Introduction
The study of southern high latitude vegetation contributes to the search for modern analogues for the sparse Cenozoic palaeobotanical record around the Antarctic margin (e.g. Edwards Reference Edwards1921, Cookson Reference Cookson1947, Askin Reference Askin2000, Raine & Askin Reference Raine and Askin2001, Thorn Reference Thorn2001, Ashworth & Cantrill Reference Ashworth and Cantrill2004, Truswell et al. Reference Truswell, Quilty, McMinn, MacPhail and Wheller2005, MacPhail & Cantrill Reference MacPhail and Cantrill2006, Prebble et al. Reference Prebble, Raine, Barrett and Hannah2006). Understanding these fossil deposits has been an important pursuit for decades, in conjunction with interest in the development of the Antarctic Ice Sheet, its processes and response to global climate change (e.g. Francis Reference Francis1999). Established ‘nearest modern relative’ or ‘nearest modern equivalent’ techniques compare the composition of the fossil flora with that of similar modern day vegetation associations which, by extrapolation, allow assumptions to be made about the past vegetation stature and environmental conditions of growth. The fern-bush communities of the southern oceanic islands, e.g. the New Zealand sub-Antarctic islands, have been suggested as modern analogues for late Eocene floras (e.g. the Cytadela and Petrified Forest Creek floras of King George Island) (Birkenmajer & Zastawniak Reference Birkenmajer, Zastawniak and Crame1989, Askin Reference Askin1992, Birkenmajer Reference Birkenmajer1997). These floras were growing in cool, humid climates in the Antarctic Peninsula region as climate deteriorated before the permanent establishment of a significant East Antarctic Ice Sheet.
This investigation contributes to the analogue reference collection of phytoliths from the high southern latitudes first described in Thorn (Reference Thorn2004) by studying phytoliths in the modern plants and soil of sub-Antarctic Campbell Island, south of New Zealand. This investigation also provides a case study for a preliminary assessment of the likely accuracy of reconstructing past vegetation using dispersed phytoliths alone.
Phytoliths are microscopic particles formed from the solidification of biogenic silica (biosilica) gel within and between the cells of many plants. They have a varied morphology depending upon the plant species and host cell type, can be isolated idioblastic forms or skeletal structures, and are commonly between 5 and 200 μm in size. Due to the multiplicity and redundancy of phytolith forms (Rovner Reference Rovner1971) the taxonomic affinity of individual phytoliths, when separated from their source plant, is complex. However, ongoing studies on phytolith production in modern plants can identify the parent plant to generic and sometimes species level (e.g. Piperno & Pearsall Reference Piperno and Pearsall1998, Carnelli et al. Reference Carnelli, Theurillat and Madella2004). Due to their siliceous composition, they can survive in oxidizing conditions and commonly provide an in situ record of vegetation composition. Phytoliths therefore have the potential to both complement and supplement plant macrofossil and terrestrial palynology records and are becoming increasingly used in palaeoecological studies.
Campbell Island is located c. 600 km SSE of South Island, New Zealand, and comprises the remains of a dissected Pliocene volcanic cone (Fig. 1; Oliver et al. Reference Oliver, Finlay and Fleming1950, Beggs Reference Beggs1978, Michaux & Leschen Reference Michaux and Leschen2005). The majority of the island is covered in vegetation and peat soils (blanket peats average up to four metres thick), which began to accumulate after the last glaciation c. 13 000 bp (Fig. 2; Campbell Reference Campbell1980, McGlone et al. Reference McGlone, Wardle and Meurk1997, McGlone Reference McGlone2002). Prevailing winds are from the west to west–north-west and the climate is generally cloudy, moist and cool with minimal sunshine. Mean monthly temperatures vary little throughout the year, ranging from 6.5°C in August to 9°C in February (De Lisle Reference De Lisle1965, New Zealand Meteorological Service 1973). The modern vegetation is a mosaic, ranging from near-pristine to communities heavily modified by grazing and burning during farming between 1895 and 1931. The island is no longer settled and domestic sheep and cattle were eliminated in 1991 (Bestic et al. Reference Bestic, Duncan, McGlone, Wilmshurst and Meurk2005). The vegetation is a relatively depauperate flora of c. 132 indigenous vascular plants and 39 naturalised adventives including Dracophyllum (up to 5 m in height), and Coprosma-Myrsine scrub and tussock grasslands beyond the immediate coastal zone (Meurk & Given Reference Meurk and Given1990, Meurk et al. Reference Meurk, Foggo and Bastow1994a, Rogers & Walker Reference Rogers and Walker2005). A notable feature of the island flora are the distinctive and impressive macrophyllous forbs, commonly referred to as “megaherbs”, which prefer the cliff tops and, except for Stilbocarpa polaris, are endemic to New Zealand's southern oceanic islands (Fig. 2; McGlone Reference McGlone2002). Floristic structure and productivity on the island has been shown to be influenced by spatial variations in marine ionic input from precipitation (Meurk et al. Reference Meurk, Foggo, Thomson, Bathurst and Crompton1994b).
Fig. 1. Campbell Island locality maps. Contour and spot heights in metres above sea level. a. Location of Campbell Island in relation to New Zealand, also indicating the Auckland Islands where two plant specimens used in this study were collected. b. Campbell Island with general topography. Boxed in area is enlarged in c. c. Locations of quadrats named in the text. GPS co-ordinates and altitudes for each are in Table I.
Fig. 2. Landscape and vegetation views of Campbell Island. a. From Col-Lyall saddle looking south across Beeman Hill to Mount Honey. Tussock grassland in the foreground. b. From Tucker Cove at the base of Beeman Hill looking west across Dracophyllum scrub towards Menhir. c. Close-up of a megaherb field association at the cliff-top near quadrat COL2. Al = Anisotome latifolia, Ps = Pleurophyllum speciosum, Br = end of season Bulbinella rossii, Pv = Polystichum vestitum.
Methods
Fieldwork for this project was undertaken during March 2004 (austral autumn). Six sites within different vegetation associations (BEE1, BEE2, BEE3, CLS, COL1 and COL2) were located across central Campbell Island from the north-eastern slopes of Beeman Hill, north–north-west to the cliff tops of Col Ridge on the western side of the island (Table I, Fig. 1). Temporary quadrats (4 m2) were used at each site (except at COL1 where a 1 m2 quadrat was more appropriate due to the small size and low-lying nature of the vegetation, Fig. 3). At each site the soil surface was sampled, plant specimens collected and canopy cover surveyed. The peat soils were of varying depths (several centimetres to over three metres), and generally poorly drained. Peat slips were avoided, and the geomorphic settings of each site appeared relatively stable, so it is assumed that there would be neglible transport of phytoliths into and out of the quadrats through aeolian or run-off processes.
Fig. 3. Photographs of the four main survey sites. Arrows indicate north. All quadrats measured 4 m x 4 m, except COL1 at 1 m x 1 m. Q, quadrat poles. a. BEE3, Coprosma/Bulbinella/Polystichum low scrub. b. CLS, Polystichum/Bulbinella tussock. c. OL1, Stony subalpine herbfield. d. COL2, Megaherb field. Al = Anisotome latifolia, Br = Bulbinella rossii, Cc = Coprosma cuneata or C. ciliata, Ds = Dracophyllum scoparium, Mg = Marsippospermum gracile, Ph = Pleurophyllum hookeri, Ps = Pleurophyllum speciosum, Pl = Poa litorosa, P2 = Poa sp.2, P3 = Poa sp.3, Pv = Polystichum vestitum.
Table I. Location and ground cover characteristics of quadrats on Campbell Island. Latitude, longitude and altitude are GPS data. Vegetation association and ground cover type, overlapping, therefore ≥ 100% at each site, as observed in the field. m.a.s.l. = metres above sea level.
Each site was characterized by estimating the overall per cent cover of vascular vegetation, moss, leaf litter, bare ground and rock. All vascular plants within the boundaries were then identified and the per cent cover of their canopies estimated to the nearest 5% in two different height bands (< 30 cm and 30–200 cm, based on Allen Reference Allen1992). Overlapping canopies allowed the total canopy cover to be greater than 100%. Per cent canopy cover of each species from both height bands was then summed, normalized and used to represent an estimate of the quantity of phytolith-producing plant organs at the site. To constrain the study, it is assumed that no plants beyond the quadrat boundaries contribute to the phytolith assemblages within. Specimens of each plant species were then collected (to include as many different organs from each plant as possible) from the site in which they were first encountered. The individuals collected are assumed to be representative of their species on Campbell Island. Voucher specimens of phytolith-producing species are stored in the Victoria University of Wellington herbarium in New Zealand (WELTU).
Soil sub-samples (c. 50 g each) were collected from the centre and all quarters of each quadrat, then amalgamated to provide a representative sample for that site. It is assumed that accumulation of phytoliths in the soil surface may be only several decades old, on the basis of work on the peaty soils at Mount Hauhungatahi (Tongariro National Park, central North Island, New Zealand; Horrocks & Ogden Reference Horrocks and Ogden2000), which suggests a soil accumulation rate of 0.8–1.5 cm per century. The soil samples were collected from 2–5 cm below the current leaf litter, which removes the bias of the immediately overlying plant.
To compare the vegetation and underlying soil phytoliths, overall phytolith production of the vegetation within each quadrat has to be in the form of a list of morphotype proportions reflecting the species canopy cover, phytolith production rate and preservation potential. This is achieved by weighting the proportion of each morphotype from each phytolith-producing species using the following formula (Thorn Reference Thorn2006):
For example, for phytolith-producing species A:
where Y = relative biosilica contribution to site (%), W = biosilica production by dry weight (%), C = canopy cover (%). Species A to D contribute phytoliths to the soil at this theoretical site.
If the same morphotype occurs in more than one species (conservative forms reflecting phytolith redundancy) the weighted proportion from each species in which it occurs are summed for that morphotype.
Phytoliths were extracted from plants and soils using a wet oxidation technique outlined in Thorn (Reference Thorn2004). Plant material was cleaned using dilute hydrochloric acid and an ultrasonic tank. Concentrated sulphuric acid and hydrogen peroxide (an exothermic and effervescent reaction) removed the organic matter from the siliceous residue. Extraction of phytoliths from the soil samples followed a modified method of Pearsall (Reference Pearsall2000). Treatment with dilute hydrochloric acid removed carbonates and concentrated sulphuric acid and hydrogen peroxide were added, as in the plant processing procedure, to remove the organic material. Inorganic siliceous grains were separated from the less dense biosilica using sodium polytungstate heavy liquid at specific gravity 2.3. Quantitative processing for both plants and soil samples allowed the calculation of per cent biosilica content by dry weight of each sample. The biosilica residues were mounted in Canada Balsam and analysed on a Leica DMLB transmitted light microscope. A count of 300 phytoliths was the target for each sample and morphotypes classified using the recent International Code of Phytolith Nomenclature (ICPN; Madella et al. Reference Madella, Alexandre and Ball2005). This resulted in a slightly different terminology than a previous study on Campbell Island phytoliths (Thorn Reference Thorn2004), in particular the grouping of all rondel forms.
Results
Vascular vegetation covered between 60 and 100% of each quadrat, with the sparser sites being those at higher altitude. Sites on Beeman Hill were at the lowest altitudes, relatively sheltered from the prevailing westerlies and consisted mainly of shrub and fern vegetation associations (Tables I & II, Fig. 3). Sites BEE1 and BEE2 were dominated by Dracophyllum scoparium together with Blechnum novae-zelandiae at BEE1, and Isolepis aucklandica and Chionochloa antarctica at BEE2 (Table II). BEE3 was dominated by Coprosma, with a high proportion of Polystichum vestitum. Site CLS was dominated by Poa sp. 3 on the Col-Lyall saddle. Site COL1 was the most exposed cliff-top site, the smallest area analysed, and had the highest diversity with 14 species, despite a low vascular plant cover of 60% and very thin soil (evident by the 20% exposure of bare rock, Table I, Fig. 3). There were no ferns or shrubs at this site, with most of the vegetation consisting of low-lying herbs and a small proportion of grass (Poa litorosa). Site COL2 was in the middle of a megaherb field, also at the cliff top and highly exposed. The canopy mainly comprised Anisotome latifolia, Pleurophyllum speciosum and Bulbinella rossii with a minor proportion of grasses.
Table II. Species list and canopy cover for each surveyed quadrat at Campbell Island. Plant organs analysed: F = non-fertile fronds, L = leaves, R = roots, Se = seeds, Fl = flowers; S = herbaceous stems, W = woody stems, MF = macrophyllous forb, * = analysed for phytoliths in Thorn (Reference Thorn2004), ? = not collected. Data in italics refer to the quadrat from which a specimen was collected (and processed if in bold italics). # = specimen collected from Enderby Island in the Auckland Islands, ∝ = specimen collected from Skua Gull Flat, south Auckland Island (Fig.1). For specimen collection numbers refer to Turnbull et al. (Reference Turnbull, Adams, Beavan, Forsyth, Garnock-Jones, Hayward, Raine, Thorn and Vajda2004).
Each quadrat had between one and four phytolith-producing species and varied in overall vascular plant diversity between seven and 14. Grass (six species), rush (two), sedge (three), herb (17, plus four macrophyllous forbs), fern (four) and shrub (six) species were recorded (Table II). In total, 30 species have been analysed for their phytolith production, of which 16 were processed for this study and 14 had been analysed in a previous study (Thorn Reference Thorn2004; Table III, Fig. 4). Some species collections were not processed due to their small size and rarity in the vegetation (e.g. Abrotanella sp.). If they later prove to be phytolith producers, it is likely they would make only a negligible contribution to the soil phytolith floras in this study. Two species had also been collected from the sub-Antarctic Auckland Islands (Table II, Fig. 1) en route to Campbell Island, so collections were not duplicated. In summary, 11 of the 30 species (37%) were found to be phytolith producers and 19 are non-producers (Table II).
Table III. Plant phytolith production summarized by morphotype category. Morphotype diversity (when > 1) within each category is indicated in brackets. Carex trifida was counted quantitatively for this study in comparison to the qualitative presence/absence data recorded for this species in Thorn (Reference Thorn2004). φ = count data from Thorn (Reference Thorn2004). Anaphalioides bellidioides, Myosotis capitata and Acaena anserinifolia produced few forms, so counts were not practical and presence only is noted*. Stomate morphotypes were noted in the Hierochloe sp. specimen beyond the formal count. Equivalent morphotype terminology used in Thorn (Reference Thorn2004) for comparison with the more recent International Code for Phytolith Nomenclature (Madella et al. Reference Madella, Alexandre and Ball2005): Acicular (this study), Trichome (hair) (Thorn Reference Thorn2004); Bilobe, Bilobate; Conical, Hat-shaped; Elongate, Elongate; Globular, Mesophyll; Polygonal, Plate (polygonal); Rondel, Rondel; Saddle, Saddle; Tabular, Plate (rectangular).
Of the grasses, all were shown to be abundant phytolith producers (between 0.61 and 4.71% biosilica by dry weight) and all produced elongate and rondel forms (Table III). Form diversity in the grasses is high (eight to 18 broad morphotype categories). Many acicular and trapeziform forms were also produced by Hierochloe sp., Poa ?antipoda, Poa sp.2 and Poa sp.3. Poa sp.2 and Poa sp.3 have similar assemblages, so may be the same species. Identification of the grasses was not always possible to species level due to the lack of flowering heads or seeds in the collected specimen. Poa litorosa differs from the other three Poa spp. by the absence of acicular, bilobe, globular and trapeziform forms in particular, producing instead relatively more elongate and saddle forms. Poa ?antipoda produces the greatest relative abundance of trapeziform forms of all the grasses. Chionochloa antarctica is markedly different from the other grasses as its phytoliths are dominantly rondel in form. Two of the three sedge species are phytolith producers. Both Carex trifida and Uncinia sp. produce characteristically high relative abundances of conical phytoliths, which distinguish the sedges from the other species. Elongate forms are also common with morphotype category diversity higher in Uncinia sp. Of the herbs only Anaphalioides bellidioides, Acaena anserinifolia and Myosotis capitata produced any biosilica, and so few morphotypes or rare phytoliths that counts were unrealistic. Despite this, acicular, globular, hair base, tabular and vascular tissue forms were noted as present with spectacular hairs and hair bases in Myosotis capitata (Thorn Reference Thorn2004). There are 17 conservative morphotypes between the phytolith-producing plants. The ferns, rushes, majority of herbs (including the macrophyllous forbs) and the shrubs were non-producers.
Fragile forms include several acicular, all conical, thin elongate, stomate, vascular tissue and thin-walled globular forms. Species with a high abundance of fragile forms are the sedges (Carex trifida, 74%; Uncinia sp., 91%) and the herb Anaphalioides bellidioides, which produces almost entirely thin-walled globular orbicular psilate forms. In comparison, the grasses have relatively low proportions of fragile phytoliths from 0% (Poa litorosa and Chionochloa antarctica) to 17% (Hierochloe sp.). The proportions of each morphotype category within the vegetation phytolith assemblage (calculated using count data weighted by relative biosilica contribution per species from Eq. (1)) are compared to the directly observed soil phytolith assemblage for sites BEE3, CLS, COL1 and COL2 (Tables IV–VI).
Table IV. Relative contribution of biosilica (Y) by each phytolith-producing species to the vegetation phytolith assemblage at each quadrat. C, normalized canopy cover. Refer to Eq. (1) in text.
Table V. Soil phytolith assemblages summarized by morphotype category. Morphotype diversity (when > 1) within each category is indicated in brackets. Soils at quadrats BEE1 and BEE2 contained rare phytoliths, so counts were not undertaken.
Table VI. Comparative table of vegetation (V) and soil (S) phytolith assemblages for each quadrat.
Sites BEE1 and BEE2 each had only one phytolith-producing species from diversities of seven and 11 species respectively (Table IV). Carex trifida had only a 1% canopy cover at BEE1 representing all of the phytolith contribution to that site. Chionochloa antarctica produced all the phytoliths for site BEE2 with a canopy cover of 21%. Vegetation reconstruction from the phytolith assemblages was not therefore possible for quadrats BEE1 and BEE2 due to insufficient phytolith content in the soil (Table V). This is unsurprising at BEE1 as the only phytolith-producing species at this site was Carex trifida, which had a canopy cover of only 1% and produces only fragile, conical forms. Chionochloa antarctica morphotypes are of generally robust morphology in the BEE2 soil assemblage, but rarely occur, despite this species having a relatively high canopy cover of 21% (Tables IV & V). This may be explained by sampling if the Chionochloa plants were sufficiently distant from the actual soil collection localities within the quadrat.
At BEE3 (diversity 12) phytolith-producing species comprised only 11% of the canopy and although there is a significant (24%) phytolith contribution from the herb canopy (Acaena anserinifolia and Anaphalioides bellidioides), grass phytoliths dominate the vegetation input (Poa spp., 76%) and comprise 99% of the soil phytoliths (Tables IV & V, Fig. 4). The remaining 1% consists of forms not produced by the plants at this site. The lack of Acaena phytoliths is probably due to their fragility, despite their abundant production (5.44% biosilica by dry weight). Taken on its own, the soil assemblage suggests a grassland source vegetation. In fact, grass comprised only 9% of the canopy cover and the soil assemblage indicates none of the shrub (58%), fern (22%) or herb (11%) components of the source vegetation, which was noted as ‘Coprosma/Bulbinella/Polystichum low scrub’. The grass morphotypes in the soil at this site are conservative forms, so recognizing the species of grass represented is not currently possible. However, the overall Poa-type assemblage dominated by elongate and rondel morphotypes is evident. Comparing the morphotype relative abundances between the vegetation and soil assemblages suggests that there is some phytolith accumulation (mainly of robust forms) as the proportions of, in particular, rondel and trapeziform phytoliths are higher in the latter (Table VI). Overall, if a Poa grass is the only species recognized in the source vegetation from the dispersed soil assemblage at this site, then 91% of the original vegetation canopy (e.g. herbs, ferns and shrubs) at the site will have no lasting phytolith record (Fig. 4).
Fig. 4. Selected phytoliths from plant species recorded within the study quadrats on Campbell Island. All scale bars = 10 µm. * = previously discussed in Thorn (Reference Thorn2004). Acicular types compare to those described by Carnelli et al. (Reference Carnelli, Theurillat and Madella2004). a. Rondel, Chionochloa antarctica*, b. Acicular [Type 1], Hierochloe sp. (CLS), c. Trapeziform polylobate, Poa ?antipoda (COL2), d. Elongate columellate, Poa litorosa*, e. Acicular [Type 1], Poa sp. 2 (BEE3), f. Conical square verrucate 3, Carex trifida (BEE1), g. Conical square verrucate 1, Uncinia sp. (COL2), h. Acicular myosotis, Myosotis capitata*, i. Acicular [Type 16] and Tabular epidermal, Acaena anserinifolia*.
Fig. 5. Graphs to compare summary aspects of the canopy cover and phytolith assemblages for the four main survey sites. a. canopy cover, b. vegetation phytolith assemblages, c. soil phytolith assemblages. Narrow bars are clarified with the following notation: R = rush, Se = sedge, H = herb, O = other.
Poa sp. 3 dominated the phytolith contribution at site CLS (diversity 12) at 94% with minor additions from Hierochloe sp., Acaena anserinifolia and Anaphalioides bellidioides (Table IV). At this site only phytoliths produced by grasses are observed in the soil (with some conservative forms that are also produced in sedges), hence a grassland vegetation community again seems the only possible source compared with the observed ‘Polystichum/Bulbinella tussock’ association (Tables II & V). Hierochloe and Poa have many common morphotypes, so the two grasses cannot be distinguished from the soil assemblage, and the soil and vegetation assemblages cannot be directly compared. There is no record in the soil assemblage of the fern (11%), herb (10%), rush (1%) or shrub (8%) canopy components due to non-production or non-preservation (Fig. 5). Higher proportions of some morphotypes in the vegetation than the soil suggests the non-preservation of some fragile forms (Table VI).
At COL1 despite the highest species diversity of all the sites (14), only two species are phytolith-producers with grass phytoliths dominating (Poa litorosa, 88%), in addition to herb production (Anaphalioides bellidioides, 12%) (Table IV). Again, only grass phytoliths from the vegetation above are observed in the soil, suggesting a grassland source, with no record of the significant herb (70%) or rush (20%) canopy cover at this ‘stony subalpine herbfield’ site. Only 47% of the phytoliths observed in the soil were forms produced by the overlying vegetation (Fig. 5). COL1 is extremely exposed suggesting that, contrary to the assumption made, there is aeolian transport of phytoliths at this site. This quadrat was also smaller than other sites including bare ground and rock, with a thin and stony soil and slow accumulation of leaf litter (Fig. 3).
Finally, COL2 is located within a megaherb field association and had the highest number of phytolith-producing species (4, diversity 11) with biosilica contributions from Poa litorosa (52%), Poa ?antipoda (27%), Uncinia sp. (18%) and Myosotis capitata (3%) (Table IV). Poa spp. and Uncinia sp. are significant phytolith contributors, despite their low canopy cover, but share many conservative forms. None of the distinctive, but fragile, Uncinia sp. conical forms survived in the soil, so the majority of the soil assemblage consists of grass morphotypes (including some conservative forms also produced by sedges) (Fig. 5). A grassland interpretation is the inevitable result, with no record of the significant macrophyllous forbs (88%) or minor other herbs (4%) and fern (2%) components of the source canopy.
Discussion
Of the species recorded in the extant vegetation at the Campbell Island quadrats only 37% were found to produce phytoliths and hence there would only ever be the potential to reconstruct up to 37% of the source vegetation diversity from dispersed soil phytoliths alone. Further, this 37% is overwhelming biased by grass phytoliths of predominantly robust morphotypes, with the sedges also producing abundant, but fragile morphotypes of a lower preservation potential. Meurk et al. (Reference Meurk, Foggo, Thomson, Bathurst and Crompton1994b) found that the high altitude herbfields had a relatively high soil pH due to a mineral-rich soil and lower levels of nutrient input from aerosols due to the height above sea level. These conditions affect phytolith preservation by enhancing dissolution. However, mechanical damage or dissolution in the study sites is likely to be low due to the acidic nature and relative stability of the peaty soils. Therefore, the soil phytolith record is inevitably diluted by lack of production in the majority of herbs, ferns and shrubs, preservation potential and in certain sedimentary settings, phytolith transport. An additional complication arises with the difficulty of differentiating taxonomically characteristic morphotypes in the dispersed soil assemblages, even at Family level. For example, conical forms reflect the presence of at least two members of the Cyperaceae, and rondel forms are consistently present in abundance throughout the Poaceae (mainly Poa spp.). Many conservative phytoliths, e.g. elongate morphotypes, are produced by two or more species, sometimes from different Families or even broader taxonomic divisions. Individual species can only easily be defined when there are no conservative forms present, e.g. in this study at COL1.
The above complications in the interpretation of dispersed soil phytolith assemblages, here including 17 conservative morphotypes, allow only the reconstruction of the overlying source vegetation to a broad grassland community level for all sites. Overall, there is disagreement between the soil phytolith assemblage and the character of the overlying source vegetation canopy, with both the vegetation phytolith production and soil phytolith content reflecting the dominance of grass phytolith production and preservation.
A pilot study on soil surface samples from Campbell Island also recorded the dominance of grass phytoliths and, although conical forms were rare in the soil samples during the current study, two samples in this previous study contained significant abundances, suggesting that soil phytolith assemblages are indeed localized with minimal homogenization of assemblages over large areas (Thorn Reference Thorn2004). The grassland interpretation for the quadrats in this study is only really applicable to that at CLS with a 70% grass canopy cover. This overall result reflects a previous observation during a similar study at Tongariro National Park, central North Island, New Zealand, that the presence of grass in the source vegetation is likely to be over-represented in general, and the proportion of grasses in the source vegetation association cannot be estimated from the dispersed phytolith assemblage alone (Thorn Reference Thorn2006).
The nature of the pollen record on Campbell Island has been investigated through studies on the modern pollen rain (McGlone & Meurk Reference McGlone and Meurk2000) and Holocene peat profiles (McGlone et al. Reference McGlone, Wardle and Meurk1997), but little is currently known about the pre-Quaternary terrestrial palynology record. The terrestrial palynology and phytolith content of Neogene samples collected during this fieldwork will be discussed in a future publication. Similarly, the pollen rain, in agreement with this phytolith study, is dominated by Poaceae, but, in contrast to this phytolith study, was found to correspond approximately with canopy cover interpreted as a result of restricted dispersal and low canopies in the central part of the island (McGlone & Meurk Reference McGlone and Meurk2000). The diversity and abundance of Poaceae phytolith morphotypes observed during this and studies by other authors suggests that, with further detailed taxonomic work on modern grasses involving cluster analysis in the future, differentiation to generic or even species level may be possible (e.g. Marx et al. Reference Marx, Lee, Lloyd and Lee2004). Therefore, dispersed phytoliths in the sedimentary record, despite the apparent poor reflection of overall source vegetation character as shown in this study, have the potential to provide a significant contribution to the interpretation of past grass-rich floras, particularly as Poaceae pollen is conservative in morphology. This study has shown that previous terrestrial palynology work provides the most useful reconstruction of source vegetation compared with that of the dispersed soil phytolith record (McGlone et al. Reference McGlone, Wardle and Meurk1997, McGlone & Meurk Reference McGlone and Meurk2000). However, in sedimentary situations where organics are absent, such a phytolith record would provide presence/absence data on plant types, at least to Family level when no other palaeobotanical information was available.
Conclusions
At four quadrats on sub-Antarctic Campbell Island phytoliths produced by the extant vegetation have been compared with the dispersed soil assemblages accumulating directly beneath. The results indicate that the soil record is dominantly composed of grass phytoliths and provides a poor representation of the overlying source vegetation. The over-representation of grass phytoliths implies ubiquitous grassland at all of the study sites suggesting the apparent absence of significant plant types, e.g. ferns and shubs, in the canopy. The abundance and diversity of phytolith production from grass species is high, even if the canopy cover of grass is low. Taphonomic problems have been highlighted, which result in the loss and blurring of ecological information through time: differential production by different plant species (in phytolith abundance and diversity of form, as well as presence or absence of any biosilica), the occurrence of conservative morphologies produced by several plant species, and the apparently biased preservation of ‘robust’ solid, idioblastic forms in preference to the fragile, thinner-walled and more planar morphotypes. However, taking these issues into account, the dispersed phytolith record can still confirm the presence of grasses (and to a lesser extent sedges and rushes). However, overall, a lack of phytolith production in many key species of the modern flora within these quadrats (e.g. shrubs, ferns and macrophyllous forbs) prevent detailed reconstructions of the source vegetation composition as is possible from the terrestrial palynology record.
The derivation of detailed source vegetation composition from dispersed fossil phytolith assemblages has not been possible at these Campbell Island localities. The implication is that similar difficulties are likely in any such study of dispersed phytoliths from assemblages in the fossil record. Such vegetation reconstructions should be treated with due caution and with reference to the particular taphonomic issues that phytoliths present. However, the dominance of grass phytoliths in the soil assemblages at Campbell Island, and their varied and often robust morphology emphasises the potential importance of phytoliths as evidence of grasses in the geological past. Future statistical analyses on grass phytolith taxonomy will also refine their modern analogue potential in Cenozoic palaeoecological studies of high latitude floras. Grass macrofossils are rare, and grass pollen is of conservative morphology, therefore attention to the fossil record of grass phytoliths can contribute significantly to our understanding of past grassland ecosystems.
Acknowledgements
The author would like to thank P.J. Garnock-Jones of Victoria University of Wellington (VUW) and J.I. Raine of GNS Science, New Zealand, for assistance in the field and botanical advice. I.M. Turnbull and Fiordland Ecology Holidays are thanked for field logistical support. The New Zealand Department for Conservation are thanked for a research permit (Thorn, SO-13639-GEO). J.E. Francis at the University of Leeds, UK is also thanked for helpful discussions. The study was supported by a New Zealand Foundation for Research, Science and Technology Post-Doctoral Fellowship, the Royal Society Marsden Fund (grant GNS201) and the VUW University Research Fund (grant 22342, award 1282). This study has been completed during an Antarctic Funding Initiative Post-Doctoral Fellowship at the University of Leeds, UK. The Antarctic Science Bursary (UK) is thanked for support towards ongoing palynological training.
