Phytolith assemblages as indicators of coastal environmental changes and hurricane overwash...

Phytolith assemblages as indicators ofcoastal environmental changes andhurricane overwash depositionHou-Yuan Lu1,2 and Kam-biu Liu2*

(1Institute of Geology and Geophysics, Chinese Academy of Sciences, P.O. Box

9825, Beijing 100029, China; 2Department of Geography and Anthropology,Louisiana State University, 227 Howe-Russell Geoscience Complex, Baton RougeLA 70803, USA)

Received 1 August 2004; revised manuscript accepted 12 January 2005

Abstract: We demonstrate that phytolith assemblages are a useful proxy for reconstructing coastal

environmental changes and for validating the overwash origin of sand layers in palaeotempestology

studies. Phytolith analysis was conducted on 50 topsoil or surface sediment samples collected from a

variety of coastal plant communities or depositional environments in the southeastern USA. The data

suggest that different coastal subenvironments can be distinguished by their modern phytolith assemblages.

For example, coastal marsh samples contain a diverse phytolith assemblage dominated by smooth-elongate

and square morpho-types and maritime forest samples are dominated by phytoliths from palms and

broadleaf dicotyledonous plants. Remarkably, the phytolith assemblages from sand dunes are

characterized by high percentages of two-horned-tower, flat-tower, spool/horned-tower and short-saddle

types. Phytolith analysis of three prehistoric sand layers in a sediment core from Western Lake,

northwestern Florida, shows that they contain a phytolith assemblage similar to those characteristic of

sand dunes and interdune meadows. These observations are confirmed by the results of principal

components analysis and discriminant analysis on the modern and fossil phytolith data sets. Our study

results support the interpretation that the sand layers in Western Lake were indeed formed by the erosion

of sand dunes during overwash processes caused by landfalling catastrophic hurricanes.

Key words: Phytoliths, hurricanes, palaeotempestology, coastal environments, sand dunes, overwash

deposition, Atlantic Coast, Gulf of Mexico, late Holocene.

Introduction

In the emerging field of palaeotempestology (Liu and Fearn,

2000a; Liu, 2004), the proxy proven to be most useful in the

reconstruction of past hurricane strikes comes from overwash

sand layers preserved in sediment cores retrieved from coastal

lakes and marshes (Liu and Fearn, 1993, 2000a,b; Zhou, 1998;

Donnelly et al., 2001a,b; Scott et al., 2003; Liu, 2004). The

principle underlying this research methodology is that when an

intense hurricane makes landfall and causes a storm surge that

exceeds the height of the sand barrier, sand will be washed over

from the beach and dunes into the backbarrier lakes and

marshes, forming an overwash fan (Liu, 2004). Although sand

layers of recent age (e.g., twentieth century) found in some

coastal sediment cores can be reliably attributed to historic

storm events and their associated overwash processes by

examining historic aerial photos of coastal sand transport

and by means of radiometric (137Cs and 210Pb) dating of these

cores (Donnelly et al., 2001a,b), independent evidence is

needed to verify the overwash origin of the older, especially

prehistoric, sand layers because the sand could have come from

other coastal or even fluvial environments rather than the sand

dunes or beach ridges.

Microfossils such as dinoflagellates, foraminifera, diatoms

and pollen have been used as indicators of saltwater intrusion

into coastal lakes and marshes resulting from past storm surges

and tsunamis (Hemphill-Haley, 1996; Zhou, 1998; Collins

et al., 1999; Hippensteel and Martin, 1999; Liu and Fearn,

2000b; Hughes et al., 2002; Liu et al., 2003; Scott et al., 2003).

However, some of these conventional proxies for environmen-

tal reconstruction, such as pollen, are not very useful for

deciphering the provenance of the sand in the sand layers and

for identifying the coastal sandy subenvironments from which

the sand is derived. In this paper we demonstrate that modern

phytolith assemblages in soil or sediment samples can be used

to differentiate different coastal subenvironments such as sand*Author for correspondence (e-mail: kliu1@lsu.edu)

The Holocene 15,7 (2005) pp. 965�/972

# 2005 Edward Arnold (Publishers) Ltd 10.1191/0959683605hl870ra

dunes, marshes, freshwater lakes, estuarine and bay bottoms

and maritime forests. In particular, phytolith analysis can be a

useful tool to test the hypothesis that the sand in the sand

layers used in palaeotempestology studies does indeed come

from the coastal sand dunes resulting from overwash processes.

Material and methods

Phytolith surface samplesFifty surface samples were collected for phytolith analysis from

three coastal locations in the southeastern USA (Figure 1A).

Cumberland Island (Figure 1B), a barrier island in southern

Georgia, is fringed by a well-developed continuous line of

beach and high dunes along the Atlantic coast. Interspersed

between and behind the beach and sand dunes is a complex

array of freshwater lakes, small Juncus/sedge fens, interdune

meadows and upland maritime forests dominated by various

species of oaks, pines, palms and other hardwoods.

Atchafalaya Marsh (Figure 1C), southern Louisiana, is a

large coastal/estuarine marsh that extends along the mouth of

the Atchafalaya River and the west shore of Atchafalaya Bay

to the Gulf of Mexico. This large coastal marsh exhibits a

salinity gradient from cypress swamp, fresh marsh, brackish

marsh, to salt marsh, whereas the Atchafalaya Bay itself offers

a distinct set of open-water, estuarine bay depositional

environments.

Western Lake (Figure 1D), a backbarrier oligohaline lake

within Grayton Beach State Park, northwestern Florida, is

situated behind a barrier beach about 150�/200 m wide and a

continuous line of sand dunes rising to 6.2�/9.3 m above sea

level (Liu and Fearn, 2000a). Vegetation around Western Lake

is a typical Gulf coastal community (Moreno-Casasola, 1988)

consisting of sea oats (Uniola paniculata) growing on the sand

dunes, cordgrass (Spartina alterniflora) growing on the seaward

edge of the dunes and marshes around the lake and pine

flatwoods, scrub oak thickets and mixed hardwood-palmetto

woodlands occurring on the uplands and hammocks.

The 50 surface samples (Figure 1) were collected from nine

different coastal vegetation communities or depositional en-

vironments representing different ecological conditions: fresh

marsh, brackish marsh, salt marsh, maritime forest, freshwater

lake, Juncus/sedge fen, interdune meadow, primary dune and

estuarine bay. Terrestrial samples were collected from the

uppermost 2�/3 cm of the topsoil (forest, marsh and dune)

by means of a small spade. Those derived from the surface mud

(lakes and estuarine bay) were collected by means of an

Ekman’s dredge.

The samples were prepared by a procedure slightly modified

from Piperno (1988) and Runge (1999). It consists of sodium

pyrophosphate (Na4P2O7) deflocculation, treatment with 30%

hydrogen peroxide (H2O2) and cold 15% hydrochloric acid

(HCl), zinc bromide (ZnBr2, density 2.35 g/cm3) heavy liquid

separation and mounting on a slide with Canada Balsam.

Phytolith counting and identification was performed using a

Nikon Optiphot microscope at 400�/ magnification. More

than 350 phytolith grains were counted in each sample.

Identification was aided by the use of reference materials

(Lu and Liu, 2003a,b) and published keys (Piperno, 1988;

Mulholland and Rapp, 1992, Kondo et al., 1994; Piperno and

Pearsall, 1998; Runge, 1999, Pearsall, 2000).

Sediment cores from Western LakeTo test if the modern phytolith data can be used to aid the

reconstruction of coastal palaeoenvironmental changes and the

identification of the provenance of lake sediments, we analysed

the phytolith contents of a short core (core 9) from Western

Figure 1 A. Location of Cumberland Island (Georgia), Atchafalaya Marsh (Louisiana) and Western Lake (Florida) on the Atlantic coastand Gulf coast of southeastern USA. B. Major habitat types and location of surface samples (dots) (C27-47) on Cumberland Island. C.Marsh salinity zones and location of surface samples (A1�/26) in Atchafalaya Marsh. D. Geomorphic setting of Western Lake and locationof surface samples (W48-50) and core 1 and core 9 (triangles). Contours are in metres

966 The Holocene 15 (2005)

Lake, northwestern Florida (Figure 1). A 6.5-m core (core 1)

from Western Lake has yielded a 5000-yr proxy record of

hurricane activity that includes 12 direct strikes by catastrophic

hurricanes over the past 3400 years (Liu and Fearn, 2000a).

The same processing method was used on the core samples as

on the surface samples for the extraction of phytoliths.

Statistical analysisBy using the statistical program CANOCO 4.0 (ter Braak,

1996), principal components analysis (PCA) was performed on

the 50 surface samples to reveal the internal structure of the

modern phytolith data set. The PCA results provided a basis to

classify the surface samples into statistically meaningful

groups, which could be interpreted in terms of the coastal

vegetation communities or depositional environments from

which the surface samples were collected. In addition, dis-

criminant analysis (sensu Liu and Lam, 1985), an inferential

statistical technique, was performed using SPSS version 10.0 to

classify the 50 phytolith samples with reference to the a

priori groups suggested by the PCA results. The discriminant

functions derived from these surface samples were then

applied to the fossil phytolith data obtained from the Western

Lake core to predict the group membership of each fossil

phytolith sample. These predicted group memberships repre-

sent the palaeovegetation communities from which the fossil

phytolith assemblages are derived and therefore shed light

on the provenance of the sediment that contain them

(Liu and Lam, 1985).

Results

Phytolith assemblages in modern coastalsubenvironmentsThirty phytolith morpho-types (including two ‘unknown’

morpho-types) were recognized in the 50 surface samples

(Figure 2). Figure 3 shows some of the more common or

characteristic types. All surface samples from coastal marshes

are characterized by high percentages of smooth-elongate

(169/6%) (Figure 3j) and square (119/6%) (Figure 3n)

phytoliths. Remarkably, marshes of different salinity environ-

ments seem to be distinguishable by the occurrence of certain

associated phytolith types. Fresh marsh samples tend to have

higher percentages of dumbbell (Figure 3a) and sinuate-

elongate (Figure 3i) phytolith types than the brackish and

salt marshes and the presence of Pteridophyte phytoliths

reflects the presence of ferns in fresh marshes. Between the

brackish marsh and salt marsh samples, brackish marsh

samples seem to have higher percentages of spool/horned

towers (Figure 3d), whereas salt marsh samples generally

have higher percentages of fan-shaped (79/5%) and Spartina-

type (69/2%) phytoliths. The Spartina-type (Figure 3f) is

a rondel/saddle ellipsoid morpho-type typically produced

by Spartina alterniflora, the dominant grass of salt marshes

(Lu and Liu, 2003b).

Phytolith assemblages from the topsoil samples of maritime

forests on Cumberland Island are characterized by very high

percentages of broadleaf-types (309/23%) (Figure 3l, m) and

Palmae-type 1 (439/28%) (Figure 3g). The broadleaf-types

include many phytolith morpho-types (Piperno, 1988; Kondo

et al., 1994; Runge, 1999) such as the elongate phytolith type

A4 (equivalent to broadleaf-type 1 in this study, Figure 3l) and

type A5 (equivalent to broadleaf-type 2 in this study, Figure

3m) of Runge (1999). They are characterized by the presence of

facets and are mainly produced by broadleaf trees or shrubs.

Palmae-type 1 (Figure 3g) is a spherical phytolith with a

spinulose surface, typically produced by palms (e.g., Piperno,

1988; Runge, 1999).

Phytolith assemblages from freshwater coastal lakes on

Cumberland Island are different from those of the adjacent

maritime forests in having very high percentages of both

Palmae-type 1 (749/6%) and Palmae-type 2 (99/6%), whereas

the broadleaf types are absent. Palmae-type 2 is a spherical

Figure 2 Diagram of phytolith percentages from 50 surface samples from Cumberland (C), Atchafalaya Marsh (A) and Western lake (W)arranged by coastal subenvironments from which they were collected

Hou-Yuan Lu and Kam-biu Liu: Phytolith assemblages in coastal palaeotempestology 967

phytolith with a regularly spinulose surface like Palmae-type 1,

but the spinule apices are round instead of sharp; they are

probably derived from a specific member of the Palmae family

or from another monocotyledonous plant (Runge, 1999).

Three surface samples from Juncus/sedge fens in Cumber-

land Island are characterized by abundant sinuate-elongate,

smooth-elongate and Cyperaceae type phytoliths (8.0�/12.1%).

The abundance of the Cyperaceae type is attributed to the

predominance of sedges (especially Carex spp.) in these small

wetlands, some of which are probably developed on dried or

ephemeral lake beds. The spikes of broadleaf-type 2 and

Palmae-type 1 in two samples collected near the edges of these

small fens are probably due to local influence from the adjacent

forest vegetation.

The phytolith composition of the surface samples from

interdune meadows shows very high frequencies of dumbbell

phytoliths (479/13%). Cross-shaped phytoliths (Figure 3b)

reach values of around 2�/8%. Dumbbell and cross phytoliths

Figure 3 Photomicrographs of the main phytolith morpho-types counted in this study: (a) dumbbell, (b) cross, (c) long-saddle, (d) spool/horned tower, (e) flat tower and two-horned tower, (f) Spartina-type, (g) palmae-type 1, (h) short-saddle, (i) sinuate-elongate, (j) smooth-elongate, (k) long-point, (l) broadleaf-type 1, (m) broadleaf-type 2, (n) square

968 The Holocene 15 (2005)

are produced at high percentages by grasses of the Panicoideae

subfamily, such as Cenchrus incertus and Panicum hemitomon

(Lu and Liu, 2003b).

The modern phytolith composition of the primary dunes in

both Cumberland Island and Western Lake is characterized by

high amounts of two-horned-tower (309/25%) (Figure 3e), flat-

tower (139/7%) (Figure 3e), spool/horned-tower (149/6%) and

short-saddle (99/6%) (Figure 3h) phytoliths. This assemblage is

produced by the dominant grasses of coastal sand dunes, such

as Uniola paniculata and Aristida desmantha (Lu and Liu,

2003b). Although, to some extent, these characteristics are also

found in samples from the coastal marshes, the dune samples

can be distinguished from the marsh samples by having very

low percentages of square, smooth-elongate, sinuate-elongate

and long-point types (Figure 3k).

The phytolith composition in the surface samples of

estuarine bay bottoms is characterized by high percentages of

fan-shaped, square, smooth-elongate and long-point phytoliths

of silt sizes (109/7%, 139/9%, 289/7%, 89/4%, respectively)

that are derived from epidermal long-cells, trichome-cells and

bulliform-cells of grasses. Not surprisingly, these samples from

the Atchafalaya Bay are generally similar to the marsh samples

collected from the adjacent Atchafalaya Marsh in terms of

their predominant morpho-types. However, the estuarine bay

bottom samples can be distinguished from the marsh samples

by having no or few dumbbell, spool/horned-tower and

Spartina-type phytoliths. The lower diversity of phytoliths in

the estuarine and bay bottom samples may be due to

differential dispersal and deposition abilities of different

phytolith morpho-types in a fluvial-estuarine environment.

Numerical analysis of modern phytolith dataPercentage data from 23 phytolith types (see Figure 4a) were

used in the principal components analysis (PCA) and dis-

criminant analysis. The PCA biplots of the 23 phytolith types

and the 50 surface samples are shown in Figure 4. The first two

principal components, axis 1 and axis 2, account for 56.1% and

18.8% of the variance in the phytolith data, respectively. On the

biplot of principal component loadings for the 23 phytolith

types (Figure 4a), Palmae-types and broadleaf-type 1 have the

highest loadings on axis 1, whereas two-horned-tower, flat-

tower and dumbbell�/cross have the highest loadings on axis 2.

On the biplot for principal component scores for the 50 surface

samples (Figure 4b), samples from freshwater lake and

maritime forest sites have the highest positive scores on axis

1, where the Palmae-types and broadleaf-type 1 also have the

highest loadings. Samples from estuarine bay and marsh sites

score most negatively on the far left of axis 1, corresponding to

highly negative loadings for fan-shaped, square, smooth-

elongate and long-point phytolith types. The second principal

component, axis 2, primarily distinguishes the samples from

sand dunes and interdune meadows from the rest of the

samples. Thus, the results of the PCA ordination analysis

suggest that the 50 surface samples can be divided into five

primary groups: (1) estuarine bay; (2) marsh (including fresh

marsh, brackish marsh and salt marsh); (3) Juncus/sedge fen;

(4) sand dune/interdune meadow; and (5) freshwater lake/

maritime forest. With only a few exceptions (e.g., A18, C37,

C46, C47), the phytolith ‘signatures’ of these five groups are

quite well-defined with little overlap among them (Figure 4a,

see also Figure 2).

After these five groups were delineated by PCA, discrimi-

nant analysis was used to validate the classification and to

generate probability estimates for statistical inference (Liu and

Lam, 1985). All (100%) of the samples were correctly classified

into their respective a priori groups (Table 1). Figure 5 shows

the five groups of surface samples plotted against discriminant

functions 1 and 2, which account for 40.1% and 33.5% of the

variance, respectively. Again, the five groups and their cen-

troids are clearly distinct from each other with very little

overlap between groups. The high degree of correct classifica-

tion suggests that the classification of the surface samples into

the five groups is statistically robust.

These surface sample results indicate that the phytoliths

deposited in the different coastal subenvironments in our study

areas are reliable indicators of the major vegetation commu-

nities at or around the site. These results can then be applied to

the fossil phytolith spectra from a sediment core from Western

Lake to reconstruct palaeoenvironmental changes and to infer

the provenance of the sediment.

Figure 4 a. PCA biplot showing coordinates of the 23 phytolithtypes (vectors) plotted along axis 1 and axis 2. Abbreviationfor phytolith types: Broad1, broadleaf-type 1; Broad2, broadleaf-type 2; Cypera, cyperaceae type; Dumb�/Cross, dumbbell andcross; Fan, fan-shaped; Ftower, flat-tower; Gymnos, gymnospermtype; LPoint, long-point; LSaddl, long-saddle; Palma1, palmae-type 1; Palma2, palmae-type 2; Pterid, pteridophyte type; Rectan,rectangle; RFan, fan-reed; Rondel, rondel; Selong, sinuate-elon-gate; SHTower, spool/horned tower; Smlong, smooth-elongate;Sparti, Spartina type; SPoint, short-point; Square, square; SSaddl,short-saddle; THTower, two-horned tower. b. PCA biplot showingcoordinates of the 50 surface samples plotted along axis 1 andaxis 2. Serial numbers of the samples are the same as in Figures 1and 2

Hou-Yuan Lu and Kam-biu Liu: Phytolith assemblages in coastal palaeotempestology 969

Phytolith record from Western LakeCore 9 from Western Lake, a short core taken near core 1,

contains 90 cm of organic lake mud (organic matter content

10�/13% as determined by loss-on-ignition) interrupted by

seven sand layers. These sand layers have been interpreted to be

formed by overwash processes occurring during strikes by

catastrophic hurricanes of category 4 or 5 intensity according

to the Saffir/Simpson scale (Liu and Fearn, 2000a). For the

phytolith analysis we selected eleven samples from the organic

lake mud and three samples from three thick sand layers

(referred to as sand layers 1, 2 and 3, respectively) occurring at

depths of 52.0�/53.0 cm, 58.5�/60.0 cm and 66.0�/68.5 cm as

measured on the core (Figure 6). These three sand layers

were selected because they were sufficiently thick to provide

clean samples of sand uncontaminated by the embedding

organic sediments. These prominent sand layers were radio-

carbon dated to 13209/50, 14109/50 and 18509/50 yr BP

(uncalibrated 14C ages), respectively (Liu and Fearn, 2000a).

Phytolith analysis of the eleven organic lake sediment

samples shows that they are dominated by the Palmae-type 2

(419/9%), Palmae-type 1 (129/3%) and smooth-elongate (9.69/

3.7%) morpho-types. Dumbbell and cross (7.49/3.9%) and

flat-tower and two-horned-tower (5.89/2.5%) morpho-types

are also present consistently in the organic mud at lower

percentages (Figure 6). The presence of broadleaf-type 1 and

broadleaf-type 2 phytoliths is also notable. This apparently

mixed assemblage suggests that the organic sediments in

Western Lake come from multiple sources, but the principal

contributors are maritime forests and freshwater wetlands

around the lake. On the other hand, the sand sampled from

the three thick sand layers contains very high percentages of

flat-tower phytoliths (at 20.1%, 18.6% and 51.1% in sand layers

1, 2 and 3, respectively), along with somewhat elevated

percentages of short-saddle and two-horned-tower phytoliths.

This unusual assemblage, especially the prominent spikes of the

flat-tower type, is similar to the modern phytolith assemblages

characteristic of the sand dunes near Western Lake and in

Cumberland Island.

The contrast in phytolith assemblages between the organic

lake sediment and the sand layers is confirmed by the results

of discriminant analysis. The same discriminant functions

derived from the surface sample data set were used to classify

the 14 fossil phytolith samples. All 11 samples from organic

lake sediment were classified as freshwater lake/maritime

forest (group 4), whereas the three samples from the sand

layers were classified as sand dune/interdune meadow (group 5)

(Table 1; Figure 6). When the fossil samples from Western

Lake are plotted against the same discriminant functions as

the surface samples (Figure 5), the three sand layer samples

are located closer to the centroid of the sand dune/interdune

meadow group than the rest of the core samples. Therefore,

the phytolith evidence from the Western Lake core suggests

that the sand in the three prominant sand layers was derived

from the sand dunes separating the lake and the Gulf. Grasses

growing on the dune ridges, such as Uniola paniculata,

Aristida desmantha and Panicum amarum, are the principal

producers of flat-tower and two-horned-tower phytoliths

Table 1 Classification results of the discriminant analysis performed on the 50 surface samples from the five vegetation groups and on 14fossil samples (bold) from core 9 of Western lake

Group no. Actual group Predicted group membership Total cases

Estuarine bay Marsh Juncus/

sedge fen

Sand dune/

interdune meadow

Freshwater lake/

maritime forest

Group 1 Group 2 Group 3 Group 4 Group 5

1 Estuarine Bay 14a (100%)b 0 0 0 0 14

2 Marsh 0 14 (100%) 0 0 0 14

3 Juncus/sedge fen 0 0 3 (100%) 0 0 03

4 Sand dune/interdune meadow 0 0 0 8 (100%) 0 8

5 Freshwater lake/ Maritime forest 0 0 11 (100%) 11

Ungrouped (Core 9 samples) 0 0 0 3 (21.4%) 11 (78.6%) 14

aNumber of samples classified as that group.bPercentage of samples classified as that group.

Figure 5 The 50 surface samples plotted against canonicaldiscriminant functions 1 and 2 and their classification into fivegroups corresponding to five vegetation communities. Also plottedare the 14 fossil samples from Western Lake core 9. Samples fromthe three prominent sand layers interpreted to be of overwashorigin are labelled as sand 1, 2 and 3

970 The Holocene 15 (2005)

(Lu and Liu, 2003b). Overwash processes associated with

catastrophic hurricane landfalls are the most likely mechanism

that washed sand from the sand dunes to the bottom of

Western Lake.

Discussion and conclusions

Previous palaeoecological applications of phytolith studies

have focused mainly on the reconstructions of broad vegetation

types at regional or continental scales, such as temperate or

tropical grasslands (e.g., Fredlund and Tieszen, 1994; Barboni

et al., 1999; Blinnikov et al., 2002) as well as Mediterranean

woodlands and tropical forests (Piperno and Becker, 1996;

Delhon et al., 2003; Piperno and Jones, 2003). Except in a few

cases (Fearn, 1998; Horrocks et al., 2000), phytolith analysis

has not been applied to the study of coastal environmental

changes, especially at the finer spatial scales that identify

coastal subenvironments such as salt marshes, maritime forests

and sand dunes.

In this paper, we demonstrate that different coastal sub-

environments can be distinguished by their modern phytolith

assemblages. Notably, surface samples from coastal marshes

contain a diverse phytolith assemblage dominated by the

smooth-elongate and square types, whereas soil samples from

maritime forests are dominated by phytoliths from palms and

broadleaf dicotyledonous plants. Moreover, surface samples

from coastal sand dunes are characterized by high percentages

of two-horned-tower, flat-tower, spool/horned-tower and

short-saddle phytoliths. These results confirm and reinforce

our previous conclusion that grasses typically growing on sand

dunes (e.g., Uniola paniculata, Aristida desmantha) produce

distinctive phytolith assemblages (Lu and Liu, 2003b). Thus,

phytoliths found in the sand layers of sediment cores taken

from Western Lake could be used to reveal the provenance of

the sand that embeds them. Data from core 9 show that the

three most prominent sand layers contain a phytolith assem-

blage similar to those derived from the grass plants and

sediment samples from the primary dunes. This finding

supports our previous interpretation that the sand layers

were formed by the erosion of sand dunes during overwash

processes caused by catastrophic hurricane landfalls (Liu and

Fearn, 2000a). This study demonstrates that, in association

with overwash sand layers, phytoliths can be used as a useful

proxy for the reconstruction of past hurricane activity from

coastal lake sediments.

Acknowledgements

We thank J. Blackburn, D. Cahoon, M. Fearn, S. Fournet,

C.A. Reese, C.M. Shen, Z. Yao and X.Y. Zhou for assistance

with fieldwork and M. Eggart for cartographic assistance. This

work was supported by NSFC (40024202, 40325002), the Risk

Prediction Initiative (RPI) of the Bermuda Biological Station

for Research (RPI-00-1-002) and the US National Science

Foundation (SES-9122058; BCS-0213884).

References

Barboni, D., Bonnefille, R., Alexandre, A. and Meunier, J.D. 1999:Phytoliths as paleoenvironmental indicators, West Side MiddleAwash Valley, Ethiopia. Palaeogeography, Palaeoclimatology,Palaeoecology 152, 87�/100.Blinnikov, M., Busacca, A. and Whitlock, C. 2002: Reconstructionof the late Pleistocene grassland of the Columbia Basin,Washington, USA, based on phytolith records in loess.Palaeogeography, Palaeoclimatology, Palaeoecology 177, 77�/101.Collins, E.S., Scott, D.B. and Gayes, P.T. 1999: Hurricane recordson the South Carolina coast: can they be detected in the sedimentrecord? Quaternary International 56, 15�/26.Delhon, C., Alexandre, A., Berger, J.F., Thiebault, S., Brochier, J.L.and Meunier, D. 2003: Phytolith assemblages as a promisingtool for reconstructing Mediterranean Holocene vegetation.Quaternary Research 59, 48�/60.Donnelly, J.P., Roll, S., Wengren, M., Butler, J., Lederer, R. andWebb, T., III 2001a: Sedimentary evidence of intense hurricanestrikes from New Jersey. Geology 29, 615�/18.Donnelly, J.P., Bryant, S.S., Butler, J., Dowling, J., Fan, L.,Hausmann, N., Newby, P., Shuman, B., Stern, J., Westover, K.and Webb, T., III 2001b: 700 yr sedimentary record of intensehurricane landfalls in southern New England. Bulletin of theGeological Society of America 113, 714�/27.Fearn, M.L. 1998: Phytoliths in sediment as indicators of grasspollen source. Review of Palaeobotany and Palynology 103, 75�/81.

Figure 6 Percentage diagram of major phytolith types in core 9 of from Western Lake. Phytolith spectra from the three prominent sandlayers are delimited by dotted lines. Radiocarbon dates are shown on the right

Hou-Yuan Lu and Kam-biu Liu: Phytolith assemblages in coastal palaeotempestology 971

Fredlund, G.G. and Tieszen, L.L. 1994: Modern phytolithassemblages from the North American Great Plains. Journal ofBiogeography 21, 321�/35.Hemphill-Haley, E. 1996: Diatoms as an aid in identifying late-Holocene tsunami deposits. The Holocene 6, 439�/48.Hippensteel, S.P. and Martin, R.E. 1999: Foraminifera as anindicator of overwash deposits, barrier Island sediment supply andbarrier Island evolution: Folly Island, South Carolina.Palaeogeography, Palaeoclimatology, Palaeoecology 149, 115�/25.Horrocks, M., Deng, Y., Ogden, J. and Sutton, D.G. 2000:A reconstruction of the history of a Holocene sand dune onGreat Barrier Island, northern New Zealand, using pollen andphytolith analyses. Journal of Biogeography 27, 1269�/77.Hughes, J.F., Mathewes, R.W. and Clague, J.J. 2002: Use of pollenand vascular plants to estimate coseismic subsidence at a tidalmarsh near Tofino, British Columbia. Palaeogeography,Palaeoclimatology, Palaeoecology 185, 145�/61.Kondo, R., Childs, C. and Atkinson, I. 1994: Opal phytoliths of NewZealand. Canterbury: Manaaki Whenua Press, 85 pp.Liu, K-b. 2004: Palaeotempestology: principles, methods andexamples from Gulf coast lake sediments. In Murnane, R. andLiu, K-b., editors, Hurricanes and typhoons: past, presentand future. New York: Columbia University Press, 13�/57.Liu, K-b. and Fearn, M.L. 1993: Lake-sediment record of lateHolocene hurricane activities from coastal Alabama. Geology 21,793�/96.____ 2000a: Reconstruction of prehistoric landfall frequencies ofcatastrophic hurricanes in northwestern Florida from lakesediment records. Quaternary Research 54, 238�/45.____ 2000b: Holocene history of catastrophic hurricane landfallsalong the Gulf of Mexico coast reconstructed from coastal lakeand marsh sediments. In Ning, Z.H. and Abdollahi, K.K., editors,Current stresses and potential vulnerabilities: implications of globalchange for the Gulf coast region of the United States. Baton RougeLA: Franklin Press, 38�/47.Liu, K-b. and Lam, N.S.N. 1985: Paleovegetational reconstructionbased on modern and fossil pollen data: an application ofdiscriminant analysis. Annals of the Association of AmericanGeographers 75, 115�/30.Liu, K-b., Lu, H.Y. and Shen, C.M. 2003: Assessing thevulnerability of the Alabama Gulf coast to intense hurricanestrikes and forest fires in the light of long-term climatic changes. InNing, Z.H., Turner, R.E., Doyle, T. and Abdollahi, K., editors.,Integrated assessment of the climate change impacts on the GulfCoast region. Baton Rouge LA: Gulf Coast Regional ClimateChange Council, 223�/30.

Lu, H.Y. and Liu, K-b. 2003a: Morphological variations of lobate

phytoliths from grasses in China and the southeastern USA.

Diversity and Distributions 9, 73�/87.____ 2003b: Phytoliths of common grasses in the coastal

environments of southeastern USA. Estuarine, Coastal and Shelf

Science 58, 587�/600.Moreno-Casasola, P. 1988: Patterns of plant species distribution on

coastal dunes along the Gulf of Mexico. Journal of Biogeography

15, 787�/806.Mulholland, S.C. and Rapp, G., Jr 1992: A morphological

classification of grass silica-bodies. In Rapp, G., Jr and

Mulholland S.C., editors, Phytolith systematics: advances in

archaeological and museum science. New York and London:

Plenum Press, 65�/89.Pearsall, D.M. 2000: Paleoethnobotany: a handbook of procedures.

Second edition. San Diego CA: Academic Press, 700 pp.Piperno, D.R. 1988: Phytolith analysis: an archaeological

and geological perspective. San Diego CA: Academic Press,

280 pp.Piperno, D.R. and Becker, P. 1996: Vegetational history of a site in

the central Amazon Basin derived from phytolith and charcoal

records from natural soils. Quaternary Research 45, 202�/209.Piperno, D.R. and Jones, J.G. 2003: Paleoecological and

archaeological implications of a Late Pleistocene/Early Holocene

record of vegetation and climate from the Pacific coastal plain of

Panama. Quaternary Research 59, 79�/87.Piperno, D.R. and Pearsall, D.M. 1998: The silica bodies of

tropical American grasses: morphology, taxonomy and

implications for grass systematics and fossil phytolith

identification. Annals of the Smithsonian Institution 85, 1�/40.Runge, F. 1999: The opal phytolith inventory of soils in central

Africa �/ quantities, shapes, classification and spectra. Review of

Palaeobotany and Palynology 107, 23�/53.Scott, D.B., Collins, E.S., Gayes, P.T. and Wright, E. 2003: Records

of prehistoric hurricanes on the South Carolina coast based on

micropaleontological and sedimentological evidence, with

comparison to other Atlantic Coast records. Bulletin of the

Geological Society of America 115, 1027�/39.ter Braak, C.J.F. 1996: Unimodal models to relate species to

environment. DLO-Agricultural Mathematics Group, Box

100,NL-7600 AC Wageningen, The Netherlands, 266 pp.Zhou, X. 1998: A 4000-year pollen record of vegetational changes,

sea-level rise and hurricane disturbance in the Atchafalaya Marsh

of southern Louisiana. M.S. Thesis, Baton Rouge LA: Louisiana

State University, 143 pp.

972 The Holocene 15 (2005)