Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download - University of Alberta
Document related concepts
Molecular ecology wikipedia , lookup
Occupancy–abundance relationship wikipedia , lookup
Biodiversity action plan wikipedia , lookup
Introduced species wikipedia , lookup
Island restoration wikipedia , lookup
Fauna of Africa wikipedia , lookup
Latitudinal gradients in species diversity wikipedia , lookup
Transcript
Vol. 136,No. 3 The American Naturalist 1990 September INTERCONTINENTAL COMPARISON OF SMALL-LAKE FISH ASSEMBLAGES: THE BALANCE BETWEEN LOCAL AND REGIONAL PROCESSES WILLIAMM. TONN, JOHNJ. MAGNUSON,MARTTIRASK, AND JORMATOIVONEN ofAlberta,Edmonton, ofZoology,University AlbertaT6G 2E9, Canada;Centerfor Department and Department ofZoology,University ofWisconsin-Madison, Madison, Limnology ofHelsinki,SF-16900Lammi, Wisconsin 53706;LammiBiologicalStation,University CentralFishCultureStation Finland;FinnishGameand FisheriesResearchInstitute, forEasternFinland,SF-58175Enonkoski, Finland SubmittedApril 15, 1987; Revised March 24, 1988, and February21, 1989; Accepted July20, 1989 The idea thatlocal and contemporary interacprocesses(e.g., interspecific and tions),combinedwithsuchregionaland historical processesas immigration shouldproducepredictablepatternsof community reorganization extinction, and convergence kindledinterestin questionsconcerning community similarity and Wilson1967;MacArthur ofthestructure (MacArthur 1972).Can knowledge be used to predicttheorganization ofone community ofa second, and function similarbutgeographically distantcommunity composedofunreenvironmentally ofourunderstanding latedspecies?Thisquestionprobesthegenerality ofcommuand the extentto whichcommunity nityorganization ecologycan become a science(A. Johnson1973;Oriansand Paine 1983;Westoby1988). predictive at an answer,however,has not been easy (Orians 1987). Several Arriving in variouscommunity have foundhighdegreesof similarity studies properties Fuentes Schluter Pianka (e.g., 1973;Cody 1974; 1976; 1986),butothershave not (e.g., Ricklefsand Travis1980;Lawton1984;Mahon1984).Two geographically in similarlocal environments distantcommunities mayor maynothave similar of or historical because factors. For example,Mahon characteristics regional that in and fish streams rivers of Polandand Ontario (1984)found assemblages the and that from in richness two in distribudiffer species species regionsdiffer and These results tionpatterns,life-history traits, morphological adaptations. in size and isolationof glacialrefugesand thepresenceand reflectdifferences inOntario,ofentirefamiliesoflake-dwelling importance, specialists.In contrast, existbetweenassemblagesofEuropeanandwesternNorth characteristics similar of Pleistoceneand Americanstreamfishes,in partbecause of greatersimilarity eventsin theseregions(Moyleand Herbold1987). post-Pleistocene of large-scalephenomena,Ricklefs(1987) chalRecognizingthe importance and historical scopesof lengedcommunity ecologiststo expandthegeographical Because community-level characteristics theirconceptsand investigations. may Am. Nat. 1990. Vol. 136, pp. 345-375 All rightsreserved. ? 1990 by The Universityof Chicago. 0003-0147/90/3603-0008$02.00. 346 THE AMERICANNATURALIST influenced processes,theseinvestigabe differentially bysmall-versuslarge-scale studiesare tionsmustincorporate phenomenafromseveralscales. Manipulative or impractical forsuchquestions,butcomparative studfrequently inappropriate multivariate analysesof large-scale ies and naturalexperiments, incorporating testsoflarge-scalehypotheses can oftenprovideappropriate (Diamond pattern, 1986;Ricklefs1987;Brownand Maurer1989). The value of such large-scalecomparisonsdependson how well the local arematched(Orians1987).Smallforestlakesinglaciatedregionsof environments theHolarcticare similarly repeatablesuitesofenviaged ecosystemscontaining conditions.Commonformsofenvironmental severity, suchas winter ronmental influencefish hypoxiaand dystrophy (stained,acidic waters),can profoundly offish-species composition and richassemblages(Rahel 1984).Distinctpatterns ness have been observedwithina numberof regions(see, e.g., Zhakov 1974; Robinsonand Tonn 1989).Giventhe insularnatureof lakes (Magnuson1976), theirisolation,as well as theirarea and habitatdiversity, mayhave important and richness(Barbourand Brown1974; effectson fish-assemblage composition Tonnand Magnuson1982;Eadie and Keast 1984;Rahel 1986). andoverallstructural Thesecommoncharacteristics simplicity, combinedwith in species composition and regionalhistory,make small inevitabledifferences comparisons andthe Holarcticlakesgoodecologicalsystemsforintercontinental oftherelativerolesoflocalandregionalprocesses.In thisstudy,we investigation withfishassemblagesin smallforest exploretheconceptofcommunity similarity questions.Do similar lakesofFinlandandWisconsin.We addressedthefollowing richnessexistinsmalllakesofFinlandandWisconsin?To offish-species patterns in each regioninfluenced whatextentare anycommunity-level by local patterns or by thepresenceofdominant or competitors? abioticcharacteristics predators ofthetworegionsassociatedin similarwayswiththe Areassemblagestructures suchthattherelations betweenassemblagestructures andthe local environment, forone regioncan be used to predictfish-assemblage environment patternsin affectintercontinental anotherregion?How do regionalor historicalinfluences comparisons? METHODS StudyAreas withthousands Wisconsincontainlakedistricts Finlandandnorthern Southern of smallforestlakes of glacial origin(Hutchinson1957;Svardson1970;Attig in bedrockorgranitic till,resulting 1984).In bothregions,lakeslie atopa granitic is lowlevelsofdissolvedminerals(Judayet al. 1938;Toivonen1972).Dystrophy commonin both regions.AlthoughFinlandis farthernorththanWisconsin, influenceof the Gulf overallclimatesare similarowingto the ameliorating at Jyvaskyla, air temperatures Streamin northwestern Europe. Mean monthly to 16.6?Cin July Finland(62?16' N, 25?50' E), rangefrom- 9.4?C in January Instituteof Finland1984);analogoustemperatures rangefrom (Meteorological - 12.4?C (January)to 19.4?C (July)at Park Falls, Wisconsin(45056' N, 90?27'W; of Commerce1984). U.S. Department INTERCONTINENTAL FISH-ASSEMBLAGE COMPARISON 347 to theirenvironmental fishfaunasofthetworegionsare similarities, In contrast dissimilar.Because of Cenozoic connectionsand subsequent compositionally dispersalbetweenNorthAmericaandEurasia,severalfamiliesoffishesfoundin Wisconsinare Holarctic(Briggs1986).Since thoseinterFinlandand northern of the Percentsimilarity changes,however,faunashave divergedconsiderably. tworegionsat thefamiliallevelis 62%, butthisdropsto 25% and 4% at generic and specificlevels,respectively. Data Sets We compiled data sets forfishassemblages and environmentalcharacteristics for113 Finnishand 51 Wisconsinlakes containingfish.Findingsof several studies were combinedto cover the rangeof smallforestlakes in both regions.(A completeset of tabulardata maybe purchasedfromtheDepositoryof UnpubandTechnicalInformation, forScientific National lishedData, CanadianInstitute ResearchCouncilCanada, OttawaKIA 0S2, Canada.) and centralFinlandwereobtained Finland.-Data on 50 lakes fromsouthern withrotenone whole-lakesampling followedbyadditional froma studyinvolving collectionby scuba divers(Toivonenet al. 1964;Tuunainen1970; systematic Sumari1971).For one or moreyearsfrom1961to 1969,46 otherlakes were withvariable-mesh gillnets;we used meanrelativeabunsampledintensively dances (by biomass)of the 1961and 1962samplings(Anonymous1961,1963). studieswerecarriedout on 5 lakes in the Evo area (Rask Mark-and-recapture 1983,1984;RaskandArvola1985;Rasketal. 1985).FiveotherlakesfromtheEvo FisheriesResearchStationweresampledwithgillnets(M. Pursiainen, personal as were3 lakesinNorthKarelia(J. Aho,personalcommunicacommunication), with5-mm tion).Finally,4 ponds,also inNorthKarelia,weresampledintensively meshtrapsand gillnetsor withrotenone(Holopainenand Pitkanen1985;I. J. Holopainen,personalcommunication). As passive samplinggear,gill nets can involvesome selectivity by size or species; however,the use of netswithmultiplemeshsizes reducesselectivity (Hubert1983). Indeed, 18 of 20 species foundover the entire113 lakes were and Miller'sthumb caughtwithgillnets.The exceptions,ninespinestickleback names),wereeach foundin onlyone of the50 lakes (see App. A forscientific thatthesespeciesoccurredinfrequently in small treatedwithrotenone, indicating withexperimental forestlakes.Intensivesampling gillnetsalso providedrelativeabundancedata comparableto thoseobtainedfromwhole-lakerotenone/scuba sampling.Eighteenlakes had been sampledwithgillnetsbeforerotenonetreatbetweengill-net catchesand ment(Toivonenet al. 1964);meanpercentsimilarity rotenonesampleswas 86%. and Our data sourcesprovidedmeasuresof pH, surfacearea, conductivity, methods.To maximum limnological depthforall 113lakes,collectedbystandard we devisedan indexoflake isolation(App. B) andcalculated evaluateinsularity, itsvalueforeach lake usingdata obtainedfromtopographical maps. Wisconsin.-Twostudies(TonnandMagnuson1982;Rahel1984),usingcombinationsof minnowtraps,fykenets,and trammel nets,provideddata on species Wisconsin. composition (presenceor absence)forfishassemblagesin northern 348 THE AMERICAN NATURALIST Percentsimilarities formultiple samplings ofsevenlakes,ranging inrichness from 1 to 11 species,averaged99% (Tonn 1980,and unpublished data). Limnological data came fromthe same publishedsources;we also calculatedour indexof isolationforeach Wisconsinlake usingmeasurements fromtopographical maps. Analyses Species richness.-Whencomparingspecies richnessbetweentwo areas, a numberofdifferent scales mustbe considered(Whittaker 1972).Listsof species composingregionalfaunas(Whittaker's gammarichness)were compiledfrom distribution maps and accountsby Becker(1983)and Koli (1984). Regionwas definedas thegeographical area boundedby lakeswithinourdata operationally sets. A regionalbut ecologicallybased faunawouldincludeonlyspeciesfrom regionalfaunasthatare able to maintainpopulationsin smallforestlakes; we definedthisfaunaas speciesnaturally withinthelakes of our data reproducing ofregionalfaunasthatcomposedthesesmall-lake sets.Proportions faunaswere used as an indirect,regionalmeasureof habitatspecialization by fishes;these werecomparedbetweenregionsvia thez-test(R. Johnson1973). proportions Because thedata set fromFinlandincludedtwiceas manylakes as thatfrom of we performed a simulation exercisethatgavecumulative numbers Wisconsin, fromone to Ni lakes(NF = 113,NW = 51). For speciesin setsoflakesranging fromamongthe each region,a lake was chosenrandomly (without replacement) studylakes,andall speciespresentinthatlakewererecorded.A secondlakewas chosenat random,and any speciesnotpresentin thefirstlake wereaddedto a list.Thisprocesswas repeateduntilalllakeswereselected.Thewhole cumulative rateof species procedurewas thenrepeated99 times.We used the resulting thatis, accumulation amonglakesas an indexofWhittaker's (1972)betadiversity, Fromtheresultsof theextentofchangein speciescomposition alonga gradient. we also calculatedtheasymptotic number ofspeciesexpectedin thesesimulations numberof smalllakes fromeach regionusingthe Walford sets of an infinite modelof growth(Ricker methodfordetermining Loofromthe von Bertalanffy 1975). Whittaker's alpharichness(or local richness)was measuredas thenumberof fishspecies per lake. Local richnesswas comparedbetweenregionsvia the test.We also comparedtherelationsoflocal to regionalrichness Kruskal-Wallis and Faaborg betweenFinlandand Wisconsin.Ricklefs(1987;see also Terborgh thatthislocal-regional richnessrelation fora pairorgroupofsites 1980)suggested could be used as an index of the relativeimportanceof local (vs. regional) If local conditionsdominate,thenan local diversity. processesin determining on local richness, increaseinregionalrichnessshouldhavelittleeffect producing relation.Iflocal diversity is sensitiveto regionaleffects, an asymptotic thenlocal richnessshouldcontinueto risewithincreasesin regionalrichness. how specificlake featuresmay have Species distributions.-Toinvestigate thedistribution of individualspecies,we analyzeddistribution affected patterns method(Schoenerand Schoener1983). For each by the occurrence-sequence valueforeachenvironmenregion,we rankedlakesalongsequencesofincreasing maximum talfactor(pH, area, isolation,conductivity, depth)and analyzedpat- INTERCONTINENTAL FISH-ASSEMBLAGE COMPARISON 349 ternsof occurrenceof speciesalongthesesequencesfororderedor haphazard via Mann-Whitney distributions U-tests.Ifthepresenceor absenceofa speciesis relatedto an environmental variable,thenthe species' distribution shouldbe at one end of theenvironmental concentrated sequence;ifthereis no relation, shouldnotreveala patternrelativeto the sethenthe occurrencedistribution theprecisionofanypresence-absence quence.Thisanalysisdetermines threshold and has theadvantageofmakingno ofa speciesalongan environmental gradient a prioriassumptionsabout the exact locationof a threshold(Schoenerand Schoener1983;Cornell1986).Because of limitations oftheMann-Whitney test, to speciesfoundin morethanfivelakes. analyseswererestricted To examinehowpairsofspeciesweredistributed amonglakesrelativeto each thepotential effects ofinterspecific other,and thusinvestigate inthe interactions two regions,we analyzedpatternsof species co-occurrence. We used 2 x 2 tables and Fisher'sexact testto assess the degreeof positiveor contingency negativeassociationbetweenpairsof speciesin each region.Fisher'stestcould to testthenullhypothesis thatthetwospeciesofa pairwere notbe used directly distributed thatall observaindependently amonglakes,becausetheassumption is violated(Bowersand Brown1982).Instead, tionsare statistically independent we used the testto classifyco-occurrences intothreegroups:(1) pairshaving overallpositiveassociationsand P < .05 fromthe Fisher'stest (positivecooccurrences); (2) pairshavingnegativeassociationsandFisher'sprobabilities less and(3) pairsineitherdirection than.05(negativeco-occurrences); havingP - .05 The distributions of species' pairsamongthesethree (neutralco-occurrences). betweenFinlandand Wisconsinvia a groupswerethentestedforindependence betweenregions.As above,analyses ofco-occurrence G-testto comipare patterns to speciesfoundin morethanfivelakes. wererestricted Assemblagecomposition.-Toexaminecompositional offishassempatterns blagesand relationsamongassemblagesin Finland,we appliedthemultivariate methodsof ordinationand classification to the lakes-bycommunity-analysis low-dimensional summaries ofcommunity speciesdatamatrix.Byproviding data, thedata's basic structure; thesewell-established analyseshelpecologistsidentify causes of identified hypothesescan thenbe generatedconcerningunderlying in appatterns(Gauch 1982; Hermy1988). Because theyare complementary andclassification ofordination proach,a combination appliedtothesamedataset toolforrevealing thegeneralstructure can be an especiallyusefulandpowerful of data (Gauch 1982;Legendreand Legendre1983). community are available; Several specifictechniquesof ordinationand classification andweaknessesofeachhavebeenthesubjectofnumerous strengths analysesand muchdiscussionover the years,withno end in sight(Pielou 1984;Kent and Ballard 1988).We used primarily principal-components analysis(PCA) on the matrix(Franeet al. 1983)forthepurposeofordering correlation assemblagesina low-dimensional space (patternanalysis).Althoughcriticizedby some workers PCA is, conceptually, undercertaincircumstances, forgivingmisleading patterns ordination themoststraightforward method,andithas provedsuitableanduseful is low and the fornumerouscommunity analyses,especiallywhenbetadiversity data set spans a shortgradientor has a linearstructure (Austinand Noy-Meir 350 THE AMERICANNATURALIST 1971;Legendreand Legendre1983;Pielou 1984;Orloci1988).Preliminary analyses suggestedthatthesetraitsappliedto theFinnishdata set; however,additional ordinationswere performed using detrendedcorrespondenceanalysis (DCA; Hill and Gauch 1980),a methoddesignedfordatawitha nonlinear strucwe also performed ture.To complement ordination, hierarchical classification (Ward'smethodwithsquaredEuclideandistanceand the RELOCATE procedure,as suggestedin Wishart1978). we omittedrare species, definedin and ordination, For both classification earlieranalysesas those occurringin fiveor fewerlakes. This eliminatesa whichcan adverselyaffectanalnumberof zeros in thedata matrix, substantial yses(Legendreand Legendre1983);withmanymultivariate methods,especially and irregular ordination, infrequent appearancesofrarespeciesalso tendto add becauseanalyses"perceiverarespeciesas outliers, morenoisethaninformation theanalysisofthedata set as a whole" (Gauch 1982,p. 214). thusobscuring The Finnishassemblageswereanalyzedon twoscales,presence-absence and relativeabundance.Community compositionpatternsamongfishassemblages fromWisconsin(presenceor absenceonly)havebeendescribedusingordination (TonnandMagnuson1982;Rahel 1984);we used a compositepattern (fromTonn et al. 1983)to classifythe51 Wisconsinassemblagesin thepresentdata set into recurrent groupsof co-occurrent species,or "assemblagetypes." in community data sets,including structure thedefinition Whereasidentifying orclassification, ofanyassemblagetypes,is accomplished environbyordination and the generationof ecologicalhypothesesabout this mentalinterpretation structurerequireadditionaldata and analyses (Gauch 1982; Legendreand distinctness Legendre1983). To evaluateenvironmental amongany identified environmental factorscontributing to thisdisgroupsof lakes and to identify we appliedmultiple discriminant andSampson1983), tinctness, analysis(Jennrich factorsas variables.AllfactorsexceptpH werelogusingthefiveenvironmental to stabilizevariances.As withPCA, thediscriminant-analysis transformed model frommultinormality itselfis notsensitiveto departures (Legendreand Legendre distinctness oflakegroupswasjudgedbythepercentage of 1983).Environmental classifiedby thediscriminant functions. lakescorrectly RESULTS LimnologicalConditions ConsiderableoverlapexistedbetweenWisconsinand Finnishlakes forall characteristics (table1). The twosetsof studylakesweresimilarwithrespectto surfacearea andpH, beingsmallandacidic.StudylakesfromFinlandwere,as a thanthosein Wisconsin. set, deeper,less isolated,and higherin conductivity ofseepagelakes; GreaterisolationinWisconsinresultedfroma higher proportion weresteeperinFinland.Ratherthaneliminate lakesone byone streamgradients to achievestatistical betweenregions,we concluded in an attempt homogeneity similarenoughtouse as wholeunits thatthetwosetsoflakeswerelimnologically in ourintercontinental comparisons. INTERCONTINENTAL FISH-ASSEMBLAGE 351 COMPARISON TABLE 1 FACTORS ENVIRONMENTAL FINLAND AND AND OF THE RICHNESS SPECIES WISCONSIN EXAMINED SMALL IN THIS X Environmentalfactors: pH Lake area (ha) Lake isolationa Conductivity(pLSat 200C) Maximumdepth (m) Species richness: Regional fishfauna (no. of species) Small-forest-lake fauna (no. of species) Single-lake species of species) richness (no. 6.2 9.5 49 28 9.6 37b 20 3.7 Range 4.3-7.5 .2-64.0 0-188 3-96 1.5-27.0 X 6.0 14.0 78 27 4.8 KRUSKAL- Range 4.3-8.0 .2-86.9 0-156 8-145 1.2-10.0 H (df = 1) WALLIS 1.68 .18 5.82* 10.2** 32.9*** 65c 23 1-10 4.4 N = 113; Wisconsin, N = 51. NOTE.-Finland, a See App. B for definition and formula. b Includes "freshwater" and "holoeuryhaline" species (Koli 1984). c Species and Vilas counties, from distribution in Forest, Iron, Oneida, * P < .05. ** P < .01. *** FROM WISCONSIN FINLAND VARIABLE LAKES FOREST STUDY 1-11 maps in Becker 2.57 1983. P < .001. Species Richness Wisconsincontains1.8timesas Ata regionallevel(gammarichness),northern fishesas does Finland(table1); theseregionalfaunas manyspeciesoffreshwater canbe treatedas "mainland"or sourcepoolsforsmalllake "islands." However, we foundnearlyas manyspecies in smallforestlakes of Finland(20) as in northernWisconsin (23; App. A). The 20 Finnishspecies representa significantly oftheregionalfaunathando the23 speciesinWisconsin(z-test, proportion higher lakes(alpharichness), no significant difference P < .01).Atthelevelofindividual in averagespecies richnessexistedbetweenFinlandand Wisconsin(table 1). forthetwo regionsalso did notdiffer (F = 0.9, df = Species-arearegressions 2,160, P = .4); the regressionequationforWisconsinwas logS = 0.30 + 0.32logA, R2 = 0.41; theequationforFinlandwas logS = 0.23 + 0.34logA,R2 = 0.40. ofspeciesaccumulation (fig.1),a setof51 small On thebasisofoursimulation a lakesin Finlandwouldhave an averageof 16.1 species,whichstillrepresents of the regionalfaunathanwhatwe foundin 51 higherproportion significantly accumulatedmorerapidlyin Wisconsinlakes (z-test,P < .01). Species initially a higherbeta richnessin Wisconsin.DeWisconsinthanin Finland,suggesting in cumulative richness,lake setsfromthetworegions spitethisinitialdifference fish convergewithregardto richnessoftheirsmall-forest-lake wouldeventually were22.9 speciesforFinnish faunas.ValuesofS. calculatedfromthesimulation lakes and 24.5 forWisconsin.Because alpha richnessdid not differbetween THE AMERICANNATURALIST 352 O 25 ui(20 000000000 C/) 1,5 u0 10 0 10ogo *g I WISCONSIN ui0 co z FINLAND 5. 0 20 60 40 100 80 120 NUMBER OF LAKES number ofspeciesina setofsmallforestlakesversusthe FIG. 1.-The total(cumulative) numberof lakesin theset,based on a random-lakes selection-simulation exercise(see the werefromthe studylakes in Finland(N = 113) and text).Data used in the simulation simulations Wisconsin(N = 51). As an exampleofthevariability amongindividual (N = limitsare givenforsetsof sevenlakes. 100),95% confidence 7 . 0 7 REGIONAL ENRICHMENT LOCAL SATURATION 6 ZI 5-...........-T < 24 co 3 ui 2-/ fl. 1-/ 0 10 20 30 40 50 60 70 REGIONAL SPECIES RICHNESS FIG. 2.-Relations betweenlocal and regionalspeciesrichnessforfishassemblagesof and Wisconsin The relation for smallforestlakesin Finland(opentriangle) (solidtriangle). was comparedwitha modeloflocal saturation Wisconsin (dashedline)andregionalenrichobservedinFinland.Valuesforlocalrichness ment(dottedline),on thebasisoftherelation intervals. are meansofindividual assemblages;verticallinesare 95% confidence regions,therelationof local to regionalrichnesswas nearlyasymptotic (fig.2), in local assemblagespredominated. thatecologicalsaturation suggesting Fish Distributions Occurrence sequences.-With theconsistent exceptionofcruciancarp,nearly orderedsequencesofoccurrence all speciesexaminedinFinlandhadsignificantly more forpH, area, isolation,and conductivity (table2). Moreover,significantly in thesame direction Finnishspeciesthanexpectedweredistributed alonggra- INTERCONTINENTAL FISH-ASSEMBLAGE COMPARISON 353 TABLE 2 OCCURRENCE SEQUENCES WITH FIVE ENVIRONMENTAL VARIABLES FOR FISHES IN SMALL LAKES OF FINLAND AND WISCONSIN SIGNIFICANT Z VALUES* TOTAL Z VALUES ENVIRONMENTAL VARIABLE Finland: pH Lake area Lake isolation Conductivity Maximumdepth Wisconsin: pH Lake area Lake isolation Conductivity Maximumdepth Positive Negative Positive 0 7 0 9 8 0 9 8 .040 .004 .004 .040 1 7 1 11 .006 0 7 0 0 1 4 0 2 7 0 7 1 5 0 4 2 1 9 0 1 2 8 4 7 Negative Pt .004 10 4 8 5 - .038 .388 .388 .744* 9; forWisconNOTE.-Species occurringin more thanfivelakes were analyzed (forFinland,N sin, N = 12). Standard normal variates (z values) were determinedfromMann-WhitneyU-tests. Positive z values indicate a tendencyfor species to be presentin lakes with smaller values of the environmental variables; negativez values show a trendtowardoccurrencein lakes withlargervalues. * P < .05. t Probabilitythat at least the observed numberof the total negative (or positive for isolation) z values occurredby chance, on the basis of the binomialdistribution. factors.Lakes inwhichtheseFinnishspecieswere dientsofallfiveenvironmental presentwereless acidic,larger,higherin conductivity, deeper,and less isolated thanlakeswithoutthesespecies. did notexistforWisA comparable,commonpatternof similardistributions of morethanhalfof thespecieswere consinspecies(table2); thedistributions orderedonlyalong the pH gradient.For lake area and isolation, significantly northern pike,black bullhead,whitesucker,and yellowperchhad orderedseinlarger,less isolatedlakes.Pike morefrequently quences,occurring significantly in found more often shallow were and bullhead lakes, whereasbluegilland in more bass occurred of largemouth frequently deeperlakes. Onlyforgradients more were Wisconsin than lake area distribpH and significantly species expected thesespecieshadhigher utedinthesamedirection (table2); lakescontaining pH's andlargerareas thanlakes lackingthem. Species co-occurrence.-Of 36 species pairsexaminedin Finland,23 (64%) showedpositiveco-occurrences,12 (33%) exhibiteda patternof neutralcooccurrence,and only 1 (cruciancarp and perch) displayednegativeco-occurrence.Each ofthe9 speciesexaminedinFinlandexhibited positiveco-occur2 most co-occurred withmore other least with at rencepatterns species; positively 66 made In of others. of the half than contrast, pairwisecomparisons among whereas50 Wisconsinspecies,only11 (17%) revealedpositiveco-occurrences, (76%) exhibitedneutralco-occurrence patternsand 5 (8%) displayednegative withpiinvolvedcentralmudminnows, co-occurrences 5 all patterns; negative 354 THE AMERICANNATURALIST bass andnorthern pikeandtheassociatedbluegill, pumpkinscivorouslargemouth differed significantly seed, and whitesucker.Overall,patternsofco-occurrence between regions(G = 22.1, df = 2, P < .001). Connorand Simberloff (1983,p. 461) pointedout thatgeneralproceduresfor failto accountforexpectedco-occurrences co-occurrence caused by examining For our Finnishdata set, such the "omnipresent"species-arearelationship. omnipresent patternsalso includedrelationsofrichnesswithpH, isolation,conrelaand depth(table2). To correctforsimilarspecies-environment ductivity, tions,we repeatedour co-occurrenceanalyses,except thatfor each pair of species,we includedonlylakes havingat least as manyspecies as the most depauperate assemblagein whichthetwospecieswerefound.In thisreanalysis, in Finlanddroppedto only9 (25%), inthenumberof positiveco-occurrences thecomofneutralpairsto26 (72%). Thesechangesreflected creasingthenumber monpatternof occurrencesequencesdescribedabove. Few changesoccurred in pairwisecomparisonsfor Wisconsin;3 additionalnegativeco-occurrences involvednorthern redbellydace withthepiscivorouslargemouth bass, northern remained pike,and yellowperch.Despite thesechanges,overalldistributions betweenregions(G = 6.4, df = 2, P < .05). different Patterns of Species Composition Wisconsin.-An earlierstudy(Tonnet al. [1983],whichused thedata set of Tonn and Magnuson[1982]and a portionof Rahel's [1984]data) dividedfish Wisconsinlakes into threeassemblagetypes assemblagesof small,northern Rahel(1984)subsequently subdivided (mudminnow, pike,bass; table3). Although themudminnow becauseof assemblagetypeintotwo(Umbra-Perca andcyprinid) theabsenceofmanycyprinid (minnow)speciesinboglakesoflowpH, thatfinerscale subdivisionwas not evidentin analysesof our largerdata set. Rahel's "centrarchid" assemblagesare equivalentto ourbass assemblagetype. classified84% ofthelakesin termsof Multiplediscriminant analysiscorrectly whichassemblagetypetheycontained(table4; see also table9). Combinations of environmental severity(low pH, low winteroxygen),bioticseverity(especially fromtheseconditions andthepresenceorabsenceofrefuges predation), appeared and occurrencepatternof assemto be largelyresponsibleforthecomposition blage types.Mudminnowassemblageswere foundin small,shallow,isolated had low levelsofdissolvedoxygenduringthewinter(Tonn lakesthatfrequently and Magnuson1982; Tonn and Paszkowski1986). These assemblageslacked specializedpiscivores(largemouth bass, northern pike)and weredominated by northern small,soft-rayed species (centralmudminnow, redbellydace, golden shiner,and othercyprinids), manyofwhichrarelyoccurredin theotherassemblagetypes.Bass assemblagesalso occurredin small,isolatedlakes,butonlyin hadhigherlevelsofwinter lakesthatweredeeperandtherefore dissolvedoxygen (Tonnand Magnuson1982;Rahel 1984).In additionto largemouth bass, these of large,spiny-rayed assemblageswerecomposedprimarily species.Pike lakes thanmudminnow werelargerand moreproductive or bass lakes,and although thewinter, theywereshallowandhadlow levelsofdissolvedoxygenduring they werewellconnectedto streamsthatcouldproviderefuges fromseasonallyharsh COMPARISON FISH-ASSEMBLAGE INTERCONTINENTAL 355 TABLE 3 RELATIVE FREQUENCIES OF OCCURRENCE OF FISH SPECIES, AND OTHER COMMUNITY THREE ASSEMBLAGE TYPES IN NORTHERN WISCONSIN LAKES CHARACTERISTICS, IN ASSEMBLAGE TYPE (N = 8) Bass (N = 23) Mudminnow SPECIES Yellow perch Centralmudminnow Largemouthbass Bluegill White sucker Black bullhead Pumpkinseed Golden shiner Northernpike Yellow bullhead Black crappie Northernredbellydace Brook stickleback Rock bass Iowa darter Blacknose shiner Finescale dace Fathead minnow Bluntnoseminnow Pearl dace Common shiner Redhorse Smallmouthbass Total no. of species Species richness: Mean no. of speciesa Range No. of unique speciesb .88 .12 .25 .38 .88 1.00 .75 .38 1.00 .25 .25 ... ... .50 .25 .12 ... ... .25 ... .12 .12 ... 17 .91 .43 .78 .52 .26 .09 .30 .17 .13 .30 .22 ... ... .04 ... ... ... ... ... ... ... ... .04 13 .60 .90 ... ... .05 .20 ... ... 12 7.5t 5-10 1 4.2t 2-7 0 3.3t 1-11 5 Pike (N = 20) ALL LAKES (N = 51) .78 .57 .39 .29 .27 .27 .25 .25 .22 .18 .14 .12 .10 .10 .10 .08 .06 .06 .04 .04 .02 .02 .02 ... .30 ... ... ... .30 .25 ... .15 .15 .15 .15 ... .10 ... 23 4.4 a different fromeach other(Tukey-Kramertest). Means withthe same symbolare not significantly Overall ANOVA, F = 11.3, df = 2,48, P < .001. b Species thatoccur in two or more lakes but in only one assemblage type. abioticconditions.Like bass assemblages,pikeassemblagesconsistedprimarily are of large,spiny-rayed species. These patternsand ecologicalrelationships discussedindetailelsewhere(TonnandMagnuson1982;Rahel1984;Magnusonet al., in press). Finland.-Ordinationof Finnishfishassemblagesbased on species' presence distinctassemblagetypes(Tonnet al., unpublished or absencedid notidentify data); rather,a hierarchicalcontinuumof species richnessbased on species we dividedthe additionwas suggested.To facilitate analysisof thiscontinuum, assemblagesintothreegroupsbased on species richness(table5): groupI included assemblageswith 1-3 species; group II assemblagescontained4-6 species; and groupIII lakes had 7-10 species. Examinationof these groups Thiscontinuum of species revealeda nestedpatternofassemblagecomposition. continuum (table6). Areawas themost composition paralleledan environmental TABLE 4 MEAN VALUES OF ENVIRONMENTAL FACTORS FOR THREE FISH ASSEMBLAGE TYPES IN NORTHERN WISCONSIN LAKES ASSEMBLAGE TYPE FACTOR Pike (N = 8) Bass (N = 23) Mudminnow (N = 20) UNIVARIATE F (df = 2,48) pH Lake area (ha) Lake isolationa Conductivity([iS at 20?C) Maximumdepth (m) 7.2 41.6 0.7 59.3 2.1 5.8 6.1 73.4 12.3 6.3 5.7 3.5 72.2 19.0 3.5 10.7*** 23.2*** 49.1*** 22.1*** 25.4*** beforeanalyses. Equations forthe two canonNOTE.-All factorsexcept pH were log-transformed ical variables (X, Y) froma discriminantanalysis are as follows:X = - 1.01 + 1.73log(isolation + 1) + O.241og(depth) - 1.04log(area + 1) - 0.621og(conductivity)- 0.03pH; Y = -3.27 0.621og(isolation + 1) + 4.881og(depth) + 1.851og(area + 1) - 2.861og(conductivity)+ 0.53pH. a See App. B forformulaand description. *** P < .001. TABLE 5 RELATIVE FREQUENCIES OF OCCURRENCE OF FISH SPECIES, AND OTHER COMMUNITY THREE SPECIES-RICHNESS GROUPS IN FINNISH LAKES GROUP SPECIES Core species: Eurasian perch Northernpike Roach Burbot Ruffe Secondary species: Crucian carp Bream Bleak Tench Satellite species: Minnow Rudd I 1-3 Species (N = 58) GROUP 4-6 Species (N = 41) GROUP III 7-10 Species (N = 14) ALL LAKES (N = 113) .86 .50 .14 .07 .14 1.00 .98 .80 .85 .56 1.00 .93 1.00 1.00 1.00 .92 .72 .48 .46 .39 .14 .07 .07 .12 .05 .57 .79 .50 .50 .17 .12 .11 .08 ... ... ... .03 ... Ninespine stickleback ... Miller's thumb Vendace ... ... .02 .07 ... ... .02 ... .02 .02 ... .14 Dace ... Smelt Ide White bream Zander Whitefish ... ... ... ... ... ... ... ... ... ... 7 1.9 0 15 4.8 1 Total no. of species Mean species richness No. of unique speciesa II CHARACTERISTICS, IN .02 ... .21 .14 .14 .14 .07 16 8.1 4 a Species thatoccur in two or more lakes but in only one species-richnessgroup. .03 .03 .01 .01 .01 .04 .03 .02 .02 .02 .01 20 3.7 FISH-ASSEMBLAGE INTERCONTINENTAL 357 COMPARISON TABLE 6 MEAN VALUES OF ENVIRONMENTAL FACTORS FOR THREE SPECIES-RICHNESS GROUP FACTOR pH Lake area (ha) Lake isolationa Conductivity([iS at 20?C) Maximumdepth (m) I 1-3 Species (N = 58) 5.9 3.1 45.5 19 7.6 GROUP II 4-6 Species (N = 41) 6.4 10.2 11.3 24 8.1 GROUPS IN FINNISH LAKES GROUP III 7-10 Species (N = 14) 6.6 13.8 7.9 40 10.2 F (df = 2,110) UNIVARIATE 18.3*** 29.7*** 19.2*** 8.8*** 1.3 beforeanalyses. NOTE.-All factorsexcept pH were log-transformed a See App. B forformulaand description. *** P < .001. the species-richness factordistinguishing groups,inenvironmental important groupIII creasingfromthedepauperategroupI lakes to themorespecies-rich variablesincludedlake isolation(greatestin groupI) and lakes. Otherimportant pH (highestin groupIII). Five corespecies(Hanski1982)appearedinall threerichnessgroupsandwere nearlyubiquitousingroupIII lakes(table5). Perchandpikewereusuallypresent in groupII lakes and in 63% of two-specieslakes; perchwerefoundin 68% of lakes. Secondaryand satellitespeciesweregenerally absentfrom single-species in progressively withgreaterfrequency richerasdepauperatelakes, occurring exceptiontothis semblages.Cruciancarp,a secondaryspecies,was an important in 21% of single-species lakes. occurring pattern, ourmultivariate comFor Finnishassemblages,we werealso able to perform munity analyseson a secondscale,usingbiomassrelativeabundance.Ratherthan each witha different threegroupswere distinguishable, a singlecontinuum, gavesimilarresults(Tonnet species(fig.3; table7); DCA ordination predominant data). The threeassemblagetypesconsistedof 58 assemblages al., unpublished dominated byroach(or,rarely,pike),and 8 byperch,47 assemblagesdominated speciesusuallycontribassemblagesdominatedby cruciancarp. The dominant species. As exutedat least twiceas muchbiomassas did the second-ranked pected fromthe presence-absenceanalyses,assemblagetypesdid not differ fewspecieswererestricted to one asemblagetype in speciescomposition; greatly (table7). characteristics of the lakes of each Finnishassemblagetype Environmental as indicatedby discriminant analysis(table8). Thisfivewererelatively distinct, on Wisconsinlakes, correctly variableanalysis,comparableto one performed classified68% of the lakes (table9). The indexof local isolationwas the most the threegroups;roachlakes were significantly factordistinguishing important variables less isolatedthaneitherperchor cruciancarplakes. Otherimportant includedmaximum depth(lowerin cruciancarplakes) and pH (higherin roach lakes). THE AMERICAN NATURALIST 358 6 /|PERCH Assemblages= O /ROACH Assemblages=0 CRUCIAN CARP Assemblages=A 0 0 P.42 '.4 0-2 -1 0 1 2 1st PRINCIPAL COMPONENT 3 FIG.3.-Ordinationoffishassemblagesin 113Finnishlakes,basedon therelativeabundanceofspecies.Threetypesofassemblagesaredistinguished, identifiable bythepredominantspeciesin theassemblages.The first threeprincipal components explained20%, 16%, and 13%ofthetotalvariance,respectively. IntercontinentalComparisons of Assemblage-Environment Relations Fromtheabove community analyses,itappearedthatfish-assemblage typesof in similarways alongenvironmental thetwo regionsweredistributed gradients oflakescontaining (cf.tables4 and 8). Characteristics perch,roach,and crucian carpassemblagesinFinlandparalleledthoseofbass,pike,andmudminnow lakes, in Wisconsin.Roach and pike lakes were generallylarger,less respectively, acidic,and less isolated,relativeto theotherlake types.Cruciancarpand mudminnowlakestendedto be small,shallow,and isolated,whereasperchand bass isolatedbutrelatively lakesweretypically deep. thisapparentpattern, we used discriminant To investigate functions fromone forthesecondregion(rather thanas descriptors regionas a prioripredictors ofthe originaldata set, as used above; see also Cody 1978;Tonn et al. 1983). We calculatedcanonicalvariatescoresforeach Finnishlake usingfunctions generated by the discriminant analysisof Wisconsinlakes (table 4) and calculated canonicalvariatescoresfortheWisconsinlakesusingdiscriminant functions from TABLE 7 MEAN RELATIVE ABUNDANCE OF FISH SPECIES, AND OTHER COMMUNITY CHARACTERISTICS, IN THREE ASSEMBLAGE TYPES IN FINNISH LAKES ASSEMBLAGE TYPE SPECIES Roach (N = 47) Eurasian perch Northernpike .23 .18 Roach .50 Ruffe Crucian carp Bream .02 Bleak Tench Perch (N = 58) Crucian Carp (N= 8) .86 .06 .06 .08 .53 .11 .ooa .20 .OOa .02 Burbot Total no. of species Species richness: Mean no. of speciesb Range No. of unique speciesc .02 ... .OOa .ooa .00a .85 .03 .00a ... .ooa .ooa ... .ooa .ooa 17 .04 16 2.7t 1-9 2 5.3t 1-10 1 ALL LAKES (N = 113) .02 .03 .05 .01 .ooa .ooa 5 2.Ot 1-5 0 20 3.7 NOTE.-Only species occurringin fiveor more lakes are listed. a Presentin at least one assemblage. b Means withthe same symbolare not significantly fromeach other(Tukey-Kramertest). different Overall ANOVA, F = 28.1, df = 2,110, P < .001. Species thatoccur in two or more lakes but in only one assemblage type. TABLE 8 MEAN VALUES OF ENVIRONMENTAL FACTORS FOR THREE ASSEMBLAGE TYPES IN FINNISH LAKES ASSEMBLAGE TYPE FACTOR Roach (N = 47) Perch (N = 58) pH Lake area (ha) Lake isolationa Conductivity([LS at 20?C) Maximumdepth (m) 9.6 8.7 30.2 8.1 6.5 5.9 6.1 40.5 19.4 8.7 Crucian Carp (N = 8) UNIVARIATE F (df = 2,110) 6.0 2.2 50.3 16.7 3.3 14.0*** 10.9*** 18.9*** 7.6*** 8.7*** beforeanalyses. Equations forthe two canonNOTE.-All factorsexcept pH were log-transformed ical variables(X, Y) froma discriminant analysis are as follows:X = 5.23 + 1.21log(isolation + 1) 0.16log(depth) - 0.31log(area + 1) - 0.241og(conductivity)- l.OOpH;Y = 1.59 + 0.52log(isolation + 1) + 2.731og(depth) + 1.341og(area + 1) + 1.04log(conductivity)- 0.67pH. a See App. B forformulaand description. *** P < .001. THE AMERICAN NATURALIST 360 TABLE 9 ANALYSES MATRICES RESULTING FROMDISCRIMINANT-FUNCTION CLASSIFICATION FISH ASSEMBLAGE TYPE LAKEENVIRONMENT TYPE Pike Bass N Pike: W F Bass: W F Mudminnow: W F % correct: W F Mudminnow 8 23 20 8 8 1 1 2 2 0 0 20 17 3 4 0 0 2 5 15 14 100 100 87 74 75 70 FISH ASSEMBLAGE TYPE LAKEENVIRONMENT Roach TYPE N Roach: W F Perch: W F Crucian carp: W F % correct: W F Perch Crucian Carp 47 58 8 34 35 12 13 1 1 9 9 33 36 1 1 4 3 13 9 6 6 72 74 57 62 75 75 NOTE.-Lakes of each fishassemblage type were classifiedinto groups accordingto discriminant functionsfromenvironmentalanalyses. Presentedare the numberof lakes of each assemblage type thatwere classifiedinto each lake environmenttype and the percentageof lakes thatwere correctly classified.W (Wisconsin) and F (Finland) are the sets of lakes fromwhichthe discriminantfunctions were derived. five-variable thecomparable analysisofthelakesinFinland(table8). Usingthese canonicalvariatescores,new discriminant analyseswere then between-region whether relations thatseparatedlakesof foreach regionto determine performed couldsuccessfully lakes different distinguish assemblagetypeson one continent containing analogousassemblagetypeson theothercontinent. analyses Discrimination amongassemblagetypesusingthesebetween-region was nearlyas good as in theoriginalwithin-region analyses(table9). Distributionsof Finnishlakes havingperch,roach,and cruciancarp assemblagesdiffromeach otherin discriminant feredsignificantly space usingthe two canonanalysisof variance, ical variablesderivedfromWisconsinlakes (multivariate MANOVA,P < .001).CanonicalvariablesderivedfromtheWisconsindiscrimiclassified73 of the 133Finnishlakes (65%), compared nantfunctions correctly derivedfromtheFinnishlakes themwith77 correct(68%) based on functions than40 selves.This was a significantly greaternumberofcorrectclassifications analysesin whicheach Finnishlake (36%), theaverageof 20 cross-continental thatthetotal was randomly assignedto an assemblagetype,withtherestriction oflakesin each typewas thesameas in theoriginalanalysis(t = 6.9, df number = 19, P < .001). ofWisconsinlakeswithbass,pike,andmudminnow assemblages Distributions fromeachotheralongcanonicalvariablesderivedfromFinnishlakes also differed classified39 of 51 (MANOVA, P < .001). These canonicalvariablescorrectly Wisconsinlakes (76%) yersus43 correctwiththe originalWisconsinvariables (84%); thiswas greaterthanthe22 correct(43%) produced,on theaverage,with INTERCONTINENTAL FISH-ASSEMBLAGE COMPARISON 361 assignedto thethreeassemblagetypes 20 analysesinwhichlakeswererandomly (t = 5.9, df = 19,P < .001). Analogous assemblage typesare indeed distributedin space in parallelwayson thetwocontinents. environmental DISCUSSION Our comparativeanalysesof fishassemblagesin a Nearcticand Palearctic of thatreflect jointbutvariableinfluences patterns regionrevealcommunity-level and are regional that and those and local contemporary are that phenomena historical.Althoughour assessmentof therelativerolesof thesetwogroupsof evidence,itis apparent andbynecessity,reliedon indirect processesfrequently, by phenomena primarily are influenced of fish assemblages features thatcertain other scale dominate the on alternative operating whereas processes on one scale, proand study of these differences be aware must ecologists features.Clearly, (Allen being investigated trait the to community scale on the appropriate cesses and Starr1982). Species Richness to differences in Historicalfactorsare generallyrecognizedas contributing manyfamiliesoffreshwater regionalspeciesrichness(Westoby1988).Although between as a resultof Cenozoicconnections fishesare Holarcticin distribution effectsof Pleistocene NorthAmerica,Europe,and Asia (Briggs1986),differing Europe than in glaciationresultedin a more depauperatefauna in northern similarregionsin centralNorthAmerica.Because of the lack of climatically occurredwithin (Banarescu1975),widespreadselectiveextinction nearbyrefuges Europeanfauna,withconditionsfavoringlarge,long-lived,and the northern Wisconspecies(Mahon1984;Moyleand Herbold1987).In contrast, migratory sin's fishfaunaenjoyeda large,nearbyrefuge(and a centerof speciation)in the MississippiRiverbasin (Burrand Page 1986),withconditionsfavoringsmall, narrowrangesof habitats(Mahon short-lived speciesor speciesusingrelatively 1984). European Mahon (1984) suggestedthatlow species richnessin the northern faunaand the absence of a numberof NorthAmericanfamilies(e.g., Centrarofbroad anddevelopment competition intraspecific chidae,Ictaluridae)promoted oftheEuropeanfaunacan be in Europe.A largeproportion habitat-use patterns foundin rivers,streams,large lakes, and smallforestlakes, whereasNorth in fewertypesof Americanspeciestendto be morespecialized,each occurring fishes habitat(MuusandDahlstr0m1978;Becker1983).Onlya thirdoffreshwater Wisconsinoccurinsmallforestlakes,andbetarichnessis significantly innorthern highertherethanin Finland.Greaterspecializationand higherspatialturnover species (Brownand Maurer1989), maybe a commonpatternforsmall-bodied partoftheWisconsin,butnotFinnish,fauna. whichmakeup a significant PostglacialFennoscandiahas been coveredby severalbodiesofbrackishand freshwater(Svardson1970,1976),in whichmostspeciesofthemodernregional faunaprobablylived (Nordqvist1903). Because smalllake basinshad already believedthat formed bythetimeofAncylusLake (8,500-8,000yrago),Nordqvist 362 THE AMERICANNATURALIST mostfishessimplyremainedin the smallerbasinsas waterlevels droppedand remainedsuitable.If so, a largesetofforestlakes persistedas longas conditions faunanotonlybecause of of theregionalfreshwater containsa highproportion from butbecause theforestlake faunais a remnant theirecologicalgenerality historicalextinctions mayhave been moreimportant AncylusLake. Therefore, ofindividthespeciesrichnessandcomposition indetermining thancolonizations Wisconsin(Attig ual lakes. The absence of largepro-glaciallakes in northern less isolatedregionally fromglacialrefuges, Wiscon1984)suggeststhatalthough thanwereFinnish of from local sources immigrants sinlakesweremoreisolated in gamma(regional)and beta (among-lake) richness, lakes. Despitedifferences fishassemblageswithinindividuallakes of thetworegionsrevealedconvergent and similarlevels of richness.Lakes of the two regionshad similarareas and bothofwhichcouldhave contributed, or indirectly, directly to waterchemistry, Greater and Keast and of local richness less levels depth 1984). similar (Eadie richnessinFinnishlakes,all else being isolationbothshouldhavefavoredgreater equal; the increasedbeta richnessin Wisconsin,suggestedby figure1, and an in theNorthAmericanfauna(Mahon increaseddegreeof habitatspecialization forthehigherregionalrichnessinWisconsin.The 1984)couldhavecompensated thatfoundin rangeoflocal speciesrichnessinourstudylakesalso approximated USSR and Jaroslavl lakes of sized (Zhakov 1974), provinces, Vologda similarly Sweden(Alm1960,1961),and Ontario(Eadie et al. 1986). relationof local to regionalrichness(fig.2) can be The nearlyasymptotic oflocal(vs. regional)processes oftherelativeimportance viewedas an indication local diversity relationalso has in determining (Ricklefs1987).The asymptotic and marinesystems(Cody 1966;Heck 1979;Bohnsack beenfoundin terrestrial and Faaborg1980),suggesting somegenerality ofthe andTalbot1980;Terborgh pattern. have stressedthatsaturation wouldoccurbecausecomStandardexplanations species(Cornell1985). petition imposesa fixedlimiton thenumberofcoexisting to increasedhabitatspecialization ofeastCompetition probablyhas contributed ernNorthAmericanfishesrelativeto thosein northern Europe(Mahon 1984). andhabitatseverity ofsmall However,theisolation,ecologicalimpoverishment, also have structured fish forestlakesandtheoccurrenceofsize-limited predation assemblagesin smallWisconsinlakes (Rahel 1984;Tonnand Paszkowski1986, than 1987;Magnusonet al., inpress).Thesefactorsareprobablymoreimportant the numberof species in individualWisconsinlakes to in limiting competition in Finland. levelssimilarto thoseoccurring relationbetweenlocal and regionalrichnessis notuniversal. An asymptotic between Lawton(1984)and Cornell(1985,1986)bothfoundstrongcorrelations systems,whichled themto conalpha and gammarichnessforhost-herbivore andthatlocalinteractions cludethattheseassemblageswerenotsaturated among Causes behindthedifferent patternsare notclear. specieswerenotimportant. on thetypeof relationobservedbetweenlocal and regional Potentialinfluences richnessinvolvethesizes oflocal assemblagesand regionalspeciespools in the Pool "exhaussystemsbeingcompared(J. Lawton,personalcommunication). INTERCONTINENTAL FISH-ASSEMBLAGE COMPARISON 363 tion' (Lawton and Strong1981,p. 326) can producean asymptotic relation; richnessin systemsreported to havelinearrelations alternatively, might level-off iflocalitieswithlargerregionalpoolsarediscovered.Finally,theresimplyappear "kinds" of communities forwhichdifferent to be different processes,operating on different scales, are important (Schoener1986a;Westoby1988).Currently, a generalexplanationaccountingfor variouslocal-regional relationscannotbe provided. Fish Distributionand Assemblage Composition A proximate factorinvolvedin thehigherbetarichnessofWisconsinwas the irregular patternsof occurrenceof Wisconsinspeciesalongenvironmental grain Finland most whereas species displayed similar, regular occurrence dients, patternsalong the same gradients.Resultsof co-occurrence analysisand the occurrencesequencesofindividual speciessuggestthatindividualistic responses ratherthanbioticinteractions to theenvironment, amongspecies,playedmajor rolesin creatingoverallpatternsof occurrenceand assemblagecomposition in Finlandat thelevel of species' presenceor absence.Strong,bioticinteractions favorable thatcan eliminate lakescouldwellhaveresulted speciesfromotherwise in haphazardoccurrencesequencesalong environmental gradients.Nearlyall orderedsequencesofoccurrence speciesexaminedinFinlandhadsignificantly for fouroffiveenvironmental low numbersof factors;combinedwithexceptionally thissuggeststhatnegativeinteractions negativeco-occurrences, werenotimporinsimilarways tant.Rather,itseemsthatspecieswereresponding independently to an overallenvironmental gradient.These patternsare consistentwiththe observation that,on the scale of species' presenceor absence,Finnishassemratherthandiscretegroups;thatis, assemblagesare blagespresenta continuum fashion(table5; see also Cornell1986;Schoener1986b). nestedin a hierarchical of species richnessand composition, Zhakov(1974) also noteda continuum withthesame fivecore speciesbeingubiquitousin richerassemblages;he also thisas a formof describeda regularpatternof speciesloss. Zhakovinterpreted intoothers," of certainichthyocoenoses succession,"regulartransformations inthenumberof andhe believedthat"thisprocessis accompaniedbyreduction ofspeciesto each speciesand a continuously increasing degreeofco-adaptation other"(1974,p. 213). can be viewedintermsoftheregionalandhistorthecontinuum Alternatively, outlinedbyNordqvist(1903),inwhichnearlyall speciespresent icalprocessfirst in thefaunagainedaccess to smalllake basinsat thetimeofAncylusLake. As individual conditions andlimnological climatic changedwithin basins,moresensior perished.In thisview,the currentorderedrichness tive species emigrated effectofprogresof speciestoleranceto thefiltering a gradient reflects gradient of small,isolatedforestlakes,comparedwith environments sivelydeteriorating to thisdeterioration in earlierlargelakes. Likelyfactorscontributing conditions thatdevelopas a includetheincreasedacidityand decreasedhabitatcomplexity floating bogmatencirclestheshoreline (Rahel1984)andtheincreasedprobability accumulatesand depthdecreases(Tonnand Magas organicmatter ofwinterkill 364 THE AMERICANNATURALIST inducedextinction, nuson1982).Thus,environmentally augmented by recolonizationinless isolatedlakes,is proposedas theprimary ecologicalprocessbehind in Finland. continuum theobservedpresence-absence In contrast,less regularand consistentpatternsof occurrencesequencesfor Wisconsinfishessuggestthatdistinct groupsof speciesare morespecializedfor and/orthatbiologicalinteractions withinor among specificlake environments in determining thesegroupsare important species' presenceor absence. The negativeco-occurrences, involving small-bodied, soft-rayed fishesand eitherpiscivoresorotherlarge-bodied, are spiny-rayed speciesassociatedwithpiscivores, withtheexistenceofdiscreteassemblagetypesin Wisconsinand also consistent of predationin determining suggestthe importance species composition (Tonn and Magnuson1982;Rahel 1984;Magnusonet al., in press).We believethata betweenWisconsinand Finlandincommunity-level majorreasonfordifferences patternson the scale of species' presenceor absence is the abilityof North Americanpiscivoresto excludemanysmallpreyspecies,forexample,central northern mudminnow, redbellydace, and others(see also Tonnand Paszkowski 1986;Robinsonand Tonn 1989).In contrast, predatorsand preycoexistin Finland,at leastin partbecause mostsoft-rayed speciesin northern Europe,includresemblewhitesuckersin Wisconsinin thattheycan growlarge ingcyprinids, enoughto acquirea size refugefrompredation(Tonnet al. 1989). in structuring can also be important Finnishassemblages, Bioticinteractions buton thescale ofrelativeabundanceratherthanpresenceor absence.In most lakes whereroachwas dominant, perchpopulationsweremuchreduced,often is theprimary mechanism making up less than10%oftotalbiomass.Competition dominanceby roachin moreproductive lakes (Sumari1971; behindthefrequent thanis perch;in addition, Persson1986).Roach is a moreefficient planktivore blue-green algae and detritusserve as exclusiveresourcesforroach (Persson 1983,1987). Both factorsgive roach a competitive advantagewherethesereabundant.In contrast, sourcesarerelatively perchlakesweremoreisolated,more thanwerelakes dominated acidic,smaller,and lowerin conductivity by roach. Manyperchlakes,in fact,containedonlyperchor perchand pike,withlow pH filter suchas roach(Overrein et apparently actingas an effective againstcyprinids characteristics ofparticular lakes maymodify or overal. 1981).Environmental and thushave broader,assemblage-wide ridebioticinteractions consequences (Perssonet al. 1988). Isolated,small,andshallowlakesdominated bycruciancarpwereenvironmenThistypeoflakefrequently severewinter tallymostdistinct. experiences hypoxia an unfavorable environment tomostfishes.However,crucian offers andtherefore well adaptedforsurviving winterhypoxia(Holopainen carpare physiologically and Hyvarinen1985;Hyvarinenet al. 1985;Piironenand Holopainen1986)and attainhighdensitiesinthesesevereanddepauperate lakes(Holopainen frequently andPitkanen1985).In benignlakeswithricherassemblages,onlysparsepopulationsoflargecruciancarpoccur(table7; Hamrin1979;Piironenand Holopainen 1988)as a resultof predationand/orinterspecific competition (Svardson1976; ofa body-sizerefugeto manyFinnishspecies, Tonnet al. 1989).The availability INTERCONTINENTAL FISH-ASSEMBLAGE COMPARISON 365 refuges (Holopainenet al. 1988),mayprevent withtheiruse ofstructural together piscivores. lakes from containing completeexclusionofpreyspecies Relations IntercontinentalComparisons of Assemblage-Environment factorsand betweenlocalenvironmental andsimilarrelations Becauseofstrong to makeand we were able both in regions, of local assemblages thecomposition ability, fish of This type. assemblage intercontinental predictions test successfully usingassemblagetypesof largelyunrelatedspecies,is ecologicallysignificant relations thatsomeoftheseintercontinental (Oriansand Paine 1983),suggesting similaroperationsoflocal bioticor abioticprocesses. reflect of winterhypoxiaand lake isolationpreventsmostspecies The combination and cruciancarp populationsin mudminnow and maintaining fromestablishing bothcentralmudmindifferences, and morphological lakes.Despitephylogenetic havedevelopedwaysofsurviving nows(Umbridae)andcruciancarp(Cyprinidae) (Gee 1981;Magnusonet al. 1983;Holopainen severity thistypeofenvironmental tobioticinterac1985).However,bothspeciesalso arevulnerable andHyvarinen and enjoytheir predation(moreso formudminnows) tionssuch as size-limited lakes (Tonn 1985;Tonnand Paszkowski1986; greatestsuccessin single-species Paszkowskiet al. 1989). speciesand organizing Somewhatless similarecologicallyare thecomponent processesof bass (Wisconsin)and perch(Finland)assemblages.Abioticfactors combinedwith diversity, and structural low levels of pH, nutrients, (primarily assemblagesfor highisolation)exclude severalspecies, creatingfundamental intorealizedassemblagesby thatcan thenbe shapedfurther theseenvironments in perchlakes is especiallyeffective filter The environmental bioticinteractions. and often ofcyprinids theestablishment (meanrichness,2.7 species),restricting perchand, tolerantand ecologicallygeneralized leavingonlythephysiologically theassemblages(RaskandArvola1985).Low toa lesserextent,piketo dominate frommanybass lakes(Rahel1984),butevenwherethe cyprinids pH also restricts bass can preventthe is morebenign,predationby largemouth environment et al. 1987).Othercore oftheseandothersmallspecies(Carpenter establishment tolerspeciesofbass assemblages,yellowperchand bluegill,are physiologically thatreducevulnerability andhavemorphologies generalized, ant,areecologically to predationby piscivores(Wahl and Stein 1988; Robinson1989). Withthe exceptionof perch,the regionalabsence of species havingall threeof these to theloweraveragerichnessofperch in Finlandmaycontribute characteristics withbass assemblages. assemblagesin comparison ecologicallyare pike(Wisconsin)and roach(Finland)assemMost dissimilar blages.Lakes ofthesetwotypesare thelargest,leastisolated,leastacidic,and shouldbe filters ofthestudylakes;theirenvironmental mostproductive probably that is it and unlikely manyspecies moreporousthanthoseofotherlake types, are also unableto excludeor are excludedat thislevel.In roachlakes,predators of evenregulatemanypreypopulations(Perssonet al. 1988);a largeproportion dominant of the competitively thefishbiomassintheseassemblagesis composed dominanceby roach.It is not clear,however,to whatextentthe competitive 366 THE AMERICANNATURALIST predation is mostlikely roachresultsintheexclusionofotherspecies.in contrast, fortheabsenceof mostminnowsand othersmall-bodied speciesin responsible local pikeassemblagesin Wisconsin(Tonnand Magnuson1982).Thus,although assemblage are similarinbothregionsand associatedwithdistinct environments and influence types,dominantspecies in each regionare ecologicallydifferent forthepresassemblagesvia different processes.Historicalfactorsresponsible in Finlandand small-bodied speciesin Wisconsin ence oflarge-bodied cyprinids observedbetweenanalogousasin the ecologicaldifferences appearimportant semblagetypes. CONCLUSIONS Our comparative,multiple-scale approachallowedus to examineimportant amenableto investigation methods.By by manipulative questionsnottypically intercontinental employing comparisonsof fishassemblagesfromhistorically similarlocal environments, in environmentally faunasoccurring we were distinct andhistorical and also able to expandthegeographical scopeofourinvestigation assess the relativeroles of local and regionalprocesses(Ricklefs1987).From multivariate analyses,we wereable to extractoverallstructural complementary data sets; by relatingthesepatternsto our mulpatternsfromthe community tivariateanalysesof the lakes' environmental features,we were thenable to and initiatetheecologicalinterpretation ofthecommunitygeneratehypotheses A seriesofanalysesoflower-level phenomena-forexample,the levelpatterns. of species pairs,and the of individualspecies, the co-occurrence distribution accumulationof species among individuallakes-aided our interpretation, and identified ourhypotheses, to commuprocessesthatcontribute strengthened associatedwith hypotheses, nity-level patterns.Some of theseprocess-oriented are amenableto direct,experimental testing(Tonnand lower-level phenomena, Paszkowski1986;Tonn et al. 1989).However,we stressthat,althoughmany in termsof the population-and community-level patternsmay be interpreted individual-level processesof autecology,it is unlikelythatautecologicalstudies Andwithoutstudiesoflarge-scalephenomwillactuallydiscoversuchpatterns. fromeven a largenumberof to drawusefulgeneralizations ena, it is difficult microecological manipulations (Brownand Maurer1989).A balanced,complementary approachis calledfor. An outcomeof thisapproachis our viewthatlocal fishassemblagesin small thatoperate,alongwithuniqueor forestlakes are productsof a seriesoffilters unpredictable events,on several spatialand temporalscales (fig.4; see also Simpson1953;SmithandPowell1971;Holmes1986;Tonn1990).Higherlevelsof withonlya portionoftheproperties act as filters, passing biologicalorganization lowerlevels.Although theselarger-scale and influencing filters directly through at the adjacentlowerlevel,theyalso can influence affectcharacteristics levels whichspecies are availableto entersubsequent down by determining farther filters. Local assemblagesare immediateproductsof local and contemporary proarestrong, ofregionalandhistorical filters thentheimpactof cesses,butifeffects FISH-ASSEMBLAGE INTERCONTINENTAL Regional Regional Processes \ / \ Lake-Type Local Processes Local Community Structure ,northernWisconsin lake district * Abiotic conditions Characteristics Small-Lake Species Pool 367 * Pleistoceneevents * Dispersal barriers * Climaticdifferences Geomorphic/edaphic limits " eg Regional Species Pool COMPARISON . Resource distribution * Habitat stability & complexity CA e.g., small lakes in northernWisconsin 0 Area * * * * Structural complexity Isolation Abioticconditions Biotic interactions e.g., Jude Lake FIG. 4.-Conceptual modelof theoriginand maintenance of fishassemblagesin small theeffects forestlakes,illustrating offilters on faunalcharacteristics andcommuoperating on different scales.Modified fromTonn1990. nitystructure spatialand temporal factorson aspectsof individual theselarger-scale assemblagescan be significant structure is (Holmes 1986;Ricklefs1987).Whethera traitlikelocal community in two regionsdepends notonlyon the matchoflocal phenomsimilaror different ena butalso on therelativeeffects(e.g., "pore size" or "selectivity")oflargerSimilarities and differences betweenfishassemblagesin smallforest scale filters. and differences of local and lakes of Finlandand Wisconsinreflectsimilarities 368 THE AMERICANNATURALIST thatis, thelakesin whichtheassemblagesoccurand thefaunas regionalfilters, fromwhichtheyare derived. SUMMARY of fishassemblagesfromFinlandand northern We comparedcharacteristics in environsimilarity to scales community-level investigate Wisconsinon several has only Finland lakes. Although small different but faunistically similar mentally differ not did richness local species pool, its in species fishes regional as many half assemblages among in composition species Variability two regions. the between of the faunacould thata largeproportion was lower in Finland,suggesting whereas conditions, maintain populationsacrossa broadrangeofenvironmental theWisconsinfaunawas composedof morespecializedspecies.Threetypesof in Wisconsinbased on species' presenceor fishassemblageswere identified a hierarchical absence,whereasonthesamescale,Finnishassemblagespresented based on species addition.These and otherpatternssuggestthat continuum habitatspecializationand the exclusionof smallpreyspecies fromlakes with assemblagetypesin to theoccurrenceofpresence-absence piscivorescontribute can affectpopulationdensitiesof Finnishspecies Wisconsin.Bioticinteractions inFinlandonlyon the exclusion,andassemblagetypescouldbe identified without by perch, basis of species' relativeabundance.Finnishassemblagesdominated similarto roach,and cruciancarp occurredin lakes thatwereenvironmentally assemblages,respectively. lakesin Wisconsinwithbass, pike,and mudminnow fromone regioncould functions Because ofthesesimilarrelations,discriminant theassemblagetypesthatoccurredin lakesfromtheother identify successfully suggeststhatfishassemblagestrucpredictability region;thiscommunity-level by commonsets of environmental forestlakes are influenced turesin northern thebalancebetween scalesreflect on different anddifferences factors.Similarities are well matched local and regionalprocesses; althoughlocal characteristics Because aredissimilar. phenomena betweenregions,manyregionalandhistorical community on severalscales,viewedas a seriesoffilters, operating ofinfluences ecologistsshouldnot be contentwithsimpleacceptanceor rejectionof the willcomefrom of communities Greaterunderstanding hypothesis. convergence ofcauses and effects. studiesthatencompassa hierarchy comprehensive ACKNOWLEDGMENTS theFinnish whoassistedus in compiling We wishto thankthehydrobiologists data:J.Aho, allowedus to use theirownunpublished datasetand/orgenerously and P. Tuunainen.S. Magnusonand A. McLain I. Holopainen,M. Pursiainen, the map workrequiredforcalculationof the isolationindexes;L. performed LeClairdraftedthefigures.W.M.T. and J.J.M.wouldliketo acknowledgethe of our hosts at the Universitiesof Helsinki,Joensuu,Jyvaskyla, hospitality Kuopio,and Turkuand especiallyof J. Syijamaki,I. Hakala, and theirstaffat contributed LammiBiologicalStation.DiscussionswithmanyFinnishscientists of smallforestlakes; special of theecologyand limnology to ourunderstanding INTERCONTINENTAL FISH-ASSEMBLAGE COMPARISON 369 thepapers to ourattention thanksgo to 0. Lindqvistandto P. Baggeforbringing criticisms by J. Holmes,J. Lawton,C. ofNordqvistand Zhakov.Constructive Paszkowski,L. Persson,F. Rahel,C. Robinson,R. Stein,and an anonymous This projecthas been reviewerimprovedearlierversionsof the manuscript. generouslysupportedby grantsfromthe NationalScience Foundation(INTResearchCouncilof Canada 8312862),the NaturalSciences and Engineering ofAlberta'sCentralResearchFund. (A2363and IC-0269),and theUniversity ; ! s.: r 2 v ? . EX = & E= E r X y s X Q E~~~~~~~~t ;~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~t - o r z : u ; N< > *gu wg < >; H; f 4 3; U r 2 ,z < _ m; G~; c; * CL- z d u u t s , INTERCONTINENTAL FISH-ASSEMBLAGE COMPARISON 371 APPENDIX B DETERMINATION OF THE INDEX OF ISOLATION ofeach studylake,one To calculatethisindex,usedinassessingthedegreeofinsularity was used: oftwoformulas fordrainagelakes:I = G, for seepage lakes: I = G + 3A + 90, (m km 1) alonga watercourse whereI is theindexofisolation,G is theverticalgradient distancefrom andA is thetotalaltitudinal fromthestudylaketothenextlakedownstream, A is direction. thestudylaketothenearestbodyofwater(lenticorlotic)inthedownstream in altitudebetweenthestudylake and thedownslopewater nearlyalwaysthedifference wereaddedto irregularities body;however,in a fewcomplexlandscapes,topographical theverticaldistances. two lakes is a simplebut The verticalgradientor slope of the streamconnecting thedifference ofisolationforan upstream lake,measuring ecologicalyardstick appropriate fish,thuscreatinga largeror in habitatthata streamwouldpresentto a lake-dwelling For studylakeslackingoutletstreamsdirectly connecting smallerbarrierto colonization. sources,totalverticaldistanceto thenearestdownslopewaterbody themto downstream moreimportant thanslopewhenconsidering pastchangesin shouldbe biogeographically waterlevelsand/orpostglacialisostaticreboundofland(Nordqvist1903;Svardson1970). relativetoanyportion thatwas Theterm3A + 90 was usedto weightthisoverlandportion To putA, whichrangesfrom0 to 30,onthesamescaleas G (0-90),A alonga watercourse. by3. We added90tothe3Atermbecausewe believedthattheleastisolated was multiplied seepagelake (A = 0) is at least as ecologicallyisolatedforfishesas themostisolated forWisconsin drainagelake (G = 90). The scalingfactor3A + 90 was also appropriate seepagelakes since,forall drainagelakes,G < 90. LITERATURE CITED University forecologicalcomplexity. perspectives Allen,T. F. H., andT. B. Starr.1982.Hierarchy: ofChicagoPress,Chicago. Res. Inst.Freshwater indenKalarne-Seen. Untersuchung Alm,G. 1960.Limnologisch-fischereiliche Rep. 41. CarlBloms,Lund. Res. Rep. 42. Inst.Freshwater derFischaussatzeindenKalarne-Seen. 1961.Die Ergebnisse CarlBloms,Lund. I. koskevakoe-jatutkimustoiminta. parantamista 1961.Sisavesientuottoa ja laadullista Anonymous. of ResearchInstitute saaliitvuonna1961[Fisheries koekalastuksen Koejarvienyleisluonneja ofAgriculture] Helsinki. theMinistry II. koskevakoe-jatutkimustoiminta. 1963. Sisavesientuottoaja laadullistaparantamista tutkimustoikalataloudellinen Kalansaaliskoejarvistavuonna1962.Maataloushallituksen of Agriculture, of theMinistry julkaisuja[FisheriesResearchInstitute misto,monistettuja Tech. Rep., Helsinki,Finland]25:1-20. of Attig,J. W. 1984.The Pleistocenegeologyof Vilas County,Wisconsin.Ph.D. diss. University Madison. Wisconsin, in ordination: with experiments Austin,M. P., and I. Noy-Meir.1971.The problemofnon-linearity models.J.Ecol. 59:763-773. two-gradient NOLIT, Belgrade. ofzoogeography. andproblems Banarescu,P. 1975.Principles in lakes.Am.Nat. 108:473-489. Barbour,C. D., andJ.H. Brown.1974.Fishspeciesdiversity Press,Madison. ofWisconsin Becker,G. C. 1983.FishesofWisconsin.University andCaribbean byreeffisheson Australian Bohnsack,J.A., andF. H. Talbot.1980.Species-packing approach.Bull.Mar. Sci. 30:710-723. reefs:an experimental Bowers,M. A., and J. H. Brown.1982.Body size and coexistencein desertrodents:chanceor structure? Ecology63:391-400. community 372 THE AMERICAN NATURALIST to thezoogeography ofNorthAmerican fishes.Pages 1-16inC. H. Briggs,J.C. 1986.Introduction ofNorthAmerican freshwater fishes.Wiley, HocuttandE. 0. Wiley,eds. Thezoogeography New York. thedivision offoodandspaceamongspecieson Brown,J.H., andB. A. Maurer.1989.Macroecology: continents. Science(Washington, D.C.) 243:1145-1150. ofthelowerOhio-upper Mississippi basin.Pages287Burr,B. M., andL. M. Page. 1986.Zoography ofNorthAmerican freshwater 324inC. H. HocuttandE. 0. Wiley,eds. Thezoogeography fishes.Wiley,New York. S. R., J.F. Kitchell, J.R. Hodgson,P. A. Cochran,J.J.Elser,M. M. Elser,D. M. Lodge, Carpenter, D. Kretchmer, X. He, andC. N. vonEnde. 1987.Regulation oflakeprimary productivity by foodweb structure. Ecology68:1863-1876. ofintraandinter-continental birdspeciescounts.Am. grassland Cody,M. L. 1966.The consistency Nat. 100:371-376. andthestructure ofbirdcommunities. Princeton University Press,Prince1974.Competition ton,N.J. territoriality amongthesylviidwarblers ofEngland 1978.Habitatselectionand interspecific and Sweden.Ecol. Monogr.48:351-396. 1983.Interspecific andspeciesco-occurrence competition patterns Connor,E. F., andD. Simberloff. on islands:nullmodelsand theevaluationofevidence.Oikos41:455-465. Am.Nat. 126:565gallwaspsarenotsaturated. Cornell,H. V. 1985.Speciesassemblagesofcynipid 569. andhostsize influence cynipine waspspeciesrichness.Ecology 1986.Oak speciesattributes 67:1582-1592. fieldexperiments, and naturalexperiments. Diamond,J. 1986.Overview:laboratory experiments, ecology.Harper& Row,New Pages 3-22 in J. Diamondand T. J. Case, eds. Community York. andfishspeciesdiversity inlakes.Can. J. Eadie,J.M., andA. Keast. 1984.Resourceheterogeneity Zool. 62:1689-1695. andK. L. Teather.1986.Lakes andriversas islands: Eadie,J.M., T. A. Hurly,R. D. Montgomerie, in thefishfaunasofOntario.Environ.Biol. Fishes15:81-89. species-arearelationships and P. Sampson.1983.Factoranalysis.Pages480-499in W. J. Dixon,ed. Frane,J.,R. Jennrich, ofCalifornia BMDP biomedicalcomputer University Press,Berkeley. programs. oflizardcommunities inChileandCalifornia. Ecology Fuentes,E. R. 1976.Ecologicalconvergence 57:3-17. ecology.Cambridge University Press, Gauch,H. G., Jr.1982.Multivariate analysisin community Cambridge. of respiratory and hydrostatic functions of the swimbladder in the Gee, J. H. 1981.Coordination Umbralimi.J. Exp. Biol. 92:37-52. centralmudminnow, lakeinsouthern S. 1979.Biomassoftench,perch,cruciancarpandpikeina small,eutrophic Hamrin, 19:57-67. Sweden.Jyvaskylan YliopistonBiologianLaitoksenTiedonantoja Oikos thecoreand satellitespecieshypothesis. Hanski,I. 1982.Dynamicsofregionaldistribution: 38:210-221. ofthecomposition andabundanceofmotilemacroinvertebrate Heck,K. L. 1979.Somedeterminants turtle meadows.J.Biogeogr. speciesintropicalandtemperate grass(Thalassiatestudinum) 6:183-197. to multivariate analysisin biology.Bull. Soc. R. Bot. Belg. 121: Hermy,M. 1988.An introduction 128-133. ordination analysis,an improved Hill,M. O., and H. G. Gauch. 1980.Detrendedcorrespondence technique.Vegetatio42:47-58. of helminth communities. Pages 203-208in M. J. Howell,ed. Holmes,J. C. 1986.The structure of theSixthInternational Congressof Parasitology. Parasitology-quovadit?Proceedings Australian Academyof Science,Canberra. of cruciancarp [Carassius Holopainen,I. J., and H. Hyvarinen.1985.Ecologyand physiology in winter.Int. Ver. Theor. carassius(L.)] in smallFinnishpondswithanoxicconditions Angew.Limnol.Verh.22:2566-2570. INTERCONTINENTAL FISH-ASSEMBLAGE COMPARISON 373 ofcruciancarp[CarayHolopainen,I. J.,and A. K. Pitkanen.1985.Populationsize and structure sius carassius (L.)] in two small,naturalpondsin easternFinland.Ann.Zool. Fenn. 22: 397-406. andA. K. Pitkanen. 1988.Habitatuse,dielactivity, I. J.,W. M. Tonn,C. A. Paszkowski, Holopainen, pond.Int.Ver.Theor.Angew.Limnol.Verh. ofcruciancarpin a manipulated andgrowth 23:1743-1750. Pages95-111in L. A. NielsenandD. L. Johnson, Hubert,W. A. 1983.Passivecapturetechniques. eds. Fisheriestechniques.AmericanFisheriesSociety,Bethesda,Md. Vol. 1. Wiley,New York. G. E. 1957.A treatiseon limnology. Hutchinson, ofcruciancarp[CarasH., I. J.Holopainen,andJ.Piironen.1985.Anaerobicwintering Hyvarinen, of glycogenreservesin nature.Comp.Biochem. silus carassiius (L.)]. I. Annualdynamics Physiol.A, Comp.Physiol.82:797-803. analysis.Pages519-537in W. J.Dixon, R., and P. Sampson.1983.Stepwisediscriminant Jennrich, ofCalifornia Press,Berkeley. University programs. ed. BMDP biomedicalcomputer Johnson,A. W. 1973. Historicalview of the conceptof ecosystemconvergence.Pages 3-7 in New York. Springer, typeecosystems. F. di CostriandH. A. Mooney,eds. Mediterranean Duxbury,NorthScituate,Mass. statistics. R. R. 1973.Elementary Johnson, ofthelakewatersofnortheastern Juday,C., E. A. Birge,andV. W. Meloche.1938.Mineralcontent Wisconsin.Trans.Wis. Acad. Sci. ArtsLett.31:223-276. intheapplication ofclassification andordination Kent,M., andJ.Ballard.1988.Trendsandproblems methodsin plantecology.Vegetatio78:109-124. Koli,L., ed. 1984.Suomenelaimet3. Weilin& G6os, Helsinki. andvacantniches: communities, non-convergent populations, Lawton,J.H. 1984.Non-competitive L. G. Abele,and ofbracken.Pages 67-100in D. R. Strong,D. Simberloff, theherbivores conceptualissuesand theevidence.Princeton A. B. Thistle,eds. Ecologicalcommunities: N.J. Press,Princeton, University and competition in folivorous insects. patterns Lawton,J. H., and D. R. Strong.1981.Community Am.Nat. 118:317-338. Legendre,L., and P. Legendre.1983.Numericalecology.Elsevier,Amsterdam. R. H. 1972.Geographical ecology.Harper& Row,New York. MacArthur, ofislandbiogeography. Princeton University R. H., andE. 0. Wilson.1967.The theory MacArthur, N.J. Press,Princeton, Magnuson,J. J. 1976.Managingwithexotics-a gameof chance.Trans.Am. Fish. Soc. 105:1-9. different J.J.,J.W. Keller,A. L. Beckel,andG. W. Gallepp.1983.Breathing gasmixtures Magnuson, forsurvivalundertheice ofa facultative fish.Science air-breathing fromair:an adaptation D.C.) 220:312-314. (Washington, Magnuson,J.J.,C. A. Paszkowski,F. J.Rahel,and W. M. Tonn.In press.Fishecologyin severe Wisconsin.In J. W. Gibbonsand R. R. of smallisolatedlakes in northern environments fromnatural,managed,and wetlandsand wildlife: perspectives Scharitz,eds. Freshwater Service, degradedecosystems.Dep. EnergySymp.Ser. NationalTechnicalInformation D.C. Washington, in fishtaxocenesof northtemperate streams.Can. J. Fish. Mahon,R. 1984.Divergentstructure Aquat.Sci. 41:330-350. reviewoftheclimateofFinland.Vol. 78. No. 13. ofFinland.1984.Monthly Institute Meteorological Helsinki. instreamfishesof andcommunity structure patterns Moyle,P. B., andB. Herbold.1987.Life-history witheasternNorthAmericaandEurope.Pages25-32 NorthAmerica:comparisons western and evolutionary ecologyof North in W. J. Matthewsand D. C. Heins,eds. Community ofOklahomaPress,Norman. Americanstreamfishes.University fishesofBritainand Europe.Collins,London. Muus,B. J.,and P. Dahlstr0m.1978.Freshwater offreshwater fishin Finland. 0. 1903.Somebiologicalreasonsforthepresentdistribution Nordqvist, Fennia20(8):1-29. Science(Washington, D.C.) 235:226-227. Orians,G. H. 1987.Ecologicalcomparisons. at thecommunity level.Pages431-458in evolution Orians,G. H., andR. T. Paine.1983.Convergent Mass. D. J. Futuymaand M. Slatkin,eds. Coevolution.Sinauer,Sunderland, 374 THE AMERICAN NATURALIST recentadvancesin numericalmethods.Can. J. Bot. organization: Orl6ci,L. 1988. Community 66:2626-2633. ofacidwateronaquaticlife.Pages 143-161 L. N., H. M. Seip,andA. Tollan.1981.Effects Overrein, on forestand fish.Final rep. 2d ed. SNSF (Sur Nedb0rs in Acid precipitation-effects virkining p?aSkogog Fisk)Project,Oslo. C. A., W. M. Tonn,andI. J.Holopainen.1989.Anexperimental studyofbodysize and Paszkowski, in cruciancarp,Carassius carassius. Environ.Biol. Fishes24:275-286. foodsize relations ofa on dynamics and size structure competition Persson,L. 1983.Effectsofintra-andinterspecific Oikos41:126-132. perchPerca fluviatilisand a roachRutilus rutiluspopulation. on resourceutilization in perch(Perca competition 1986.Effectsof reducedinterspecific fluviatilis).Ecology67:355-364. interactions ofhabitatandseasononcompetitive betweenroach(Rutilusrutilus) 1987.Effects andperch(Perca fluviatilis).Oecologia(Berl.)73:170-177. 1988.Predator andprimary S. F. Hamrin,andL. Johansson. regulation Persson,L., G. Andersson, of temperate lake ecosystems.Pages 45-65 in alongtheproductivity gradient production in lakecommunities. ed. Complexinteractions New York. S. R. Carpenter, Springer, Annu.Rev. Ecol. Syst.4:53-74. oflizardcommunities. Pianka,E. R. 1973.The structure ofecologicaldata.Wiley,New York. Pielou,E. C. 1984.The interpretation in anoxiatoleranceof cruciancarp Piironen, J.,and I. J. Holopainen.1986.A noteon seasonality Ann.Zool. Fenn.23:335-338. [Carassius carassius (L.)] in thelaboratory. ofcruciancarp(Carassius carassius (L.)) andreproductive potential 1988.Lengthstructure in somesmallforestponds.Ann.Zool. Fenn.25:203-208. populations fishassemblages gradient. Ecology alonga boglakesuccessional Rahel,F. J.1984.Factorsstructuring 65:1276-1289. influences on fishspeciescomposition ofnorthern Wisconsin lakeswith 1986.Biogeographic forlakeacidification studies.Can. J.Fish.Aquat.Sci. 43:124-134. applications in growthof perch(Perca fluviatilisL.) in two smallforestlakes. Rask, M. 1983. Differences 101:139-144. Hydrobiologia oflowpH on perch,PercafluviatilisL. III. Theperchpopulation 1984.The effect ina small, humicforestlake. Ann.Zool. Fenn.21:15-22. acidic,extremely ofpike,perchandwhitefish intwosmall Rask,M., andL. Arvola.1985.Thebiomassandproduction Finland.Ann.Zool. Fenn.22:129-136. lakesin southern 1985.Rotenonkasittleyn vaikutuksista pienen Rask,M., L. Arvola,K. Salonen,and M. Pursiainen. 92:12-15. eliostoon.Suom.Kalastuslehti metsajarven and interpretation ofbiologicalstatistics offishpopulations. Bull. Ricker,W. E. 1975.Computation Fish.Res. BoardCan. 191:1-382. relativerolesof local and regionalprocesses.Science diversity: Ricklefs,R. E. 1987.Community D.C.) 235:167-171. (Washington, approachto thestudyof aviancommunity Ricklefs,R. E., and J. Travis. 1980.A morphological Auk97:321-338. organization. survivaloffourpreyin thepresenceofnorthern pike.Can. J. Robinson,C. L. K. 1989.Laboratory Zool. 67:418-420. ofenvironmental factorsandpiscivory in Robinson,C. L. K., and W. M. Tonn.1989.The influence fishassemblagesofsmallAlbertalakes.Can. J.Fish.Aquat.Sci. 46:81-89. structuring andconvergence offinch communities. D. 1986.Testsforsimilarity Ecology67:1073-1085. Schluter, becomespluralistic. Schoener,T. W. 1986a.Overview:kindsof ecologicalcommunities-ecology ecology.Harper& Row,New Pages467-479in J.DiamondandT. J.Case, eds. Community York. versusarthropod do systematic communities: differences in 1986b.Patternsin terrestrial exist?Pages 556-586in J. Diamondand T. J. Case, eds. Community ecology. regularity Harper& Row,New York. ofvertebrates on someverysmallislands.I. Schoener,T. W., and A. Schoener.1983.Distribution Occurrence sequencesofindividual species.J.Anim.Ecol. 52:209-235. CondonLectures,OregonStateSystemof Higher Simpson,G. G. 1953.Evolutionand geography. Education,Eugene,Oreg. INTERCONTINENTAL FISH-ASSEMBLAGE COMPARISON 375 ofBrierCreek,MarshallCounty, fishcommunities Smith,C. L., andC. R. Powell.1971.Thesummer Oklahoma.Am.Mus. Novit.2458:1-30. of somepondsin Finland.Ann.Zool. Fenn. of theperchpopulations Sumari,0. 1971.Structure 8:406-421. incoregonid ofintrogression evolution. Pages33-59inC. C. Lindsey Svardson,G. 1970.Significance fishes.University ofManitobaPress,Winnipeg. andC. S. Woods,eds. Biologyofcoregonid of Scandinavian lakes. Inst. populationdominancein fishcommunities 1976.Interspecific Res. Drottningholm Rep. 55:144-171. Freshwater ofbirdcommunities in theWestIndies.Am.Nat. J.W., andJ.Faaborg.1980.Saturation Terborgh, 116:178-195. oflargeoligotrophic glaciallakesin Europe(about Toivonen,J. 1972.The fishfaunaandlimnology 1800A.D.). J.Fish.Res. BoardCan. 29:629-637. andL. Peippo.1964.Rotenonmyrkytysten avullasaatujatietojaeraiden Toivonen,J.,P. Tuunainen, Suom.Kalastuslehti 71:156-164. tekijoista. ja niihinvaikuttavista lampienkalakannoista offishcommunities innorthern in theassemblyanddiversity Wisconsin Tonn,W. M. 1980.Patterns ofWisconsin, Madison. lakes.Master'sthesis.University in Umbra-Perca fishassemblagesofnorthern Wisconsin lakes. 1985.Densitycompensation Ecology66:415-429. a conceptualapproach.Trans.Am. Fish. Soc. 1990.Climatechangeand fishcommunities: 119:337-352. in thespeciescomposition and richnessof fish Tonn,W. M., and J. J. Magnuson.1982.Patterns Wisconsinlakes.Ecology63:1149-1166. assemblagesin northern andtheorganization of winterkill, predation, Tonn,W. M., andC. A. Paszkowski.1986.Size-limited fishassemblages.Can. J. Fish.Aquat.Sci. 43:194-202. Umbra-Perca (Umbralimi)andyellowperch(Percaflavescen,) 1987.Habitatuse ofthecentralmudminnow in Umbra-Perca andtheabioticenvironpredation, assemblages:therolesofcompetition, ment.Can. J.Zool. 65:862-870. analysisin fishery management: Tonn,W. M., J.J.Magnuson,andA. M. Forbes.1983.Community withnorthern Wisconsin lakes.Trans.Am. Fish. Soc. 112:368-377. an application andI. J.Holopainen.1989.Responsesofcruciancarppopulations to Tonn,W. M., C. A. Paszkowski, differential pond.Can. J.Zool. 67:2841-2849. predation pressurein a manipulated P. 1970.Relationsbetweenthebenthicfaunaand two speciesof troutin some small Tuunainen, Finnishlakestreatedwithrotenone.Ann.Zool. Fenn.7:67-120. of theUnitedStates.No. 20. NationalClimatic of Commerce.1984.Climatology U.S. Department Data Center,Asheville,N.C. bythreeesocids:theroleofpreybehaviorand Wahl,D. H., andR. A. Stein.1988.Selectivepredation Trans.Am. Fish. Soc. 117:142-151. morphology. Australian ecosystemsto thoseelsewhere.BioScience38:549-556. Westoby,M. 1988.Comparing ofspeciesdiversity. Taxon21:213-251. R. H. 1972.Evolutionandmeasurement Whittaker, D. 1978.ClustanIC user'smanual.University College,London. Wishart, in relation to thespecific successionsoflakeichthyocoenoses Zhakov,L. A. 1974.The composition offishes. J.Ichthyol. offaunistic features Vopr.Ikhtiol.)14:208-218. (Engl.transl., complexes
