance of plant microfossils insediments builds onto our understanding of sedimentaryhistory, being good indicators of plant biomass, rates ofsediment deposition, and levels of soil development.Following phytolith extraction from the 42 sedimentsamples, residues were dried and weighed. The totalnumber of particles was counted in every field of viewexamined for the phytolith analysis and the proportionof phytoliths to total particles was calculated. Phytolithabundance was estimated by multiplying this value bythe weight of each residue sample. Starch granules wereextracted from 16 sediment samples (see Table 3) using

Fig. 4. Results of SEMmicroprobe analysis showing concentration of calcium oxide vs. iron oxide for each of the tephra layers (Layers 10, 9, 7, 5, 3 and 2)in FAO 1000/1000 (cf. Fig. 3, Torrence et al., 2000, p. 231).


Fig. 4 (continued ).



the protocol described in Lentfer et al. (2002). Due torelatively low yields, total numbers of starch granuleswere counted for each starch residue sample (Therin,1994; Therin et al., 1999).

The resulting phytolith and starch granule distributions(see Fig. 5) show considerable variation through thesedimentary sequence, with marked differences betweenunconsolidated tephra layers and palaeosol/accretionlayers. High concentrations of phytoliths occur in thetephra accretion layers (Layers 11 and 8) and lowconcentrations occur in the unconsolidated tephra hori-zons (Layers 10 and 7) where there is likely to have beenlow vegetation biomass. The depth profile of phytolithdistribution showing distinctive peaks in the buriedpalaeosol/accretion layers is typical of the Type-2 phytolithdepth function (PDF) identified by Stace et al. (1968) anddiscussed by Hart and Humphries (2003). A small degreeof downwashing or vertical displacement, possibly frombioturbation, is evident from the presence of phytoliths inthe unconsolidated tephra horizons (note especially therelatively high percentage value of burnt phytoliths insample 16 in the W-K2 tephra). However, since the verylow concentrations of phytoliths occur in the unconsoli-dated tephra samples relative to samples derived from thelayers with weathered tephra accretion (i.e. palaeosol/accretion layers), good phytolith/sediment integrity isindicated. This has also been noted by similar phytolithstudies on Garua sediments (Parr, 1999; Parr et al., 2001)and is further supported by experiments dating occludedcarbon in phytoliths (Lentfer and Boyd, in preparation).Consequently, for each major tephra accretion horizonwhere there is gradual downward displacement ofphytoliths, a pattern of distribution most similar to theType 1 PDF (but with less downward displacement) isindicated (Stace et al., 1968; Hart and Humphries, 2003).Notably, experimental studies examining starch grainmovements through soils have found the same pattern ofdistribution (e.g. Therin, 1998, Table 3, 66).

Within the upper part of the sequence, it is alsosignificant that Layer 9 representing theW-K3? tephra isdissimilar to the primary Dk and W-K2 airfall tephralayers (Layers 10 and 7). Rather, it has a similar patternof deposition to the W-K2 and Dk palaeosol/accretionlayers, Layers 8 and 11. Marked by high phytolithconcentration and a relatively high coarse sand and finesand and silt component, the depositional processes inwhich very small amounts of airfall tephra are added at aslow rate that would enable vegetation biomass to bemaintained, would likely to be similar for all threelayers.

Similar taphonomic processes further down thesequence are indicated by similar distribution patterns

of phytoliths. As with patterns observed in the W-K2and Dk unconsolidated tephra layers, low phytolithconcentrations indicate relatively rapid sediment accu-mulation between samples 30 to 41 in Layers 5 to 2.Importantly, when phytolith concentrations are com-pared with the results of the particle size analysis (seeFigs. 5 and 3), Layers 5 and 2, representing rapidlydeposited airfall tephra, can be clearly distinguished.Given the equally low phytolith concentration, it ispossible that the tephra represented by Layer 3 wasdeposited in a similar fashion. The depositional pro-cesses associated with Layer 4, exhibiting higher clayand silt fractions indicative of soil development but lowphytolith concentration, are more difficult to determine.Tephra accretion deposited over a relatively short periodof time but still of sufficient duration to allow weath-ering and soil development is one possible explanation.It is notable also that all phytolith concentrations ofsamples in the homogenous clay, Layer 6, are high inrelation to the unconsolidated tephra layers (Layers 10and 7) and the tephras identified in the lower part of thesequence. As there is high clay content concurrent witha very low coarse sand and gravel component, tephrahorizons within this layer, if indeed they occur, are notobvious. Given the characteristics of this particularlayer, therefore, it is likely to be representative of aslower rate of tephra accretion and a higher degree ofsoil development than other accretion layers.

There is a marked difference between the patterns ofphytolith and starch deposition within the sedimentarysequence. In contrast to the variable phytolith abun-dance, with high concentration in the palaeosol/ac-cretion layers and low concentrations in unconsolidatedtephra layers, the concentration of the N5 m starchgranules is relatively stable throughout the sequence ofthe selected samples. However, a large variation in thedistribution of b5 m granules is apparent and differsfrom phytoliths. Rather than following an expectedpattern of peaking in soil horizons, numbers of smallstarch granules peak in association with tephra layers10, 5 and 2. Clearly this pattern of distribution isindicative of a previously unidentified taphonomic pro-cess, possibly involving impacts of airfall tephra onvegetation, as discussed in more detail below.

3.5. Summary of sedimentary history

The analyses outlined above have improved ourunderstanding of depositional processes in the FAO1000/1000 sedimentary sequence. The tephras wereclearly differentiated by the geochemical analysis andthe absence of the W-K1 tephra in the sequence at FAO

Fig. 5. Summary diagram of FAO 1000/1000 sediment samples including phytolith concentrations, starch frequencies, proportion of burnt phytoliths, As, Fe and Mn concentrations, density of chippedstones, levels of weathering, density of chipped stones and summary of mineralogy. F=Feldspar; H=?Halloysite; K=Kaolinite. Lower case indicates that only small amount of the specified mineral ispresent. Valuesb1 marked by ●.

92C.Lentfer,

R.Torrence

/Review

ofPalaeobotany

andPalynology

143(2007)

83–105


was confirmed, suggesting a period of erosion followingits emplacement at 6200–5740 cal BP. This is not sur-prising given its re-deposited context elsewhere onGarua and on the nearby mainland (Torrence et al.,1990, 2000). It is notable that the tephras in the lowerpart of the sequence are different to those known tooriginate from either Witori or Dakataua. It is highlylikely that volcanoes other than Witori and Dakatauahave impacted upon the site within the early deposi-tional history of the sedimentary sequence examined.Rapid tephra emplacement, most likely in the initialplinian eruptive phase of each volcanic event identifiedin the sequence, has been indicated in Layers 2, 5, 7, 10and possibly 3. Slower tephra emplacement consistingof much finer particles, referred to as tephra accretion, isindicated in the alternating layers. It is apparent from thedistributions of phytoliths that emplacement of tephra inLayer 4 was relatively rapid. In contrast to Layer 4, thetephra accretion in Layer 6 appears to have occurredover a long period, possibly thousands of years. This isfurther supported by the high degree of weatheringshown by the elevated frequencies of pitted phytolithsand the presence of kaolinite. Finally, it is significantthat the distribution of phytoliths, particle size and claymineralogy in the W-K3? tephra layer (Layer 9) does notconform with expected patterns for rapid tephra em-placement. This layer is more characteristic of a re-latively slow rate of tephra accretion.

4. Vegetation history

Given the reconstruction of the sedimentary historyof the site we would expect to see episodes of massivedestruction followed by regeneration. Following themajor volcanic events represented in Layers 2, 3, 5, 7and 10, regeneration would typically favour pioneergrasses and ferns suited to exposed situations whichwould be subsequently replaced by more advanced re-growth gradually merging into primary forest. Wherethe volcanic impact was less, regeneration would typ-ically commence at a more advanced level. Importantly,each successional process would be broken by the nexteruptive episode. The plant microfossil record shoulddocument these differential revegetation regimes.

Diagnostic morphotypes from 42 phytolith residuesamples were identified by making comparisons withthe modern phytolith reference collection for the flora ofPapua New Guinea (Lentfer, 2003) and previouslypublished material (e.g. Kealhofer and Piperno, 1998).In this analysis, phytolith morphotypes are describedaccording to plant taxon, but additional descriptors forshape and anatomical origin are sometimes included

(Lentfer et al., 2000; Wallis et al., 2000; Lentfer, 2003;Madella et al., 2003).

4.1. Impacts

The patterns of variation in the distributions of phy-tolith taxa (Figs. 6, 7) show a complex distribution ofarboreal/herbaceous phytoliths and grasses, character-ized by seral vegetation regimes, which alternate bet-ween pioneer plant communities and more advancedregrowth. The four primary tephra layers (Layers 2, 5, 7and 10) are clearly associated with peaks of panicoidand other general grass morphotypes indicating that theimpacts of each of these eruptions were of sufficientmagnitude to decimate existing vegetation and encour-age new growth of light tolerant pioneer grasses. Ad-ditional measures for gauging levels of volcanic impactcan be gained from a closer examination of small starch-granule distributions (Fig. 5) and phytolith morphotypeassemblages (Figs. 6A,B and 7). Tephra layers 2, 5 and10 display high frequencies of small starch grains aswell peaks in non-grass and/or dicotyledonous phytolithmorphotypes. Although starch analyses were notundertaken for the basal samples of the W-K2 tephra(Layer 7) where starch accumulation would be expected,the pattern of phytolith concentration in this part of thesequence is typical of the other tephra layers. Thus, thereis clear evidence for taphonomic processes very dif-ferent to expected depositional processes characteristicof a stable environment.

Importantly, for the FAO site, this pattern is highlysuggestive of slight to severe vegetation damage fol-lowing catastrophic volcanic eruptions, where plantbiomass has accumulated in the unconsolidated tephrahorizon itself, due to limb breakage and defoliation ofexisting vegetation (cf. Table 3, Lentfer and Boyd, 2001,pp. 36–39). Peaks in grasses follow the starch and dicot/non-grass phytolith peaks in every instance. It is ap-parent that vegetation damage was moderate to severe inthe four tephra layers (2, 5, 7 and 10), resulting in burialof ground layer elements, massive limb breakage and ahigh level of tree loss. This accords with expecteddamage levels from thick ash deposits (Lentfer andBoyd, 2001). Significantly, there is one instance whereepidermal silica skeletons of one of the most vigorousand common pioneer grasses in the region, Imperatacylindrica, occur in association with primary tephra fall(Fig. 8). This is indicative of good preservation, but isfound in sample 40, Layer 3, towards the base of thesection where a high degree of weathering causing dis-articulation of phytoliths would be expected. The oc-currence of these well-preserved microfossils in this


otherwise anomalous context provides evidence for therapid burial of the ground layer vegetation of Layer 2 bysediment with high bulk density— possibly wet tephra.

4.2. Patterns of recovery

The plant microfossil record clearly demonstratesthe expected patterns of regeneration, in which earlyregrowth dominated by pioneer plants is gradually re-placed by less light tolerant forest plants. In Layer 3, forexample, after the deposition of what is likely to havebeen wet ash, recovery of regrowth forest is indicatedby the presence of ferns, palms, gingers and banana/Heliconia (Musaceae) morphotypes (Fig. 9).Subsequently the presence in Layer 4 of ginger/palm,banana/Heliconia, and palm phytoliths and relativelyhigh frequencies of non-grass morphotypes suggests thedevelopment of a closed forest with a high proportionof regrowth elements. Grass frequencies increase in theupper levels of this sequence, but the presence of theshade tolerant grass Scotochloa urceolata and treespecies including morphotypes derived from Canar-ium, Syzygium, figs (Moraceae) and Meliaceae species

Fig. 6. A: Phytolith diagram of monocot morphotypes. Percentage valuesb1%morphotypes. Percentage valuesb1% marked by ●.

towards the top of the layer is diagnostic of a trendtowards the development of a closed-canopy forestenvironment. The plant microfossil data concurs withthe sedimentary analysis outlined previously. Althoughthe low phytolith abundance indicates a period of rapidtephra accretion within this layer, given the plant taxaassemblage, tephra deposition at any one time musthave been thin enough to allow continuation of forestgrowth.

Similar patterns of initial recovery are seen in theearly part of the regeneration sequences in subsequentlayers. However, unlike Layers 3 and 4 there is signi-ficant disruption to forest recovery, ostensibly unrelatedto impacts of airfall tephra. In Layer 6, for example,vegetation in the initial stages of regeneration is dom-inated by grasses and regrowth forest with Euphorbia-ceae and Canarium species. In the mid to upper part ofthe sequence, however, where early regrowth elementswould be expected to decline in favour of closed forestelements, forest development ceases and a sharp in-crease in cuneiform morphotypes found in panicoidgrasses, particularly Imperata cylindrica (Fig. 6A), oc-curs. There is a subsequent decline in grasses after this

marked by●. B: Phytolith diagram of dicot, non-grass and redundant

Fig.6(con

tinued).


Fig. 7. Summary diagram of phytolith morphotype assemblages recovered from FAO 1000/1000 sediment samples. Percentage valuesb1% marked by ●.

96C.Lentfer,

R.Torrence

/Review

ofPalaeobotany

andPalynology

143(2007)

83–105

Fig. 8. Silica skeletons from the epidermis of Imperata cylindrica. Awas found in sample 40 at the base of Layer 3. B was extracted from the leaf of amodern plant specimen (WNB931) collected from New Guinea.


as arboreal and herbaceous regrowth elements againbecome more common. Similarly, the early part of thesequence in Layer 8 is dominated by panicoid grassesand pioneer regrowth forest elements. Above this, peaksin non-grass globular and dicot morphotypes in sample16 show that forest gradually regenerates. As in Layer 6,however, a marked break in this recovery evidenced by asharp increase in panicoid grasses and a decline in palmsand gingers follows, indicative of a period of severeforest degradation. This has also been noted by Parr

Fig. 9. Phytolith morphotypes found in Layer 3 of FAO 1000/1000. A: robglobular morphotypes very commonly found in palms and gingers. C: troughepalm morphotype.

et al. (2001) in the same horizon elsewhere in theFAO site. After this, forest recovery begins again andcontinues into Layer 9, where, in accordance with theresults of the sedimentary analysis, it appears that nodiscernible damage to vegetation resulted from theemplacement of the W-K3? tephra. In the latter part ofthis particular sequence, however, the forested envi-ronment declines as panicoid grasses again becomemore dominant. Finally, in the most recently depositedLayer 11, grasses dominate the entire sequence although

ust, faceted, elongate morphotype common in Selaginella species. B:d morphotypes commonly found inMusa species. D: echinate, globular

Fig. 10. Sample and vector plots of the 1st and 2nd principal components. The 1st principal component explains 19% of the total variation in thephytolith assemblages and the second principal component explains 14%. PT=Pteridophytes; S=sedge; BMB=bamboos; PAN=Panicoid grasses;GRAM=general grass morphotypes; ORCH=Orchidaceae; ZM=Zingiberaceae /Marantaceae; ZMS=Zingiberaceae/Musaceae; ZP=Zingiberaceae/Palm; P=Palm; C=Cordyline; MOR=Moraceae; COMP=Compositae; Eu=Euphorbiaceae; CTSM=Canarium/Trema/Syzygium/Meliaceae; ODIC=other dicots. (NB. The relatively low level of variance explained by the first 2 principal component scores is due to the mixed vegetation in Zone IIIsamples).


there is an increasing presence of figs and palms andgingers in the topsoil.

4.3. Long-term trends

The plant microfossil record provides tangible evi-dence for repeated cycles of change throughout the

sequence in association with each of the major tephraeruptive episodes. This is in keeping with our expecta-tions and signifies regular and fluctuating environmentalperturbation that typically characterizes an unstablevolcanic environment. However, anomalies in recoveryprocesses, and in particular the lack of expected pat-terns of forest recovery in Layers 6, 8 and 9, cannot be

Fig. 11. Canarium leaf morphotypes. A: fossil phytolith found in sample 22, Layer 6. B: phytoliths extracted from leaf of Canarium indicum (plantspecimen WNB926).


explained simply in terms of volcanism. The microfossilplant data point to strong influences from other envi-ronmental determinants. In addition, over the long termwe witness unidirectional changes that also cannot beexplained solely in terms of volcanism. The long-termtrend runs parallel to, and seemingly overrides, theshorter-term volcanic influences, particularly in theupper levels of the sequence.

Importantly, this trend, as seen in the summaryphytolith curves (Fig. 7) and in the results of a PrincipalComponents Analysis (PCA) (Fig. 10), is characterizedby an overall reduction in tree, shrub and other non-grass phytolith morphotypes throughout the sequencetogether with an increase in grasses. Four major zoneshave been identified according to the PCA. Zone IV,comprises samples from Layer 1 to the lowermost sam-ples in Layer 6 and has the strongest association withnon-grass (NG) morphotypes. Zone III, comprisingsamples from the upper levels of Layer 6 and Layers 7and 8, is representative of the weathered clay and W-K2layers. The position of this zone in the middle of the plot

Fig. 12. Phytolith morphotypes of Caryota rumphiana. A: fossil phytolith fourumphiana (plant specimen WNB220).

signifies mixed association with a broad group of plants,including grasses and non-grasses, indicative of vege-tation dominated mainly by early to middle stages ofregrowth. Zone II, comprising W-K3? samples 8, 9 and10 in Layer 9, clusters towards the top right side ofthe plot and provides the strongest association withMusaceae (MS, ZMS), Zingiberaceae and Marantaceae(ZM, ZMS), Canarium and other dicotyledonous trees(CTSM, ODIC), palms (P) and ferns (PT). From theposition on the first Principal Component axis, a strongassociation with grasses (GRAM, BMB) and figs(MOR) is also indicated. Zone I, representing the mostrecent period, comprises Dk samples from the upper-most layers, 10 and 11. These have strong associationswith figs (MOR), grasses (GRASS), bamboos (BMB)and sedges (S). Notably, sample 1, the uppermostsample representing the modern topsoil, stands out ashaving the strongest association with these elementswithin this zone.

This vegetation history provides strong evidence forinterruption to recovery processes and displays a long-

nd in W-K2 soil, Layer 8. B: phytoliths extracted from leaf of Caryota

Fig. 13. Bamboo phytolith morphotypes. A: fossil phytolith found in the Dk soil, Layer 1. B: phytoliths extracted from leaf sheath of Bambusa sp.(plant specimen WNB399).


term trend towards grass dominance and a more openlandscape. Our next task is to determine the cause. Is ithumans?

5. Human land use

The distribution of stone artefacts throughout theFAO stratigraphic profile makes it clear that humanswere present throughout its stratigraphic history (Fig. 5).The density of stone artefacts suggests that human use ofthe site was infrequent during the earlier part of thesequence. From Layer 6 onwards artefact densitiesincrease suggesting a more intensive use of the site. Amarked increase in artefact density first occurs in thelower level of Layer 6. There is another consider-able increase in artefact density in the upper levelimmediately below the unconsolidated W-K2 tephra. Incontrast, there is a marked drop-off in stone discardimmediately following the W-K2 eruption at 3640–3360 cal BP. although artefacts are present throughoutboth the W-K2 and W-K3? palaeosol/accretion layers.Subsequently, there is a sharp rise in Layer 11, followingthe Dakataua eruption of 1410–1240 cal BP.

Additional evidence for human presence can be seenin the burning record. Examination of the distribution ofburnt phytoliths, stone artefacts and grasses in the se-quence (Figs. 5 and 7) shows an overall positive rela-tionship between all three. This association establishesthe inference that burning was mostly related to humanactivity, rather than the result of natural causes. Thebasal layers show a noticeable increase in grasses asso-ciated with moderate burning events, especially inLayers 2 and 4. Major burning episodes in Layer 6,evidenced by burnt phytoliths, coincide with a suddenincrease in the density of stone artefacts. At the sametime interruption to forest recovery occurs and grasses

become more dominant. Major burning periods andpeaks in grasses also occur in Layer 8 in associationwith the first appearance of pottery at the FAO site, andsubsequently but to a lesser degree in Layer 9. Finally,along with a marked increase in density of stone arte-facts in Layer 11, there is another major peak in burntphytoliths and grasses.

Given these data, therefore, it is very likely, thathuman use of fire together with other forms of deliberateclearance were the primary cause for interruption to theregrowth regime, driving the shift towards a more openlandscape. The patterns of burning, artefact discard, andvegetation change evidence a major shift towards landuse intensification commencing in Layer 6, prior to theW-K2 eruption and the first occupation of people usingLapita-style pottery. Notably, this is coincident with thefirst appearance of bamboo (Figs. 5, 6A and 7) (cf.Kealhofer et al., 1999) and the highest accumulation ofCanarium nutshells in the sequence, as evidenced bymacroremain and phytolith assemblages (Fig. 11).Subsequent to Layer 6, the major burning episodes inLayer 8, the W-K2 layer associated with the first ap-pearance of Lapita, sees the first appearance of thefishtail palm Caryota rumphiana (Fig. 12), possibly aplant introduction signifying cultivation. This palm isrecognised as being one of the major starch producingplants in Southeast Asia (Yen, 1990, pp. 260). Impor-tantly, palm phytolith residues have also been found onstone tools in the corresponding layer of the FAO 970/1000 trench (Kealhofer et al., 1999, p. 538). Bambooalso persists in this layer, and again, is present asresidues on stone tools at FAO (Table 6, Kealhofer et al.,1999, p. 542). It is notable too, that in the same layerMusaceae phytolith frequencies increase, albeit slightly,and may be signaling the introduction of useful Musaand Heliconia species to the site. Subsequently, after the


Dakataua eruption, there is a sharp rise in burning levelsand a dominance of small grasses — and bamboos(Fig. 13). Such a major change suggests the commence-ment of another new and major period of land usemanagement, innovation and intensification.

6. Discussion

The sedimentary, palaeobotanical and archaeologicalstudies of the FAO 1000/1000 sequence combine toprovide a detailed insight into the Holocene of GaruaIsland and its human history. The vegetation history onthe island has been subject to the effects of catastrophicvolcanic eruptions for the entire period spanning theHolocene, and perhaps much longer. In several instancesthe wholesale destruction of the vegetation forced aperiod of regeneration. At no time during this period didthe environment maintain lengthy periods of stabilitycharacterized by mature tropical forest. Instead, theenvironment has fluctuated between early pioneer vege-tation and various stages of regrowth. In each of thesecycles panicoid grass phytoliths, including morphotypescommonly found in invasive grass species, such asImperata cylindrica, dominated the early stages ofregrowth after the volcanic event. Morphotypes ofparticular tree species also indicate the presence of re-growth forests. Increases in the frequency of ginger,palm and dicotyledonous tree and shrub morphotypessignal the development of canopy and more advancedregrowth forest.

The overall pattern reconstructed from the FAO sitematches that witnessed following single volcanic erup-tions in the region during recent times (Paijmans, 1973;Thornton, 1996, 1997; Whittaker et al., 1997; Thorntonet al., 2000; Lentfer and Boyd, 2001), but there are alsosignificant differences due to the long historical per-spective provided by this deep stratigraphic sequence.As one moves towards the present time, revegetation onGarua Island was stopped progressively earlier withinthe regeneration sequence and the vegetation returned toan increasingly immature state characterised by thedominance of grasses. Combined with evidence for thepresence of burning and the association of stone arte-facts within the same stratigraphic units, the most plau-sible explanation for the occurrence of this long-termtrend is increasing levels of human interference throughlandscape manipulation and modification. These resultshighlight the complex history of interactions on GaruaIsland within the dynamic context of its volcanicactivity, vegetation history, and human settlement.

Although the data support a primary role for humanmodification as an explanation for the truncated and

retroverted regeneration sequences, an additional factorshould be considered. Since the frequency of the vol-canic disasters has increased during the Holocene, itseems possible that the overall character of the regen-erated forest was gradually altered through time. Due toa range of stochastic factors, such as the nature of thedispersal of seeds into the region, a different primaryforest would have resulted from each individual regen-eration event. Changes in the availability of potentialplant foods resulting from this process could have had afollow-on influence on human subsistence patterns aswell as affecting the potential for forest managementstrategies through weeding, burning, etc. Given the de-gree to which phytolith morphotypes are specific toplant taxa, however, the contribution of random factorsin causing variability in the regeneration process cannotbe evaluated with the data at hand and requires furthertesting, perhaps through pollen analysis.

Another possible scenario is that the increased fre-quency of volcanic activity prevented the regenerationof climax vegetation to a greater degree in the morerecent periods. Could this help explain the long-termtemporal pattern toward a reduction in mature vegetationrather than increased levels of human interference? Thisalternative suggestion is negated by the pattern of re-generation, particularly in Layers 6, 8, and 9. In thesecases the change from early colonisers to forest begins asexpected but is interrupted. Rather than becoming ma-ture forest before the next eruption takes place, thevegetation reverts to grasses and palms. Clearly someexogenous variable is required to explain the breakdownin succession.

One possible explanation is that climate change led toincreased frequency in natural burning. Haberle et al.(2001) have argued on the basis of a regional synthesisof pollen data that the increase in climatic instability andparticular in El Nino activity during the mid to lateHolocene has contributed to higher levels of naturalburning within Malesia. Although climatic factors mighthave played a role in the patterns observed on GaruaIsland, the close correlation between degree of burningand numbers of stone artefacts in Layers 6, 8, 9 (Fig. 5)indicates that anthropogenic factors were the key agentsin reverting the process of forest succession and increating local environments dominated by grasses.

Recent phytolith studies of archaeological test pitsfrom the mainland of West New Britain about 20 km tothe south provide additional support for the Garua se-quence. In a study of 6 identical stratigraphic sequences,Parr (2003) used a reduced set of phytolith morpho-types to monitor vegetation histories during the past6000 years. Although the major changes detected by the


phytolith analyses were caused by localised shifts in sealevel, Parr (2003) also argued that changes in vegetationcould be ascribed to an increase in burning which was inturn caused by an intensification in human land use.

6.1. Changes in land management

The reconstruction of vegetation history has impor-tant implications for previous archaeological reconstruc-tions of human land use patterns on Garua Island. Basedon temporal variations in the nature of stone artefacts,Torrence (1992) proposed a trend toward increased se-dentism during the past 6000 years (cf. Fullagar, 1992).Additional data on the spatial distribution of stone andceramic artefacts across the island were used to supportthis model and to argue that the change in mobilitypatterns was caused by an increase in the intensifi-cation of land use, although no explicit reconstructionconcerning deliberate cultivation or domestication wasmade. Additional studies by Pavlides (2004, 2006)in southwest New Britain have confirmed the pro-posed trend toward an increase in land management (cf.Torrence et al., 2000). Within the overall trend on GaruaIsland, a major change in land use coincided with the re-colonisation of the island by people using Lapita potteryfollowing the W-K2 volcanic event. This was also cor-related with the loss of a significant stone artefact type,called stemmed tools, and a shift toward more casual useof stone artefacts (Kealhofer et al., 1999; Torrence et al.,2000; Torrence and Stevenson, 2000; Torrence, 2002a,b).The arrival of Lapita pottery users within Papua NewGuinea and the wider Pacific region has commonly beendepicted by archaeologists as bringing about a transfor-mation in land use caused by the initiation or significantintensification of agriculture (Bellwood, 1996; Kirch,1997) and the Garua Island archaeological sequence hasbeen cited to support this view (e.g. Spriggs, 1997, p. 88).

The new data fromFAOpresent a significant challengeto this widely accepted view about past human subsis-tence patterns. Firstly, a rapid rise in burning in the lowerpart of Layer 6 has been recorded. It indicates that thepeople who reoccupied the island following the volcanicevent in Layer 5 brought with them a system of landmanagement practices that incorporated burning andclearance. Furthermore, the scale of the burning in Layer 6(more than 4000 years ago) is much larger than waspresent previously. Secondly, the resulting nature of land-scape alteration as represented by the composition of thephytolith assemblages suggests a fairly intensive patternof clearance, which was intensified through time. Withinthe upper portion of Layer 6, representing a time just priorto the W-K2 eruption, the pattern of vegetation – mostly

dominated by grasses – is not unlike what one mightexpect in the presence of cultivation. It is also interestingthat bamboo morphotypes appear in this layer for the firsttime, perhaps signalling change in subsistence linked tocultivation in the vicinity of the site.

Thirdly, the scale of landscape modification after theW-K2 event, during the time when Lapita pottery wasintroduced, is not significantly different to the previouspattern, except for the slight increase in the presence ofMusaceae and the appearance of the Caryota palm.Fourthly, although these changes may suggest a slightshift in land management strategies, there are no sig-nificant differences that would indicate a qualitativechange in subsistence related to the introduction of aparticular horticultural regime. A sudden and majorchange in subsistence or land use patterns is therefore notassociated with the arrival of the colonising population,which introduced Lapita pottery to Garua Island.

In contrast, a quantum shift in the degree of humaninterference through clearance and burning occurs muchlater in time in the period after the Dk volcanic eruption.The summary of the vegetation changes provided by theprincipal component analysis (Fig. 10) shows that theintensification of land management accelerated in Layer10 and most probably represents sedentary communitiessupported by cultivation. Along with a high proportionof panicoid grasses, at this time we also observe a largeincrease in bamboos and in fig trees, both plant types thatare commonly planted around villages today. If peoplelived in relatively sedentary settlements during the timeof Lapita pottery or even in the pre-Lapita period ofLayer 6, then the structure of these villages was probablyquite different from those known in recent times.

6.2. Future studies

The significant results derived from this integratedapproach to environmental reconstruction could beenhanced further by additional research. Unfortunately,the lower two-thirds of the sequence is as yet undated,but it is likely that it spans at least the whole of theHolocene period, and judging from the nearby Kaponana Dari site (Torrence et al., 2004a), it may be con-siderably older. At this site, artefact bearing layers sim-ilar to the highly weathered tephras which underlie theHolocene Witori/Dakataua sequence at FAO have beendated to at least 35–40,000 BP at a site located on thenearby mainland (Torrence et al., 2004a,b). The pre-servation of phytoliths is much better at FAO, however,suggesting a somewhat younger date. Dating the lowerportions of the FAO section to obtain more preciseinformation about the volcanic events as well as the


initiation of burning and the timing of human interven-tion would be highly desirable, but the highly weatheredclays will demand new techniques since preservation oforganics or material suitable for luminescence dating isvery poor. Possibly radiocarbon dating of phytolithsmay play a role in this process.

Results obtained from the FAO section are alsosomewhat limited by its small spatial scale. Phytolithassemblages provide a localised record of vegetationthat may not be entirely representative of a whole re-gion. Although the temporal trends in the FAO se-quence are probably a good proxy for a larger area,small variations may be ascribed to spatial variations inhuman behaviour. For example, Parr et al. (2001) havedetected differences in the phytolith assemblages datingto Layers 8 and 9 and derived from various areas of theFAO site. The 1000/1000 pit, which is located in thecentral portion of the hilltop, recorded a higher level ofhuman disturbance than pits located on the periphery ofwhat was proposed to have been a settlement on thebasis of the spatial patterning of artefacts. Expandingthe phytolith study beyond the 1000/1000 pit to otherareas of the island would therefore be desirable. Usingthis approach it might also be possible to use intra-sitepatterning as a measure of site use and by proxy ofsedentism.

7. Conclusions

Temporal patterning in vegetation change on GaruaIsland raises important questions about the nature andscale of volcanic and anthropogenic impacts on a trop-ical ecosystem. Evidence for burning in association withthe presence of rare stone artefacts is present in theearliest part of the sequence, which is not yet dated butrepresents the earliest record for human settlement onthe island. Even at this early period humans were ac-tively manipulating and modifying their landscape tosome degree, although there is no evidence at this stagewhether or how forest management was used to enhancesubsistence patterns. Intensive modification throughburning and clearance was initiated earlier than theoldest radiocarbon dates at the site and by c. 5–4000 BPit seems likely that some form of cultivation was takingplace. The results illustrate that different reconstructionscan be obtained from plant microfossils as opposed to thepresence/absence or spatial patterning of particular kindsof material culture. Since plant microfossils are morelikely to be direct indicators of subsistence patterns, thenthe changes associated with the arrival of Lapita potteryon Garua Island and elsewhere in Near Oceania can beassumed to be related to social or ideological changes

rather than to different forms of land use or subsistence ashas been suggested previously.

Although phytoliths have long been used to interpretenvironmental change, this study shows that they canalso have a useful role to play in the interpretation ofsedimentary sequences. Grave and Kealhofer (1999)demonstrated this general approach through their anal-ysis of phytolith data to assess the degree of bioturbationin ditch fills. Our analysis found that the distribution ofphytoliths within the section helped confirm the positionof tephra horizons that were difficult to detect in thefield due to the high degree of weathering. The quantityand assemblage composition of phytoliths in the de-posits also helped clarify the differentiation of coarseprimary airfall tephras, slow accretions of fine tephradusts, and soils formed over more stable conditions.When integrated into a wider study, microfossil data canplay a useful role in interpreting sedimentary history.

The integrated approach used here – in which data onsediments, microfossils, and stone artefacts have beenbrought together – has provided a much more detailedand comprehensive picture of landscape change thanwould be possible by depending on a single analyticaltechnique. The seminal contribution of phytolith anal-ysis for understanding environmental change and inparticular human/environment interactions on GaruaIsland shows that this relatively new technique is avaluable addition to the normal battery of techniquesused in palaeobotany. In addition, the inclusion of lim-ited data on starch granules within the overall micro-fossil analyses has also strengthened the results,although a fuller study using one of several populationapproaches recently proposed for starch granules couldenhance the interpretations gained from these data (e.g.Lentfer et al., 1999; Torrence et al., 2004b). Additionalstudies of this type should have an important place in thestudy of how volcanic activity and natural disasters ingeneral as well as human history have shaped moderntropical environments.

Acknowledgements

This research was funded by the Australian ResearchCouncil, Pacific Biological Foundation, AustralianMuseum and Southern Cross University and receivedsubstantial assistance from the National Museum andArt Gallery (PNG), National Research Institute (PNG),West New Britain Provincial Cultural Centre (especiallyJohn Namuno and John Normu), and Walindi Planta-tion. For help on Garua Island we especially thank theKimbe Bay Shipping Agencies, Bob Wilson, Jo Bola, ateam of international volunteers led by site supervisors


Laurie Victor and Glenn Summerhayes and the localcommunity for their hospitality and friendship. We alsothank Bill Boyd for his support, advice and participationin the project, Jeff Parr for help with phytolith extrac-tions, Peter Jackson for the SEM analyses, CameronPetrie and Vince Neall for specialist advice, and PeterWhite, Robert Neal and Simon Haberle for commentson the manuscript.

References

Bellwood, P., 1996. The origins and spread of agriculture in the Indo-Pacific region: gradualism and diffusion or revolution and colonisa-tion? In: Harris, D.R. (Ed.), The Origins and Spread of Agricultureand Pastoralism in Eurasia. UCL Press, London, pp. 465–498.

Boyd, W.E., Lentfer, C.J., Luker, G., 1999. Environmental impacts ofmajor catastrophic volcanic eruptions in New Britain, P.N.G.: apreliminarymodel for palaeoenvironmental change. Proceedings of theInstitute of Australian Geographers Conference 1998, pp. 361–372.

Bronk Ramsey, C., 2003. OxCal Program, Version 3.9. University ofOxford Radiocarbon Accelerator Unit, Oxford.

Fullagar, R., 1992. Lithically Lapita. Functional analysis of flakedstone assemblages from West New Britain Province, Papua NewGuinea. In: Galipaud, J.-C. (Ed.), Poterie Lapita et Peuplement.ORSTOM, Noumea, pp. 135–143.

Grave, P., Kealhofer, L., 1999. Assessing bioturbation in archaeologicalsediments using soil morphology and phytolith analysis. J. Archaeol.Sci. 26, 1239–1248.

Haberle, S.G., Hope, G.S., van der Kaars, S., 2001. Biomass burningin Indonesia and Papua New Guinea: natural and human inducedfire events in the fossil record. Palaeogeogr. Palaeoclimatol.Palaeoecol. 171, 259–268.

Hart, D.M., Humphries, G.S., 2003. Phytolith depth functions in surfaceregolith materials. In: Roach, I.C. (Ed.), Advances in Regolith. CRC,Leme, pp. 159–163.

Kealhofer, L., Piperno, D.R., 1998. Opal phytoliths in Southeast Asianflora. Smithsonian Contributions to Botany, vol. 88. SmithsonianInstitution Press, Washington D.C.

Kealhofer, L., Torrence, R., Fullagar, R., 1999. Integrating phytolithswithin use-wear/residue studies of stone tools. J. Archaeol. Sci. 26,527–546.

Kirch, P.V., 1997. The Lapita Peoples: Ancestors of the OceanicWorld. Blackwell Publishers, Oxford.

Lentfer, C. 2003. Plants, people and landscapes in prehistoric PapuaNew Guinea: A compendium of phytolith (and starch) analyses.Ph.D. Thesis, Southern Cross University, Lismore, Australia.

Lentfer, C.J., Boyd, W.E., 1998. A comparison of three methods forthe extraction of phytoliths from sediments. J. Archaeol. Sci. 25,1159–1183.

Lentfer, C.J., Boyd, W.E., 1999. An assessment of techniques for thedeflocculation and removal of clays from sediments used inphytolith analysis. J. Archaeol. Sci. 26, 31–44.

Lentfer, C., Boyd, W.E., 2001. Maunten Paia. Volcanoes, People andthe Environment: The 1994 Rabaul Eruptions. Southern CrossUniversity Press, Lismore.

Lentfer, C., Boyd, W.E. in preparation. AMS dating of opaline silicaphytoliths from volcanic palaeosols.

Lentfer, C., Wallis, L.A., Bowdery, D., Hart, D.M., 2000. A universalphytolith key 3: Prismatic/ellipsoid/trapezoid (PET) class. Poster

presentation at the 3rd International Meeting for Phytolith Research,Brussels.

Lentfer, C.J., Therin, M., Torrence, R., 2002. Starch grains andenvironmental reconstruction: a modern test case from West NewBritain, Papua New Guinea. J. Archaeol. Sci. 29, 687–698.

Machida, H., Blong, R., Moriwaki, H., Hayakawa, Y., Talai, B.,Lolock, D., Specht, J., Torrence, R., Pain, C., 1996. Holoceneexplosive eruptions of Witori and Dakataua volcanoes in WestNew Britain, Papua New Guinea and their possible impact onhuman environment. Quat. Int. 35/36, 65–78.

Madella, M., Alexandre, A., Ball, T., 2003. International code forphytolith nomenclature 1.0. Phytolitharien 15, 7–16.

Neall, V.E., 1977. Genesis and weathering of andosols in Taranaki,New Zealand. Soil Sci. 123, 400–408.

Paijmans, K., 1973. Plant succession on Pago and Witori Volcanoes,New Britain. Pac. Sci. 27, 260–268.

Parr, J. 1999. The spatial patterning of palaeo-forest disturbance at anarchaeological site on Garua Island, West New Britain, as indicatedby fossil phytoliths. Honours Thesis, Southern Cross University,Lismore, Australia.

Parr, J. 2003. A Study of Landscapes in Numundo Region of WestNew Britain, Papua New Guinea, as Indicated by Fossil PhytolithAnalysis. Ph.D. Thesis, Southern Cross University, Lismore,Australia.

Parr, J., Lentfer, C.J., Boyd, W.E., 2001. Spatial analysis of fossilphytolith assemblages at an archaeological site in West NewBritain, Papua New Guinea. In: Clark, G.R., Anderson, A.J.,Vunidilo, T. (Eds.), The Archaeology of Lapita Dispersal inOceania. Pandanus Books, Canberra, pp. 125–134.

Pavlides, C., 2004. From Misisil Cave to Eliva hamlet: rediscoveringthe Pleistocene in interior West New Britain. In: Attenbrow, V.,Fullagar, R. (Eds.), A Pacific Odyssey: Archaeology andAnthropology in the Western Pacific. Papers in Honour of JimSpecht. Records of the Australian Museum, Supplement, vol. 29.Australian Museum, Sydney, pp. 97–108.

Pavlides, C., 2006. Life before Lapita: new developments inMelanesia's long-term history. In: Lilley, I. (Ed.), Archaeologyof Oceania: Australia and the Pacific Islands. Blackwell, Oxford,pp. 205–227.

Petrie, C., Torrence, R. in preparation. Chronology of volcaniceruptions, human abandonment and reoccupation of theWillaumezPeninsula, West New Britain, Papua New Guinea.

Piperno, D.R., Becker, P., 1996. Vegetational history of a site in theCentral Amazon Basin derived from phytolith and charcoal recordsfrom natural soils. Quat. Res. 45, 202–209.

Spriggs, M., 1997. The Island Melanesians. Blackwell Publishers,Oxford.

Stace, H.C.T., Hubble, G.D., Brewer, R., Northcote, K.H., Sleeman, J.R.,Mulcahy, M.J., Hallsworth, E.C., 1968. A Handbook of AustralianSoils. Rellim Technica Publications, Glenside, South Australia.

Therin, M. 1994. Subsistence through starch: The examination ofsubsistence changes on Garua Island, West New Britain, Papua NewGuinea, through the extraction and identification of starch fromsediments. B.A. Honours Thesis. University of Sydney, Australia.

Therin, M., 1998. The movement of starch grains in sediments. In:Fullagar, R. (Ed.), A Closer Look: Australian Studies of StoneTools. University of Sydney Archaeological Computing Labora-tory, Sydney, pp. 61–72.

Therin, M., Fullagar, R., Torrence, R., 1999. Starch in sediments: anew approach to the study of subsistence and land use in PapuaNew Guinea. In: Gosden, C., Hather, J. (Eds.), Prehistory of Food.Routledge, London, pp. 438–462.


Thornton, I., 1996. The origins and development of island biotas asillustrated by Krakatau. In: Keast, A., Miller, S.E. (Eds.), TheOrigins and Development of Island Biotas as Illustrated byKrakatau. Academic Publishing, Amsterdam, pp. 67–90.

Thornton, I., 1997. Krakatau: The Destruction and Reassembly of anIsland Ecosystem. Havard University Press, London.

Thornton, I., Mawdesley, N.A., Partomihardjo, 2000. Persistence ofbiota on Anak Krakatau after a three-year period of volcanicactivity. Trop. Biodivers. 7, 25–43.

Torrence,R., 1992.What is Lapita about obsidian?Aview from theTalaseasources. In: Galipaud, J.-C. (Ed.), Poterie Lapita et Peuplement.ORSTOM, Noumea, pp. 111–126.

Torrence, R., 2002a. Cultural landscapes on Garua Island, Papua NewGuinea. Antiquity 76, 766–776.

Torrence, R., 2002b. What makes a disaster? A long-term view ofvolcanic eruptions and human responses in Papua New Guinea. In:Torrence, R., Grattan, J. (Eds.), Natural Disasters and CulturalChange. Routledge, London, pp. 292–312.

Torrence, R., Stevenson, C., 2000. Beyond the beach: changing Lapitalandscapes on Garua Island. In: Murray, T., Anderson, A. (Eds.),Australian Archaeologist: Papers in honour of Jim Allen. CoombsAcademic Publishing, Canberra, pp. 324–345.

Torrence, R., Specht, J., Fullagar, R., 1990. Pompeis of the Pacific.Aust. Nat. Hist. 23, 3–16.

Torrence, R., Jackson, P., Pavlides, C., Webb, J., 2000. Volcanicdisasters and cultural discontinuities in the Holocene of West NewBritain, Papua New Guinea. In: McGuire, B., Griffiths, D.,Stewart, I. (Eds.), The Archaeology of Geological Catastrophies.Geological Society of London, London, pp. 225–244.

Torrence, R., Neall, V., Doelman, T., Rhodes, E., McKee, C., Davies,H., Bonetti, R., Guglielmetti, A., Manzoni, A., Oddone, M., Parr,J., Wallace, C., 2004a. Pleistocene colonisation of the BismarckArchipelago: new evidence from West New Britain. Archaeol.Ocean. 39, 101–130.

Torrence, R., Wright, R., Conway, R., 2004b. Identification of starchgranules using image analysis andmultivariate techniques. J. Archaeol.Sci. 31, 519–532.

Wallis, L.A., Bowdery, D., Hart, D.M., Lentfer, C., 2000. A universalphytolith key 1: lobate class. Poster presentation at the 3rdInternational Meeting for Phytolith Research, Brussels.

Whittaker, R.J., Jones, S.H., Partomihardjo, 1997. The rebuilding of anisland rain forest assemblage: how disharmonic is the flora ofKarakatau? Biodivers. Conserv. 6, 1671–1696.

Yen, D.E., 1990. Environment, agriculture and the colonisation of thePacific. In: Yen, D.E., Mummery, J.M.J. (Eds.), Pacific ProductionSystems: approaches to Economic Prehistory. Department ofPrehistory, School of Pacific Studies. Australian NationalUniversity, Canberra, pp. 258–277.

holocene volcanic activity, vegetation succession, and ancient human land use: unraveling the...

Documents