Land cover, riparian zones and instream habitat influence ......monitoring of the hydrographic basin, instream habitat and aquatic fauna is needed to test for potential legacy effects

Ecol Freshw Fish 201928317ndash329 wileyonlinelibrarycomjournaleff emsp|emsp317copy 2018 John Wiley amp Sons AS Published by John Wiley amp Sons Ltd

Received28May2018emsp |emsp Revised27September2018emsp |emsp Accepted28September2018DOI 101111eff12455

O R I G I N A L A R T I C L E

Land cover riparian zones and instream habitat influence stream fish assemblages in the eastern Amazon

Luciano F A Montag12 emsp|emspKirk O Winemiller2 emsp|emspFriedrich W Keppeler2 emsp|emsp Hiacutengara Leatildeo3 emsp|emspNaraiana L Benone14 emsp|emspNaiara R Torres5 emsp|emspBruno S Prudente6 emsp|emsp Tiago O Begot3 emsp|emspLuke M Bower2 emsp|emspDavid E Saenz2 emsp|emspEdwin O Lopez-Delgado2 emsp|emsp Yasmin Quintana2 emsp|emspDavid J Hoeinghaus7 emsp|emspLeandro Juen1

1LaboratoacuteriodeEcologiaeConservaccedilatildeoInstitutodeCiecircnciasBioloacutegicasUniversidadeFederaldoParaacuteBeleacutemBrazil2DepartmentofWildlifeandFisheriesSciencesTexasAampMUniversityCollegeStationTexasUSA3ProgramadePoacutes-graduaccedilatildeoemZoologiaMuseuParaenseEmiacutelioGoeldiUniversidadeFederaldoParaacuteBeleacutemBrazil4ProgramadePoacutes-graduaccedilatildeoemEcologiaInstitutodeCiecircnciasBioloacutegicasUniversidadeFederaldoParaacuteBeleacutemBrazil5ProgramadePoacutes-graduaccedilatildeoemEcologiaAquaacuteticaePescaInstitutodeCiecircnciasBioloacutegicasUniversidadeFederaldoParaacuteBeleacutemBrazil6UniversidadeFederalRuraldaAmazocircniaCapitatildeoPoccediloBrazil7DepartmentofBiologicalSciencesAdvancedEnvironmentalResearchInstituteUniversityofNorthTexasDentonTexasUSA

CorrespondenceLucianoFAMontagLaboratoacuteriodeEcologiaeConservaccedilatildeoInstitutodeCiecircnciasBioloacutegicasUniversidadeFederaldoParaacuteBeleacutemBrazilEmaillfamontaggmailcom

Funding informationCoordenaccedilatildeodeAperfeiccediloamentodePessoaldeNiacutevelSuperiorGrantAwardNumber888810684252014-01and888811190972016-1USNationalScienceFoundationgrantGrantAwardNumberDEB1257813ConselhoNacionaldeDesenvolvimentoCientiacuteficoeTecnoloacutegicoGrantAwardNumber12820143050172016-03075972016-4and4493142014-2ConservationInternationalofBrazil(CI-Brazil)AgropalmaGroupBiopalmaValeCIKELLtda33ForestCapitalUnitedStatesAgencyforInternationalDevelopment(USAID)theestateofGeorgeandCarolynKelsoviatheInternationalSportfishFund

AbstractTheAmazon rainforesthasexperienced rapid land-usechangesover the last fewdecades including extensive deforestation that can affect riparian habitats andstreamsTheaimofthisstudywastoassessresponsesofstreamfishassemblagestodeforestationandlandcoverchangeintheeasternAmazonWeexpectedthatper-centageofforestinthecatchmentiscorrelatedwithlocalhabitatcomplexitywhichin turn determines fish assemblage composition and structure We sampled 71streams in areas with different land uses and tested for relationships betweenstreamfishassemblagesandlocalhabitatandlandscapevariableswhilecontrollingfortheeffectofintersitedistanceFishassemblagecompositionandstructurewerecorrelatedwith forestcoveragebut localhabitatvariablesexplainedmoreof thevariation inbothassemblagecompositionandstructure than landscapevariablesIntersitedistancecontributedtovarianceexplainedbylocalhabitatandlandscapevariablesandthepercentageofvarianceexplainedbytheuniquecontributionoflocalhabitatwasapproximatelyequivalenttothesharedvarianceexplainedbyallthreefactorsinthemodelInthesestreamsoftheeasternAmazonfishassemblagesweremoststronglyinfluencedbyfeaturesofinstreamandriparianhabitatsyetin-directeffectsofdeforestationonfishassemblagecompositionandstructurewereobservedeventhoughintactriparianzoneswerepresentatmostsitesLong-termmonitoringofthehydrographicbasininstreamhabitatandaquaticfaunaisneededtotestforpotentiallegacyeffectsandtimelagsaswellasassessspeciesresponsestocontinuingdeforestationandland-usechangesintheAmazon

318emsp |emsp emspensp MONTAG eT Al

1emsp |emspINTRODUC TION

Land-use change is an important driver of biodiversity change infreshwaterecosystems(Allan2004)Smallstreamsareparticularlysensitivetolandscapealterationduetotheimportanceofcatchmentinputsformaintenanceofinstreamhabitatandfoodwebdynamics(egBrejatildeoHoeinghausPeacuterez-MayorgaFerrazampCasatti2018Lealetal2017Sweeneyetal2004)Forexampledeforestationcan increase sedimentation and reduce woody debris inputs re-sulting insimplifiedandhomogenisedhabitatsandalteredhydrol-ogyandwaterchemistry(CardinalePalmerSwanBrooksampPoff2002) Intact riparian zones can decrease some negative effectsof catchment deforestation on streams (Chen etal 2017 Puseyamp Arthington 2003) Removal of riparian vegetation destabilisesstreambanksdisruptsfluxesbetweenterrestrialandaquaticfoodwebs increases solar radiation and water temperature and altersecosystemmetabolismandnutrientdynamics(NakanoampMurakami2001 Teresa Casatti amp Cianciaruso 2015 Zeni Hoeinghaus ampCasatti2017)

Streamdiversityoftenrespondsstronglytochangesininstreamconditions caused by catchment and riparian deforestation (egBrejatildeoetal2018Lealetal2017TeresaampCasatti2012)Forex-ampleasinstreamhabitatsaresimplifiedandhomogenisedpopu-lationsofspecialistspeciesoftendeclineorareextirpatedwhereasgeneralistspeciestendtoincreaseinabundance(Brejatildeoetal2018Zeni etal 2017) Furthermore stream longitudinal connectivitymaybeaffectedbyanthropogenicdisturbanceoccurringatdiffer-entspatialscalesanddisturbanceatone locationmay impact theconditionandbiodiversityofaconnectedstreamreach(Allan2004Benone Esposito Juen Pompeu amp Montag 2017 WinemillerFlecker amp Hoeinghaus 2010) Alternatively an impacted streammayactasasinkhabitatthat is repeatedlyrecolonisedbyspeciesfrom an unimpacted location (Noss 1990) From a metacommu-nityperspectivetheaforementionedprocesseswouldrepresentabalance between environmental filtering and dispersal leading toacombinationofspeciessortingandmasseffectsacrossahetero-geneouslandscape(Leiboldetal2004reviewedforloticsystemsbyWinemilleretal2010)Howeversomestudieshavefoundthatstreamfishassemblagesappearedtobeminimallyaffectedbyde-forestation and in some cases richness and density increased inimpactedareas(BojsenampBarriga2002Burcham1988TohamampTeugels1999)Thesefindingssuggestthattheeffectofforestcovermaybecontextdependentvaryaccordingtoscaleorbeinfluencedbyotherfactors(egtemperaturehabitatcomplexity)anyofwhichcouldhinderdiscoveryofgeneralrelationships

Effectsofdeforestationonbiodiversityareespeciallyconcern-ing in megadiverse tropical regions (Laurance Sayer amp Cassman2014) At the global scale approximately 13million hectares of

forestswereclearedannuallybetween2000and2010withmostofthatdeforestationoccurringindevelopingcountriesofthetropicsespeciallyinSouthAmerica(BlaserSarrePooreampJohnson2011)Thisisunsurprisinggiventhatacceleratedland-usechangeisdrivenby growth of human populations economic activities andor de-mandforfoodtimberproductsandenergy(LauranceampBalmford2013Loboacuten-CerviaacuteMazzoniampRezende2016)ForexampletheAmazonregionhastheworldrsquoslargestremainingtropicalforestbutalsohighratesofdeforestationduetoarapidlygrowingpopulation(currently gt30million people) that is largely supported by naturalresource extraction and land conversion for agriculture (Gardneretal 2013)Despite the fact that deforestation rates are highestintropicalregionsimpactsofdeforestationonaquaticbiodiversityarepoorlydocumentedinthetropicscomparedwithtemperatere-gionsGivenconsiderabledifferences inclimate land coverbiodi-versityandotheraspectsbetweentropicalandtemperateregionsitisunclearwhethergeneralisationsfromtemperateresearchdirectlyapplytotropicalsettingsThusresearchoneffectsofdeforestationontropicalsystemsisurgentlyneeded

Herein we report findings from an extensive survey of 71streamsintheeasternAmazonbasinthatsharethesameregionalspecies pool and similar climate native vegetation topographyandgeologybutdifferinthetypeandintensityofcatchmentlandcoverchangeWehypothesisedthat theregional fishmetacom-munityisstructuredbyacombinationofspeciessortingandmasseffects(Leiboldetal2004Winemilleretal2010)Specificallywe expected that local habitat (primarily habitat complexitywhichwepredict is associatedwith percentageof forest in thecatchment)istheprimarydriveroffishassemblagestructure(iespeciessorting)andthatdistancebetweensitesmayaccountforsomeadditionalvariationduetopotentialmovementamongsites(iemasseffects)Localhabitat featurespreviously reportedtoinfluencestreamfishassemblagesweremeasured (ie instreamand riparian zone) and analyses included factors at differentlandscapescales(ieimmediateandtotalcatchments)whilecon-trollingforspatialautocorrelationVariancepartitioningwasusedtoassesstherelativeimportanceofuniqueandsharedcontribu-tionsoflocalandlandscapefactorstofishassemblagecomposi-tionandstructure

2emsp |emspMATERIAL S AND METHODS

21emsp|emspStudy area

The study area lies within a region of the eastern Amazon (ParaacuteState Brazil) that historically was covered by dense ombrophil-ous (tolerantofwetconditions) forestTheclimate isclassifiedasequatorialldquoAfrdquointheKoumlppensystem(PeelFinlaysonampMcMahon

K E Y W O R D S

aquaticbiodiversityforestcoverhumanimpactlandscapelocalhabitatordinationvariancepartitioning

emspensp emsp | emsp319MONTAG eT Al

2007)Mean annual temperature is 272degC and themean annualprecipitation is 1800mm with rainfall well-distributed through-outtheyearandexceeding50mmeveninthedriestmonths (ieAugustthroughOctoberWatrinampRocha1992)Inrecentdecadestheregionhasexperiencedincreaseddeforestationfortimberhar-vest(PereiraZweedeAsnerampKeller2002)andclearingforpas-turesandcropsPresentlyoilpalmproductionisthemaincropbutinterestinsugarcaneisincreasingMostoftheremainingforestsaresubjectedtoreduced-impactloggingthataimstominimisedamagetosoilandforestorconventionalloggingthatinvolvesminimalcon-siderationofforestresilience(PrudentePompeuJuenampMontag2017)Withinourstudyregionconventionalloggingstartedinthe1970soilpalmhasbeencultivatedsincethe1980sandreduced-impactloggingwasestablishedin2000underguidelinesproposedbythemandatory forestmanagementplanof theBrazilianForestCode (Law 477165 Brasil 1965) Some of the sites sampled inthisstudyweresurveyed inpreviousstudies (Benoneetal2017FerreiraBegotPrudenteJuenampMontag2018Juenetal2016PrudentePompeuampMontag2018Prudenteetal2017)Relatedstudies from the project Sustainable Amazon Network (RedeAmazocircniaSustentaacutevel)byLeitatildeoetal(2018)Gardneretal(2013)

andLealetal (20162017)arefromthesamebasinbutdifferentsamplingsites

22emsp|emspStudy design

We sampled the fish assemblages at 71 sites in 1st- to 3rd-orderstreams inareaswithdifferent land-usepatternswithintheAcaraacuteandCapimcatchments (Figure1)Weselectedeach streambasedonlandcover(loggingcropspastureandforest)accessibilityandrelativeindependenceofsitesacrosssubdrainages(ieaminimumfluvialdistanceof2kmbetweensamplingsiteswasusedtoavoidpseudoreplication)Thedrainagenetworkwasconstructed fromaSRTMDigital ElevationModelwith a spatial resolution of 30me-tres(httpsearthexplorerusgsgov)Weavoidedsamplingstreamsnearroadsorbridgesinsituationswhereitwasnotpossibletoavoidthesebuiltstructureswesampledfishatleast500mupstreamofthese structuresWemeasured or estimated local and landscapevariables (defined below) at each site and fluvial distance wasmeasuredbetweenallpairsofsitesusing1100000scale imagesFishesweresurveyedandlocalenvironmentaldatawerecollectedduring thedryseason (AugustndashOctoberbetween2012and2015)

F IGURE 1emspStudyareaintheAcaraacutendashCapimbasin(easternAmazonBrazil)Surveysites(n=71)aredenotedbyblackcirclesandland-usecategories(ieprimaryforestsecondaryforestagriculturepastureandbaresoil)aredenotedbycolourswithineachcatchment(solidblacklines)

(a) (b)

(c)

(d)

320emsp |emsp emspensp MONTAG eT Al

whendischargeislowestandsamplingefficiencyisgreatestandtoavoidinfluencefromseasonalvariation(Espiacuterito-SantoMagnussonZuanonMendonccedilaampLandeiro2009)Accordingtohistoricaldata(Meteorological station BeleacutemmdashPAmdashOMM 82191 INMET 2018)therewereno abnormal periodsof precipitation and temperaturein the study area during 2012ndash2015We observed no significanthumandisturbance in thestudyareaduring thestudyperiodandinterannual differences are not expected to influence results andconclusions

23emsp|emspFish assemblage structure

Wesampledthefishassemblageswithtwocircular55-cm-diameterdipnetswith2-mmmeshThisactivesamplingmethodwasusedalongstreamchannelbanksandsubstratesandenabledefficientsamplingofmostmicrohabitatswithinthesesmallstreams(Uiedaamp Castro 1999) The sampling unit for this study was a 150-mstretch of each stream subdivided into 10 subsections of 15meachDuringeachstreamsurveytotalsamplingeffortwas18minforeachsubsectiontotalling3hrofeffortpersite(Ferreiraetal2018Prudenteetal2017)CollectedfisheswereeuthanisedwithlethaldosesoftheanaestheticEugenolfixedina10formalinso-lutionandtransferredto70ethanolafter48hrInthelaboratoryspecimenswere identified to the lowest feasible taxonomic levelusingpublishedkeys(egAlbert2001ReisKullanderampFerraris2003 Van der Seen amp Albert 2018) and in some cases exami-nations by specialists Specimenswere deposited in the ichthyo-logicalcollectionoftheMuseuParaenseEmiacutelioGoeldi (MPEG) inBeleacutemBrazilSurveyswereconductedunderpermit4681ndash1fromthe Instituto Chico Mendes de Conservaccedilatildeo da Biodiversidadeof the Ministeacuterio do Meio Ambiente (Brazilian Ministry of theEnvironment)

24emsp|emspEnvironmental variables

Instreamhabitatandriparianzonevariables(hereaftercitedasldquolocalvariablesrdquo)weremeasuredfollowingthemethodproposedbyPecketal (2006)withdatareductionandmetriccalculationsfollowingKaufmannLevineRobisonSeeligerandPeck(1999)Wecollecteddataforenvironmentalvariablesthathaveshownsignificantasso-ciationswithstreamfishassemblages intheeasternAmazon (egFerreira etal 2018 Juen etal 2016 Prudente etal 2017)Wemeasured 26 physical (eg substrate discharge and flow regime)andchemical(egpHanddissolvedoxygen)variablesthatcomprisesevengeneralcategorieschannelmorphologysubstrateflowlargewoodydebris fishcover riparianvegetationcoverandhuman im-pact (Table1)Human impactwas estimated as the proportion ofthe stream reachdirectly affectedby thepresenceof channelisa-tiondamsbuildingsroadspipelinesrubbleandrubbishThiswascalculatedbydividingthenumberofhuman-influencedsubsectionsbythetotalnumberofsubsections(10perstreamreach)Thusthevariablerangedfrom0instreamreacheswithoutanyhumaninflu-enceto1incompletelymodifiedstreamreaches(humaninfluence

present along the entire stream stretch studied for details seeKaufmannetal(1999))

Wequantified land covervariablesfortheentiredrainagenet-work upstream of the sampling site (hereafter ldquototal catchmentrdquo)and within a 30-m buffer extending 300m upstream and down-streamfromthesamplingsite(hereafterldquoimmediatecatchmentrdquoseeSupportingInformationFigureS1)tocapturepotentiallandscapein-fluencesonstreamfishassemblages(Frimpongetal2005MolinaRoa-FuentesZeniampCasatti2017)Bufferzonesattheimmediatecatchmentscalearecomparabletothewayriparianbuffersaremea-suredaccordingtoBrazilianlawLandcoverclassificationsfollowedZhuLiWangChuandYan(2017)and includedthefollowing (a)primary vegetation characterised by areaswith dense ombrophi-lousforest(b)secondaryvegetationwithvegetationresultingfromnatural succession processes after total or partial suppression ofprimaryvegetationbyanthropogenicactionsornaturalcauses (c)agriculturalareaswithareasoccupiedbyagriculturalactivitiesin-cludingbothmonoculturesandpolycultures(d)pasturewithareasoccupied by intensive andor extensive livestock farming and (e)baresoilwithareasofunprotectedsoilmainlycontainingroadnet-workssuchasdirtroadsandhighways

Land cover was classified through digital image processing oftheRapidEyeEarthImagingSystem(REIS)opticalsensorREISsen-sorimagescoveranarea77kmwideby1500kmlongwitha5-mspatial resolutionand5-bandspectral resolutionwhichallowsforgreaterprecisionintheclassificationoflanduseandgroundcoverImageswereselectedfromthesameyearthatthefieldsurveyswereconductedAllimagesweregrainedandorthorectifiedandthensub-jectedtoatmosphericcorrectiontoattenuatetheeffectsoftheat-mosphereonthespectralresponseofthetargetsinthescenesandconvert thedigitalnumberpixel values to reflectance (RichardsampJia1999)ThisimageprocessingwasperformedwithPCIGeomatics2015softwareusingtheATCORGroundReflectancemoduleAfteratmosphericcorrectionoftheimagesamosaicofREISsceneswascompiledusingtheOrthoEnginemoduleandtheREISmosaicwassubmittedtoobject-orientedclassificationusingeCognition9soft-wareValidationoftheclassificationwasperformedwithTerraClass2014 images provided by the Instituto Nacional de PesquisasEspaciais(Almeidaetal2016)

25emsp|emspData analysis

Analyticalprocedurestoassesstheuniqueandsharedcontributionsof local habitat landscape and intersite hydrographic distance onstream fishassemblagesgenerally followed themethoddescribedby Borcard and Legendre (2002) which is based on redundancyanalyses(RDALegendreOksanenampterBraak2011)AbundancedataweretransformedusingaHellingertransformationinordertoavoid biases caused by the species abundance paradox (Legendreamp Gallagher 2001) A principal coordinates of neighbor matrices(PCNM)analysiswasconductedusingthematrixofintersitefluvialdistancestogeneratevariablesthatrepresentdistancecomponents(Landeiro Magnusson Melo Espiacuterito-Santo amp Bini 2011) Two

emspensp emsp | emsp321MONTAG eT Al

TABLE 1emspSummaryofamong-streamvariationinlocalhabitatandlandscapevariables(SD=standarddeviation)

Factor Variable Abbreviation RDA Mean SD Min Max

Local habitat

Channel morphology

Mean depth of thalweg (cm) DEPTH C S 305 716 1853 4443

SD depth of thalweg (cm) SD_DEPT S 1203 353 575 2434

SD incision height (m) SD_INC C S 134 108 0 398

Mean longitudinal section width times depth ratio (mm) WD_RAT C 912 238 459 1632

Substrate

Siltmuckclay() SILT 1095 893 0 3067

Totalorganicmatter(litterwoodrootsandalgae)() ORGAN 5872 1642 28 100

Wood() WOOD 852 727 0 40

Roots() ROOTS 941 911 0 4857

Coarselitter() LITTE 262 1566 0 6154

Meanresidualpools(m2150mofchannel) POOLS 162 072 05 535

Flow

Riffle() RIFFL 125 304 0 1867

Fastchannelhabitats() FAST 2649 2484 0 94

Sequencefastflowslowflowandpools(index) SEQ_FLW 012 006 001 031

Riparian vegetation cover

SD canopy density banks () SD_BANK S 677 618 089 372

Mean area covered by herbaceous vegetation layer gt05 cm HERB S 27 571 0 2784

Mean woody cover WOODY_C C S 11181 5582 1727 23625

SD woody cover SD_WD_C 2632 833 1013 4977

Large woody debris

NumberLWDinsidebankfullchannelm2ndashsizeclass1 LWD1 012 009 0 037

NumberLWDinside+abovem2ndashsizeclass3 LWD3 015 118 0 1005

Fish cover

Meanlargewoodydebrisarealcover WD_DB_C 1304 1108 046 5114

Proportionoftreesandroots TREE_ROO 092 013 046 1

Proportion of undercut banks UND_BAN C S 033 034 0 1

Human impact

Proportion of human impact IMPAC S 034 035 0 1

Water quality

pH PH 485 053 345 692

Electrical conductivity (μScm) COND 002 001 001 004

Dissolved oxygen mgL DO C 76 324 298 149

Landscape Total catchment ()

Primaryforest PRI_T 6426 3957 0 100

Secondary forest SEC_T a 983 2003 0 9905

Agriculture AGR_T a 1278 2939 0 100

Pasture PAS_T a 476 1462 0 9399

Baresoil BAR_T a 721 1682 0 9378

Immediate catchment ()

Primary forest PRI_I a 6035 4365 0 100

Secondary forest SEC_I a 934 1986 0 9504

Agriculture AGR_I a 1569 3156 0 100

Pasture PAS_I a 39 132 0 6979

Baresoil BAR_I a 1072 2403 0 971

NotesBoldprintdenotesvariablesretainedafterforwardselectioninRDAsforfishassemblagecompositionandstructureandCandS intheldquoRDArdquocolumndenotewhichanalyses(compositionandstructurerespectively)thevariablewasincludedinaLandscapevariableswerereducedinPCAsbeforebeingincludedintheRDAanalysis(seeTable2)

322emsp |emsp emspensp MONTAG eT Al

PCNMeigenvectorswereextractedandusedasconditional(ietocontrolforeffectsofintersitedistance)orpredictorvariablesinsub-sequentanalysesThesignificanceofthePCNMaxeswastestedbyMoranrsquosIstatistic(Landeiroetal2011)

IndependentRDAswereconductedusingeachofthethreeenvi-ronmentaldatasets(ielocalhabitatlandscapedistance)toexplainvariationinfishassemblagecomposition(iespeciespresencendashab-sence) and structure (ie species relative abundances) Varianceinflationfactors (VIF)wereassessedto identifypotentialmulticol-linearity among environmental variables within total catchmentandimmediatecatchmentscalesValuesgt10(CurtoampPinto2010)indicated that landscape variables for both total catchment andimmediatecatchmentwerehighlyautocorrelatedThereforeprin-cipal components analysis (PCA)was used to generate compositegradients (iePC1andPC2)representing landscapeattributesforbothcatchmentscalesthatcouldbeincludedinsubsequentRDAsTosimplifytheRDAmodelsgeneratedweusedforwardselectionbasedonpermutationtests (9999randomisations)to includeonlythosevariablesthatsignificantlycontributedtovarianceexplainedTheadjustedcoefficientofmultipledetermination(adjusted-Rsup2)foreachglobalmodelwascalculatedandusedasasecondarystoppingcriterionforforwardselection(BorcardampLegendre2002)

To visualise the relationships between fish assemblages andlocalhabitatand landscapevariables selected in theRDAmodels

acombinedRDAwasconductedwithPCNMaxesincludedascon-ditional factors to control for effects of intersite distance NextvariancepartitioningwasconductedusingpartialRDA(iepRDA)to identify the unique and shared contributions of local habitatlandscapeanddistanceforexplainingfishassemblagecompositionandstructure(Peres-NetoLegendreDrayampBorcard2006)OnlyvariablesselectedintheindependentRDAanalyseswereincludedsignificantvariableswithloadingsgt|06|wereconsideredimportantandsignificanceofuniquecontributionsoflocalhabitatlandscapeanddistancevariableswasassessedusingpermutationtests

AllanalyseswereperformedusingR(RCoreTeam2013)PCNMRDAandpRDAwereimplementedusingthevegan(Oksanenetal2018) and packfor packages (Dray Legendre amp Blanchet 2011)andthefarawaypackagewasusedtocalculateVIF(Faraway2016)Significancewasassessedasple005

3emsp |emspRESULTS

31emsp|emspFish assemblages and significant environmental variables

Ninety-one fish species represented by 27733 specimens werecollected during surveys Approximately half of the species wereconsidered rare (ie 46 species with lt15 individuals collectedSupportingInformationTableS1)Sixspeciesrepresentedapproxi-mately70ofthetotalsampleabundanceThosesixspecieswereHyphessobrycon heterorhabdus (8460 specimens captured from70 sites)Microcharacidium weitzmani (4372 specimens 56 sites)Apistogrammagrregani (2405specimens59sites)Copella arnoldi (1975specimens50sites)Iguanodectes rachovii(1170specimens49sites)andPyrrhulina aff brevis(1034specimens67sites)

Only6ofthe26localhabitatvariableswereselectedforinclu-sion in the assemblage composition RDA and eight local habitatvariableswereselectedfortheassemblagestructureRDA(Table1)Surprisingly noneof the variables associatedwith substrate floworwoodydebriswereincludedineitherRDAAllchannelmorphol-ogyvariableswereselectedineitherthefinalmodelforassemblagecomposition or structure (Table1)Other variables included in ei-thermodelwerethreevariablesassociatedwithriparianvegetation(standarddeviationofcanopydensitymeanareacoveredbyherba-ceousvegetationlayergt05cmandmeanareaofwoodycover)onefish cover variable (proportion of undercut banks) human impactanddissolvedoxygenconcentration(Table1)

All landscapevariableshadhighamong-sitevariation (Table1)andPCAyieldedsimilarpatternsforthetotal(T)andimmediate(I)catchmentscalesAt the totalcatchmentscale the first twoprin-cipal components explained 831 of the variation in landscapefeatures(Table2)PC1_T(626varianceexplained)waspositivelyassociatedwithprimaryvegetationcoverandnegativelyassociatedwithagriculturewhereasPC2_T (205)waspositivelyassociatedwith secondaryvegetationandnegativelyassociatedwithagricul-tureFortheimmediatecatchmentscalePC1_IandPC2_Iexplained835ofthetotalvariation(Table2)PC1_I(641)waspositively

TABLE 2emspResultsoftheprincipalcomponentanalysisforlandscapevariablesLoadingsindicatethecontributionofeachvariabletothefirstandsecondprincipalcomponentsSuperscriptldquoCrdquoindicatesaxesthatwereimportantforfishcompositiondataandsuperscriptldquoSrdquoforfishstructure

Landscape variables Abbreviation

Loading

PC1_T PC2_TC

Totalcatchment

Primary forest PRI_T 085 minus025

Secondaryforest SEC_T minus016 049

Agriculture AGR_T minus047 minus075

Pasture PAS_T minus008 023

Baresoil BAR_T minus012 026

explained 626 205

Eigenvalues 21215 6966

PC1_ICS PC2_IC

Immediatecatchment

Primary forest PRI_I 084 020

Secondaryforest SEC_I minus014 minus024

Agriculture AGR_I minus046 072

Pasture PAS_I minus002 minus006

Baresoil BAR_I minus021 minus061

explained 641 194

Eigenvalues 26317 7983

Notes Variablesinboldareimportantforprincipalcomponentinterpretation

emspensp emsp | emsp323MONTAG eT Al

associatedwithprimaryvegetationandnegativelyassociatedwithagricultureandPC2_I(194)waspositivelyassociatedwithagricul-tureandnegativelyassociatedwithsecondaryvegetationForwardselection included both PC axes from the immediate catchmentscale(PC1_IandPC2_I)andPC2_TfromthetotalcatchmentscaleintheRDAforassemblagecompositiononlyPC1_Ifromtheimme-diatecatchmentwasselectedintheRDAforassemblagestructure

Both PCNM axes were significant (Moranrsquos I p-value lt0001)andexplained659of thevariation in the intersitedistancema-trix (PCNM1387varianceexplainedMoranrsquos I=083PCNM2272MoranrsquosI=072)OnlyPCNM1wasselectedforinclusioninthe finalmodel for theRDAforassemblagecompositionwhereasbothPCNMaxeswereincludedintheRDAforassemblagestructure(SupportingInformationTableS2)

F IGURE 2emspRedundancyanalysisbiplotsforfishassemblagecomposition(aandb)andstructure(candd)PCNMaxeswereusedasconditioningvariablestocontrolforspatialrelationshipspriortoexplainingvariationattributedtoinstreamhabitatandlandscapevariablesBiplotsshowRDAscoresfor71streamsspanningagradientofforestcoverthatrangesfromyellowrepresentingnonforesttogreenrepresentingadenselyforestedinwatershedSpecieswithgoodnessoffitlowerthan02wereomittedfromtheplotforeaseofinterpretationSomespeciesabbreviationsinplots(b)and(d)wereremovedtoimprovelegibilityseeSupportingInformationTableS1forspeciesRDAaxisloadingsCodesforenvironmentalvariablesandspeciesaregiveninTables1and2respectively

324emsp |emsp emspensp MONTAG eT Al

32emsp|emspFish assemblagendashenvironment relationships and variance partitioning

The first two RDA axes (F = 284 df = 9 plt0001) constrainedby local and landscape variables and conditioned by distance ac-counted for 169 of the total variation in fish assemblage com-positionamong sitesAxis1explained108of thevariationandwasassociatedwithSDofincisionheightPC1_Iandproportionofundercutbanks (Figure2a Supporting InformationTableS3)Axis1wascorrelatedwithagradientof forest cover (Pearsoncorrela-tion=minus068)withhigh forestcover sitesassociatedwithnegativescores and low forest cover siteswith positive scores (Figure2a)Axis 2 explained 606 of the variation andwas associatedwithproportion of undercut banks width-to-depth ratio and depth(Figure2a)Theprimaryfishspeciesassociatedwithforestedsitesand greater variation in bank incision height and greater propor-tion of undercut banks were Denticetopsis epa Brachyhypopomus sp2Characidium cf etheostomaCrenicichlagrsaxatilis Erythrinus erythrinusGymnotusgrcoropinae and Ggrpantherinus(Figure2bSupportingInformationTableS4)

For assemblage structure (species relative abundances) the firsttwoRDAaxes(F = 436 df = 9 plt0001)explained272ofthevari-ation The first axis accounted for 148 of the variation explainedand similar to the assemblage composition analysis was negativelycorrelatedwith forest cover (Pearsoncorrelation=minus073Figure2b)IncisionheightSDproportionofundercutbanksandPC1_Iwereas-sociatedwithnegativevaluesonAxis1andhighforestcoverwhereashuman impact loaded positively on the first axis and is associatedwith low forest cover (Figure2c Supporting Information Table S3)Depth was positively associated with Axis 2 Species distinguishingassemblagestructureofforestedsitesincludeCrenicichlagrsaxatilis Characidium cf etheostoma Erythrinus erythrinus Gymnorhamphichthys petiti Gymnotus gr coropinae G gr pantherinus and Crenicichla grsaxatilis whereas species distinguishing assemblage structure ofnonforested sites with human impacts were Hemigrammus ocelliferPotamoglanis hasemaniNannacara cf taenia Carnegiella strigata and Bunocephalus coracoideus(Figure2dSupportingInformationTableS4)

Local habitat variables explained greater proportions of uniquevariation in assemblage composition and structure than either land-scapevariablesorfluvialdistance(Figure3)Specificallylocalhabitat

factorsaloneexplained8ofthevariationinassemblagecomposition(F = 211 plt0001)and14ofthevariationinassemblagestructure(F = 276 plt0001) Variance explained by landscape and distancevariablesaloneneverexceed3whereasthesharedcomponentsofvarianceexplainedweresimilartothatoflocalhabitatalone(Figure3SupportingInformationTableS5)Variationexplainedbylocalhabitatandfluvialdistancetogetherwas4and8forassemblagecomposi-tionandstructurerespectivelyandthesharedvariationexplainedbyallthreepredictorcategoriestogetherwasanadditional4forassem-blagecompositionand7forstructure(Figure3)

4emsp |emspDISCUSSION

AsexpectedfishassemblagesofeasternAmazonianstreamsweresignificantly associatedwith both landscape representedmainlyby per cent forest cover and local habitat variables Howeverfishassemblagesweremorestronglyrelatedtolocalhabitatthanlandscape features Similarly Terra Hughes and Arauacutejo (2016)and Roa-Fuentes and Casatti (2017) found that fish assemblagestructurewasmore influencedby local variables thancatchmentandspatialvariables Inaseparatestudy involvingdifferentsiteswithinourstudybasinLealetal(2017)alsofoundthatfishassem-blagesweremorestronglyassociatedwith localhabitatvariablesthan landscapefeatures Incontrasttoreportsofstrongassocia-tionsbetweencatchmentcharacteristicsinstreamhabitatandfishassemblages in other regions (eg LorionampKennedy 2009)wefound relativelyweak relationships between landscape and localhabitatvariableswith fishassemblagecompositionandstructure(lt15ofvariationinassemblagecompositionorstructurewasex-plainedbyeachsetofvariables)Inclusionofthevariableldquointersitedistancerdquointhemodelnearlydoubledthepercentageofvariationexplainedbysetsoflandscapeorlocalhabitatvariables(Figure3)Similar findings were reported from other stream fish studies(eg Grenouillet Pont amp Heacuterisseacute 2004 Magalhatildees Batalha ampCollares-Pereira 2002) Furthermore approximately half of thevariance explained for assemblage composition and structure isshared among the combination of instream habitat and distanceplusall threepredictor-variablecategories (distance localhabitatandlandscape)

F IGURE 3emspVenndiagramsummarisingvariancepartitioning(pRDA)amonglocalhabitatlandscapeanddistancevariablesforfishassemblagecomposition(a)andstructure(b)indicatesplt005forthetestablecomponents(uniquecontributionsoflocalhabitatlandscapeanddistance)

emspensp emsp | emsp325MONTAG eT Al

Basedonpreviousresearchwehypothesisedthatlandscapefac-torswouldindirectlyaffectfishassemblagesthroughinfluencesonlocalhabitatLandcovervariedalongagradientrangingfrom100primaryvegetationtodeforestedagriculturalareasTerrestrialland-scapesinfluenceprocessessuchasrainwaterretentionandinputsofsedimentsandallochthonousfoodresourcesinsmallstreamssuchasthoseinthisstudy(Lealetal2017)Rainwaterretentiondependsonhumiditytemperatureandsoilporosityallofwhicharestronglyinfluenced by vegetation cover Many energy sources supportingtropicalstreamfoodwebs(egterrestrialarthropodsleavesseedsfruits)originatefromriparianvegetation(BrejatildeoGerhardampZuanon2013)Landcoverwasonlyweakly(thoughsignificantly)correlatedwithassemblagecompositionandstructureandstandarddeviationofincisionheightandproportionofundercutbanksexhibitedsimi-larrelationshipsintheRDAsbutwithgreatervariationexplainedinpartitioninganalysesTheseresultssupportourexpectationthattherelationship between land cover and fish assemblage compositionandstructureisprobablyderivedfromindirecteffectsoflandscapeon local habitat (including the shared variance explained by land-scapeandlocalfactors)

Thestrengthoftherelationshipbetweenlandcoverandfishas-semblagesishighlyvariableamongregions(egTiburcioCarvalhoFerreiraGoiteinampRibeiro2016)Over the last fewdecades theAmazon region has undergone changes in forest cover with im-pactstobothterrestrialandaquaticecosystems(Juenetal2016Lauranceetal2014)IntheeasternAmazonthereissomeevidencethat certain features of stream ecology appear to be maintaineddespite impacts from deforestation (Ferreira etal 2018) In ourstudymostfishspecieswerenotstronglyassociatedwithlandscapevariables initiallysuggestingsomedegreeofresiliencetoland-useimpactsatthecatchmentscale(UiedaampPinto2011)Forexamplestreamnetworkconnectivitycouldallowspecieswithhighdisper-salcapacitytopersistindegradedhabitatsviamasseffects(PerkinampGido2012)Thecontributionofdistancetothesharedvarianceexplainedbylocalhabitatandallthreepredictorcategoriestogetherprovides some support for this interpretation but there was nosharedvariationexplainedbydistanceandlandscapetogether(iewithout inclusionof local instreamhabitat) and thereforehabitatfilteringispresumedtoplayadominantroleinstructuringthefishassemblagesOneplausibleexplanationforthelimitedrelationshipbetweenlandscapevariablesandfishassemblages isthepresenceofprimaryandsecondaryforestwithinlocalcatchmentsorriparianzonesForestcoverintheriparianzonecouldminimisetheimpactsof reduced forestcover in thecatchmenton instreamhabitatandfish assemblages (Terra etal 2016)Aminimumwidthof riparianbufferasrequiredbyBrazilianlawwaspresentatmostofoursur-vey siteswhichmay havemitigated the negative effects of land-usechangesonlocalcatchmentsHowevertheprimaryaxisinRDAanalyseswasessentiallyagradientofprimaryforestcoverandthelocal habitat variables selected in the RDA are likely affected bycatchmentlanduseandhumanimpactThusdeforestationappearstobeindirectlyaffectingstreamfishassemblagesthroughalterationofinstreamhabitateven thoughriparianzonesaremostlyintactThis

interpretationcorroboratesfindingsbyarecentstudyconductedbyLealetal(2017)whichconcludedthattheminimumriparianbufferestablishedbyBrazilianlawwasnotsufficienttoprotectstreamfishbiodiversity

Thehabitat variableswith the strongest correlationswithpat-ternsofspeciescompositionandassemblagestructure inforestedsites were undercut banks and the standard deviation of incisionheightIncontrastsiteswithmostlydeforestedcatchmentstendedtobedeeper(withlowvariationinincisionheight)andhavegreaterpercentofhumanimpactRiparianvegetationpromotessteeperandmorestablebankswithundercutsthatprovidecoverforaquaticor-ganisms(FlorsheimMountampChin2008)IncisionheightvarianceisanimportantmetricofchannelmorphologyaswellasanindicatorofstreambeddegradationHigh incisionheightvariancecanresultfromnaturalerosionprocesses(DuncanGoodloeMeyerampProwell2011) but couldalso reflect recenterosion in response tohumanimpacts(egdischargeofurbanstormwaterorrapidrun-offfroma degraded catchment) (Rogger etal 2017) Low incision heightvariancehasbeenassociatedwithstreamchannelsimplificationandlow habitat complexity (Roni Pess Beechie amp Hanson 2014) Inourstudyhighervaluesofthestandarddeviationofincisionheightwereassociatedwithgreaterforestcoversuggestingthatthesefor-estedareasmaysupportnaturalprocessesoferosionanddeposi-tionthatpromotestreamchannelcomplexity(egcausedbyahighfrequencyofsmall-scaledisturbancessuchas individualtreefalls)Lowincisionheightvariationandgreaterdepthindeforestedsitessuggestthatdeforestationaffectsstreamfishassemblagesthroughlossofgeomorphiccomplexityduetoerosionanddown-cutting

The fish species that were strongly associated with under-cut banks and forested areas were those that inhabit structur-ally complex microhabitats (eg Characidium cf etheostomaErythrinus erythrinusGymnotusgrcoropinaeGgrpantherinus and Brachyhypopomus sp 2) Gymnotiformes are nocturnal fishes thatoften take refuge in undercut banks duringdaytime (Brejatildeo etal2013MaximeampAlbert2009)Becausetheyarestronglyreliantontheirelectrosensorysystemgymnotiformesareparticularlysensi-tive to changes inwater quality and have been promoted for useas bioindicators (Thomas Flroion ampChretien 1998) CharaciformfishesofthegeneraCharacidium and Erythrinusarediurnalbenthicspeciesthatgenerally inhabit leafpacks twigsandotherstructur-ally complexmicrohabitats fromwhere they ambush prey (Brejatildeoetal2013)ConsequentlythesespeciesbenefitfromaccumulationofdebrisfromriparianforestsinstreamchannelsFishespositivelyassociatedwithnon-forestedareasweremostlydiurnalomnivorousspeciesSomeofthesespeciesarebenthic(egBunocephalus cora-coideus and Microcharacidium weitzmani) andothersswimactivelywithin the water column (eg Carnegiella strigata Copella arnoldiHemigrammus ocellifer Laimosemion cf strigatus and Nannacara cf taeniaBrejatildeoetal2013)

Direct human influence on local habitat affected fish assem-blage structure and was negatively associated with forest coverExpansionofroadnetworksdirectlyincreasesstreamchannelero-sionandsedimentationpromotesfurtherexpansionofagriculture

326emsp |emsp emspensp MONTAG eT Al

andaccess to streamsanddisrupts connectivity (Lealetal2016Leitatildeoetal2018WantzenampMol2013)Disruptedconnectivitylikelyaffects thepotential fordispersal and the importanceof in-tersite distanceon assemblage composition and structureWithinourstudyareanon-forestedareasusuallywereassociatedwithdirtroads Inadditiontorun-offvehiclesandwindsuspenddustfromthedirtroadnetworkanddepositionofsedimentinstreamsmod-ifieshabitatandecologicaldynamics(CrokeampMockler2001)Forexampleexcessivesedimentationandsiltingreducetheavailabilityofstructurallycomplexhabitatforbenthicfishesresultinginlocalassemblagesdominatedbyhabitatgeneralists(Brejatildeoetal2013)

A recent studybyBrejatildeoetal (2018) in thewesternAmazonfound that many stream fishes exhibited significant negativethreshold responses to low levels of catchment deforestationwithin justafewyearsofthe impactwhereaspositiveresponsesbygeneralist speciesoccurredmanyyears afterdeforestation Inadditionpast land-usechangecouldhavea legacyeffectoncon-temporarybiodiversitythustheremaybetimelagsforobservableresponses to impacts (Iwata Nakano amp Inoue 2003 Zeni etal2017)ThusanimportantconsiderationisthetemporalsequenceofanthropogenicimpactsForexamplethelackofstrongrelation-shipsbetweencatchment landuse and fish assemblage composi-tionand structure in this studycouldbepartlydue toa time laginresponseofinstreamhabitattochangesinlanduseandoralaginassemblageresponsetoinstreamhabitatchangeSuchtimelagscouldmaskimpactsofland-usechangeonstreamfishassemblagessuchthatinterpretationsofanalysesusingcontemporarylandscapefeatures(egdeforestedcatchments)suggestresilienceoffishas-semblagestolandscapechangewheninfactthetimescalewastooshorttodetectaresponse (Brejatildeoetal2018 Iwataetal2003Zenietal2017)Thuslonger-termstudiesmayberequiredtoelu-cidateeffectsofland-usegradientsonassemblagestructureespe-ciallyifcatchmentsaresubjectedtovariableperiodsandtypesoflandusesandifspecialisttaxahavealreadybeenextirpated(Zenietal2017)

Given the strong interrelationships between terrestrial andaquatichabitats(TambosiVidalFerrazampMetzger2015)manage-mentofriparianvegetationhasbecomethemainstrategytomain-taintheecologicalqualityandintegrityofstreamecosystems(WhiteampGreer2006)ForthisreasonBrazilrsquosforestcodedesignatedripar-ianzonesthroughoutthecountryasPermanentPreservationAreasHoweverinareassuchastheAmazonwhereagro-industryactivi-tiesarerapidlyexpandingthelegallyestablishedlimitsforriparianzonesareoftendisrespected(BarakaampKatz2015)Ourresultsindi-catethatalthoughmoststreamsinthestudyareapresentrelativelyintactriparianzonesevenacrossdifferentlevelsofdeforestationinthebasinfocusingsolelyontheconservationofriparianvegetationmaybeinsufficientformaintainingthebioticcomponentsofstreams(Lealetal2017)Itisalsoimportanttoconsiderthedendriticnatureof stream systems and fluvial connectivity (ie intersite distancecontributedtosharedvarianceexplainedbylocalhabitatandland-scapevariablesinourstudy)aswellasthecatchmentlandscapeForexampledeforestationincreasesthevulnerabilityofriparianzones

toedgeeffectscompromisingtheintegrityofthestreamasawhole(Heartsill-ScalleyampAide2003)

Although theAcaraacuteandCapim rivercatchmentshaveexperi-encedmoderate deforestation there are stillmany streamswithintact riparian vegetation several large intact forest fragmentsandhigh-qualityhabitatsupportingdiversefreshwaterfishassem-blagesHoweverthisfindingdoesnotconsiderpotentialtimelagsor legacy effects (Harding Benfield Bolstad Helfman amp Jones1998Leitatildeoetal2018Zenietal2017)andlandcoverchangeinthebasin is rapidandevolving (egsecondaryforestandnewagricultural crops such as sugarcane likelywill expand over thecomingyears)Thishighlightsachallengeandanopportunity forunderstandingeffectsofdeforestationand landcoverchangeonthe integrity of Amazonian stream systems The challenge is theneed for long-termmonitoring of landscape attributes instreamconditions essential formaintaining biodiversity and species dis-tributionsandabundancestoaddresstimelagsandlegacyeffectsFor relatively unimpacted areas initiating long-term monitoringnowshouldbeviewedasanopportunitytogeneratedataontherelationshipsbetween landscape instreamconditionsandstreamdiversitytoserveasbaselinedataforcontinuingresearchBaselinedata that represent relatively ldquopristinerdquo conditions are lacking inmost instanceswhichcanunderminetheabilityofstudies to in-terpretecologicaldynamicsinresponsetolandcoverchange(Zenietal2017)Long-termmonitoringstudiesshouldincorporatekeyindicatorvariablesforhabitatevaluation (egundercutbankandSD incisionheight from this studywoodvolume as proposedbyLeal etal (2017) and Leitatildeo etal (2018))Other physical habitatvariables (channel morphology substrate water velocity sub-mergedwoodstructure riparianvegetationcoverandhuman in-fluence)shouldalsobemonitoredbecausetheyhavebeenshowntoinfluenceaquaticbiotainvariouswaysdependingontheregionand associated environmental conditions (Kaufmann etal 1999Pecketal2006)Lastlylong-termmonitoringstudiesshouldalsoincorporatedataonspatialrelationships(egthisstudy)aswellasthehistory(egZenietal2017)andrate(egBrejatildeoetal2018)oflandcoverchange

ACKNOWLEDG EMENTS

WethankCarolineCArantesandBinsongJinforhelpfuldiscussionsabouttheresearchandJoshuahPerkinforsuggestionstoimprovethemanuscriptFinancialandlogisticsupportwasprovidedbyConservationInternational of Brazil (CI-Brazil) Agropalma Group BiopalmaValeCIKELLtda33ForestCapitalUnitedStatesAgencyforInternationalDevelopment (USAID) and Conselho Nacional de DesenvolvimentoCientiacuteficoeTecnoloacutegico(CNPq)projects4493142014-2(CNPq)and1282014(FAPESPACNPq)KOWreceivedfundingfromUSNationalScienceFoundationgrantDEB1257813andtheestateofGeorgeandCarolynKelsovia the InternationalSportfishFundWethankCNPqfor graduate scholarships to HL NLB NRT and BSP and researchproductivity scholarships to LJ (process 3075972016-4) and LFAM(process3050172016-0)WearealsogratefultotheCoordenaccedilatildeode

emspensp emsp | emsp327MONTAG eT Al

AperfeiccediloamentodePessoaldeNiacutevelSuperior(CAPES)forPROCADCAPES funding (project no 888810684252014-01) the graduatescholarshipforTOBandseniorinternshipscholarshipforLFAMtocon-ductresearchatTexasAampMUniversity(process888811190972016-1) The manuscript was improved during the review process byaddressing the thoughtful comments from Philip Kaufmann and ananonymousreviewer

CONFLIC T OF INTERE S T

Theauthorshavenoconflictofintereststodeclare

AUTHORSrsquo CONTRIBUTION

LFAMKOWandLJconceivedanddesignedtheinvestigationHLBSPandTOBperformedfieldandorlaboratoryworkFWKNLBNRTandLMBanalysedthedataDESEOLandYQMcontributedmaterials reagents andor analysis tools LFAM KOW FWK andDJHwrotethepaper

ORCID

Luciano F A Montag httporcidorg0000-0001-9370-6747

Kirk O Winemiller httporcidorg0000-0003-0236-5129

Friedrich W Keppeler httporcidorg0000-0002-5165-1298

Hiacutengara Leatildeo httporcidorg0000-0002-9938-7967

Naraiana L Benone httporcidorg0000-0003-4155-9938

Naiara R Torres httporcidorg0000-0003-1446-6760

Bruno S Prudente httporcidorg0000-0003-4226-2431

Tiago O Begot httporcidorg0000-0002-4879-3869

Luke M Bower httporcidorg0000-0002-0739-858X

David E Saenz httporcidorg0000-0003-0238-0919

Edwin O Lopez-Delgado httporcidorg0000-0002-4010-1880

Yasmin Quintana httporcidorg0000-0002-1051-0202

David J Hoeinghaus httporcidorg0000-0003-0363-8723

Leandro Juen httporcidorg0000-0002-6188-4386

R E FE R E N C E S

Albert J S (2001) Species diversity and phylogenetic systematics of American knifefishes (Gymnotiformes Teleostei) Ann ArborM0049Miscellaneous Publications Museum of Zoology University ofMichigan

Allan JD (2004) Landscapes and riverscapes The influence of landuse on stream ecosystems Annual Review of Ecology Evolution and Systematics 35 257ndash284 httpsdoiorg101146annurevecolsys35120202110122

Almeida C A Coutinho A C Esquerdo J C D M Adami MVenturieriADinizCGGomesAR(2016)Highspatialreso-lutionlanduseandlandcovermappingoftheBrazilianLegalAmazonin2008usingLandsat-5TMandMODISdataActa Amazonica46291ndash302httpsdoiorg1015901809-4392201505504

BarakaBampKatzD (2015)Valuing instreamandriparianaspectsofstreamrestorationndashAwillingnesstotaxapproachLand Use Policy45204ndash212httpsdoiorg101016jlandusepol201501023

BenoneNLEspositoMCJuenLPompeuPSampMontagLFA(2017)RegionalcontrolsonphysicalhabitatstructureofAmazonstreams River Research and Applications 33 766ndash776 httpsdoiorg101002rra3137

Blaser J Sarre A Poore D amp Johnson S (2011) Status of tropical forest management 2011 (38thed)YokohamaJapan InternationalTropicalTimberOrganization

BojsenBHampBarrigaR(2002)Effectsofdeforestationonfishcom-munitystructureinEcuadorianAmazonstreamsFreshwater Biology472246ndash2260httpsdoiorg101046j1365-2427200200956x

Borcard D amp Legendre P (2002) All-scale spatial analysis of eco-logical data by means of principal coordinates of neighbour ma-trices Ecological Modelling 153 51ndash68 httpsdoiorg101016S0304-3800(01)00501-4

Brasil (1965) Forest Code Federal LawNo 477165Official FederalGazetteoftheFederativeRepublicofBrazilSeptember15th1965

Brejatildeo G L Gerhard P amp Zuanon J (2013) Functional trophiccomposition of the ichthyofauna of forest streams in easternBrazilianAmazonNeotropical Ichthyology11361ndash373httpsdoiorg101590S1679-62252013005000006

BrejatildeoGLHoeinghausDJPeacuterez-MayorgaMAFerrazSFBampCasattiL(2018)ThresholdresponsesofAmazonianstreamfishestotimingandextentofdeforestationConservation Biology32860ndash871httpsdoiorg101111cobi13061

BurchamJ(1988)FishcommunitiesandenvironmentalcharacteristicsoftwolowlandstreamsinCostaRicaRevista de Biologia Tropical36273ndash285

Cardinale B J PalmerM A Swan CM Brooks S amp Poff N L(2002) The influence of substrate heterogeneity on biofilm me-tabolism in a streamecosystemEcology83 412ndash422 httpsdoiorg1018900012-9658(2002)083[0412TIOSHO]20CO2

ChenKHughesRMBritoJGLealCGLeitatildeoRPOliveira-JuacuteniorJMBZuanonJ(2017)Amulti-assemblagemulti-metricbiologicalcondition indexforeasternAmazoniastreamsEcological Indicators7848ndash61httpsdoiorg101016jecolind201703003

CrokeJCampMocklerS(2001)Gullyinitiationandroad-to-streamlink-age in a forested catchment southeasternAustraliaEarth Surface Processes and Landforms26205ndash217httpsdoiorg1010021096-9837(200102)262lt205AID-ESP168gt30CO2-G

Curto JDampPinto JC (2010)ThecorrectedVIF (CVIF) Journal of Statistics38 1499ndash1507 httpsdoiorg101080026647632010505956

DraySLegendrePampBlanchetG (2011)Packfor Forward selection with permutation (Canoco p46)Rpackageversion00-8r100

DuncanWWGoodloeRBMeyerJLampProwellES(2011)Doeschannel incisionaffect in-streamhabitatExaminingtheeffectsofmultiplegeomorphicvariablesonfishhabitatRestoration Ecology1964ndash73httpsdoiorg101111j1526-100X200900534x

Espiacuterito-SantoHMVMagnussonWEZuanonJMendonccedilaFPamp Landeiro V L (2009) Seasonal variation in the composition offish assemblages in small Amazonian forest streams Evidence forpredictable changes Freshwater Biology 54 536ndash548 httpsdoiorg101111j1365-2427200802129x

FarawayL(2016)Functions and datasets for books by Julian Faraway R packageversionn107

FerreiraMCBegotTOPrudenteBSJuenLampMontagLFA(2018)EffectsofoilpalmplantationsonhabitatstructureandfishassemblagesinAmazonstreamsEnvironmental Biology of Fishes101547ndash562httpsdoiorg101007s10641-018-0716-4

Florsheim J LMount JFampChinA (2008)Bankerosionasade-sirable attribute of rivers BioScience 58 519ndash529 httpsdoiorg101641B580608

328emsp |emsp emspensp MONTAG eT Al

FrimpongEASuttonTMLimKJHrodeyPJEngelBASimonTP LeMasterDC (2005)ADeterminationofoptimal ripar-ianforestbufferdimensionsforstreambiota-landscapeassociationCanadian Journal of Fisheries and Aquatic Sciences62 1ndash6 httpsdoiorg101139f05-020

GardnerTA Ferreira JBarlow J LeeAC Parry LVieira ICGampZuanonJ(2013)Asocialandecologicalassessmentoftrop-icallandusesatmultiplescalesTheSustainableAmazonNetworkPhilosophical Transactions of the Royal Society of London Series B Biological Sciences 368 20120166 httpsdoiorg101098rstb20120166

GrenouilletGPontDampHeacuterisseacuteC(2004)Within-basinfishassem-blagestructuretherelativeinfluenceofhabitatversusstreamspatialpositionon localspeciesrichnessCanadian Journal of Fisheries and Aquatic Sciences6193ndash102httpsdoiorg101139F03-145

HardingJSBenfieldEFBolstadPVHelfmanGSampJonesEBDIII(1998)StreambiodiversityTheghostoflandusepastProceedings of the National Academy of Sciences of the United States of America9514843ndash14847httpsdoiorg101073pnas952514843

Heartsill-Scalley T amp Aide T M (2003) Riparian vegetation andstreamcondition ina tropicalagriculture-secondary forestmosaicEcological Applications13225ndash234httpsdoiorg1018901051-0761(2003)013[0225RVASCI]20CO2

INMET-InstitutoNacionaldeMeteorologia(2018)BDMEP ndash Banco de Dados Meteoroloacutegicos para Ensino e Pesquisa Available on httpwwwinmetgovbrportalindexphpr=bdmepbdmep LastAccessedApril152018

IwataTNakanoSampInoueM(2003)Impactsofpastripariandefor-estationonstreamcommunities inatropicalrainforest inBorneoEcological Applications13 461ndash473 httpsdoiorg1018901051-0761(2003)013[0461IOPRDO]20CO2

JuenLCunhaEJCarvalhoFGFerreiraMCBegotTOAndradeA L Montag L FA (2016)Effectsofoil palmplantationsonthehabitatstructureandbiotaofstreamsinEasternAmazonRiver Research and Applications322081ndash2094httpsdoiorg101002rra3050

Kaufmann P R Levine P Robison G E Seeliger C amp Peck D V(1999) Quantifying physical habitat in wadeable streams (1st ed)WashingtonDCUSEnvironmentalProtectionAgency

Landeiro V L Magnusson W E Melo A S Espiacuterito-Santo H MV amp Bini L M (2011) Spatial eigenfunction analyses in streamnetworks Do watercourse and overland distances produce dif-ferent results Freshwater Biology 56 1184ndash1192 httpsdoiorg101111j1365-2427201002563x

LauranceWFampBalmfordA(2013)LanduseAglobalmapforroadbuildingNature495308ndash309httpsdoiorg101038495308a

LauranceWFSayerJampCassmanKG(2014)AgriculturalexpansionanditsimpactsontropicalnatureTrends in Ecology and Evolution29107ndash116httpsdoiorg101016jtree201312001

LealCGBarlowJGardnerTAHughesRMLeitatildeoRPMacNallyRPompeuPS(2017)Isenvironmentallegislationconserv-ingtropicalstreamfaunasAlarge-scaleassessmentoflocalriparianandcatchment-scaleinfluencesonAmazonianfishJournal of Applied Ecology551312ndash1326httpsdoiorg1011111365-266413028

LealCG PompeuP SGardner TA LeitaoR PHughesRMKaufmannPRBarlowJ(2016)Multi-scaleassessmentofhuman-inducedchangestoAmazonianinstreamhabitatsLandscape Ecology311725ndash1745httpsdoiorg101007s10980-016-0358-x

Legendre P ampGallagher E D (2001) Ecologicallymeaningful trans-formationsforordinationofspeciesdataOecologia129271ndash280httpsdoiorg101007s004420100716

Legendre P Oksanen J amp ter Braak C J F (2011) Testingthe significance of canonical axes in redundancy analy-sis Methods in Ecology and Evolution 2 269ndash277 httpsdoiorg101111j2041-210X201000078x

LeiboldMAHolyoakMMouquetNAmarasekarePChaseJMHoopesMFampGonzalesA(2004)ThemetacommunityconceptAframeworkfor largescalecommunityecologyEcology Letters7601ndash613httpsdoiorg101111j1461-0248200400608x

LeitatildeoRPZuanonJLealCGPompeuPSGardnerTBarlowJhellipMouillotD(2018)DisentanglingthemultipleeffectsoflanduseonthefunctionalstructureoffishassemblagesinsmallAmazonstreamsEcography41219ndash232httpsdoiorg101111ecog02845

Loboacuten-CerviaacuteJMazzoniRampRezendeCF(2016)EffectsofriparianforestremovalonthetrophicdynamicsofaNeotropicalstreamfishassemblageRipariancoverandstreamdwellingfishesJournal of Fish Biology8950ndash64httpsdoiorg101111jfb12973

LorionCMampKennedyBP (2009)Riparianforestbuffersmitigatethe effects of deforestation on fish assemblages in tropical head-water streams Ecological Applications 19 468ndash479 httpsdoiorg10189008-00501

Magalhatildees M F Batalha D C amp Collares-Pereira M J (2002)Gradients in stream fish assemblages across a Mediterraneanlandscape Contributions of environmental factors and spa-tial structure Freshwater Biology 47 1015ndash1031 httpsdoiorg101046j1365-2427200200830x

Maxime E L amp Albert J S (2009) A new species of Gymnotus(GymnotiformesGymnotidae) from the Fitzcarrald Arch of south-eastern Peru Neotropical Ichthyology 7 579ndash585 httpsdoiorg101590S1679-62252009000400004

Molina M C Roa-Fuentes C A Zeni J O amp Casatti L (2017)The effects of land use at different spatial scales on instreamfeatures in agricultural streams Limnologica ndash Ecology and Management of Inland Waters6514ndash21httpsdoiorg101016jlimno201706001

NakanoSampMurakamiM(2001)ReciprocalsubsidiesDynamicinter-dependencebetweenterrestrialandaquaticfoodwebsProceedings of the National Academy of Sciences of the United States of America98166ndash170httpsdoiorg101073pnas981166

Noss R (1990) Indicators for monitoring biodiversity A hierarchi-cal approach Conservation Biology 4(3) 55ndash364 httpsdoiorg101111j1523-17391990tb00309x

OksanenJBlanchetFGFriendlyMKindtRLegendrePMcGlinnDhellipWagnerH (2018)Vegan Community ecologyRpackagever-sion24-6

PeckDVHerlihy A THill BHHughes RM Kaufmann P RKlemmDJhellipCappaertMR(2006)Environmental monitoring and assessment program-surface waters Western pilot study field operations manual for wadeable streamsEPA620R-06003WashingtonDCUSEnvironmentalProtectionAgency

PeelMCFinlaysonBLampMcMahonTA (2007)Updatedworldmap of the Koumlppen-Geiger climate classification Hydrology and Earth System Sciences 11 1633ndash1644 httpsdoiorg105194hess-11-1633-2007

Pereira R Jr Zweede J C Asner G P amp Keller M (2002)Forest canopy damage and recovery in reduced impact andconventional selective logging in eastern Para Brazil Forest Ecology and Management168 77ndash89 httpsdoiorg101016S0378-1127(01)00732-0

Peres-NetoPRLegendrePDraySampBorcardD(2006)VariationpartitioningofspeciesdatamatricesEstimationandcomparisonoffractions Ecology 87 2614ndash2625 httpsdoiorg1018900012-9658(2006)87[2614VPOSDM]20CO2

Perkin J S amp Gido K B (2012) Fragmentation alters stream fishcommunity structure in dendritic ecological networks Ecological Applications222176ndash2187httpsdoiorg10189012-03181

PrudenteBSPompeuPSJuenLampMontagLFA(2017)Effectsofreduced-impactloggingonphysicalhabitatandfishassemblagesin streams of Eastern Amazonia Freshwater Biology 62 303ndash316httpsdoiorg101111fwb12868

emspensp emsp | emsp329MONTAG eT Al

PrudenteBSPompeuPSampMontagL (2018)UsingmultimetricindicestoassesstheeffectofreducedimpactloggingonecologicalintegrityofAmazonian streamsEcological Indicators91 315ndash323httpsdoiorg101016jecolind201804020

PuseyBJampArthingtonAH(2003)Importanceoftheriparianzoneto theconservationandmanagementof freshwater fishAreviewMarine and Freshwater Research54 1ndash16 httpsdoiorg101071MF02041

RCore Team (2013)R A language and environment for statistical com-puting Vienna Austria R Foundation for Statistical ComputingRetrievedfromhttpwwwR-projectorg

ReisREKullanderSOampFerrarisCJ(2003)Check list of the fresh-water fishes of South and Central America (1st ed) Porto AlegreBrazilEdipucrs

RichardsJAampJiaX(1999)Remote sensing digital image analysis An introduction(3rded)HeidelbergBerlinSpringer-Verlaghttpsdoiorg101007978-3-662-03978-6

Roa-FuentesCAampCasattiL(2017)Influenceofenvironmentalfea-turesatmultiplescalesandspatialstructureonstreamfishcommu-nities inatropicalagriculturalregionJournal of Freshwater Ecology32281ndash295httpsdoiorg1010800270506020171287129

RoggerMAgnolettiMAlaouiABathurstJCBodnerGBorgaM Bloumlschl G (2017) Land use change impacts on floods atthe catchment scale Challenges and opportunities for future re-search Water Resources Research 53 5209ndash5219 httpsdoiorg1010022017WR020723

RoniPPessGRBeechieTJampHansonKM(2014)Fish-habitat re-lationships and the effectiveness of habitat restoration(1sted)SeattleWANOAATechnicalMemorandum

SweeneyBWBottTLJacksonJKKaplanLANewboldJDStandleyLJampHorwitzRJ(2004)Ripariandeforestationstreamnarrowing and loss of stream ecosystem services Proceedings of the National Academy of Sciences of the United States of America10114132ndash14137httpsdoiorg101073pnas0405895101

Tambosi L R Vidal M M Ferraz S F B ampMetzger J P (2015)Funccedilotildees eco-hidroloacutegicas das florestas nativas e o CoacutedigoFlorestalEstudos Avanccedilados29151ndash162httpsdoiorg101590S0103-40142015000200010

TeresaFBampCasattiL(2012)Influenceofforestcoverandmesohab-itattypesonfunctionalandtaxonomicdiversityoffishcommunitiesinNeotropicallowlandstreamsEcology of Freshwater Fish21433ndash442httpsdoiorg101111j1600-0633201200562x

Teresa F B Casatti L amp Cianciaruso M V (2015) Functional dif-ferentiation between fish assemblages from forest and defor-ested streams Neotropical Ichthyology 13 361ndash370 httpsdoiorg1015901982-0224-20130229

Terra B D F Hughes RM amp Arauacutejo F G (2016) Fish assemblagesinAtlanticForest streamsThe relative influenceof local andcatch-ment environments on taxonomic and functional speciesEcology of Freshwater Fish25527ndash544httpsdoiorg101111eff12231

ThomasMFlroionAampChretienD(1998)Anewwarningbiomonitorusingaweaklyelectric fishApteronotus albifrons (Gymnotiformes)and the effect of temperature on the bioelectric responsesEnvironmental Monitoring and Assessment51605ndash620httpsdoiorg101023A1005700519636

TiburcioGSCarvalhoCSFerreiraFCGoiteinRampRibeiroMC (2016) Landscape effects on the occurrence of ichthyofauna

in first-order streams of southeastern Brazil Acta Limnologica Brasiliensia28e2httpsdoiorg101590S2179-975X2515

TohamAKampTeugelsGG (1999)Firstdataonan IndexofBioticIntegrity (IBI)basedonfishassemblagesfortheassessmentoftheimpact of deforestation in a tropical West African river systemHydrobiologia39729ndash38httpsdoiorg101023A100360580187

Uieda V S ampCastro RM C (1999) Coleta e fixaccedilatildeo de peixes deriachos InEPCaramaschiRMazzoniampPRPeres-Neto (Eds)Ecologia de Peixes de Riachos (pp 1ndash22) Rio de Janeiro BrazilPPGE-UFRJ

Uieda V S amp Pinto T L F (2011) Feeding selectivity of ichthyo-fauna in a tropical stream Space-time variations in trophic plas-ticity Community Ecology 12 31ndash39 httpsdoiorg101556ComEc12201115

Van der Seen P amp Albert J S (2018)Field guide to the fishes of the Amazon Orinoco amp GuianasOxfordshireUK PrincetonUniversityPressPrincetonandOxford

WantzenKampMolJ(2013)SoilerosionfromagricultureandminingA threat to tropical stream ecosystems Agriculture 3 660ndash683httpsdoiorg103390agriculture3040660

WatrinOSampRochaAMA(1992)Levantamento de vegetaccedilatildeo natural e uso da terra no Municiacutepio de Paragominas (PA) utilizando imagens TMLandsatBeleacutemBrazilEMBRAPA-CPATU

WhiteMDampGreer KA (2006) The effects ofwatershed urban-ization on the stream hydrology and riparian vegetation of LosPentildeasquitos Creek California Landscape and Urban Planning 74125ndash138httpsdoiorg101016jlandurbplan200411015

WinemillerKOFleckerASampHoeinghausD J (2010)Patchdy-namicsandenvironmentalheterogeneityinloticecosystemsJournal of the North American Benthological Society29 84ndash99 httpsdoiorg10189908-0481

Zeni J O Hoeinghaus D J amp Casatti L (2017) Effects of pastureconversion to sugarcane for biofuel production on stream fish as-semblagesintropicalagroecosystemsFreshwater Biology622026ndash2038httpsdoiorg101111fwb13047

ZhuRLiOWangWChuLampYanY(2017)Effectsoflocalriver-networkandcatchmentfactorsonfishassemblagesintheheadwa-terstreamsoftheXinrsquoanbasinChinaJournal of Freshwater Ecology32309ndash322httpsdoiorg101080027050602016127840

SUPPORTING INFORMATION

Additional supporting information may be found online in theSupportingInformationsectionattheendofthearticle

How to cite this articleMontagLFAWinemillerKOKeppelerFWetalLandcoverriparianzonesandinstreamhabitatinfluencestreamfishassemblagesintheeasternAmazonEcol Freshw Fish 201928317ndash329 httpsdoiorg101111eff12455