The environmental impact of the Minoan eruption of Santorini (Thera): statistical analysis of palaeoecological data from Golbisar, southwest Turkey

(1)

http://hol.sagepub.com

The Holocene

DOI: 10.1191/0959683602hl557rp 2002; 12; 431 The Holocene

W. J. Eastwood, J. Tibby, N. Roberts, H. J.B. Birks and H. F. Lamb

analysis of palaeoecological data from Golbisar, southwest Turkey

The environmental impact of the Minoan eruption of Santorini (Thera): statistical

http://hol.sagepub.com/cgi/content/abstract/12/4/431 The online version of this article can be found at:

Published by:

http://www.sagepublications.com

can be found at:

The Holocene Additional services and information for

http://hol.sagepub.com/cgi/alerts Email Alerts:

http://hol.sagepub.com/subscriptions Subscriptions:

http://www.sagepub.com/journalsReprints.nav Reprints:

http://www.sagepub.com/journalsPermissions.nav Permissions:

http://hol.sagepub.com/cgi/content/refs/12/4/431 SAGE Journals Online and HighWire Press platforms):

(this article cites 33 articles hosted on the Citations

at Universitetsbiblioteket i on December 12, 2007 http://hol.sagepub.com

Downloaded from

(2)

The environmental impact ^of the ^Minoan eruption ^of Santorini ^(Thera): statistical

analysis of palaeoecological data ^from

Golbisar, southwest Turkey

W.J. Eastwood,l* J. Tibby,2 N. Roberts,3 H.J.B. Birks4'5 and H.F. Lamb6

('School of Geography and Environmental Sciences, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK; 2School of Geography and

Environmental Science, Monash University, Melbourne, 3800, Victoria, Australia; 3Department of Geographical Sciences, University of Plymouth,

Plymouth, PL4 8AA, UK; 4Botanical Institute, University of Bergen, Allegaten 41, N-5007 Bergen, Norway; 5Environmental Change Research Centre,

University College London, 26 Bedford Way, London WCJH OAP, UK;

6lnstitute of Geography and Earth Sciences, University of Wales, Aberystwyth, Ceredigion, SY23 3DB, UK)

Received 10August2000;revised manuscript accepted27November 2001

A

HOLOCENE

RESEARCH PAPER

Abstract: Atephralayeroriginating fromthemid-second millennium BC(-3300'4CyrBP) 'Minoan' eruption ofSantorini(or Thera)in theAegean has been found in lake sediments atG6lhisarin southwestTurkey.

Microstratigraphicanalyses oftephra shardconcentration (TSC), pollen,diatoms, sponge spiculesand non- siliceousmicrofossilsinsediments fromGtlhisarpermittheimpactofthismajorvolcaniceruptiononterrestrial andaquatic biota to beinvestigatedquantitatively. Partial redundancy analysisand associated Monte Carlo permutation tests suggest thatTSCalonecannot be shown to have had a demonstrableindependentand statistically significant effect onterrestrialpollen,non-siliceousmicrofossilordiatomassemblages.The lack of any clear, discerniblechangein the terrestrialpollen composition following tephra deposition suggeststhat there was minimalimpact onregional vegetationoverdecadal-to-centurytimescales. However,evidence that the depositionofSantorini tephra may havehad animpactonthe lakesystemcomesfromthecombinedeffect oflithologyand TSC(whichsignificantly covary)thatexplainsasignificantamountof variance in theaquatic datasets.Inparticular, diatoms and non-siliceous algae show increasesinconcentrationfollowingtephra deposition,exhibitingwhat appear to be±decadalresponsetimestoperturbation. These implyenhancedlakepro- ductivitydue to acceleratedinputofsilicaand othernutrientsfollowingtephradissolution.

Key words:Santorini, Thera,Minoan,volcanicimpact, pollen,diatoms, redundancy analysis,variancepar- titioning,MonteCarlopermutation tests,late Holocene.

Introduction

Volcanic eruptions have thepotential to cause substantial impacts onhuman and natural ecosystems.Directimpactsmay occur due tothedirectdepositionoftephraandotherpyroclastics, including gaseous compounds in the troposphere (Grattan and Charman,

*Authorforcorrespondence(e-mail:

wj.eaistwoodabham.ac.uk)

C

Arnold 2002

1994; Camuffo and Enzi, 1995; Sadler and Grattan, 1999).

Indirectimpactsoccur when eruption products (especially SO2) areejectedinto thestratospherewhich mayproduceasulphuric acid aerosol thatcanhavethecapacitytointerferewith theplanet- ary albedo, thereby causing short-term climatic perturbations (RobockandMao, 1995).When tephra shardsaredepositedas layersinlakes,miresanddeep-seasediments,andareaccurately characterized, they have the potential to provide a record of 10.1

191/0959683602hl557rp

Downloaded from

(3)

volcanism as well as producing excellent time-synchronous marker horizonswhich facilitate correlation of sitesacrosslarge

regions (Knox,

1993;Eastwoodetal.,1998b). Tephra layersalso provideaunique opportunitytoinvestigatethe effect of past erup- tionsonbiotaand their environment(LotterandBirks,1993),and thetiming ofrecovery ofperturbatedecosystems(Barkeretal., 2000).

Santorini volcanic eruption: dynamics and postulated environmental effects

One of the most powerful volcanic eruptions known to have occurredduringtheHolocenewasthemid-secondmillenniumac (-3300'4CyrBP) 'Minoan' eruptionof Santorini(orThera)in the southernAegeanSea(Figure IA). Volcanologicalreconstructions suggest that theeruptionofSantorini volcanocommenced witha plinian phase and then went through phreatomagmatic and

ignimbrite

emplacement phases producing a total of around 30 km3 denserockequivalent (DRE)ofmainly rhyolitic ejecta (Pyle, 1990; Sigurdssonetal., 1990; SparksandWilson, 1990).Previous estimates oftephradistribution derived from theinvestigationof deep-seacoressuggestedapredominantly southeasterly dispersal plume (Watkins

etal.,

1978; Vinci, 1985;Federman and Carey, 1980; Ninkovich and Heezen, 1965; StanleyandSheng, 1986), but later discoveries of Santorini tephrain deep-seacores from

Figure1 (A)Locationmapof theeasternMediterraneanshowingthe site ofGolhisartogetherwithsites whereSantorinitephrahasbeendiscovered (figures in parenthesesdenotesthickness oftephra in cm). Dashed line depictsthecurrently accepteddirection ofmajor tephrafallout(datafrom Pyle,1990;McCoyandHeiken, 2000;Momigliano, 2000). (B) Geological mapof theGolhisarcatchment(modifiedafterSenel,1997).

the Black Sea(Guichardetal., 1993) and substantial deposits in westernAnatolian lakedeposits (Sullivan, 1988; 1990) suggest a predominantly northeasterly axis of dispersal (McCoy and Heiken, 2000; Figure IA).Therecentdetection ofa4 cmthick layer of Santorini tephra in lake deposits from Golhisar Goli in southwestTurkey (Robertsetal., 1997) supports a northeasterly axis oftephra dispersal (Pearceetal.,2002).

Thepossibleeffects that theSantorinieruption may have had onnatural and culturalenvironments has attracted much attention sinceMarinatos

(1939)

first

hypothesized

thatit may have caused the destruction of the Minoan civilization based on Crete. It is indisputablethat the Santorinieruption impacted directly on the late BronzeAgesettlement of AkrotirionSantoriniisland, bury- ing it Pompeii-style in several metres of pyroclastic deposits.

However, archaeologicalexcavations andtephrastratigraphy at the archaeologicalsiteatMochlosonCrete have shown that the eruption occurred towards the end of the Late Minoan IA period, whereasthecollapseof the Minoan civilization isrelatively dated toLate MinoanIB(Solesetal., 1995).Othercatastrophetheorists (e.g.,Baillie, 1989; Burgess, 1989;White andHumphreys, 1994;

LaMoreaux, 1995)have suggested widespreadculturalimpacts associated with the Santorinieruption; unfortunatelythese hypoth- esesare notbasedonany directcause/effect evidence.Similarly, the Santorinieruptionhas been implicatedin many widespread environmentalimpacts. Severalrecentattemptshave been made torelate inferred mid-secondmillennium Bc climaticvariations,as manifested in anomaloustree-ring growthratesandaciditypeaks registeredin icecoresfrom GreenlandtotheSantorinieruption.

One suchacidity peakrecordedin theDye3 icecorefromGreen- land is datedto 1645±20BC and equatesto -200 milliontonnes ofatmospheric sulphur (Hammeretal., 1987; Sigurdssonetal., 1990). The Minoan eruption of Santorini, assigneda volcanic

explosivity

index(VEI) of 6 by Newhall and Self(1982),was deemedthe^mostlikelycandidatefor correlation with thisacidity peakonthebasisthatitwasthelargestknowneruptionfor this time

period. However,

otherlarge-scale eruptionsencompassthe sametimeperiodand make suitable candidates(Mullineauxetal.,

1975;

Miller and

Smith,

1987; Vogeletal., 1990; Beget etal.,

1992).

Furthermore,aSantoriniprovenancefor the 1645 BCacid-

ity peak,

basedongeochemical analysesofice-embeddedtephra, has been

questioned by

Zielinski andGermani

(1998).

While acidity peaksin ice cores arehighly likely to have a volcanicorigin,anomalousgrowth rings in treeswill have been causedbyclimateperturbationswhich mayormaynothave any connection with volcanism.Narrowgrowth rings indicatingfrost damagehavebeenfound inbristleconepinesinwesternUSA for theyears1628-26BC(LaMarche andHirschboek, 1984),while bogoaks inwestern

Europe

display severely restricted growthfor the decadecommencing1628 BC,possiblyas aresultof increased waterlogging (BaillieandMunro, 1988).An environmental dis- ruptionis alsoregisteredin a floatingSwedishtree-ringrecord datedto 1635 BC ±65 years(Grudd etal., 2000). Acomposite

floating

tree-ring chronology using juniper,cedar andpinetimbers from

archaeological deposits

in westernTurkey showsasignifi- cant

positive

anomaly (200% of normal) at 1641 +76/-22 BC whichindicatesenhanced,notreduced,treegrowthforaperiod which lasted not more than 10 years (Kuniholm etal., 1996).

Kuniholmetal.

(1996)

correlatetheir

tree-ring

chronologywith those from the USA and western Europe and, althoughthere appearstohave beensomehemispherical-orglobal-scale climatic

perturbation during

the mid-seventeenthcentury BC,no definite cause/effect

relationship

between this and the Minoaneruptionof Santorinicanpresentlybe demonstrated.

Aims

and objectives

The presentstudy is the first attempttoinvestigatethepossible distaleffects of theeruptionofSantorini(Thera),and forthiswe

Downloaded from

(4)

usethe lake site ofGdlhisarin southwest Turkey, -400 kmENE of Santorini volcano. The investigation adopts a statistical approach to detect thepossible impacts that this majormid-second millennium Bceruption may have had on terrestrial and aquatic biota. It is conducted in direct association with a distal tephra layer firmlyattributed tothe Santorini(Minoan) eruption and takes place within the zone of tephra deposition in southwest Turkey as delimited by tephra iso- pachmaps (Pyle, 1990; Sigurdssonetal., 1990; McCoy and Heiken, 2000; Pearce etal., 2002; Figure IA). In terms of proxy data that might reflect suchimpacts, pollen can elucidate short-lived perturbations to terrestrial vegetation, as well as reflecting longer-term landscape and climate dynamics. Within aquatic ecosystems, diatom assemblagecomposition and abundance provide a sensitive index of alterations to limnological conditions, including nutrient status, pH, conductivityAalinity andlight climate. However, diatoms represent only a proportion of the biomass in lake ecosystems and changes in otherindicators can provide both alternative and complementary lines of evidence (e.g.,Sayeretal.,1999). Inaddition to diatomanalysis,it is possible to assess changesin other components of lake ecosystems, including aquatic macrophyte composition and abundance (aquatic pollenand other microfossilremains),algae(coenobiaof Coelastrum and

Pediastrum)

andspongespicules.

Thesediments of the lake ecosystematG6lhisararewellsuited to assessing the regional impact of tephra deposition, as they include the distal component of the Santorini tephralayer in a continuous, replicated sequence (seeFigure 2 in Eastwoodetal., 1999a), thereby allowing possiblecause/effectrelationshipstobe established directly from proxy palaeoecological data. The tephra wasdeposited atadepth of between-245 and 275 cm (seeEast- woodetal., 1 999a, Figure 2 and text, for explanation) and consists of transparent,colourless, vitricshards and pumicefragments from submicronto>200pm in size. Morphologically,thetephra consists ofplatey shards together with fluted, elongate-shaped pumicefragmentsand vesicularpumice fragments characterized byflattenedspheroidalorellipsoidalvesicles andtypicallyrhyol- itic in nature. Geochemical studies undertaken on the glass shards from the tephradepositat

Gtlhisar

by Eastwoodetal. (1998a;

1999a) and Pearce etal. (2002) show unequivocally that the provenance of the tephra is the -3300'4Cyr BP ormid-second millennium BC 'Minoan' eruption of Santorini or Thera. Radio-

carbon age determinations on the peat underlying the tephra layer at G6lhisar (3300±70 and 3225 ± 45 yr BP: Eastwood etat., 1999a) support the geochemical results on a Santorini provenance for the tephra.These '4C ages do not, however, help to break the present impasseconcerning an exact and much-needed calendrical date for theeruption due to the assigned errors and the nature of the calibration curve for this period (Housley etal., 1999). The lakesediments atG6lhisar also contain well-preservedpollen,dia- tomandsponge-spicule records. We have carried out fine-interval stratigraphic investigationsof these palaeoecological indicators fromsediments associated with this tephra, and here test the null hypothesis that the Santorini eruption and its subsequent tephra deposition had no effect on either the terrestrial vegetation or the aquaticecosystematG6lhisar in southwestTurkey.

Although the evaluation of competing hypotheses is not new in palaeoecology(e.g., Flower and Battarbee, 1983), they have usually been tested through the falsification or verification of multiple working hypotheses (cf. Chamberlain, 1965) and/or are reliantonsite-specific conditions(e.g., PeglarandBirks, 1993).

The evaluation of multiple causal factors, though a useful approach,ishampered by thepossibilityofconfounding effects (e.g.,vegetation succession, lake ontogeny). However, with the adventofappropriatestatisticaltechniques(terBraak and Pren- tice, 1988; ter Braak and Smilauer, 1998)and associated com- puter-intensiverandomizationand permutation tests (Birks, 1998), confounding effects may be partialled out to evaluatestatistically the responses ofpalaeoecologicalvariablestoenvironmentalper- turbations.For example,Lotterand Birks(1993),Birks and Lotter (1994)andLotteretal.(1995)used(partial) redundancyanalysis (RDA;terBraak, 1994)to investigatetheeffects of the -11 500 14C yrBP Laacher Seeeruption in Germany on terrestrialand aquatic biota. Similarly, Barkeretal. (2000) investigated diatom responsestotephradeposition in crater-lakesediments from Tan- zania usingvariancepartitioningandrate-of-changeanalysis. In the G6lhisarstudy a series of (partial)redundancy analyses is applied using the compositionandconcentration ofkey terrestrial and aquatic ecosystemassemblages as response variables, and withchanges in Santorini tephra, changes in local depositional environment and time aspredictororexplanatory (co)variables.

Mv- C7 | zzr.

244^XX_ ,MEEl

2I

Xz'g1

250 s 251

F

254- 255A

257s 20 40 s0 so 13 200 40W 30 s00

Losson nfltior 10granscm

Ca1oreois silt-ch1 O'gonic-r6chsitt-clay Tephro Pect & ephrN Peat

Analysts WKEAJCT

Figure2Lithology,loss-on-ignitionandmicrofossil concentrationdata for theGolhisarshort coreGHE.93-6 (sed=sediment).

ICN

11: 1-1(p

.4t.f

41P NO

'14§. ^\11

..e

A

. I

-x Ii

.=ft-

Downloaded from

(5)

The study

area

Golhisar Golti (3708'N, 29°36'E; elevation 930 m) is a small (-4 km2), shallow (-2.5 m) lake located in the Oro-Mediterranean vegetation zone ofthe Lycian Taurus Mountains in southwest Turkey (Figure 1A). A substantial ancient fortifiedstructureident- ifiedasthe historic siteof'Sinda' by Hall(1994) is situatedon apeninsulaprotruding into the lake from its northeasternshore (Figure1B). The lake's hydrological catchment is -88km2,and risesto an elevation of 2095m. Thecatchmenttolake area (z) ratio is 22:1,or 17:1 when the alluvial lowlands tothe north of the lakeareexcluded.Themodernlakewater is alkaline andoligo- saline(Table 1). The slopes around the southeast of the basincon-

sist mainly of Mesozoic limestone. Ultrabasic rocks (peridotite, serpentinite)areexposedtotheeast,whereas Neogenemarlout- crops tothe southwest and chertylimestonetothe southof the lake (Figure lb). Average annual precipitation for Golhisar is -600mm, ofwhich -50% fallsduring DJF and 12% during JJA (Meteroroloji Bulteni, 1974). Present-day land use immediately surrounding the lake comprises mostly cultivated fields, while the hills and rocky outcropsare generally barren, probably largely through subrecent deforestation and overgrazing. Degraded oak scrub (Quercus coccifera) ^{occurs on} the slopes surrounding the lake with pine (Pinus brutia,P.

nigra)

athigher elevations.

Bottema andWoldring (1984) describedthepollen stratigraphy ofa-2mlong lake marginalcorefromGolhisar.Thebasinwas

reinvestigated in greater detail in 1992 and 1993 when further

coresandpeatsectionswereobtained. One of thesecores(GHA:

8.13mlong)hasabasalageof about9500 '4CyrBPand records thedevelopment of Holocene woodland comprising oak, pine and juniper, followed by a period of human impact (the

Beysehir

Occupation,orBO phase) which is characterized by fruiticulture (olives,manna,pistachio, walnut), cereal-growing and pastoralism (Eastwood etal., 1998b; 1999b). Radiocarbon dating indicates that the BO phaseatGolhisar Golu beganat -3160 '4C yrBP (cal. -1400 BC) - shortly after the deposition of the Santorini tephra layer (datedto -3300 '4C yrBP; cal. -1613 Bc)- and continued until -1300 '4C yr BP (cal. -AD 700) when pine becamethedominantpollentype.

DiatomsarepreservedinmostofcoreGHA andcomprisepre-

dominantly freshwater alkaliphilous periphytic or benthic taxa (seeFigure11inEastwoodetal., 1999b).Thissuggestsrelatively shallowbutpermanent waters atthislake-marginal site. The early- Holocene record includes diatoms such as Cymbella spp. and Cocconeisplacentulawhich live attachedtoaquaticmacrophytes.

Towards 6000 '4C yr BP some taxa (e.g., Nitzschia spp.) are presentwhich tolerate slightly brackishwaters, and thisis fol- lowedbya zoneofpoordiatompreservation. Above the Santorini tephra layer (STL) diatoms are better preserved and species assemblages are lessvariable stratigraphically, being dominated by speciesof smallFragilaria.

Thetime interval studiedwasdeliberately selected after analysisof thelongsequences, so astobe 'nested' within millennial- scale environmental changes, and witha sampling interval fine enoughtopermit annual-decadalresponsesto tephra depositionto

beevaluated. Themid-secondmillenniumBCeruption of Santorini (Thera)occurredpriortomajorhuman-induced deforestation in the Golhisar catchment (Eastwood etal., 1999b), which ^means that any volcanicimpacts would have been registeredon what

was alargely natural landscape. The short duration of the interval under studyalsoimplies that variables whicharesignificanton a

Holocene timescale (e.g., lake ontogeny,forest succession)are

unlikelytohave beenimportantovershort times and their influ-

ence canappropriatelybe 'partialled out' statistically.

Sampling, analytical and statistical methods

The GHE series of sediment cores (see Figure 2in Eastwood

etal., 1999a), obtained specificallytoprovidematerialwith which totestthehypotheses advanced in this study,wereretrievedusing

a 30cm long, 3cm diameter Dachnowsky corer. Cores were

extruded in thefield, wrappedinclingfilm, placed inlabelledsec-

tions of PVC guttering cut lengthways, and then wrapped in heavy-duty plastic sleeving. Upon return to the laboratory the

cores werestoredat-4°C. The sedimentsweredescribed in the field and inthelaboratory usingamodified version of the Troels- Smith(1955)schemeasproposed by Aaby and Berglund (I 986).

Theupper16.5cmof shortcoreGHE.93-6, takenatthe southeast-

ernlakemargin, startingata depth of 240cm, wasselectedfor detailed investigation.Unlike lake-centre cores from Golhisar, wherebioturbationand othermixingprocesseshave ledtovertical diffusionandblurringof theSantorini tephra layer inapredomi- nantly unconsolidatedlake-mudmatrix,the 4cmthick layer of tephra at the GHE core site appears to have been deposited directlyon apeatsurface and iswellpreservedas adistinct hor- izon,and is thussuitable fortestingthepalaeoecologicalhypoth-

esesoutlined in thispaper.

Theorganic matter and carbonate contentsofthe sediments (Figure 2) have been estimated using loss-on-ignition (Dean, 1974). Samplesforpollen (1 cm3) anddiatomsweretakenevery

0.5cm above the tephra layer andevery Icmbelow it. Pollen extraction follows the standardproceduresofFaegriandIversen (1989).ExoticLycopodiumtabletsofaknown concentrationwere

added in order to estimate pollen concentrations (Stockmarr, 1971). Pollen grainsand other non-siliceous microfossils were

countedusingaNikonLabophot-2 microscopeuntilaland-pollen

sumof 350grainswasreached(exceptintheSTL) with critical identificationsbeing conducted under oil immersion at X1000, togetherwithphase-contrast microscopy. Aquatic pollenandnon-

siliceous microfossilsare expressedas apercentageof thetotal

Table1 Water-qualitydata forGclhisarG6tu

pH Cond. Datain molarequivalents(meq1 ') Anionsum Cationsum

(u Scm ')

Mg Ca Na K Cl S04 Carbonate

July 1992* 8.95 1350-1500 nd nd nd nd nd nd nd nd nd

August 1996 8.3 920 8.58 1.69 1.30 0.09 0.82 1.66 7.01 9.46 11.66

April 1997 9.0 660 8.38 1.66 0.91 0.07 0.76 1.25 6.41 8.42 11.03

September 1999 8.1 765 nd nd nd nd nd nd nd nd nd

July2000 8.4 697 nd nd nd nd nd nd nd nd nd

July2001 8.0 867 nd nd nd nd nd nd nd nd nd

*Lake

edge

sample.

nd=not

deternmned.

Downloaded from

(6)

number ofpalynomorphs.Generally, pollenpreservationisgood, but the Pinaceae undiff. category includes grains that couldnot confidently be assignedtoahigher taxonomic resolution and larg- elycomprisesgrainsthat aredegradedand crumpled. Fossil dia- tomsampleswerepreparedusing standard techniques(Battarbee, 1986) and counted in transects at X1000 magnification with a ZeissAxioscopecompound microscope equipped withdifferential interference contrast optics, and identified by reference to Krammer andLange-Bertalot (1986, 1988,1991) and other taxonomic floras. Diatom andsponge-spiculeconcentrations were cal- culated usinga known area of coverslip counted and a known

proportion

of

sample

addedtothe

coverslip. Tephra-shardconcen-

tration(TSC; Figure 2) wasquantified by contact with two pre- selected microscopegraticule pointsin 200 fields of view on the diatom slides alongpreviously selected regularly spacedtransects using thepoint count method (Clark, 1982).

Local pollen and diatom assemblage-zone boundaries were delimited on the basis ofstratigraphically constrainedincremental sum-of-squares clusteranalyses (CONISS; Grimm, 1987) usinga square-roottransformationandchord-distance dissimilarity meas- ure forthepollen anddiatom types thatoccur atgreater than2%

abundance; differentpalaeoecologicalvariables(pollen, diatoms, etc.)wereeachzonedseparately. Summary pollen and diatom dia- grams were constructed using Tilia and Tilia-graph (Grimm, 1991) and show thepollenandalgal types and diatoms usedin the numerical analyses, while uncommon diatom species are groupedat genus levelin thediagram.Numerical analysescon- ducted onrelativeabundance dataincludedall taxa which had a representationof >1% in the data setsexamined.

Detrended canonical correspondence analysis (DCCA), with detrendingby segmentsand non-linearrescalingwithdepthasthe solepredictor variable,wasusedtoassessthegradient lengthof variation in the stratigraphical data. This technique provides a measureof thecompositionalturnoverof the data sets inrelation to depth (Hill andGauch, 1980;ter Braak andPrentice, 1988).

Gradient lengthisfundamentalto the choice of ordination technique because, over short gradients, taxa respond in a linear fashion,while overlongergradientstaxarise and fall inaunim- odal manner(terBraak, 1994;terBraak andPrentice, 1988). All gradientsinthe data wereshort (< 1.9 standarddeviations) and therefore RDA, theconstrainedform of linearprincipalcompo- nents analysis(PCA) ordination (terBraak andPrentice, 1988), wasused.RDA enablestheeffect ofoneor moreenvironmental forcingson multivariatedatasets tobemodelled andevaluated statistically (ter Braak, 1994). Throughthe use of Monte Carlo permutationtests,RDAcanassesswhetherbiotic shiftsassociated with

(palaeo)environmentdl

phenomena (such as tephra

deposition)

areno morelikelythanwouldbeexpected bychance.

Furthermore (where relevant data exist) the influence ofcon- foundingvariables(termed 'covariables')canbeallowed for and

'partialled

out'(terBraak, 1994).

Relative abundance data for pollen and diatom assemblages were square-root transformed, whiletotal concentration datafor pollen,diatom andspongespiculeswere

log,()

transformedprior tonumericalanalysis.Theexplanatoryvariables used for the RDA were depth, lithology and tephra shard concentration (TSC;

Table2).

The explanatory variable'depth'wasused hereas a surrogate fortime. Variance

explainedby depth

maythereforebeassociated withlong-termforcingmechanisms suchasunidirectional climate change,soilmaturationand lake ontogeny.

Thevariable'lithology' includedpercentageorganic,carbonate andminerogenicmattertogetherwithfourvisible sedimenttypes

(marl,

clay, tephra, peat; Figure

2)

each coded (0 or 1) as a separate dummy variable (cf. ter Braak, 1990a). Sediment type

(lithology)

isreflectiveofthe environment in whichmicrofossils weredeposited(e.g., lakelevel,anoxia). Lithologyisapotentially

Table2 Explanatory variables and covariables used in RDA and the rationale for their selection

(Co)variable Factorsattributableto the (co)variable

Lithology Limnological conditions,

depositionalenvironments,changing taphonomic conditions

Depth Unidirectionalclimatechange, soil

maturationandlakeontogeny

Tephrashardconcentration(TSC) Santorini tephra effects

importantexplanatoryvariable for theaquaticecosystemdata,as changes inlake environment may affectprimaryproductivityand species composition. Sediment

lithology

may alsoreflectcatch- ment landscape conditions. However, changes in lithology over the shorttimescaleanalysedinthisrecordare unlikelytoreflect processesaffecting thecomposition ofterrestrial vegetation not accountedforbythe variabledepth. Lithologywasthereforenot used as anexplanatory variable for the terrestrialpollen percentage dataascatchmentvegetationrelativecompositionshould be unrelatedto lake sedimentary facies. Lithology was used, however,asacovariablefor all data sets,includingpollen percentage data, in ordertopartialoutthe effect ofdepositionalenvironment onmicrofossil taphonomy.

TSC is likely to bemore

representative

of the effect of the Santorinitephraon ecosystems thanasimpleexponential decay functionof shard abundance(cf.LotterandBirks, 1993).Theuse ofconcentrationvalues enablesvariation inthe influenceof the Santorinitephra layer (and therefore its effect) to be assessed moreaccurately.

In orderto testthenullhypothesisthat theSantorinieruption hadnoeffectonecosystemsat

G6lhisar,

thestatisticalsignificance of thenumericalrelationshipbetweenthebiologicalvariables and TSC wasassessed using restrictedMonteCarlopermutationtests ofRDAaxes.Inthisanalysis,TSCwastheonlyexplanatory variable and the variance in the biological data explained by depth and/or lithology (representative of a range of other factors;

Table2) was

partialled

out (i.e.,not attributedtothe Santorini eruption), becausethe (co)variance associated with changes in depthandlithology neednotberelatedto thedepositionof the Santorini tephra.

The evaluationof the significanceof TSC with the variation associated

with lithological

changespartialledout(using lithology as a covariable) is a stricttest of the effect of the

eruption's

impact, as many lithological variables covary with TSC. For example, TSC and %mineral residuearesignificantlycorrelated

(i2

⁼0.37, p⁼0.003).Furthermore,aquaticbiota may contribute tochangesin lithology,rather than beaffectedby them. Monte Carlo

permutation

tests of the significance of the explanatory variablesinvolved 199 restricted

permutations

fortime-series (ter Braak,

1990b;

ter Braak and Smilauer,

1998),

appropriate to

stratigraphical

data

(Birks,

1998).The significanceofRDA axis I wastested for eachsingle explanatory variable,whileanoverall significancetest wasusedwheremorethanoneexplanatory vari- ablewaspresent. The RDAprocedureused doesnotconsider any significant lagsinecosystemresponsetoTSC variance.Using this technique,ecosystem response is assumedtobesynchronousover the timeperiod

represented

byasingle sample.Forpartial RDA, restricted

permutations

under thefullmodel (ter BraakandSmi- lauer, 1998)were used.

Inadditionto

evaluating

the

independent

effectoftheSantorini eruption,wetested several null hypothesesregardingthe com- binedeffect on the range of

biological

responses ofTSC and changes indepth(all indicators)and the combined effect ofTSC

Downloaded from

(7)

Table3 Significance levels of relationships between biostratigraphicaldata sets fromGolhisar,with differentexplanatoryvariablesandcovariables(see methodssection).Significancelevels areestablishedusing restrictedMonte Carlopermutationtests(n=199)

Explanatory Covariable(s) Terrestrial pollen Aquaticpalynomorphs Aquatic pollen Spongespicules Diatom variable(s)

% Conc. % Conc. Conc. % Conc.

All 0.005** 0.005** 0.005** 0.005** 0.005* 0.005* 0.005*

TSC depth 0.005** na na na na na na

TSC depth+lithology 0.20ns 0.03* 0.O6ns 0.06ns O.lOns 0.33ns O.8Rns

TSC+depth lithology 0.005** 0.04* 0.005** 0.25ns 0.04* 0.12ns 0.69ns

TSC +lithology depth na 0.005** 0.005** 0.03* 0.005* 0.005* 0.01*

TSC =tephra shard concentration.

%=relative abundance data.

Conc.=concentration.

**=significantatp<0.01.

* =significantat0.01<p '0.05.

ns =notsignificant.

na=notanalysed.

andchangesin sedimentlithology (all indicatorsexceptterrestrial pollen composition). In eachoftheseanalyses, the varianceasso-

ciated with lithology and depth,respectively,werepartialledout as covariables to separate their influence. The variance in the

sevendifferent palaeoecological datasetswaspartitioned following the approachof Borcardetal.(1992) usingaseries of(partial) RDA. The modelswere designedtoestimatethe variance in the

response data sets that is explained by all predictor variables, explained by TSC independent of time and sediment lithology, explained by TSC and time independent of sedimentlithology, explained by TSC and sedimentlithology independent of time, andunexplained variance (Table 4).

Allordinationanalysesweremade withtheprogramCANOCO version 3.12 (ter Braak, 1990a; 1990b) and were re-run with CANOCO version 3.12a with strictconvergencecriteria.

Results

Litbostratigraphyandtime duration

Theupper16.5cmof shortcoreGHE.93-6 comprisesfourlitho- logical units (Figure 2). The lowest part of unit GE-l (256.5- 252.5cm)iscomposedofblackhumifiedpeat(tipto70%organic matter;Figure 2),withtrace amountsoftephra shards. The slight

decreasein percentageorganic matterduringthe upperpart of this unit(252.5-250.5 cm) correspondstohigher concentrations oftephra (and is labelledseparately on Figure2 as 'peat and tephra').Unit GE-2(249-245cm)iscomprised predominantlyof tephraashighlighted bythe marked increase in TSC and mineral residue(-90%).InGE-2,terrestrialpollendata exhibitamarked decrease in concentration.For the purposeofclarity, the sedi- mentsoflithological unit GE-2 are hereafter referredto as the Santorinitephra layer (STL).Theapparentverticaldisplacement oftephrashards isaphenomenonthat isnotuniquetothedeposits atGolhisarandhaspreviouslybeenreported inpeatdeposits and lakesedimentsfromScandinavia,Faroe Islands, Iceland, Northern Ireland and Scotland (Persson, 1971; Pilcher and Hall, 1992;

Thompson etal., 1986; Dugmore et al., 1995; Charman etal., 1995).Unit GE-3(245-243 cm)iscomposedoforganic-richsilt- claywithamarked decrease inTSC, whilelithologicalunit GE- 4 (243-240.5cm) is made up of calcareous silt-clay (-10%

carbonate) correspondingto a considerable increase in diatom concentrations andrelatively low TSC.These visiblelithological unitscorrespond closelytoquantitative analysesofsedimentcom-

position, with some minor differences at unitboundaries(e.g., measured TSCstayshigh in the basal samplefrom unitGE-3).

In the absence ofannuallylaminatedsediments,it isnotposs-

ibletoprovideapreciseestimate of the timeperiodcoveredby

Table 4ProportionofGdlhisarbiostratigraphicdata setvariance explained by differentsetsof externalfactors(seemethodssection). Explanatoryvariables usedforterrestrial pollenpercentagedata were TSC+depth only;for allotherdata sets themodelsused were TSC+depth+lithology. Lithologywas usedasacovariable withall datasets. Entriesshown inparentheseshave ap-valueof>0.05(Table 3)and areconsiderednotsignificant

Sourceof variance Explanation Terrestrial pollen Aquaticpalynomorphs Aquatic pollen Sponge spicules Diatom

% Conc. % Conc. Conc. % Conc.

Temporal andlithological change All 31.3 73.0 82.7 55.8 88.0 65.0 84.8

and Santorinieffects

Santorini effectsindependentof TSC-D-L (9.0) 26.2 (10.6) (8.0) (43.5) (7.3) (0.8)

time andsediment change

Santorini and time effects TSC+D-L 20.8 38.2 18.5 (12.9) 60.9 (15.8) (7.5)

independentofsedimentchange

Santoriniand sedimentchange TSC+L-D 19.5 72.3 70.5 52.3 87.7 48.9 58.2

effectsindependentof time

Unexplainedvariance 68.7 27.0 17.3 44.2 12.0 35.0 11.2

TSC:tephrashard concentration.

D:depth.

L:lithology.

Conc:=concentration.

Downloaded from

(8)

short core GHE.93-6. However, a broad time envelope can be provided by the '4C dated age-depth scale which exists for the long Holocene core GHA, taken from the same sampling site (Eastwoodetat., 1999b). Overall sediment-accumulationratesfor thiscore range between 0.4 and 0.8 mm yr-',except for a short phase of increased sedimentationsoon aftertheonset of theBO phase of forest clearance. In addition, for short core GHE.93-6, the tephra needs to be removed from sedimentation-rate calcu- lations, and on this basis the 16.5 cm profile of GHE is estimated tocover a timespan of between 120 and 250 years.

Terrestrial pollen

Generally,theterrestrial pollen percentage data (Figure 3) do not show anysignificant changes. There is very little overall change inAP values before(-85%-95%)or after(-92-96%) the deposition of the STL. Although AP values attain -96% before the deposition of theSTL, this is in just one pollen spectrum. Of the main AP types,QuercusandCedrus show only slight increases (attaining values of -12% and -8%, respectively), while Pinus hasaslightdecreaseinthesamples containing the STL. Degraded andbroken grains increase by-5% and-10%, respectively, after thedepositionof theSTL(Eastwood, 1997)whichmay be afunc- tionofdegradationasaresult of increased runoff afterdeposition of thetephra layer. Of thenon-arborealpollen (NAP)types,Gra- mineae showsaslight increase (to -20%)in thesamplescontain- ingelevated TSC and thendeclinestoits lowest valuesfollowing tephradeposition (-5%). In short, percentage pollen data do not suggest any significant effecton terrestrialvegetation.A -14%

decrease inPinu.spollen, which is anotoriously high pollen pro- ducer, and which constitutes most of theAP, is not considered to be representative of any significant ecological changes. In addition,firmecological interpretationsarefurtherhampered due to the limitationsimposedonpalynologicaldatato recordaccu- rately ecological plant associations(Bottema, 1982),as well as limitations imposeddue to dispersal, representationand preser- vation.

RDA shows that all relevantexplanatoryvariables

(in

this case, TSC anddepth; Tables 3 and4) incombinationexplain a statistically significant (p=0.005) amountof variance (31.3%) in the relative abundance terrestrialpollendata. However, analysisof the STLeffects,withthe variance associatedwithdepth andlithology partialled out, shows that TSC independently explains a much smaller andstatisticallyinsignificantamount(p⁼0.20)of the variance

(9.0%).

TSC with only the unidirectional variance associatedwithdepth partialledout,onthe other hand, explains a significant (p= 0.005) amountof the variance(19.5%) in the pollenpercentage data.

A much greateramountofvariance inthetotal terrestrialpollen concentration (73%,p=0.005)isexplainedusing all the explanatory variables incombination(TSC, depthandlithology) than in the relative abundance data. In contrasttothe relative abundance data, TSCindependently (i.e., with the effects of depth andlith- ologypartialledout) explainsastatistically significant(p=0.03) amountofvariance interrestrialpollenconcentration(26.2%). A significant (p=0.005) amount of variance in terrestrial pollen concentrationisexplainedbyTSC andlithology,withdepth par- tialledoutas acovariable(72.3%).

Aquatic palynomorphs and sponge spicules

Of the aquaticpalynomorphs(i.e., aquatic pollen+aquatic non- siliceousmicrofossils; Figure 4),lowpercentages of totalaquatic pollen(<5%)hinder anyfirmpalaeoecologicalinferences.How- ever,Cyperaceaeandnon-siliceouspalynomorphs (typology after van Geel etal., 1989), Type 610 and Sigmopollis (Type 128B) recordincreases insamples in and above the STL (zones A-3 to A-5). Ceratophyllumleafspines have relatively high percentage values in the basal zone

(A-1),

and also a minor increase in

samplesabove theSTh (zone A-A),while coenobia ofPediastrum and, to a lesser extent,Coelastrum show marked increases above theSTL during zone A-5.

Aquatic palynomorph assemblages respond significantly (p=0.005) to the combined effects of TSC, depth and lithology (variance explained: 82.7%). However, the independent influence of TSC (with depth andlithology partialled out) on aquatic palynomorph variance (10.6%) is not statistically significant (p=0.06).

Variance explained by TSC and lithology combined, with the effects of depth partialled out, is statistically significant (70.5%

p=0.005).

The combined explanatory variables explain only around half the variance in aquatic pollen concentration(55.8%), and TSC as an independent variable does not account for a significant (p =0.06) amount of variance (8.0%). In accordance with all other data sets evaluated, changes in TSC and lithology explain a statistically significant amount of variance in the aquatic pollen concentration, even when the effects of depth are removed as a covariable (52.3% p = 0.03).

A large and significant (p = 0.005) amount of sponge-spicule concentration variance (88%) isexplainedby TSC, depth and lithology combined. Interestingly, the proportion of sponge-spicule variation explainedby TSC and lithology with the effect of depth partialled out (87.7%) is almost the same as that explained by all variables (Table 3). Despite explaining approximately two-fifths ofthe variance inspicule concentration, the effect of TSC alone is not statistically significant (p = 0.10). TSC and depth explain a significantportion of variance in sponge spicules (p=0.04; 60.9%

variance explained) with lithology included as a covariable.

Diatoms

The diatom record forGHE.93-6 is dominated by small Fragilaria taxa (F. brevistriata, F.pinnata and F. construens) which, as a group, increase in abundance towards the top of the core (Figure 5). Within these, F. brevistriata and F. construens increase towards the top ofthe sequence, with F. pinnata being most abundant in the middleof the sequence. Apart from the increased representation ofNitzschiainconspicuiain zone D-5 and thevirtual elimination ofEpithemiaadnatafrom therecord above D-2, there are few other unidirectional shiftsin diatom species composition.

Planktonic Cyclostephanos dubius and Aulacoseira

granulata,

along with the littoralAmphorapediculu.s, have bimodal relative abundance curves; C. dubius is common in zones D-1 and D-3 whileA.pediculus is wellrepresented in D-2and D-4.For much of the record the representationofC. dubius andCyclotellaocel- latais broadly similar;however, C. ocellata (15.2%) is consider- ably more abundantin zonesD-4 and D-5 thanCyclostephanos dubius (max 6.7%, mean 2%). These shifts in diatom assemblage compositionmirror rather closely the core lithological units. Over- all, however, the most notable feature of the diatom record is the marked increase in diatom concentration above the STL, especially in zone D-5 (Figure2).

TSC, depth andlithologyin combination explain a somewhat lowerproportionof variation(65%) of the diatom compositional data than for most otheraquatic indicators; however, this relationship is still significant (p=0.005; Tables 3 and 4). When the effects ofdepth and lithology are partialled out, a low (7.3%) and astatistically non-significant(p=0.33) amount of the diatom variance is explainedby TSC. However, TSC and lithology with depth as a covariable explain a significant (p=0.005)proportion of the variance in thediatomdata(48.9%). TSC and depth, with lithologyas acovariable,arenotsignificant (p=0. 12) explanatory variables for the diatomvariance (15.8%).

Similar results to the species composition data were derived from RDA of the diatom concentrations. Only combinations of all the variables,orofTSC+lithologywith depth partialled out,

Downloaded from

(9)

C=; cr

...

... ....

... ... ...

...

... I... 2:1

... ......

...

'.l.-.l.'.-.'.'.'.l.l.1.1.1111l

...

... ... E

...0...

......

....

ri

...

7.

... ...

...

l:.l.-.I.I.I.I.I.l.-.-.l.W...

...;.

...

... ...

...

. ... .

...

CL cl -0

.C=o

Q4

cr.

...

... ......

... ... CL

Go

CN CL

...

Er

(U

... 04

co

IC) "M

(wo) 41dl(l

Li

;(Mtu ,0 1Li

>1 In,

750

4:6 -I ..$ W-' --r le,

* -* -t -r -P .4

" -4 C-4 C.4 C14 f.

Downloaded from

(10)

SE:

.Q.e

r......-... e

v;

c

0 0

v

+

o0 0

00 0 0:

0

3

0:

0L

0

._

-c

.0

0 0

.

0J

0

v

0

v 0 0

0 0 0L

C C

0 v

0 1

00

1~~~~~~~~~~~--- ...

___ vOK.M

(wa)qjdaa

*"

p _

E

-6

Downloaded from

(11)

-O0

C) (A 4o

Ct-=

0)

,c:;p co

40.

,C),C's

...

.........

C14

...

......

...

...... Cl

...... C4 CL

...

CL.

.. ...

... 64

C'4

... CL

... ...

.....

... C14

...

... %

...

cn

QD

......

V. C*

CD

...

... cn

...

r

Ln 4n %m WI L" tn L"

C14 C4 C14 eq C14 fl-4 C-4 C-4 C14 eq eq r4 C4

(w3) qldao

4f

lao

Downloaded from

(12)

explained significant (p=0.005, p=0.01,respectively)amounts of the variance in the diatom concentrations (84.8%, 58.2%, respectively; Table 4).

Discussion

The main potential impacts of volcanic eruptions on terrestrial biota include physical damageto plants by tephraorashfall,the

presenceofvolatile acidsin the atmosphere (direct impacts), and regional or global climate perturbation (indirect impact). The direct effects of volcanic eruptions usually have the greatest

impacts inareasproximaltothe volcanic eruption, although there

are exceptions and these sometimes involve the deposition of acidicaerosols (Grattan and Charman, 1994; SadlerandGrattan, 1999). Palynological investigationsonthe direct impacts of volcanic eruptionsby Wilmshurstand McGlone (1996)inNew Zea- landshowed that therewassignificant damagetoproximal forests following the 1850 BP Taupo eruption but only minor damage in distalareas, with forestrecovery in proximal areas being com-

pleted within -200 '4Cyears.Modern studiesonvolcanic impacts caused by the 1980 Mount St Helens eruption (Mack, 1981)report thatdamageto vegetation outside the actual blastzone wasonly slight and was related mainly to plant morphology. Tephra- coating of trees was largely washed off afterthe first rainfall, whileprostrateherbsorthosewith claspingleavessufferedmost.

Plantswithlong slender leaves (e.g.,grasses)werehardlyaffec- ted, presumably duetotheir abilitytoshed quicklyanydeposited tephra (Mack, 1981). The Golhisar catchment (-400 km ENEof Santorinivolcano)appearstohavereceivedamaximum of-4cm

ofdistaltephra fallout. Potentially, this thickness of tephra might havebeen enoughto causesomephysical damagetolow-lying herbs similartothosecausedbythe MountSt Helenseruptionas

reported by Mack (1981). Such herbs are presumably more

strongly affected than tall and robust plants. This might explain the slight increase in Gramineae (and possiblyCyperaceae) pollen valuesasrecordedatG6lhisar during and after the STL. A similar increase in Gramineae and Cyperaceae was also recorded by Lotter and Birks(1993), Birks and Lotter (1994) and Lotteretal.

(1995) in their investigationson thepalaeoecological effectsof theLate-glacial Laacher See eruption.

Research intotwentieth-centuryvolcanically inducedclimatic perturbations suggests that temperaturelowering occurs in the

range 0.3-0.50C and lasts for a period of 3-4 years (Mass and Portman, 1989; Robock and Mao, 1995) but the relationship between volcaniceruptions and climate fluctuations is complex (SadlerandGrattan, 1999). Kuniholmetal.(1996) have suggested thataseries ofanomalously widetree-ringsrecordedin the Porsuk tree-ring sequence from south-central Turkey can be correlated with the Minoan eruption of Santorini. In climatic terms this would indicate cooler summers, higher precipitation and/or increasedcloudinessresulting in enhancedtreegrowth, rather than thereducedtreegrowthinferredfromIrishbog oaks and Californ- ianpines atthistime (Baillie and Munro, 1988; La Marche and Hirschboek, 1984). There is no othersignificant anomaly in the Porsuk series, positive or negative, during the first half of the second millenniumBC. If, as seems feasible, this dendrological anomalyrepresentstheclimaticlegacy of theSantorini eruption, the perturbation lasted only for 6-7 years based on tree-ring counts.Aneventofsuch shortdurationwill be barely detectable in most lake or peat stratigraphies, except when annual laminationsarepresent.Evenwith fine-intervalsampling of lake- sedimentcores,subdecadalclimaticchangeis normallylikelyto berepresented by onlyafew individualspectra.

Our analysesindicatethatno major changecanbediscerned interrestrialpollen assemblagesfromG6lhisar,andhencein the surrouLnding terrestrial vegetation, following tephra deposition.

Furthermore, even when time and STL-related effects are included in ouranalyses, more than two-thirds of the stratigraphic variance in terrestrial pollen composition remains unexplained (Table 4).

Thisrules out any significant direct impacts of the Santorini tephra on regional vegetation over the timescales considered here.

Indirect impacts related to any subdecadal climatic perturbation areharder to evaluate in the absence of a high-resolution chronos- tratigraphy. In contrast to the lack of change in pollen composition, total pollen concentration shows a clear decline during the STL, and is the only palaeoecological indicator analysed that shows a statistically significant synchronous response to TSC. We infer that the decline in pollen concentration during the STL is likely to be a dilution effect during a period of acceleratedsedi- ment accumulation, associated with the fallout and/or inwash of tephra. As such,the statistical relationship between total land pollen concentration and TSC is unlikely to have any real environmental significance.

In terms of the response of the aquatic ecosystem, some of the main potential impacts of thevolcanic eruption are related to increased nutrient input, from chemical weathering of tephra or changes in lake physicalcondition such as sealing the sediment- waterinterface or burial of lake-marginal plant communities. Sea- ling of the lake bed by deposition of a substantial thickness of tephra canprevent release of phosphorus into the water-column from the uppermost sediments. This, in turn, can alter the Si:P ratio of thelakewaters, favouring diatom genera such as Synedra over Aulacoseira (Barkeretal., 2000). However, shallow-water conditions would argue against this being a significant contribu- toryfactor at the core site analysed here. AtG6lhisar,there are marked changes in bothcomposition and concentration of diatoms

during

the time

period representedby

thecoresequence. Concen- trations are low in theperiod before the main tephra deposition, although slightly elevated diatomconcentrations in samples from 251-252 cm(Figure2) are associated with increased TSC during zoneD-2(Figures 2 and 5). This minor increase in diatom concentration may represent initial fertilization effects by bioavailable nutrients deposited directly onto the lake surface. A possible mechanism for this would be the almost immediate dissolution of thesubmicron size fraction of the tephra (e.g., <2 p m) as it filters down through the water-column (personal communication, J.A.Westgate,University of Toronto), but more work is needed tosubstantiate this. Interestingly, despite the deposition of the Si- rich Santorini tephra (Eastwood eta!., l999a), diatom concentrations show no significant directcorrelation with TSC and depth, if the effects of lithology are partialled out. Low diatom concen- trationsin the Si-rich environment of theSTL may have resulted fromtephra particles inducing light limitation, particularly given the non-planktonic habitat of many taxa in the record, but that may also be adilution effect, as with terrestrial pollen.Cyclosre- phanos dubiushas some of its highest abundance within the main tephra peak itself and,along witha peak inCyclotella ocellata, contributes to a highproportionof planktonic taxabeing associa- tedwith the maximum TSC values. High proportions of planktonic taxa may result from abioticshadingofnon-planktonictaxa such asEpithemia adnata.

In contrast, diatom concentrations increase markedly above the STLtowardsthe top of the short core. A similar diatom increase following tephra deposition has been recorded in lake-sediment records from other parts of the world, particularly in systems which arenormally silica-deficient (Abella, 1988; Hickman and Reasoner, 1994). Unpublished modern water data for Golhisar Golaindicate fairly high values of silica (3.81 mg 1- ') which sug- gestthat silica is not limiting in the modern system (Jane Reed, written communication, August, 2001). However, it is unlikely thatsilica or other nutrients in the modern lake were the same as thosepriortomajorhuman catchment disturbance c. 3200 years ago, shortly after the deposition of the STL. Diatom concen-

Downloaded from