Isopod assemblage in response to management techniques and …web.pdx.edu/~scm6/Isopod_Magic.pdf ·...
Transcript of Isopod assemblage in response to management techniques and …web.pdx.edu/~scm6/Isopod_Magic.pdf ·...
IsopodassemblageinresponsetomanagementtechniquesandenvironmentalvariablesonNantucketIsland AndrewMcCandless,StevenMurschel,AshleySmithers 9December2014 Introduction:
ThesandplaingrasslandandheathlandofthecoastalNortheastarepartofarareandprotectedhabitattype.Manyspeciesofplants,birds,andinsectsrelyontheseremainingsandplainhabitats.Managershavefocusedonusingburningandmowingaswaystocontrolinvasivespeciesandmaintaincurrentplantcommunities.Withanyrarelandscape,asthearearecedesandgoesthroughlandscapechanges,thespeciesendemictotheareadecline(Foster&Motzin2003).Researchonlandscapemanagementimpactsonbiotaprimarilyfocusesontheimpactstocharismaticcreaturessuchasbirds(Zuckerberg&Vickery2006;VanDykeetal.2004);lessresearchisfocusedonthelowerorderspeciessuchasinvertebrates.
Researchontheroleofsoilinvertebratesisimportantinalteredgrasslandecosystemsbecauseoftheirroleindecompositionprocessesandpotentialassistanceinrestorationofdegradedlands(Snyder&Hendrix2008).Terrestrialisopodsarearthropods(class:Crustacea)thatliveinmoisthumidenvironmentsinsoilandleaflitter(Heeley1941).Theyareimportantforcreatingmicrobiomesandmicroclimateswithingroundlitter(Souty‐Grossetetal.2005)byeatingplantmaterialaswellasmovingdecayingmaterialfromthesoilsurfacetodeepermoisturemicrositesinthesoil(Hassalletal1987).Ingrasslandcommunities,isopodsimprovesoilandmaintainlitterdecomposition(Curry1994).
Isopodsactasbioindicatorsforhabitatqualityofgrassland(Longcore2003;Souty‐Grossetetal2005;SnyderHendrix2008).Invertebrateresponsestoburningandmowingmanagementactivitiesingrasslandshavebeenreportedinseveralstudies(Andersonetal.1989;ChambersandSamways1998;Dunwiddie1991).Thesedifferentmanagementtechniquescanbeusedtochangethemicroclimatesofisopodhabitats.Springett(2006)identifiedchangesinlittermicroclimatesinresponsetofireandreductionofthenumberofforestisopods.AbetterunderstandingoftheimpactsonisopodcommunitiesofNantucketIslandwillhelpdrivemanagementdecisionsonthesandplaingrasslandandheathland.
Figure1:Aconceptualmodelofthestudy.Wehypothesize(HypothesisA)thatmowing,burning,andcontroltypemanagementsaffectisopodassemblagewithmowingandburningaffectingisopodassemblagesignificantlydifferentlyfromthecontrol.Wefurtherhypothesize(HypothesisB)thatthenon‐managementvariables(litter,woodysteamcount,%groundcoveranddistancetoocean)affectisopodassemblagewithlitterbeingthebestpredictor.
UsingunpublisheddatafromMcKenna‐Foster(2009)wherespiderandisopodassemblage
insitesrepresentingthreemanagementtypes:burning,mowing,andnotreatmentweremeasured,weaddressedthequestion:whatistheeffectoftreatment(fire,mowing,andnotreatment)onisopodassemblage?Wetestedthehypothesisthatisopodassemblageinareasmanagedwithburningandmowingwouldbesignificantlydifferentfromareaswithnotreatment(Figure1,HypothesisA).Thenwelookedattheimpactthatplantlitter,percentgroundcover(grass,heath,andbareground),distancetoocean,andwoodystemcounthadonisopodassemblage.Thesecondquestionweaddressedwas:Whatistherelativeimportanceofthesemeasuredenvironmentalvariablesonisopodassemblage?Thesecondhypothesiswetestedwasthatlitterwouldhavethelargesteffectonisopodassemblageoftheenvironmentalvariables(Figure1,HypothesisB).
Methods: Studylocation
SmoothHummocksCoastalPreserve(SHCP)locatedonNantucketIslandisasandplaingrasslandandheathlandpreservemanagedbytheNantucketIslandLandBank.NantucketIslandislocated30milessouthofCapeCod,Massachusetts(Figure2).SandplaingrasslandandheathlandarefoundonNantucket,Tuckernuck,andMartha’sVineyardislands.ThestudysiteSHCPwasapproximately345hectaresofsandplaingrassland,sandplainheathland,andpitchpinestands.
Eightmanagementsites(Figure2)locatedonSHCPwereselectedforthisstudy.The
siteswereselectedbasedonthemanagementtheyreceived,theyvaryinsize,andtheyhaveunpavedroadswithsomemowedvegetationbreaks.Onsiteswithmanagementregimesofmowingandburning,themanagementactivityhadbeendoneatleastoncebetween1998‐2008.Datacollectedforresearchonarthropodswerecollectedin2008fromtheresearchplotsplacedineachmanagementsite.Threeofthesiteshadburningasthemainmanagement
Figure2:LocationofNantucketIslandoffthecoastofmainlandMassachusetts(topleft),thelocationofSHCPonNantucketIslandisoutlined(topright),andthelocationoftheeightmanagementsitesatSHCP(bottom).Circlesindicateplotlocationsfoundineachmanagementsite.MapbyAndrewMcKenna‐Foster.
activity,fourofthesitesweremanagedwithmowing,oneofthemowedsiteswasmowedeveryyear,andtheeighthsitehadnotreatmentsandactedasthecontrol.
Figure3.Experimentaldesign.Therewerethreetypesoftreatmentsite:burn(n=3),mow(n=4),andcontrol(n=1).Eachtreatmentsitehadthreeplots.Eachplothadonetransectandeachtransectconsistedofsevenpitfalltraps.Thetotalnumberofpitfalltrapspermanagementtypevaried,burn(n=63),mow(n=84),andcontrol(n=21).Anesteddesignwithmanagementsites,plots,transects,andpitfalltraps
Threeplotswereplacedineachoftheeighttreatmentsites(Figures2and3).Plotsusedforthestudywerechosenbyvisitingrandomlyorderedsitesandchoosingonesthathadspecificcharacteristics:80%grasscoverand20mfromthenearestroad.Toensureindependencebetweensites,eachplothadtobewithinitsownsandplaingrasslandpatchandseparatedbyshrubbyvegetationfromtheothersites.Eachoftheplotshadonetransect(n=24)andsevenpitfalltraps(n=168).Thesevenpitfalltrapsweresetthreemetersapartinatransectparalleltothesoutherncoastline.Thepitfalltrapswerefivecmdiameterpolypropylenecupsfilledwith50:50propyleneglycol.PitfalltrapsweresetandsampledforsevendaysinbothMayandAugust2008.
Eachtransectwasplacedneararandomlyselectedpointintheplot.Eventhoughrandom,theresearchersrequiredthatoneendofeachtransectwasatleast0.5morgreaterfromthenearestheathplant.Sixenvironmentalvariablesweremeasuredalongeachtransect:dryleaflitteringrams,grass,heath,andbaregroundpercentcover,andwoodystemcountpersquaremeter.DistancetotheoceanwascalculatedusingGISandmeasuredinmeters.
Investigatingtheeffectsofmanagementtreatmentonisopodassemblage
Todeterminewhethermanagingsandplaingrasslandandheathlandbymowingorcontrolledburninghadasignificantimpactonisopodassemblage,aKruskal‐Wallisranksumonewayanalysisofvariancetestwasperformed.Thistestwasappropriateduetothenon‐
parametricnatureofthedataandbecausethistestcouldhandleourvaryingsamplesizes.Toaccountforthenestedexperimentaldesign,themeanaverageoftheisopodcountsforallpitfalltrapswithinamanagementsitewasusedastheisopodcountforthatmanagementsite.
Modelingisopodassemblagecountbasedonmeasuredenvironmentalvariables
Alinearmodelwasgeneratedtopredictpooledisopodassemblageacrosssitesbasedonsixpredictorvariables(drylittermass,percentcoverofgrass,heath,bareground,distancetotheocean,andwoodystemcount).Priortomodelconstruction,dataweresplitrandomlyintotwosubsets.Thefirstsubsetcontained~75%ofthedata(131observations)andthesecond~25%(44observations).Buildingthemodelbasedon75%ofthedata(themaindataset)enabledmodelvalidationusingtheremaining25%(thetestdataset).
Diagnosticplotsforafullmodelofthemaindataset(includingresidualsvs.fittedvalues,standardizedresidualsvs.theoreticalquantiles,standardizedresidualsvs.fittedvalues,andstandardizedresidualsvs.leverage)wereusedtovisuallydeterminetheneedtotransformtheresponsevariable‐isopodassemblage.Basedonvisualanalysisofacorrelationmatrixfromthemaindataset(AppendixFigureA),distancetooceanandwoodystemcountwerelogtransformedtoreducetheimpactofoutliers.ABox‐Coxpowertransformationwasusedtodeterminethetransformationmostlikelytoresultinnormallydistributedresidualsfortheresponsevariable(AppendixFigureB).Thedataweretransformedanddiagnosticplotsforasecondfullmodelwerecheckedvisually(AppendixFigureC).Havingdeterminedthemodelmetnecessaryassumptions,twomethodswereemployedtodetermineaminimumadequatemodel. Anallsubsetscriterion‐basedprocedurewasusedfollowedbyastepwise/criterion‐basedhybridproceduretotakeadvantageofthecriteriaMallow’sCP,adjustedR2,theBayesianInformationCriterion(BIC),andtheAikaikeInformationCriterion(AIC).Weselectedthenumberofvariablesandtheparticularvariablesforaminimumadequatemodelbasedonthesefourcriteria.AnANOVAtestwasusedtodeterminewhethertheminimumadequatemodelwassignificantlydifferentfromthefullmodel.Varianceinflationfactors(VIF)werecalculatedtotesttheminimumadequatemodelforissuesofmulticollinearity.Theminimumadequatemodelwasscaledusingz‐scores,andascaledminimumadequatemodelwasproduced.Z‐scoringwasusedtostandardizethepredictorcoefficients.Oncestandardized,predictorcoefficientswerecompareddirectly(whereasthiswaspreviouslyimpossibleduetounitdifferences)enablingananalysisoftherelativepredictivepowerofeachpredictorvariable.Lastlythepredictionsofthemodelweretestedusingthetestdatasetandassessedbycalculatingtherootmeansquaredeviation. Results: Theeffectsofmanagementtreatmentonisopodassemblage(Figure1,HypothesisA) Isopodassemblagevariedwidelywithintreatmentgroupswithstandarddeviationsbeinggreaterthanthemeansinallthreetreatmenttypes(Table1).Sitestreatedbymowingappeartohavesignificantlyhigherlevelsofisopodsthancontrolsitesorthosetreatedbyburning(Figure4).Sitesmanagedbyburninghadameanisopodassemblageof78thatwasveryclosetothecontrolmeanof81.Mowedsiteshadameanapproximately60%greaterwithacountof126isopods.Thestandarddeviationforsitesmanagedbyburningwas176,162formowedsites,and117forthecontrolsite(Table1).
Table1:Descriptivestatisticsofisopodassemblage(numberofindividuals)givendifferentmanagementtreatments(Burnn=3,Mown=4,Controln=1).
Burn Mow Control Minimum 0.0 0.0 5.0 Maximum 902.0 747.0 419.0 Median 14.0 68.0 22.0 Mean 78.4 126.4 80.8 Standarddeviation 176.4 162.0 117.1 Variance 31112.9 26255.8 13714.2
Figure4:Barchartofthemeanofisopodassemblage(numberofindividuals)givendifferentmanagementtreatments(Burnn=3,mean=78.4,SE=+/‐22.2,Mown=4,mean=126.4,SE=+/‐17.7Controln=1,mean=80.8,SE=+/‐22.1). Thenon‐parametricKruskal‐Wallistestdidnotindicateasignificantrelationshipbetweenisopodassemblageandanyoneofthemanagementtypes(Χ2=1.22,df=2,p=0.54).Thenullhypothesisthattherewasnorelationshipbetweenburningandmowingmanagementandisopodcouldnotberejectedbasedonthedatacollected. Modelingisopodassemblagebasedonmeasuredenvironmentalvariables(Figure1,HypothesisB) Histograms,scatterplots,andSpearmancoefficientsofeachvariableindicatedstrongcorrelationsbetweenthefourpredictorsl.stem(thelogofwoodystemcount),heath,grass,litterandtheresponsevariableisopod.Potentialmulticolinnearitywasseenbetweengrassandheath,grassandl.ocean(thelogofthedistancetotheocean),grassandl.stem,andheathandl.stem(Figure5).VisualassessmentofFigure5showsapotentiallystrongpositivecorrelationbetweenl.isopod(1+thelogofisopod)andlitter,andanotherpositivecorrelationbetweenl.isopodandl.stem.
Thefullmodelshowingl.isopodasafunctionoflitter,grass,heath,ground,l.ocean,and
l.stemyielded:
. 0.04 ∙ 0.02 ∙ 0.01 ∙ 0.12 ∙ 0.12 ∙ . 0.57 ∙ . 2.18
Thismodelindicatedthat1+naturallogofisopodcountwouldincreaseaslitter,thenaturallogofthedistancetotheocean,and1+naturallogofthewoodystemcountincreased.1+naturallogofisopodcountwoulddecreaseifthepercentcoverofgrass,heath,orbaregroundincreased (R2=0.4953, p=2.2x10‐16, α=0.05). Thefullmodelpasseddiagnosticsillustratingresidualswerenormallydistributed,withoutunderlyingstructure,andtherewerenoinfluentialoutliersaccordingtotheCook’sdistancemetric(AppendixFigureC).
Anallsubsetscriterion‐basedprocedureusingthecriteriaMallow’sCP,adjustedR2,andtheBICindicatedthatamodelwiththreepredictorsandoney‐interceptshouldbeselected(AppendixFigureD).ThiswasdeterminedbytheindicationthatamodelwithfourparameterswouldhavethelowestMallow’sCP,highestadjustedR2,andlowestBIC.Astepwise/criterion‐
Figure5:Correlationmatrixshowinghistograms,scatterplots,andSpearmancorrelationsbetweentheresponsevariableisopodassemblageandthesixpredictorvariables.
basedhybridmethodalsosuggestedthatamodelwiththreepredictorsandoney‐interceptshouldbeselectedandbothmethodsindicatedtheselectionofaminimumadequatemodelwithl.stem,bareground,andlitterasthethreepredictors.
Theminimumadequatemodelbasedonanallsubsetscriterion‐basedprocedureandastepwise/criterion‐basedhybridapproach:
. 0.56 ∙ . 0.10 ∙ 0.05 ∙ 1.46
Theminimumadequatemodel(R2=0.4989)wassignificantoverallwithap‐valueof2.2x10‐16(α=0.05).Thismodelindicatedthatl.isopodwillrisewithanincreaseinl.stemandlitteranddecreasewithanincreaseinground.Thetestformulticollinearitybetweenthevariablesl.stem(VIF=1.1),ground(VIF=1.0),andlitter(VIF=1.1)intheminimumadequatemodelrevelednoissuesofmulticollinearity.Normalityofresidualsandtheabsenceofhighleverageoutliersweredeterminedbyvisualinspectionofdiagnosticplots(AppendixFigureE).TheANOVAbetweenthefullandminimumadequatemodel(p=0.7369)indicatedthatthenullhypothesisofthetest(thatthetwomodelsarenotdifferent)couldnotberejectedthusthesimplerminimumadequatemodelwasmaintained.
Thez‐scoredminimumadequatemodelenabledadirectcomparisonoftherelativeinfluenceofeachpredictor:
. 0.51 ∙ . 0.23 ∙ 0.28 ∙ 1.34 10 16
Thez‐scoretransformationoftheminimumadequatemodelhasthesameR2valueandsignificanceastheminimumadequatemodel.Thisstandardizationofthemodelcoefficientsshowedl.stemwasthemostimportantofthemeasuredpredictorsofl.isopodandwasapproximatelyasimportantasthetwootherpredictorscombined.Groundandlitterwereverysimilarintheirpredictivepoweronl.isopodwithanincreaseingroundleadingtoadecreaseinl.isopodwhileanincreaseinlitterleadstoanincreaseinl.isopod.Ourmodelshowedtherelativeimportanceratiooftheenvironmentalvariablesonisopodassemblagetobe0.51woodystemto0.23baregroundpercentageto0.28dryleaflitter(i.e.a0.51:0.23:0.28proportion). Theminimumadequatemodelpredictedthetrendseeninboththemaindata(subsetof131observations)andthetestdata(subsetof44observationsnotusedtoconstructthemodel)(Figure6).Therootmeansquaredifferenceofl.isopodwas1.50andtheuntransformedrootmeansquaredifferencebetweentheminimumadequatemodelpredictionsandobservedisopodassemblagewas4.48.
Discussion: Managementtreatmentonisopodassemblage Whilemanagementwasfoundtonotimpactisopodassemblage,thisislikelytheresultofalackofanalyticalpowerduetothesmallsamplesize,unequaltreatmentreplication,andtheuseofarankbasedtest.Theapparentdifferencesinmeanbetweenmowedtreatmentareasandcontrolledorburnedareas(Figure4)providesfurtherevidencethatfindingnorelationshipbetweenisopodassemblageandtreatmentwasduetoalackofstatisticalpower.Otherresearchconductedacrossarthropodtaxasuggestsotherarthropodpopulationswereeithernotgreatlyinfluencedbymowingand/orburningtreatmentsortherewashighvariationofresponsestotreatmentsacrosstaxaandseason(Dunwiddie1991).Currentresearchshowsarthropods,includingisopods,generallyadapttomanagementactions(Warrenetal.1987;Dunwiddie1991).Further,observedmanagedsitesweremowedandburnedindifferentyearspotentiallyprovidingisopodassemblagetheopportunitytoreboundfromtreatmentmakingtheaffectsundetectablebythe2008observation.
Theknownrelationshipbetweenmanagementandenvironmentalpredictorsofisopodassemblagefurthersupportsthesuppositionthatalackofsignificanteffectswasduetolowpowerofanalysis.Mowingandburningaretoolsusedtomaintainthevegetationonsandplaingrasslands(Neill2006).Burningconsumesleaflitteranddestroysisopodmicrohabitats(Springett2006).Mowingshouldleadtoanincreaseinlitterthatmayalsoimpactisopodhabitatavailability.Usingmowingandburningmayindirectlybenefitisopodcommunitiesbyimprovingthehabitatrequirementsthatisopodsrelyon.
Figure6:Left:Minimumadequatemodelpredictionwiththemaindatashowingtheobservedversuspredictedvaluesofl.isopod.Thelineshowsthepredictedl.isopodvaluesbasedontheminimumadequatemodel.Right:Minimumadequatemodelpredictionwiththetestdataof44pointsinitiallysubsettedfromtheobservationsandnotusedintheconstructionoftheminimumadequatemodel.Thelineshowsthepredictedl.isopodvaluesbasedontheminimumadequatemodel.
Isopodassemblageinrelationtosignificantenvironmentalvariables Whilethehypothesisthatleaflitterwouldbethestrongestpredictorofisopod
assemblagewasnotsupported,leaflitterwasthefoundtobeofsecondhighestimportanceafterwoodystemcount.Woodystemsweremainlyrepresentedbylivingblackhuckleberryandsomebayberry.Understandingthewoodystem/isopodassemblagerelationshipisakeytoanalyzingisopodpopulations.Otherstudieshavehighlightedtheimportanceofhabitatstructure,includingconnectivelandscapepropertiesandvariationsinfaunaandsoiltypes,onisopodpopulationdynamics(Davis1984;Souty‐Grossetetal.2005).Isopodspeciesmayrelyonwoodystemsforprotectionfromsolarradiationandcapturedrainfall(Volketal.2000).Balancingthehabitatmosaicofheathandshrubswithingrasslandsmaybenefitisopodsandthegrasslandcommunity.FurtherupkeepofinterspersedwoodystemswithintheNantucketmanagedecosystemwouldlikelybenefitfloraandfaunainteractionsandisopodpopulations.
Thesignificantinfluenceoflitterandbaregroundcomparedtoothermeasuredenvironmentalfactorshighlighttheimportanceofdetritusasafoodsourcefortheisopodcommunity.Isopodsshowverystrongpreferencesfordetritus(Hassalletal.1987;Zimmer2002),andinthisecosystemgroundlitteriscomprisedmostlyofdecomposinggrasses(Burkeetal.1998).ThenegativerelationshipbetweenbaregroundandisopodassemblagefurtheremphasizestheimpactofvegetationandsubsequentgroundlitteronNantucketisopodassemblage.Maintaininganaturallydecomposingecosystemwithhighflorarichnessandscatteredvegetationdiversitywillmostbenefitisopodassemblageinthisecosystem. Althoughmorethan90%ofisopodscountedappearedtobeofthesamemorphologicaltype,thefewuniqueindividualssuggestsidentifyingisopodstothespecieslevelmayidentifyisopodspeciesindicativeofparticularimportantmicrohabitats.Specieslevelidentificationcouldhelpimproveunderstandingofecosysteminteractions.Rarepredatoryisopodshavebeenshowntobestrongbioindicatorsofsuccessfulrestorationeffortsinterrestrialshrublands(Longcore2003),andknowingisopodspeciescouldbetterinformlandscapemanagement.Identificationofvegetationspeciesmayalsoshowisopodpreferenceandcontributetoabetterunderstandingoftheimportanceoflittercomposition.
Theroleofenvironmentalvariablesnotmeasuredinourstudy,suchassoilmoisture,whichhasbeenshowntohaveasignificanteffectonisopodassemblage(Warburg1987),mayhavelimitedthepredictivepowerofourlinearmodel.Awiderangeofinteractingfactorscontributetothecreationofmicrohabitatsandmicrobiomeswithinshrubsscatteredthroughoutgrasslands.Shrubswithingrasslandsactasareasofintensenutrientfluxfromroot/leafturnoverandexcessexcrementfromsurroundingfauna(Vetaas1992).Terrestrialisopodshaveundergoneevolutionarychangestodwellonland,anddespitesomespecieslivinginaridregions,allspeciesneedprotectionfromdesiccation(Schmidt&Wägele2001).Scatteredshrubs(woodystems)ingrasslandscreatemoist,hospitablemicroclimatessimilartotheeffectscatteredtreescanhaveinalandscape(Manningetal.2006).
Themodelvalidationindicatedthatthemodeldoesagoodjobofdescribingrelatedindependentdata.Therootmeansquaredifferenceof4.48forisopodassemblageisassessedbasedonitscomparisonwiththemeanisopodassemblageof101.Adifferenceof4.48issmallincomparisontothemeanof101indicatingtheapplicabilityofthelinearmodeltoindependentdata.
Study‐wideconsiderations
Asstrongpredictorsofisopodassemblage,alterationsinmanagementthatimpactleaflitterandwoodystemsshouldaffectisopodassemblage.Furtherstudyonagreaternumberof
independentgrasslandandheathlandsitesthathavebeenmowedorburnedshoulduncovertherelationshipbetweenisopodassemblage,managementaction,andthesignificantpredictorvariablesfromourminimumadequatemodel.Managementaffectsleaflitter(Andrewetal.2000),groundcover,andwoodystems,andthosethreepredictorsindicateisopodassemblage.Thattheaffectofmanagementonisopodassemblagewasnotfoundinourstudywhileleaflitter,percentbareground,andwoodystemswerefoundtobesignificantpredictorsofisopodassemblageisalmostcertainlyduetooursamplingdesignandprovidesaclearcalltofurtherresearch.Bibliography: Anderson,R.C.,Leahy,T.,&Dhillion,S.S.(1989).Numbersandbiomassofselectedinsectgroupsonburnedandunburnedsandprairie.AmericanMidlandNaturalist,151‐162.
Andrew,N.,Rodgerson,L.,&York,A.(2000).Frequentfuel‐reductionburning:theroleoflogsandassociatedleaflitterintheconservationofantbiodiversity.AustralEcology,25(1),99‐107.
Burke,I.C.,Lauenroth,W.K.,Vinton,M.A.,Hook,P.B.,Kelly,R.H.,Epstein,H.E.,Aguiar,M.R.,Robles,M.D.,Aguilera,M.O.,Murphy,K.L.&Gill,R.A.(1998).Plant‐soilinteractionsintemperategrasslands.InPlant‐inducedsoilchanges:Processesandfeedbacks(pp.121‐143).SpringerNetherlands.
Curry,J.P.(1987).Theinvertebratefaunaofgrasslandanditsinfluenceonproductivity:thecompositionofthefauna.GrassandForageScience,42(2),103‐120.
Curry,J.P.(1994).Grasslandinvertebrates:ecology,influenceonsoilfertilityandeffectsonplantgrowth.Springer.
Chambers,B.Q.&Samways,M.J.(1998).Grasshopperresponsetoa40‐yearexperimentalburningandmowingregime,withrecommendationsforinvertebrateconservationmanagement.Biodiversity&Conservation,7(8),985‐1012.
Davis,R.C.(1984).Effectsofweatherandhabitatstructureonthepopulationdynamicsofisopodsinadunegrassland.Oikos,387‐395.
Dunwiddie,P.W.(1991).Comparisonsofabovegroundarthropodsinburned,mowedanduntreatedsitesinsandplaingrasslandsonNantucketIsland.AmericanMidlandNaturalist,206‐212.
Foster,D.R.,&Motzkin,G.(2003).InterpretingandconservingtheopenlandhabitatsofcoastalNewEngland:insightsfromlandscapehistory.ForestEcologyandManagement,185(1),127‐150.
Hassall,M.,Turner,J.G.&Rands,M.R.W.(1987).Effectsofterrestrialisopodsonthedecompositionofwoodlandleaflitter.Oecologia,72(4),597‐604.
Heeley,W.(1941,October).ObservationsontheLife‐HistoriesofsomeTerrestrialIsopods.InProceedingsoftheZoologicalSocietyofLondon(Vol.111,No.1‐2,pp.79‐149).BlackwellPublishingLtd.
Longcore,T.(2003).Terrestrialarthropodsasindicatorsofecologicalrestorationsuccessincoastalsagescrub(California,USA).RestorationEcology,11(4),397‐409.
Manning,A.D.,Fischer,J.,&Lindenmayer,D.B.(2006).Scatteredtreesarekeystonestructures–implicationsforconservation.BiologicalConservation,132(3),311‐321.
Neill,C.,VonHolle,B.,Kleese,K.,Ivy,K.D.,Colllins,A.R.,Treat,C.,&Dean,M.(2007).HistoricalinfluencesonthevegetationandsoilsoftheMartha’sVineyard,Massachusettscoastal
sandplain:implicationsforconservationandrestoration.BiologicalConservation,136(1),17‐32.
Paris,O.H.(1963).TheecologyofArmadillidiumvulgare(Isopoda:Oniscoidea)inCaliforniagrassland:food,enemies,andweather.EcologicalMonographs,1‐22.
Schmidt,C.,&Wägele,J.W.(2001).MorphologyandevolutionofrespiratorystructuresinthepleopodexopoditesofterrestrialIsopoda(Crustacea,Isopoda,Oniscidea).ActaZoologica,82(4),315‐330.
Souty‐Grosset,C.,Badenhausser,I.,Reynolds,J.D.&Morel,A.(2005).Investigationsonthepotentialofwoodliceasbioindicatorsofgrasslandhabitatquality.Europeanjournalofsoilbiology,41(3),109‐116.
Springett,J.A.(1976).TheeffectofprescribedburningonthesoilfaunaandonlitterdecompositioninWesternAustralianforests.AustralianJournalofEcology,1(2),77‐82.
VanDyke,F.,VanKley,S.E.,Page,C.E.&VanBeek,J.G.(2004).Restorationeffortsforplantandbirdcommunitiesintallgrassprairiesusingprescribedburningandmowing.RestorationEcology,12(4),575‐585.
Vetaas,O.R.(1992).Micro‐siteeffectsoftreesandshrubsindrysavannas.Journalofvegetationscience,3(3),337‐344.
Vickery,P.D.,Zuckerberg,B.,Jones,A.L.,Shriver,W.G.&Weik,A.P.(2005).InfluenceoffireandotheranthropogenicpracticesongrasslandandshrublandbirdsinNewEngland.StudiesinAvianBiology,30,139.
Volk,M.,Niklaus,P.A.,&Körner,C.(2000).SoilmoistureeffectsdetermineCO2responsesofgrasslandspecies.Oecologia,125(3),380‐388.
Warburg,M.R.(1987).Isopodsandtheirterrestrialenvironment.Advancesinecologicalresearch,17,187‐242.
Warren,S.D.,Scifres,C.J.,&Teel,P.D.(1987).Responseofgrasslandarthropodstoburning:areview.Agriculture,ecosystems&environment,19(2),105‐130.
Zimmer,M.(2002).Nutritioninterrestrialisopods(Isopoda:Oniscidea):anevolutionary‐ecologicalapproach.BiologicalReviews,77(4),455‐493.
Zuckerberg,B.&Vickery,P.D.(2006).EffectsofmowingandburningonshrublandandgrasslandbirdsonNantucketIsland,Massachusetts.TheWilsonJournalofOrnithology,118(3),353‐363.
Appendix:
FigureA:Correlationmatrixshowinghistograms,scatterplots,andSpearmancorrelationsbetweentheresponsevariableisopodassemblageandthesevenpredictorvariables.
FigureB:ResultsofBox‐Coxpowertransformationonisopodasafunctionoflitter,grass,heath,ground,l.ocean,andl.stemindicatingalogtransformationwouldbeappropriate(asseenbythe95%confidenceintervallinesenveloping0).
FigureC:Diagnosticstatisticsoffullmodel.Topleftshowsthattheresidualsarenormallydistributedaboutazerolineandshownostructure,thetoprightshowsthenormalityofthestandardizedresiduals,thebottomleftshowsthestandardizedresidualshavenounderlyingstucture,andthebottomrightshowsthattherearenohighleverageoutlierswithCook’sdistancesofgreaterthan0.5.
FigureD:Plottingtheresultsofanallsubsetscriterion‐basedproceduremodelingl.isopodasafunctionofbetween1and6variablesand1y‐intercept.TopleftshowsthelowestMallow’sCpformodelswith1‐6variables+1y‐intercept.ToprightshowsthehighestR2andbottomleftthelowestBayesianInformationCriterionformodelswith1‐6variables+1y‐intercept.Eachcriterionindicatesthatamodelwith3predictorsand1y‐interceptwouldbethebest.
FigureE:Diagnosticplotsoftheminimumadequatemodel.Topleftshowsthattheresidualsarenormallydistributedaboutazerolineandshownostructure,thetoprightshowsthenormalityofthestandardizedresiduals,thebottomleftshowsthestandardizedresidualshavenounderlyingstucture,andthebottomrightshowsthattherearenohighleverageoutlierswithCook’sdistancesofgreaterthan0.5.