Ecology and Evolution. 2019;00:1–16.  | 1www.ecolevol.org

Received:1May2019  |  Revised:16July2019  |  Accepted:19July2019DOI: 10.1002/ece3.5538

O R I G I N A L R E S E A R C H

Old divergence and restricted gene flow between torrent duck (Merganetta armata) subspecies in the Central and Southern Andes

Luis Alza1,2,3  | Philip Lavretsky4  | Jeffrey L. Peters5 | Gerardo Cerón6 | Matthew Smith3 | Cecilia Kopuchian7,8 | Andrea Astie9 | Kevin G. McCracken1,2,3,10,11

1DepartmentofBiology,UniversityofMiami,CoralGables,FL,USA2DivisióndeOrnitología,CORBIDI,Lima,Peru3InstituteofArcticBiology,DepartmentofBiologyandWildlife,UniversityofAlaskaFairbanks,AK,USA4DepartmentofBiologicalSciences,UniversityofTexasatElPaso,ElPaso,TX,USA5DepartmentofBiologicalSciences,WrightStateUniversity,Dayton,OH,USA6LaboratoriodeZoología‐CRUB,UniversidadNacionaldelComahue,Bariloche,Argentina7CentrodeEcologíaAplicadadelLitoral(CECOAL‐CONICET),Corrientes,Argentina8DivisiónOrnitología,MuseoArgentinodeCienciasNaturales(MACN‐CONICET),BuenosAires,Argentina9InstitutoArgentinodeInvestigacionesdelasZonasÁridas(CCTMendoza‐CONICET),Mendoza,Argentina10RosenstielSchoolofMarineandAtmosphericSciences,UniversityofMiami,CoralGables,FL,USA11UniversityofAlaskaMuseum,UniversityofAlaskaFairbanks,Fairbanks,AK,USA

ThisisanopenaccessarticleunderthetermsoftheCreativeCommonsAttributionLicense,whichpermitsuse,distributionandreproductioninanymedium,providedtheoriginalworkisproperlycited.©2019TheAuthors.Ecology and EvolutionpublishedbyJohnWiley&SonsLtd.

CorrespondenceLuisAlza,DepartmentofBiology,UniversityofMiami,CoralGables,FL,USA.Email:luis.alza@corbidi.org

Funding informationAgenciaNacionaldePromociónCientíficayTecnológica,Grant/AwardNumber:PICT‐2016‐3712;NationalScienceFoundation,Grant/AwardNumber:NSF0949439;ConsejoNacionaldeInvestigacionesCientíficasyTécnicas,Grant/AwardNumber:FondoiBolArgentinaD3657andPIP11220130100803CO

AbstractAim: Toinvestigatethestructureandrateofgeneflowamongpopulationsofhabitat‐specializedspeciestounderstandtheecologicalandevolutionaryprocessesunder‐pinning theirpopulationdynamicsandhistoricaldemography, including speciationandextinction.Location: PeruvianandArgentineAndes.Taxon: Twosubspeciesoftorrentduck(Merganetta armata).Methods: Wesampled156individualsinPeru(M. a. leucogenis;ChillónRiver,n = 57 andPachachacaRiver,n = 49) andArgentina (M. a. armata; ArroyoGrandeRiver,n=33andMalargüeRiver,n=17),andsequencedthemitochondrialDNA(mtDNA)controlregiontoconductcoarseandfine‐scaledemographicanalysesofpopulationstructure.Additionally, to test fordifferencesbetweensubspecies,andacrossge‐neticmarkerswithdistinctinheritancepatterns,asubsetofindividuals(Peru,n = 10 andArgentina,n=9)wassubjectedtopartialgenomeresequencing,obtaining4,027autosomaland189Z‐linkeddouble‐digestrestriction‐associatedDNAsequences.Results: Haplotype and nucleotide diversitieswere higher in Peru thanArgentinaacrossallmarkers.PeruvianandArgentinesubspeciesshowedconcordantspecies‐leveldifferences(ΦSTmtDNA=0.82;ΦSTautosomal=0.30;ΦSTZchromosome=0.45),

2  |     ALZA et AL.

1  | INTRODUC TION

Establishing how populations are structured is fundamental forunderstandingevolutionaryprocesses(Hartl&Clark,2007;Hey&Machaco,2003;Ma, Ji,&Zhang,2015;Wright,1969).Aspopula‐tions subdivide through time, variation in the rateof geneticdriftand gene flow, as well as selective pressures, will define the ge‐neticdiversity,divergence,orextinctionofeachspecies(Bowler&Benton,2005;Frankham,2005;Hey,2010;Lenormand,2002;Maetal.,2015;Shaffer,1981;Slatkin,1987).Amongtypesofspecies,habitatspecialistsareparticularlypronetoincreasedlevelsofpopu‐lationstructureasaresultofisolationonpatchyhabitats(Kawecki&Ebert,2004;Kaweckietal.,2012;Nei,2013;Orr,2005).Inaddition,patchyhabitatsoftenharborsmallpopulationswithlowgeneticdi‐versity,whichmay reduce theeffectivenessof selectionandcon‐tributetohighratesof localextinction(Charlesworth,2010;Hartl&Clark,2007).Thevulnerability toextinctionby isolatedpopula‐tions alsooccursbecauseof lackof suitablehabitat, environmen‐tal pressures (e.g., development, predation, climate change), and/orstochasticperturbations(i.e.,survivalandreproductivesuccess,habitat variation, genetic drift, catastrophes) (Callaghan, 1997;Fahrig,2013;Naranjo&Ávila,2003;Shaffer,1981;Soulé&Mills,1998).Therefore, characterizingpopulationstructureand the rateofgeneflowamongpopulationsofhabitatspecialists iscrucialforunderstanding the ecological and evolutionary processes under‐pinning population dynamics and historical demography, includingspeciation andextinction (Hartl&Clark, 2007; Lenormand,2002;McCauley,1991;Neigel,1997;Slatkin,1993).

Thetorrentduck (Merganetta armataGould,1842) isaspecial‐ized waterfowl of fast‐flowing whitewater mountain rivers thatoccupies12%oftheapproximately1,142rivers(between500and4,100ma.s.l.)onbothslopesoftheworld'slongestmountainrange,theAndes(eBird,2016;Fjeldså&Krabbe,1990).Acrosstheirrange,torrentducksdisplayplumagecolorationandbodysizedifferences

(Fjeldså & Krabbe, 1990; Gutiérrez‐Pinto et al., 2014; Johnsgard,2010;Naranjo&Ávila,2003), resulting inat least threedescribedsubspecies:M. a. colombiana(VenezuelatoEcuador),M. a. leucogenis (Peru,Bolivia,ArgentinaandChile),andM. a. armata(ArgentinaandChile).Here,wefocusonthetwosubspeciesdistributedinthehigh‐est regionsof theAndes (Brack,2000;Burkart,Bárbaro,Sánchez,& Gómez, 1999; Capitonio, Faccenna, Zlotnik, & Stegman, 2011;Fjeldså&Krabbe,1990):specifically,thesubspeciesM. a. leucogenis (300–520g)intheCentralAndesofPeru,andthelarger(360–580g)anddarker‐coloredsubspeciesM. a. armata intheSouthernAndesofArgentina(Johnsgard,2010).Althoughthesubspeciesclassifica‐tionissupportedbymorphologicaldifferences,geneticinformationthatsupportsthisclassificationislacking.Furthermore,itisunclearwhether thesesubspeciesweredifferentiatedbyancientvicariantevent(e.g.,upliftoftheAndes,desertificationetc.)oracontempo‐rary founder event (e.g., post glaciation). Additionally, the extentofgeneflowamongmanypatchilydistributedriverinepopulationsacrosstheAndesremainsunknown.

Intheriversoccupiedbytorrentducks,theadultsshowahighbreedingsitefidelity(Alzaetal.,2017;Cardona&Kattan,2010)toterritoriesthatreportedlyrangefrom0.7to2kmduringthebreed‐ing season (Cerón & Trejo, 2009; Colina, 2010; Eldridge, 1986;Johnsgard, 2010; Moffet, 1970; Naranjo & Ávila, 2003; Ubeda,Cerón,&Trejo,2007).However,immatureandadultindividualshavebeenreportedmovingawayfromriverslikelysearchingfornewhab‐itats(i.e.,fouropportunisticrecordsofmale‐biaseddispersal;Cerón&Capllonch,2016).Thusgiventhattorrentduckpairsareterritorialwithahighrateofriver‐specificsitefidelity,weexpecttorrentduckstobemoregeneticallysimilaramongthenearestriversbutisolatedbydistanceamongriversdependingonthedispersalcapacityoftheducks (Hanski&Gilpin, 1991;Harrison&Hastings, 1996;Nielsen&Slatkin,2013;Wright,1943).Alternatively,differentriverinesub‐populationsmayfollowametapopulationmodel(Hanski&Gaggiotti,2004;Levins,1968),inwhichpopulationsorsubpopulationsisolated

including no shared mtDNA haplotypes. Demographic parameters estimated formtDNAusing IMand IMa2 analyses, and for autosomalmarkers using∂a∂i (isola‐tion‐with‐migrationmodel),supportedanolddivergence (mtDNA=600,000yearsbefore present (ybp), 95%HPD range = 1.2Mya to 200,000 ybp; and autosomal∂a∂i=782,490ybp),betweenthetwosubspecies,characteristicofdeeplydivergedlineages.Thepopulationswerewell‐differentiatedinArgentinabutmoderatelydif‐ferentiatedinPeru,withlowunidirectionalgeneflowineachcountry.Main conclusions: WesuggestthattheSouthAmericanAridDiagonalwaspreexist‐ingand remainsacurrentphylogeographicbarrierbetween the rangesof the twotorrentducksubspecies,andtheadultterritorialityandbreedingsitefidelitytotheriversdefinetheirpopulationstructure.

K E Y W O R D S

Andes,geneflow,geneticdiversity,Merganetta armata,populationstructure,timesincedivergence

     |  3ALZA et AL.

inside watersheds are still connected by limited gene flow, thatwouldplayakeyroleinrecolonization(sporadicdispersal)aftertheextinctionofasubpopulation.

Here, we reconstruct population structure for mitochondrialDNA (mtDNA) and thousands of nuclear markers isolated usingpartialgenomeresequencing(i.e.,double‐digestrestrictionsite‐as‐sociatedDNAsequencing (ddRAD‐Seq)), attained fromaltitudinalsampling transects in Peru and Argentina. Specifically, we esti‐matedthetimesincedivergenceandgeneflowbetweenthetwosubspecies, andalsocharacterizedpopulationstructureandgeneflowbetweenriverinepopulationswithinsubspecies.Furthermore,we calculated genetic diversity across the different populationsto infer whether specific riverine populations have experienceddemographicprocesses such aspopulationexpansionor contrac‐tion.Becausethespatiallinearconfigurationoftheriversrestrictssuitablehabitat and constrains thepopulation,we also examinedevidence for altitudinal clines in mtDNA haplotype frequency.Finally, this genetic characterization across altitudinal gradientspermitsustoexplorewhethergeneticvariationisassociatedwithrecentlydescribedmorphologicalandphysiologicaldifferencesoftorrent ducks at high elevations, including changes in body size(Gutiérrez‐Pinto et al., 2014), function of key enzymes (Dawsonet al., 2016), changes in insulative properties of feathers (Cheek,Alza,&McCracken,2018),andchangesinthehypoxicventilatoryresponse(Ivyetal.,2019).

2  | MATERIAL S AND METHODS

2.1 | Sampling and DNA extraction

Blood samples were collected from 156 captured and releasedtorrent ducks, during the 2010–2011 dry seasons, from two riv‐ers in the western and eastern slopes of the Central Andes ofPeru (ChillónRiver,n = 57 andPachachacaRiver,n = 49; Table 1andFigure1b,c),andtworivers intheeasternslopeofthesouth‐ernAndesofArgentina(ArroyoGrandeRiver,n=32andMalargüeRiver,n=18;Table1;Figure1b,c).Eachriverwassurveyedalonganaltitudinaltransect(maximumrange,900–4,000m)usinganactivemistnetmethodbywhichtorrentducksweredriventonets(Alzaetal.,2017).Bloodsamples(3mlofwholebloodperindividual)werestoredinliquidN2inthefieldbeforebeingplacedinlong‐termstor‐ageat−80°C.BloodsamplesarearchivedattheUniversityofAlaskaMuseum(Fairbanks,Alaska),MuseoArgentinodeCienciasNaturales

“Bernadino Rivadavia” (Buenos Aires, Argentina), and CORBIDI(Lima,Peru).DNAwasextractedusingaDNeasyBlood&Tissuekitand following the manufacturer's protocols (Qiagen). Extractionswere quantified using a NanoDrop 2000 Spectrophotometer(ThermoFisherScientific Inc.) toensureaminimumconcentrationof20ng/µl.

2.2 | Mitochondrial DNA

A634‐base‐pair(bp)fragment(domainsIandII)ofthemtDNAcon‐trolregion(CR)wasamplifiedandsequencedusingspecificprimersfor torrent ducks L100 (5′‐CATACATTTATGCGCCCCATAC‐3′) andH774 (5′‐CCATACACGCCAACCGTCTC‐3′) that prevented ampli‐ficationof a knownnuclear copy (K.G.McCracken, unpublished).PCRotherwise followedstandardprotocols (McCracken, Johnson,& Sheldon, 2001). Sequences were edited using Sequencer v.4.7 (GeneCodesCorporation)andalignedbyeyeusingSe‐Al v.2.0a11 (Rambaut,2007).AllsequencesaredepositedinGenBank(AccessionNumbersMN196318:MN196473).

2.3 | ddRAD‐Seq library preparation

While mtDNA was obtained across all samples, for more exten‐sivegenomicanalysisofthetwosubspecies,asubsetfromChillónRiverinPeru(n=10)andMalargüeRiverinArgentina(n=9)weresubjectedtopartialgenomeresequencingviaddRAD‐Seq.Samplepreparation for ddRAD‐Seq followed the double‐digest protocoloutlinedinDaCostaandSorenson(2014).Inshort,~1μgofgenomicDNAwas digested using 10 U of each SbfI and EcoRI restrictionenzymes. Adapters containing sequences compatible for IlluminaTruSeq reagents and barcodes for de‐multiplexing reads were li‐gated to the sticky ends generated by the restriction enzymes.The adapter‐ligated DNA fragments were then size‐selected for300–450 bp using gel electrophoresis (2% low‐melt agarose) andpurifiedusingaMinEluteGelExtractionKit(Qiagen).Size‐selectedfragmentswerethenPCRamplifiedwithPhusionHigh‐FidelityDNAPolymerase (Thermo Scientific), and the amplified products werecleanedusingAMPureXPmagneticbeads(BeckmanCoulter, Inc.).Quantitative PCR using a Library Quantification Kit for Illumina(KAPABiosystems)wasusedtomeasure theconcentrationofpu‐rified PCR products, and samples with compatible barcode com‐binationswerepooled in equimolar concentrations.AmultiplexedlibrarywassequencedonanIlluminaHiSeq2,500usingsingle‐end

TA B L E 1  Thecountry,latitudinaldistribution,physicalcharacteristics,andelevationrangeofthefourriverswherethetwotorrentduck(Merganetta armata)subspeciesweresampledintheAndesofPeruandArgentina

Country River Latitude (°) Watershed area (ha)River + Tributary length (km)

Elevation range (m)

PERU Chillón 11.5 161,280 87.08 1,000–4,000

Pachachaca 13.5 671,844 326.80 2,700–3,100

ARGENTINA ArroyoGrande 33.5 53,016 42.63 1,800–3,200

Malargüe 35.5 69,652 37.59 1,600–1,900

4  |     ALZA et AL.

150 bp chemistry at the TuftsUniversity CoreGenomics Facility.RawIlluminareadsaredepositedinNCBI’sSequenceReadArchive(SRAdata:SRP215991).

2.4 | Bioinformatics of ddRAD‐Seq data

RawIlluminareadswereprocessedusingacomputationalpipelinedescribedbyDaCostaandSorenson (2014;http://github.com/BU‐RAD‐seq/ddRAD‐seq‐Pipeline). First, readswere assigned to indi‐vidualsamplesbasedonbarcodesequencesusingtheddRADparser.pyscript.Readspersamplewerethencollapsedintoidenticalclus‐ters using theCondenseSequences.py scriptwith low‐quality reads(i.e.,sequencesthatfailedtoclusterwithanyotherreads (–idset‐tingof0.90)andwithanaverageper‐basePhredscore<20)filteredoutwiththeFilterSequences.pyscript.Condensedandfilteredreads

fromall sampleswere thenconcatenatedandclusteredwithan–idsettingof0.85inucluST. MuSclev.3(Edgar,2004)wasusedtoalignandclusterreads,andsampleswithineachalignedclusterwere genotyped using the RADGenotypes.py script. HomozygotesandheterozygoteswereidentifiedbasedonthresholdsoutlinedinDaCostaandSorenson(2014;Lavretskyetal.,2015),withindividualgenotypes falling into threecategories: “missing” (nodata), “good”(unambiguouslygenotyped),and“flagged”(recoveredheterozygousgenotype,butwithhaplotypecountsoutsideofacceptablethresh‐oldsorwith>2allelesdetected).Lociwith<20%missinggenotypesand≤6flaggedgenotypeswereretainedfordownstreamanalyses.Moreover, alignments with end gaps due to indels, ≥2 polymor‐phisms in the last five base pairs, and/or a polymorphism in theSbfI restriction site were either automatically trimmed or flaggedformanual inspectioninGeneiouS (BiomattersInc.).Manualediting

F I G U R E 1   (a)Phylogeneticnetworkofthe27mtDNAhaplotypes(634basepairsoftheCRmtDNA)foundamongfourriversfortorrentducks(Merganetta armata)intheAndesofPeruandArgentina,basedonthestatisticalparsimonyprocedureimplementedinTCS.Circlesizesareproportionaltohaplotypefrequency(seeinset,lowerleft);missingintermediatehaplotypesareshownassmallopendots.(b)Rivernetworkmorphologyandsizeswiththespatialdistributionofthe156individualoftorrentducks(M. armata),blackdots,sampledalongaltitudinalgradients.DotsizesrepresentdifferentmtDNAhaplotypes.Thesehaplotypesdonotpresentspatialsegregation,ΦSTvalues,alongaltitude.Whitedashlinesmarkthe2,500mofelevationineachriversystem.ChillónRiverrunswest,andPachachaca,ArroyoGrandeandMalargüeriversruneast.(c)GeographicdistributionofthetworiversinPeru(M. a. leucogenis)andtworiversinArgentina(M. a. armata)alongtheAndes.Circlesizesrepresentthenumberofindividualssampledperwatershedpopulation,andcolorrepresentsahaplogroupbyoriginassociatedtoeachriver.BlackdashlineprovidesareferenceoftheAridDiagonalspatialdistribution

,

,

,

,

,

,

(a) (b) (c)

     |  5ALZA et AL.

increasedthetotalnumberofretainedmarkersby~7%,whilereduc‐ingbias resulting fromdiscarding lociwith indelsorhigh levelsofpolymorphisms.Finally,outputfilesfordownstreamanalyseswerecreatedwithcustompythonscripts(Lavretskyetal.,2016)thatin‐corporatedsequencingdepth.Tolimitanybiasesduetosequencingerrorand/orlowdepth,alleleswerecalled“missing”unlesstheymetourthresholdsofaminimumof5Xcoverageforhomozygotes,andthusatleast10Xcoverageforheterozygotestocallbothalleleswith>99%predictedsequenceaccuracy.

Finally,autosomalandZchromosome‐linkedlociwereidentifiedasdescribed inLavretskyetal. (2015),withassignmentsbasedondifferencesinsequencingdepthandhomozygositybetweenmalesand females. Because females have only one Z chromosome, Z‐linkedmarkersinfemalesareexpectedtoappearhomozygousandberecoveredatabouthalfthesequencingdepthofmales.

2.5 | Summary statistics

For mtDNA, we calculated summary statistics between rivers in‐cludinghaplotypediversity,nucleotidediversity(π/site),Tajima'sD,andΦSTinArlequinv.3.5.1.2(Excoffier&Lischer,2010).Tajima'sDwas calculated to test for thedeparture fromneutrality in sce‐narioscharacterizedbyanexcessofrarealleles.Negativevaluesforthisteststatisticmayindicateapopulationevolvingundernonran‐domprocessessuchasdirectionalselectionorrecentdemographicexpansion, whereas positive values may be indicative of popula‐tion decline or balancing selection. For autosomal and Z chromo‐some‐linkedmarkers,Tajima'sDestimateswerecalculatedintheR(http://cran.r‐project.org/) program “pegas” v.0.10 (Paradis, 2010).Analysis of molecular variance (AMOVA) was run to examine ge‐neticdifferentiationwithinandamongthefourriverssampledfrom

the two countries. Population pairwiseΦST was estimated underthe Tamura and Nei (1993) model of nucleotide substitution. ForddRAD‐SeqautosomalandZchromosome‐linkedmarkers,pairwiseΦST (i.e., "nuc.F_ST")andnucleotidediversity (i.e., “nuc.div.within”)estimateswerecalculatedintheRprogram“PopGenome”(Pfeifer,Wittelsbürger,Ramos‐Onsins,&Lercher,2014);indelpositionswereexcludedfromanalyses.

2.6 | Population structure

FormtDNA, the population structurewas visualized by recon‐structingahaplotypenetworkusingTcSv.1.21(Clement,Posada,& Crandall, 2000). TcS illustrates all connections that have a95%probabilityofbeing themostparsimonious.Networksaremoreappropriateforintraspecificgenegenealogiesthanrootedtree algorithmsbecausepopulation genealogies areoftenmul‐tifurcated, descendant genes coexist with persistent ancestralsequencesandinthecaseofnuclearDNArecombinationeventsproducereticulaterelationships,wheretraditionalphylogenetictrees treat all sequences as terminal taxa (Posada & Crandall,2001).

Analysis of population subdivision for ddRAD‐Seqdatawasdone using two methods. First, we used a principal compo‐nentanalysis (PCA)as implemented in theadegenetRprogram(i.e., “dudi.pca”;Dray&Dufour,2007; also see Jombart,2008).Next, maximum likelihood‐based individual assignment prob‐abilities were calculated in ADMiXTure (Alexander & Lange,2011;Alexander,Novembre,&Lange,2009).Todo so,biallelicsingle‐nucleotidepolymorphisms(SNPs)foreachautosomalandZ‐linked (males only) cluster were formatted for ADMiXTure analysisandthenprocessedthroughPLINK(Purcelletal.,2007)

F I G U R E 2   (a)FrequencydistributionofΦSTestimates>0.01across4,027autosomaland189ZlocibetweenPeruandArgentina.InsetprovidesthewholeviewofthefrequencydistributionofΦST.(b)Scatterplotofthefirsttwoprincipalcomponentsfor4,027autosomalddRAD‐seqmarkersoftorrentduck(Merganetta armata)fromChillónRiverinPeru(M. a. leucogenis, n=10)andMalargüeRiverinArgentina(M. a. armata, n=9)

–40

–30

–20

–10

0

10

20

30

40

–30 –20 –10 0 10 20 30

ARGENTINA

PERU

0

0.005

0.010

0.015

0.020

0.025

0.030

0.035

0.040

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.75

0.80

0.85

0.90

0.95

1.00

Autosomal

Z chromosome

(b)

ΦST PC 1 (12.4%)

PC

2 (6

.9%

)

% o

f loc

i

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.75

0.80

0.85

0.90

0.95

1.00

(a)

6  |     ALZA et AL.

following steps outlined in Alexander, Novembre, and Lange(2012). For eachADMiXTure analysis, a 10‐fold cross‐valida‐tion was performed, with a quasi‐Newton algorithm employedtoaccelerateconvergence(Zhou,Alexander,&Lange,2011).Foreachnumberof populations (K =1–5) tested,weused ablock‐relaxation algorithm for the point estimation, with analysesterminated once the change (i.e., delta) in the log‐likelihood ofthepointestimationsincreasedby<0.0001.Finaloutputswerebasedonadmixtureproportions(Qestimates;thelog‐likelihoodofgroupassignment)perindividual.Allanalyseswereperformedwithoutaprioriassignments.

2.7 | Historical population demography of mtDNA

For mtDNA, we used the coalescent genealogy sampler soft‐wares, iM and iMA2, which implement isolation‐with‐migra‐tion models (Hey, 2005, 2010; Hey & Nielsen, 2004; Nielsen&Wakeley, 2001). Thesemodels use aBayesianMarkov chainMonteCarlo(MCMC)methodthatsimultaneouslyestimatesde‐mographic parameters associated with the genetic divergencebetweenpopulations.Using IMa2,we assumed a 4‐populationgenealogy of two subspecies with two populations each andestimated: effective population sizes (Θ) of eachof the extant

F I G U R E 3   (a)Isolation‐with‐migrationmodelfortwosampledpopulationsofPeruviantorrentduck(M. a. leucogenis)inPachachacaandChillón,andtwopopulationsofArgentinetorrentduck(M. a. armata)inArroyoGrandeandMalargüe,andposteriorprobabilitydistributiongraphs(b,candd)forhistoricaldemographicparametersestimatedusingIMa2analysisinapairwisecomparisonoflocalities.Heightofthecurvescorrespondstotheestimatedprobabilitythatagivenparametervalueistrue,giventhedata(95%confidenceintervalsarereportedineachgraph).(b)Effectivepopulationsizes(Θ)scaledtotheneutralmutationrateestimatesforeachoneofthefourlocalitiessampled.(c)Scaledmigrationrates(m)betweenfourpairsoflocalities.(d)Scaledtimesincedivergence(t)estimatesbetweenthreepairsoflocalitiesandthescaledtimetomostrecentcommonancestor(TMRCA)

(a)

(d)

(b)

Θ anc

Θ 0 Θ 1 Θ 2 Θ 3

t

mm

0 10 20 30 40 50

0.00

0.05

0.10

0.15

0.20

Population size (Θ)

Pos

terio

r pro

babi

lity

Pachachaca = 19.3 (9.9-35.0)Chillón = 8.9 (3.2-18.8)Arroyo Grande = 7.6 (1.4-21.1)Malargüe = 1.9 (0.2-15.4)

0 1 2 3 4 5

0.0

0.5

1.0

1.5

2.0

Migration rate (m)

Pos

terio

r pro

babi

lity

Pachachaca > Chillón = 0.28 (0.0-1.2)Chillón > Pachachaca = 0.01 (0.0-0.3)Arroyo Grande > Malargüe = 0.45 (0.0-5.6)Malargüe > Arroyo Grande = 0.01 (0.0-2.6)

0 10 20 30 40

0.00

0.05

0.10

0.15

0.20

Scaled time since divergence (t)

Arroyo Grande-Malargüe = 3.58 (0.3-14.3)Chillón-Pachachaca = 11.78 (3.3-20.2)Peru - Argentina = 17.32 (9.3-24.5)TMRCA = 19.25 (12.25-27.75)

Chillón

(c)

Pachachaca ArroyoGrande

Malargüe

     |  7ALZA et AL.

populations, the ancestral population size (Θ) prior to popula‐tiondivergence,levelsanddirectionofgeneflow(immigration)intoeachpopulation (m) scaled to themutation rate, and timesince divergence for the populations (t) (Figure 3a). The IMa2analysiswasrun intriplicate,eachwith20Metropolis‐coupledchains of five million steps using a geometric heating schemeandaburn‐inperiodof500,000 steps.Priorsweredefinedasfollows:Θ=150,m=10,andt=50,basedonavailableinforma‐tion of density, and previous runs of the isolation‐with‐migra‐tionmodel.

Meanwhile,weusedthetwo‐population iMmodel toestimatelevelsanddirectionofgeneflow(m)scaledtothemutationrate,andtime since population divergence (t) between the two subspecies

inhabitingPeruandArgentina, respectively (Figure4a).Thediver‐genceparameter,t,isscaledtotheneutralmutationrateandcanbeconvertedtotimeinyearswithanaccurateestimateofthemutationrate(Hey&Nielsen,2004).ForthemtDNA,weusedthepointes‐timateof the substitution rate and confidence intervals publishedbyPeters,Gretes,andOmland(2005):4.8×10–8substitutions/site/year[95%confidenceinterval(CI)=3.1–6.9×10–8substitutions/site/year].TheiManalysiswasrunintriplicatetocheckforconsis‐tencyamongruns,eachwith20Metropolis‐coupledchainsofthreemillion steps using a geometric heating scheme and a burn‐in pe‐riodof500,000steps.Priorsweredefinedasfollows:ΘPERU=80,ΘARGENTINA=60,ΘANCESTRAL=150,mPER.‐ARG.=10,mARG.‐PER.=10,t=40,ands = 0–1.

F I G U R E 4   (a)Isolation‐with‐migrationmodelfortwotorrentduck(Merganetta armata)subspeciesintheAndesofPeruandArgentinaandposteriorprobabilitydistributiongraphs(bandc)forhistoricaldemographicmodelparametersestimatedusingIManalysisinpairwisecomparisons.Heightofthecurvescorrespondstotheestimatedprobabilitythatagivenparametervalueistrue,giventhedata(95%confidenceintervalsarereportedineachgraph).(b)Effectivepopulationsizes(Θ)scaledtotheneutralmutationrateestimatedforeachcountryandancestralpopulationsize.(c)Scaledmigrationrates(m)betweentwopairsofcountries

(a) (b)

(c)

Θ 0 Θ 1

t

m

0 10 20 30 40 50

0.000

0.002

0.004

0.006

0.008

0.010

Population size (Θ)

Pos

terio

r pro

babi

lity

Peru = 29.72 (20.60-45.24)Argentina = 12.45 (6.57-24.15)

0 1 2 3 4 5

0.0

0.1

0.2

0.3

0.4

0.5

Migration rate (m)

Pos

terio

r pro

babi

lity

Peru > Argentina = 0.01 (0.0-0.1)Argentina > Peru = 0.01 (0.0-0.2)

Peru Argentina

8  |     ALZA et AL.

2.8 | Historical population demography of nuclear DNA

WefurtherestimatedratesanddirectionalityofgeneflowfromtheddRAD‐Seq datawith the program ∂a∂i (Gutenkunst, Hernandez,Williamson, & Bustamante, 2009, 2010).∂a∂i implements an effi‐cientdiffusionapproximation‐basedapproachtotestempiricaldataagainst specified evolutionary models (e.g., isolation‐with‐migra‐tion).Using∂a∂i,asitefrequencyspectrumwasderivedfromallbi‐allelicRAD‐SeqautosomalSNPs.Lociwereconcatenated,andSNPsextracted and formatted for ∂a∂i using the custom python scriptnex_mo.py(Lavretskyetal.,2016).Becausewelackedanoutgroup,site frequencyspectrumdatawere folded,withonlyminorallelesconsideredinthefrequencyspectrum.Variantsobservedinzeroorinallsampleswere ignored(masked),asdescribedbyGutenkunst,Hernandez, Williamson, and Bustamante (2010). Finally, for ∂a∂i to accommodatemissing data anddifferences in sample sizes be‐tweenPeruvian(n=10individualsor20alleles)andArgentine(n = 9 individuals or 18 alleles) torrent duck populations, datasets wereprojecteddown toa totalof18allelesper subspecies.We testedtheempiricaldataagainstan isolation‐with‐migrationevolutionarymodelasimplementedin∂a∂i(Gutenkunst,Hernandez,Williamson,&Bustamante,2009;Gutenkunstetal.,2010).Demographicparam‐eterswereestimated,includingpopulationsizes(ni =(Ni/Nref)*NAnc; Nref=referenceeffectivepopulationsize;NAnc=ancestraleffectivepopulationssize),migrationrates(Mi←j = 2NAncmi←j),anddivergencetimes(t = T/2Nref: T=timesincedivergenceingenerations).

To convert the parameter estimates from ∂a∂i to biologicallyinformative values, we estimated generation time (G) and muta‐tionrates(μ,perlocus).First,generationtime(G)wascalculatedasG = α+(s/(1−s)),whereαistheageofmaturityandsistheexpectedadultsurvivalrate(Sætheretal.,2005).Althoughsexuallyactivebythefirstgeneration,mostduckspeciesreachsexualmaturityintheirsecondyear(α=2)withanaverageadultsurvivalrateof0.37esti‐mated for torrentducks frommark‐recaptureddataof theChillónRiver (pers. obs.). Together,weestimateda generation time tobe2.5years.Next,toobtainamutationratefornucleargenes,wemul‐tipliedarateof1.2×10–9substitutions/site/year,previouslycalcu‐lated fornuclear genes inotherducks (Peters,Zhuravlev,Fefelov,Humphries,&Omland,2008)bygenerationtimetoattainarateof3×10–9substitutionssite−1generation−1(ss−1 g−1).Afinalmutationratewascalculatedastheproductoftheabovemutationrateandthetotalnumberofbasepairs.

3  | RESULTS

3.1 | Genetic diversity & population structure—mtDNA

Nucleotide (π/site) and haplotype diversities of themtDNA CRdiffered among riverine populations and countries (Table 2).M. a. leucogenis inPeruhadhighernucleotidediversitythatwasfive times greater thanM. a. armata in Argentina, although theTA

BLE

2 NucleotideandhaplotypediversityformtDNAcontrolregion(ChillónRiver,

n=57;PachachacaRiver,

n=49;ArroyoGrandeRiver,

n=33andMalargüeRiver,

n=17),autosomal

andZ‐linkedmarkers(Peru‐Chillón,n=10andArgentina‐Malargüe,n=9)fromfourpopulationsandtwosubspeciesoftorrentduck(M

erga

nett

a ar

mat

a)intheAndesofPeruandArgentina

M

itoch

ondr

ial h

aplo

type

s

Nuc

leot

ide

dive

rsity

(π/s

ite)

Hap

loty

pe d

iver

sity

Mito

chon

dria

lA

utos

omal

Z Ch

rom

osom

eM

itoch

ondr

ial

Aut

osom

al (S

D)

Z Ch

rom

osom

e (S

D)

PERU

160.

015

0.00

110.

0004

0.83

60.139(0.26)

0.057(0.15)

ChillónRiver

70.

008

0.54

3

PachachacaRiver

100.

015

0.86

8

ARGENTINA

110.

003

0.00

070.

0003

0.83

70.093(0.21)

0.033(0.13)

ArroyoGrandeRiver

80.

004

0.78

4

MalargüeRiver

40.

001

0.48

4

Abbreviation:SD,standarddeviation.

     |  9ALZA et AL.

haplotype diversity was similar between countries (Table 2).ThesegeneticdiversitydifferencessuggestadeepercoalescenthistoryinthePeruviansubspecies,whilepossiblytheArgentinesubspecies’ shallower coalescence is a result of either demo‐graphicorselectiveprocessesthatdeterminethe lowergeneticdiversity.

A deep genetic distance corresponding to 4.1% uncorrectedmtDNACRdivergencewasobservedbetweenpopulationsinPeruand Argentina, defining two monophyletic lineages. Specifically,among the four independent riverswe recovered27 haplotypescharacterized by 53 variable sites. Moreover, 16 private hap‐lotypeswere found in Peruvian rivers (Chillón and Pachachaca),and11privatehaplotypeswerefoundinArgentinerivers(ArroyoGrandeandMalargüe)(Figure1a;TableS1),resultinginarelativedifferentiation(ΦST)of0.82andconsistentwiththeircurrentsub‐speciesdesignations(Figure1c;Table3).WithinPeruorArgentina,we found evidence of strong population structure between thetwo riverine populations in Peru (ΦST PERU = 0.43; Figure 1c), aswellasbetweenthetworiversinArgentina(ΦSTARGENTINA = 0.31; Figure 1c). First, we recovered a single mtDNA haplotype thatwas shared by torrent ducks from the Peruvian Pachachaca andChillónrivers,whereasthosefromthePachachacaRiverhadthehighesthaplotypediversityandnumberofprivatehaplotypesofallsampledrivers (Figure1a;Table2,TableS1). InArgentina,weonceagainfoundasinglemtDNAhaplotypesharedbetweentor‐rentducksfromtheArroyoGrandeandMalargüerivers,withtheArroyoGrandeRiverhavingahighernumberofprivatehaplotypesascomparedtothosefromtheMalargüeRiver(Figure1a;Table2,TableS1).Despiterecoveredstructureamongtorrentduckpopula‐tions(Table3),therelativedifferentiationwithineachriverpopu‐lationwaslow(ΦST<0.05;Figure1b)anddidnotfindthatmtDNAhaplotypefrequenciessegregatedbyelevationontheChillónRiver(ΦST=0.045)andArroyoGrandeRiver(ΦST=−0.037),considering

a2,500mdividealongtherivers,oronthePachachacaRiversys‐tem(ΦST=0.029).

3.2 | Genetic diversity & population structure—ddRAD‐seq loci

For ddRAD sequencing, a total of 22,662,951 raw readswere re‐coveredfromtheHiSeqIlluminarunof10Peruvianand9Argentinetorrent duck samples. After quality filtering, we recovered 4,216ddRAD‐seqloci,with4,027loci(505,475basepairs)assignedtoau‐tosomesand189locitotheZchromosome(24,242basepairs);twogametologswerealsoidentifiedandexcludedfromanalyses.Onav‐erage,therewasamediandepthof283sequencesperindividualperlocus(range=18–2,423sequences/individual/locus).

First, we recovered a higher autosomal and Z chromosomenucleotide diversity (on average 1.4‐fold higher) and haplotypediversity (on average 1.6‐fold higher) inM. a. leucogenis of Peruascompared toM. a. armataofArgentina (Table2).Consideringagenomic perspective, the overall ΦST across all ddRAD‐Seq lociwas 0.30 between the two torrent duck subspecies, with higherestimates atZ‐linked (ΦST=0.45) than for autosomal (ΦST=0.30)markers (Table 3; Figure 2a). A single biallelic SNPwas randomlychosenacrossloci,withatotalof2,356and49biallelicautosomaland Z‐linked SNPs, respectively, used for ADMIXTURE analyses.TheoptimalKwastwoforbothautosomalandZ‐linkedmarkers,andADMIXTUREresultslargelydistinguishedbetweenthetwosubspe‐cieswith99%ofassignmentprobabilitiesandwereconcordantwiththePCAresults,plottedwiththefirsttwoprincipalcomponents,inwhichthetwosubspeciesweredifferentiatedintwodiscreteclus‐ters,whereasindividualsfromPeruaremuchmoredispersedalongPC1thantheindividualsfromArgentina(Figure2b;FigureS1).NoadditionalstructuralresolutionwasattainedwhenanalyzinghighervaluesofK.

TA B L E 3  PairwiseΦSTvalues(allp‐values<.0001)andTajima'sDformtDNAcontrolregion(ChillónRiver,n=57;PachachacaRiver,n=49;ArroyoGrandeRiver,n=33andMalargüeRiver,n=17),autosomalandZ‐linkedmarkers(Peru‐Chillón,n=10andArgentina‐Malargüe,n=9)fromfourpopulationsandtwosubspeciesoftorrentduck(Merganetta armata)intheAndesofPeruandArgentina

ΦSTa Tajima's D

Mitochondrial Autosomal Z chromosomeMitochondrial (p‐value)

Autosomal (p‐value)

Z chromosome (p‐value)

PERU 0.82 0.3 0.45 1.18(.91) −1.25(.2) −1.04(.3)

ARGENTINA −0.29(.43) −1.3(.19) 0.09(.93)

ChillónRiver 0.43 0.13(.6)

PachachacaRiver 1.00(.85)

ArroyoGrandeRiver 0.31 −0.28(.42)

MalargüeRiver −1.38(.07)

ChillónRiver 0.89

AnyArgentineriver

PachachacaRiver 0.84

AnyArgentineriver

aDistancemethod:TamuraandNei(1993).

10  |     ALZA et AL.

3.3 | Historical population demography

Theeffectivepopulation sizeparameters,Θ, scaled to theneutralmutationrate in IMand IMa2formtDNAwerehigher for the riv‐ers inPeru thanArgentina (Figures3band4b).The largestΘwasforthePachachacaRiverwithapeakat19.3(95%HPD=9.9–35.0;Figure3b)inPeru;thesmallestΘwasanorderormagnitudeloweron theMalargüe River with a peak at 1.9 (95%HPD = 0.2–15.4;Figure3b).Betweensubspecies,thePeruviantorrentduckpopula‐tionwasestimatedtohavethelargestoveralleffectivepopulationsize (29.72,95%HPD=20.60–45.24;Figure4b) compared to theArgentinepopulation(12.45,95%HPD=6.57–24.15;Figure4b).AstobeexpectedformtDNA,becausethetwocladeswererecipro‐callymonophyletic,theancestralpopulationsizesestimatedwithIMand IMa2were flatanddidnot showclearpeaks (posteriorprob‐abilitiesnotshowninthefigures).

The ∂a∂i analyses utilizing the genomic data in the isolation‐with‐migrationmodelprovidedresultsthatagreethemtDNA‐esti‐matedresults.Usingautosomalmarkers,weestimatedanaveragedmutationrateof1.52×10–3ss−1 g−1(3×10–9ss−1 g−1×505,475basepairs)thatwasusedtoconvert∂a∂iresults.Giventheisolation‐with‐migrationmodel,parameterestimatessuggesteddifferenteffectivepopulation sizes (Ne) for the two subspecies of torrent ducks (Ne

PERU=49,896andNeARGENTINA=19,008).Tajima'sDvaluesformtDNAforthepopulationsinArgentina

and Peru (Table 3) were not significantly different from zero(higherp‐values).Therefore,wedidnotobserveasignificantex‐cessofrarenucleotidesitevariantscomparedtowhatwouldbeexpected under a neutralmodel of evolution, and each riverinepopulation may be evolving as per mutation‐drift equilibrium.In contrast, Tajima'sD values were negative for autosomal andZ‐linked markers for both subspecies populations in Peru andArgentina,althoughbothmarkershadnotstatisticallysignificantp‐values(Table3).

Estimating gene flow, the four‐population IMa2 analysis sug‐gests that thenumberofmigrants (m)betweenriverinepopulationswithineachcountry iseffectively lowandunidirectional (Figure3c).Importantly,wewereunabletorejectthehypothesisofnogeneflow.Theposteriordistributionofm (scaledeffectivemigration rate rela‐tivetothemutationrate,m/μ)withinPerupeakedat0.28fromthelargerPachachacapopulationtothesmallerChillónpopulation(95%HPD=0.0–1.2;Figure3c),andinthecaseofArgentinapeakedat0.45from the larger Arroyo Grande population to the smallerMalargüepopulation (95%HPD=0–5.6; Figure3c).By contrast,we failed torejectthehypothesisofnogeneflow intheoppositedirectionsbe‐tweenriverswithineachcountry.Moreover,theIManalysisbetweenthesubspeciessuggests that thenumberofmigrantsbetweenPeruandArgentina isnotdifferent fromzero (0.01,95%HPD=0.0–0.1;Figure4c).However, the∂a∂i resultsof thenucleardata supportedslightlygreater,yetsignificantgeneflowintoM. a. armatainArgentinafromM. a. leucogenis in Peru (2Nm21 = 0.30 migrants/generations)as compared to gene flow fromM. a. armata intoM. a. leucogenis (2Nm21=0.21migrants/generations).

Fortimesincedivergence,theposteriorprobabilitydistributionoft(scaledtimesincedivergence)withtheIMa2analysisbetweenthetwotorrentducksubspeciespeakedat17.32(95%HPD=9.3–24.5; Figure 2d). This value converted to time in years (Peters etal.,2005)suggeststhatPeruvianandArgentinetorrentduckssub‐speciesbegandivergingabout600,000yearsbeforepresent(ybp)(95% HPD range = 1.2Mya to 200,000 ybp). TMRCA peaked at19.25(95%HPD=12.24–27.75;Figure2d),indicatingthatallsam‐pledhaplotypescoalesceatapproximately650,000ybp.(95%HPDrange=1.4Myato280,000ybp).Convertingthe∂a∂iresultsofthenucleardataalsoindicatesdivergencewithinthesameapproximatetimeframe(782,490ybp).

4  | DISCUSSION

Weprovidethemostcomprehensivemolecularanalysisof torrentduckstodateandclearlyidentifylimitedornogeneflowandstrongdifferentiation across the mitochondrial and nuclear genomes oftorrentducks found in the riversofPeruversusArgentina.Theseresults support the current taxonomic subspecies designation ofPeruvian and Argentine torrent ducks. Moreover, we are able toconcludethatthesetwosubspeciesprobablyoriginatedduringtheearly‐tomid‐PleistoceneasaresultofanolddivergentnonvicarianteventsubsequenttotheoriginoftheSouthAmericanAridDiagonalphylogeographicbarrier(Burniard,1982).Finally,wealsorecoveredstrongpopulationstructurewithineachcountrythatisprobablyex‐plainedby both a strong territoriality andbreeding site fidelity inadults, and the geographic isolation by distance between riverinesubpopulations.

4.1 | Genetic diversity and small effective population size

Ingeneral,diversityestimatesacrossmtDNAhaplotypesforthetwotorrentducks subspecieswerehigheror similar to those found inotherAndeanduckspecies(Table2;Bulgarellaetal.,2012;Wilson,Peters, &McCracken, 2012).Moreover, despite the constraint onNe due to their exclusivepreference for torrential riverine (linear)habitats (Ellegren & Galtier, 2016; Moffet, 1970), torrent duckshavehighermtDNAdiversityestimatesascompared toother tor‐rential riverine specialist birds (e.g., New Zealand endemic blueducks (Hymenolaimus malacorhynchos; Grosser, Abdelkrim, Wing,Robertson,&Gemmell,2016);white‐throateddipper(Cinclus cinclus; Hourlayetal.,2008)).Infact,torrentducksfromPeruhadvaluesofmtDNAdiversitymoresimilartothosefoundinotherduckswithNe several ordersofmagnitude largerNe (e.g.,mallards; Lavretskyetal.,2015).Therefore,becausemutationratesarelikelytobesimilaracrossduckspecies,wecanhypothesizethattorrentduckmtDNAgenetic diversity probably is influenced by the structure of theirhabitat leading to extensive branched and subdivided, structuredriverinehabitat,suchthateachrivermaybeisolatedbydistanceandgeneticallydistinct fromotherwatersheds. Incontrast tomtDNA,

     |  11ALZA et AL.

diversityestimatesfromnuclearDNAforthetwotorrentducksub‐specieswerelowercomparedtootherwaterfowlspecies(Lavretskyetal.,2015;Lavretskyetal.,2016;Petersetal.,2016;Wilsonetal.,2012).Aprobableexplanationforthedisparityindiversityestimatesofmitochondrial andnuclear geneswithin torrent ducks couldbeacombinedeffect resulting fromhighmtDNAmutation rate (fourorders ofmagnitude), population subdivisiondue to thebranchedstructureriverinehabitat,theadultterritorialbehavior,andnaturalselection associated to high elevation (e.g., low temperatures andhypoxia)affectingnuclearDNA.

Finally,betweencountries(Argentinavs.Peru)andwithincoun‐tries(e.g.,ChillónRivervs.PachachacaRiver),wefoundthatpopu‐lationsinhabitingshorterriverscomparedtopopulationsinhabitinglongerorbranchedriverssystemshadthelowestlevelsofmolec‐ulardiversity(Figure1;Tables1and2).Thislowgeneticdiversityassociatedwith river length ismoreevident in the southern lati‐tudinal riverssampled inArgentina,whichmayhaveexperiencedfoundereventsfollowingglaciationperiods.Ingeneral,theseriversare shorter than the Peruvian rivers andmore likelymaintainingsmallNethatmightbemoresusceptibletolossofdiversityduetogeneticdrift(Table1;Figure3b).Moreover,currentandhistoricalchangesintheriverwaterregimesduetoextremedroughtanddra‐matic climatechanges (Rivera,Penalba,Villalba,&Araneo,2017)canproducebottlenecks,orriverinepopulationextinctions.Theseeventscanestablishrecurrentpopulationextinction–colonizationormetapopulationdynamicsthatwouldreduceNeandextendtheeffectsofgeneticdrift.VariableandsmallestimatesofNeacrossriverine populations, as well as mostly negative Tajima's D (i.e.,population growth), although not significant, possibly suggest ametapopulation dynamic. Specifically, our findings for Argentinepopulations prompt that these likely have been smaller and pos‐sibly have gone throughmultiple extinction–colonization events,whereasPeruvianpopulationshavelikelybeenlargeandresidenttoriversthathavehadaprolongedperiodofwaterregimestabilityascomparedtothoseinArgentina(Figures3band4b;Table3).

4.2 | Deep genetic divergence between subspecies

WefoundmtDNAmonophylyandnonsignificantestimatesofgeneflowbetweenthetwotorrentducksubspecies,M. a. leucogenisandM. a. armata (Figures 1a and 4a,c; Table 3; also for nuclear DNAFigure2b,FigureS1).Thereducedgeneflowbetweenthetwotor‐rentduckssubspeciesismostlikelyduetotheirhabitatspecializa‐tionandlackofsuitableintermediatehabitat,particularlywheretheAndesshowthehighestmountainrangesandwidestextensionthatharborsacomplexriverinespatialstructure(Capitonioetal.,2011;Gonzalez&Pfiffner,2011).Incontrast,twosubspeciesofspeckledteal(Anas flavisrostris)withsimilardistributionrangeanddivergencetimeshowrelativelylessgenomicdivergenceandasymmetricgeneflow(Grahametal.,2017).Long‐standingandrecurrentgeneflowinspeckledtealmightbefacilitatedbythelackofhabitatspecialization(i.e., lakes, pond, and also rivers), and a larger number of habitatspresentinthelowlandsduetoabroaderflattopographythaninthe

highlandsintheSouthernAndes.Therefore,whenandwheregeo‐graphicbarriersisolatethesetorrentducksubspeciesislikelytightlyconnectedwith theorogenyof theAndes, and the formationandpersistenceoftorrentialriverinesystems(Fjeldså,1985).

Finally, we estimate that the two torrent ducks likely di‐verged in the early‐ to mid‐Pleistocene (mtDNA divergencetime=600,000ybp,95%HPD range=1.2Mya to200,000ybp,Figure3d;andnucleardivergencetime=782,490ybp).Comparedtootherspecies found in theAndesofArgentinaandPeru,diver‐genceestimatesfortorrentducksaresomeoftheoldest,althoughcomparabletothoseestimatedbetweenspeckledteal(Bulgarellaetal.,2012;Grahametal.,2017;McCrackenetal.,2009;Wilsonetal.,2012).Therefore,torrentduckscanbeconsideredanolderAndeanresident probably tightly associated to the formation of river sys‐tems.However,divergenceofthesetwosubspecieslikelyfollowedthe establishment of allmajor river systems as theAndes startedtoriseduringtheLateCretaceous(~60–80Mya)andwereattheirfinalheightsbytheEarlyPliocene(~3–4Mya)(Capitonioetal.,2011;Clapperton,1983;Garzioneetal.,2008;Hartley,2003),whicharewellbeforeestimateddivergencetimes.

4.3 | South American Arid or Dry Diagonal

HistoricalandcurrentfieldrecordsshowthattwodryAndeanre‐gions serve as a disjunction between the distribution ranges ofM. a. leucogenisandM. a. armata,probablyduetothelackorextremeseasonality of torrential rivers in these regions (Conover, 1943;eBird, 2016; Fjeldså & Krabbe, 1990; Johnsgard, 2010; Johnson,1963).Oneoftheseregionsislocatedinthejunction(above28°S)ofboththeAtacama(Chile)andMonteDeserts(Argentina),hyperaridandaridsystems,respectively.Asecondregion(below18°S)isalsolocatedinthedryPunaintheEasternCordillerainSouthernBolivia.BothregionsarepartoftheSouthAmericanAridorDryDiagonalthat traverses the continent from theNorthCoastofPeru to theArgentinePatagonia.BelievedtohaveoriginatedwiththeorogenyandestablishmentoftheAndesMountainsandreinforcedbyoldandrecentglaciations (Blisniuk,Stern,Chamberlain, Idleman,&Zeitler,2005;Hartley,2003;Rabassa,Coronato,&Martínez,2011;Rabassa,Coronato, & Salemme, 2005; Zachos, Pagani, Sloan, Thomas, &Billups, 2001), theAridDiagonal is definedby its dry climate andabsence of glaciers, a defining feature dividing theCentralAndesfromSouthernAndesandPatagonia(Figure1c;andBurniard,1982;Seltzer,1990).Thus,theAridDiagonalislikelyabarriertomovementfortorrentducksasithasbeenforavarietyofotherbirds(aquaticandterrestrial;Fjeldså,1985;Ridgely&Tudor,2009;Valqui,2008;Voelker,2002;Vuilleumier,1998),mammals(González,Samaniego,Marín,&Estades,2013;Marinetal.,2007),insects(habitatofend‐emisms;Roig‐Juñent,Flores,Claver,Debandi,&Marvaldi,2001),andplants(Chacón,CamargodeAsis,Meerow,&Renner,2012;Martins,Scherz,&Renner,2014;Murillo‐A,Stuessy,&Ruiz,2016;Quiroga,2010)foundinSouthAmerica.

Moreover, given that thesehistorical climateconditionsof theAridDiagonallikelyreducedtheriverhabitatsuitabilitylongbefore

12  |     ALZA et AL.

the estimated time since divergence between these two torrentducksubspecies,wehypothesizethatcolonizationthroughdisper‐sal and subsequent isolation likely best explains their divergence.Althoughrare,torrentducksareknowntomakelongdistancemove‐ments(Cerón&Capllonch,2016)thatwouldpermitthemtotraverseacrosslargelandscapesliketheDryDiagonal.Ingeneral,wehypoth‐esizethatitisverylikelythatthesetwosubspecieswereprobablyoriginated by an old divergence throughmultiple founder events,withindividualsdispersingfromtheCentralAndestotheSouthernAndes,becausethelargeNeandgeneticdiversityrecoveredinPerucomparedtoArgentina,duringtheearly‐tomid‐Pleistocene.

4.4 | Population structure among rivers, “extended family” within rivers and metapopulation

BasedonmitochondrialDNA,weestimatedloworzerogeneflowbetweenriverswithineachcountry,suggestingwell‐differentiatedandmoderatelystructuredriverpopulationsinPeruandArgentina,respectively.Theseresultsindicatethateachriversystem,insideofawatershed, contains aparticular groupof geneticallymore simi‐lar individuals.Wehypothesizethat theseresultsare likelyduetothree nonmutually exclusive hypotheses: (1) Strong territorialityandbreedingsitefidelityinadultsincreasethesimilarityoftheindi‐vidualswithinariver,(2)geographic(e.g.,watershedboundary,highmountainranges)orclimaticbarriers(e.g.,largedeserts)isolateriverspreventingthedispersaloftheindividuals,anddrivinganindepend‐ent evolution of each riverine population by genetic drift, and (3)individualdispersalcapacity(i.e.,frequencyanddistance)describesanisolation‐by‐distancepatternthatinshortdistancedispersalre‐semblesametapopulationmodel.Forexample,theyear‐roundterri‐torialbehavior,thelong‐termpairbonds,andthestrongsitefidelityinadulttorrentducks,asmentionedinthefirsthypothesis,canpre‐ventgeneflowamongrivers,butalsocontributetothesimilarityoftheindividualsofagrowingpopulationinvacantorimprovedhabi‐tat,andthus increasetherelatednesswithineachriverinepopula‐tion anddevelop an “extended family” (Alza et al., 2017;Cardona& Kattan, 2010; Eldridge, 1986; Hartl & Clark, 2007; Pernollet,Estades,&Pavez,2012).Additionally,asreferredinthesecondhy‐pothesis,torrentialriversystemsinsidewatershedsareanalogoustoislands(islandsonmountainslopes)thatisolateandsustainpopula‐tionsandcommunities(Black,1997;Naiman,Magnuson,McKnight,Stanford,&Karr,1995;Omernik&Bailey,1997;Sullivan,Watzin,&Keeton,2007),similarlyobservedinlakes(Barbour&Brown,1974),marshes (Brown & Dinsmore, 1988), caves (Culver, Holsinger, &Baroody,1973),mountaintops(Nores,1995),orwoodlots(Holland,1978).Thus,asnewriversarecolonized(e.g.,byasmallgroupofin‐dividualsorforalongterm),theseisolatedpopulationsmaybehaveindependentlyoftheotherorlargerpopulation,growingasdistinctpopulations and leading to increased population structure amongrivers(Mayr,1963;Phillimore&Owens,2006).Finally,itispossiblethatdifferentwatershed(sub)‐populationsmaybegeneticallymoreconnectedorcompletelydifferentiateddependingonthedistanceamongthemandthedispersalcapacityoftheducks.Amongtorrent

duckriverine(sub)‐populationsitappearsthatdispersalismostlikelytooccurat close range (Cerón&Capllonch,2016;Paradis,Baillie,Sutherland,&Gregory,1998),whichcouldalsoreducetheirdiffer‐entiationandresembleametapopulationmodel(Hanski&Gaggiotti,2004; Levins, 1968). Similar population structure patterns havebeendescribed intheblueduck(Hymenolaimus malacorhynchos) inNewZealandwitharegionalgeneticdifferentiation(regionaround100 km) associated with isolation‐by‐distance pattern (Grosseretal.,2016).To test thesehypotheses, futureworkwouldbenefitfrommeasuresofrelatednessonincreasedsamplingof individualsacrossrivers,includingmultipleadjacentriversystemsstratifiedbyelevation. We note that complete haplotype admixture (mtDNA)alongeachstudyriversystemsuggests that torrentducksarenotsegregatedbyelevation (consideringa2,500mdivide,Figure1b).Previous studieshave reported seasonal altitudinalmovementsoftorrentducks,andthesemovementscanbeinfluencedbyreproduc‐tiveseason,foodavailability,variation inthestreamflow,andwin‐terconditions,especially intheSouthernAndes (Johnsgard,1966;Johnson, 1963; Pernollet et al., 2012; Ramírez, Botero, & Kattan,2014).AlsoduringourbandingactivitiesintheChillónRiver,were‐coveredtwoindividualsthathadmovedatleast4kmalongtheriver.These direct observations of altitudinal movements corroboratetheabsenceofpopulationstructurealongelevationgradientwithineachriver.Thus,thephysiologicalrestrictionimposedbyhypoxiaatelevations >2,500m appears to not constrainmitochondrial geneflowbetweenlow‐andhigh‐altitude localitiesonanyoftheriverswesampled,despiterecentlyreportedelevationalvariationinbodysize(Gutiérrez‐Pintoetal.,2014),functionofkeyenzymes(Dawsonetal.,2016),changesininsulativepropertiesoffeathers(Cheeketal.,2018), andchanges in thehypoxicventilatory response (Ivyetal.,2019).

4.5 | Conservation relevance

Torrentducksarevulnerabletosystematicpressureandstochasticperturbationsbecauseoftheirspecializationtowhitewaterstreams,low population densities, patchy distribution, low reproductivepotential,andrelianceonwaterqualityandaquatic insectpopula‐tions(Alvarez,Astie,Debandi,&Scheibler,2014;Carboneras,1992;Shaffer,1981;Zwick,1992).Indeed,populationslivinginriversandstreamshabitatsaresusceptibletoextinctionasaresultofthehighvulnerability that freshwater systemshavepollution, siltation,andanthropogenicdisturbance.Thus, torrentduckshadsuffered localextinction and decreasing population trends of many river popu‐lations (Callaghan,1997;Cerón&Trejo, 2012;Pernollet, Pavez,&Estades, 2013;Weller, 1975). In particular, the recognized causesoftheirriverineextinctionareasfollows:predationonoffspringbyexoticinvasivespecies(mink;Cerón&Trejo,2012),competitionforaquaticinsectfoodsources(trout;Eldridge,1986),lossofterritoriesandhabitatsduetourbanizationandcontaminationofriversinPeruandColombia(J.M.Barbarán,personalcommunication;Cardona&Kattan, 2010),management of thewatershed to provide drinkingwater, irrigation and hydropower production for mining activities

     |  13ALZA et AL.

andhumancommunities (Pernolletetal.,2013),andhuntingwith‐outproperregulation(pers.obs.).Whereby,thethreesubspeciesre‐gionallyhadbeenclassifiedunderthreatenedcategories(VulnerableandEndangered) (Cerón&Trejo,2012;Green,1996),eventhoughthe overall species is still classified as a “LeastConcern” (BirdLifeInternational,2016).Ourstudyalsoemphasizesthenecessitytopro‐tectandmanagethesubspeciesandriverinepopulationsasseparateconservationunitsduetothestronggeneticstructureandverylowgeneflow.Therefore,theestablishmentofmonitoringprogramsofthesubspeciesandmultiplepopulationsisaprioritytoevaluatetheirmetapopulationstructureandunderstand theactual statusof thespecies(Phillimore&Owens,2006),evenmore,underthethreatofdeglaciationand regionaldroughtconditionsassociated toclimatechange in theAndes. In general, the torrentduckprovides anex‐cellent opportunity to examine the patterns and themechanismsrelated to strong population structure among small and isolatedpopulations,andtheseresultscanbecontrastedwithotherriverineducks,suchastheAfricanblackduck(Anas sparsa),andSalvadori'sduck(Salvadorina [Anas] waigiuensis)inNewGuinea,orhabitatspe‐cialistspecies.

ACKNOWLEDG EMENTS

This study was funded by National Science Foundation (NSF0949439) grant and Kushlan Endowment forWaterbird BiologyandConservationattheUniversityofMiamitoKGMandANPCYT(PICT‐2016‐3712) and CONICET (PIP 112 201301 00803 CO;FondoiBolArgentinaD3657)grantstoCK.PermitstoconductthisresearchinArgentinawereprovidedbySecretaríadeAmbienteyDesarrolloSustentabledeArgentina, andDireccióndeRecursosNaturalesRenovables deMendoza; and inPeru byDGFFS, nowSERFOR, (Permit No. 0378‐2011‐AG‐DGFFS‐DGEFFS).Wewishtothankthepeoplewhohelpedtocollectthesedatainthefield:E.Bautista,R.delaCruz,D.Wilner,F.Hernández,A.Repetto,N.Wright,C.Sánchez,N.Gutiérrez‐Pinto,M.Lozano‐Jaramillo,J.M.Ríos,J.Barbarán,N.García,M.González,I.León,L.Alza,B.Alzaand E. V. Alza.We are also gratefulwith Centro deOrnitologíay Biodiversidad (T. Valqui, V. Vinces, W. Nañez and A. Valqui),Zoocriadero“ElHuayco”(J.Otero)andstaffofSoldeSantaRosa,FundoHuanchuy,FundoTambo, andHotelTampumayu for sup‐port inPeru;MuseoArgentinodeCienciasNaturales(P.Tubaro),andConsejoNacionaldeInvestigacionesCientíficasyTécnicasforlogisticalsupportinArgentina.Finally,wewouldliketothankthetwo anonymous reviewers for valuable comments that improveandclarify thismanuscript, aswell asD.Sikes,N.Kerhoulas,N.Takebayashi,K.Winker,I.Ilhan,F.Angulo,H.Lanier,D.DeAngelis,K.Keenan,and“Butch”Keenanfortheirsupportanddiscussionsduring theconceptionof ideas,andhelpfulcommentsonearlierdraftsofthismanuscript.

CONFLIC T OF INTERE S T

Nonedeclared.

AUTHOR CONTRIBUTIONS

L.A.analyzedthedataanddraftedthemanuscript;L.A.,M.S.,C.K.,andK.G.M.collectedandacquiredthedata;P.L.acquiredandana‐lyzedthegenomicdataandhelpedtorevisethemanuscript;L.A.andG.C.conceivedtheideas;J.P.interpretedtheresultsandrevisedthemanuscript;K.G.M.designedthestudy,interpretedtheresults,andhelpedtowrite themanuscript;andall theauthorshavereadandcommentedonthefinalmanuscript.

DATA AVAIL ABILIT Y S TATEMENT

DNA sequences: GenBank (Accession Numbers MN196318: MN196473). Raw Illumina reads: NCBI’s Sequence Read Archive(SRAdata:SRP215991).

ORCID

Luis Alza https://orcid.org/0000‐0002‐5926‐4460

Philip Lavretsky https://orcid.org/0000‐0002‐5904‐8821

R E FE R E N C E S

Alexander,D.H.,&Lange,K.(2011).EnhancementstotheADMIXTUREalgorithmforindividualancestryestimation.BMC Bioinformatics,12,246.https://doi.org/10.1186/1471‐2105‐12‐246

Alexander,D.H.,Novembre, J.,& Lange,K. (2009). Fastmodel‐basedestimationofancestryinunrelatedindividuals.Genome Research,19,1655–1664.https://doi.org/10.1101/gr.094052.109

Alexander, D. H., Novembre, J., & Lange, K. (2012). Admixture 1.22 software manual.

Alvarez,L.M.,Astie,A.A.,Debandi,G.O.,&Scheibler,E.E.(2014).EffectoffoodavailabilityandhabitatcharacteristicsontheabundanceofTorrentDucks during the early breeding season in the central Andes, PAGN:ProvidevisiblespaceArgentina.The Wilson Journal of Ornithology,126(3),525–533.https://doi.org/10.1676/13‐143.1

Alza, L., Bautista, E., Smith, M., Gutiérrez‐Pinto, N., Astie, A., &McCracken,K.G.(2017).Captureefficiencyofthetorrentducksbytheactivemist‐netmethod.Wildlife Society Bulletin,41(2),370–375.

Barbour,C.D.,&Brown,J.H.(1974).Fishspeciesdiversityinlakes.The American Naturalist,108,473–489.https://doi.org/10.1086/282927

BirdLife International (2016). Merganetta armata. In: IUCN Red list of threatened species 2016.Retrievedfromhttp://www.iucnredlist.org.

Black, P. E. (1997). Watershed functions. Journal of the American Water Resources Association, 33(1), 1–11. https://doi.org/10.1111/j.1752‐1688.1997.tb04077.x

Blisniuk,P.M.,Stern,L.A.,Chamberlain,C.P.,Idleman,B.,&Zeitler,P.K.(2005).ClimaticandecologicchangesduringMiocenesurfaceup‐lift in the SouthernPatagonianAndes.Earth and Planetary Science Letters,230,125–142.https://doi.org/10.1016/j.epsl.2004.11.015

Bowler,D.E.,&Benton,T.G.(2005).Causesandconsequencesofanimaldispersalstrategies:Relatingindividualsbehaviortospatialdynam‐ics.Biological Reviews,80,205–225.

Brack,A.(2000).Ecología del Perú.Lima,Peru:AsociaciónEditorialBruño.Brown,M.,&Dinsmore,J.J.(1988).Habitatislandsandtheequilibrium

theoryofislandbiogeography:Testingsomepredictions.Oecologia,75,426–429.https://doi.org/10.1007/BF00376947

Bulgarella, M., Peters, J. L., Kopuchian, C., Valqui, T., Wilson, R. E.,& McCracken, K. G. (2012). Multilocus coalescent analysis of

14  |     ALZA et AL.

haemoglobindifferentiationbetweenlow‐andhigh‐altitudepopula‐tionsofcrestedducks(Lophonetta specularioides).Molecular Ecology,21(2),350–368.

Burkart, R., Bárbaro, N., Sánchez, R. O., & Gómez, D. A. (1999).Ecorregiones de la Argentina.BuenosAires,Argentina:PRODIA.

Burniard, E.D. (1982). La diagonal árida argentina:Un límite climáticoreal. Revista Geográfica,95,5–20.

Callaghan, D. A. (1997). Conservation status of the Torrent DucksMerganetta.Wildfowl,48,166–173.

Capitonio, F. A., Faccenna, C., Zlotnik, S., & Stegman, D. R. (2011).Subduction dynamics and the origin of Andean orogeny and theBolivianorocline.Nature,480,83–86.https://doi.org/10.1038/nature10596

Carboneras,C.(1992).Familyanatidae(ducks,geese,andswans).InJ.delHoyo,A.Elliot,&J.Sargatal(Eds.),Handbook of the birds of the world. vol. I: Ostrich to ducks(pp.528–628).Barcelona,Spain:LynxEditions.

Cardona,W.,&Kattan,G. (2010).Comportamiento territorialy repro‐ductive del pato de torrentes (Merganetta armata) en la cordilleracentraldeColombia.Ornitología Colombiana,9,38–47.

Cerón, G., & Capllonch, P. (2016). Observaciones de eventos de dis‐persión del pato de los torrentes (Merganetta armata).Ornitología Neotropical,27,303–306.

Cerón, G., & Trejo, A. (2009). Descripción de la técnica de buceo delPatodeTorrente(Merganetta armata)enelParqueNacionalNahuelHuapi.Argentina. Hornero,24(1),57–59.

Cerón, G., & Trejo, A. (2012). Torrent Duck (Merganetta armata) pop‐ulation trend in Northwestern Patagonia, Argentina. Ornitología Neotropical,23,407–415.

Chacón,J.,CamargodeAsis,M.,Meerow,A.W.,&Renner,S.S.(2012).FromEastGondwanatoCentralAmerica:HistoricalbiogeographyoftheAlstroemeriaceae.Journal of Biogeography,39,1806–1818.https://doi.org/10.1111/j.1365‐2699.2012.02749.x

Charlesworth,B.(2010).Effectivepopulationsizeandpatternsofmolec‐ularevolutionandvariation.Genetics,10(3),195–205.

Cheek,G.R.,Alza,L.,&McCracken,K.G.(2018).Downfeathersstructurevariesbetweenlow‐andhigh‐altitudeTorrentDucks(Merganetta ar‐mata).Ornitología Neotropical,29,27–35.

Clapperton,C.M.(1983).TheglaciationoftheAndes.Quaternary Science Reviews,2,83–155.https://doi.org/10.1016/0277‐3791(83)90005‐7

Clement,M.,Posada,D.,&Crandall,K.A.(2000).TCS:Acomputerpro‐gramtoestimategenegenealogies.Molecular Ecology,9(10),1657–1660.https://doi.org/10.1046/j.1365‐294x.2000.01020.x

Colina,U.(2010).Descripcióndelaetologíareprodcutivaynidificacióndelpatode torrente (Merganetta armata berlepschi)enelNoroesteArgentino.Nótulas Faunísticas,2(54),1–6.

Conover,B.(1943).AstudyoftheTorrentDucks.Zoological Series of Field Museum of Natural History,24(31),345–356.

Culver,D.,Holsinger,J.R.,&Baroody,R.(1973).Towardapredictivecavebiogeography:TheGreenbriervalleyasacasestudy.Evolution,27(4),689–695.https://doi.org/10.1111/j.1558‐5646.1973.tb00716.x

DaCosta,J.M.,&Sorenson,M.D.(2014).Amplificationbiasesandcon‐sistent recoveryof loci inadouble‐digestRAD‐seqprotocol.PLoS ONE,9,e106713.https://doi.org/10.1371/journal.pone.0106713

Dawson,N.J.,Ivy,C.M.,Alza,L.,Cheek,R.,York,J.M.,Chua,B.,…Scott,G. R. (2016).Mitochondrial physiology in the skeletal and cardiacmusclesisalteredintorrentducks,Merganetta armata,fromhighal‐titudesintheAndes.Journal of Experimental Biology,219,3719–3728.https://doi.org/10.1242/jeb.142711

Dray,S.,&Dufour,A.‐B.(2007).Theade4package:Implementingthedu‐alitydiagramforecologists.Journal of Statistical Software,22,1–20.

eBirdBasicDataset(2016).Version: EBD_relAug‐2015.Ithaca,NY:CornellLabofOrnithology.

Edgar,R.C.(2004).MUSCLE:Multiplesequencealignmentwithhighac‐curacyandhighthroughput.Nucleic Acids Research,32,1792–1797.https://doi.org/10.1093/nar/gkh340

Eldridge,J.L.(1986).ObservationsonapairofTorrentDucks.Wildfowl,37,113–122.

Ellegren, H., & Galtier, N. (2016). Determinants of genetics diversity.Nature Reviews Genetics,17,283–291.

Excoffier, L., & Lischer, H. E. L. (2010). Arlequin suite ver 3.5: A newseries of programs to performpopulation genetics analyses underLinuxandWindows.Molecular Ecology Resources,10(3),564–567.

Fahrig,L.(2013).Rethinkingpatchsizeandisolationeffects:Thehabitatamounthypothesis.Journal of Biogeography,40,1649–1663.https://doi.org/10.1111/jbi.12130

Fjeldså,J.(1985).Origin,evolution,andstatusoftheavifaunaofAndeanwetlands. Ornithological Monographs, 36, 85–112. https://doi.org/10.2307/40168279

Fjeldså, J., & Krabbe, N. (1990). Birds of the high Andes. Copenhagen,Denmark:ZoologicalMuseum,UniversityofCopenhagen.

Frankham, R. (2005). Genetics and extinction. Biological Conservation,126,131–140.https://doi.org/10.1016/j.biocon.2005.05.002

Garzione,C.N.,Hoke,G.D.,Libarkin,J.C.,Withers,S.,MacFadden,B.,Eiler,J.,…Mulch,A.(2008).RiseoftheAndes.Science,320,1304–1307.https://doi.org/10.1126/science.1148615

González,B.A.,Samaniego,H.,Marín,J.C.,&Estades,C.F.(2013).Unveilingcurrent guanacodistribution inChilebaseduponniche structureofphylogeographiclineages:AndeanPunatosubpolarforests.PLoS ONE,8(11),e78894.https://doi.org/10.1371/journal.pone.0078894

Gonzalez, L., & Pfiffner, O. A. (2011). Morphologic evolution of thecentral Andes of Peru. International Journal of Earth Sciences, 101,307–321.

Graham,A.M.,Lavretsky,P.,Muñoz‐Fuentes,V.,Green,A.J.,Wilson,R.E.,&McCracken,K.G.(2017).Migration‐selectionbalancedrivesge‐neticdifferentiationingenesassociatedwithhigh‐altitudefunctioninthespeckledteal (Anas flavirostris) intheAndes.Genome Biology and Evolution,10(1),14–32.https://doi.org/10.1093/gbe/evx253

Green, A. J. (1996). Analyses of globally threatened Anatidae in rela‐tion to threats, distribution, migration, patterns, and habitat use.Conservation Biology, 10(5), 1435–1445. https://doi.org/10.1046/j.1523‐1739.1996.10051435.x

Grosser,S.,Abdelkrim, J.,Wing, J.,Robertson,B.C.,&Gemmell,N. J.(2016). Strong isolation by distance argues for separate popula‐tion management of endangered blue duck (Hymenolaimus mal‐acorhynchos). Conservation Genetics, 18(2), 327–341. https://doi.org/10.1007/s10592‐016‐0908‐4

Gutenkunst,R.N.,Hernandez,R.D.,Williamson,S.H.,&Bustamante,C.D.(2009).Inferringthejointdemographichistoryofmultiplepopu‐lationsfrommultidimensionalSNPfrequencydata.PLoS Genetics,5,e1000695.https://doi.org/10.1371/journal.pgen.1000695

Gutenkunst,R.N.,Hernandez,R.D.,Williamson,S.H.,&Bustamante,C.D.(2010).Diffusionapproximationfordemographicinference:DaDi.Nature Precedings.http://hdl.handle.net/10101/npre.2010.4594.1

Gutiérrez‐Pinto,N.,McCracken,K.G.,Alza,L.,Tubaro,P.,Kopuchian,C.,Astie,A.,&Cadena,C. (2014).Thevalidityofecogeographic rulesiscontextdependentinawidespreadAndeanduck:TorrentDucksfollowBergmann's rule by latitude but not by elevation.Biological Journal of the Linnean Society,4,850–862.

Hanski,I.,&Gaggiotti,O.E.(2004).Metapopulationbiology:Past,pres‐ent,andfuture.InI.Hanski,&O.E.Gaggiotti(Eds.),Ecology, genet‐ics, and evolution of metapopulations (pp. 3–22). Amsterdam, TheNetherlands:ElsevierAcademicPress.

Hanski, I.,&Gilpin,M. (1991).Metapopulationdynamics:Briefhistoryandconceptualdomain.Biological Journal of the Linnean Society,42,3–16.https://doi.org/10.1111/j.1095‐8312.1991.tb00548.x

Harrison, S., & Hastings, A. (1996). Genetic and evolutionary conse‐quencesofmetapopulationstructure.Trends in Ecology and Evolution,11(4),180–183.https://doi.org/10.1016/0169‐5347(96)20008‐4

Hartl,D. L.,&Clark,A.G. (2007).Principles of population genetics, 4thedn.Sunderland,MA:SinauerAssociates.

     |  15ALZA et AL.

Hartley, A. J. (2003). Andean uplift and climate change. Journal of the Geological Society, London, 160, 7–10. https://doi.org/10.1144/0016‐764902‐083

Hey, J. (2005). On the number of new world founders: A populationgeneticportraitofthepeoplingoftheAmericas.PLoS Biology,3(6),e193.https://doi.org/10.1371/journal.pbio.0030193

Hey,J.(2010).Isolationwithmigrationmodelsformorethantwopopu‐lations.Molecular Biology and Evolution,27(4),905–920.https://doi.org/10.1093/molbev/msp296

Hey,J.,&Machaco,C.A. (2003).Thestudyofstructuredpopulations:Newhopeforadifficultanddividedscience.Nature Reviews Genetics,4,535–543.https://doi.org/10.1038/nrg1112

Hey,J.,&Nielsen,R. (2004).Multilocusmethodsforestimatingpopu‐lationsizes,migration ratesanddivergence time,withapplicationsto the divergence of Drosophila pseudoobscura and D. persimilis. Genetics,167,747–760.

Holland,P.G. (1978). Species turnover indeciduous forestvegeta‐tion. Vegetatio, 38(2), 113–118. https://doi.org/10.1007/BF00052042

Hourlay,F.,Libois,R.,D’Amico,F.,Sarà,M.,O’Halloran,J.,&Michaux,J.R. (2008).Evidenceofahighlycomplexphylogeographicstructureonaspecialistriverbirdspecies,thedipper(Cinclus cinclus).Molecular Phylogenetics and Evolution,49,435–444.https://doi.org/10.1016/j.ympev.2008.07.025

Ivy, C.M., Lague, S. L., York, J.M., Chua, B. A., Alza, L., Cheek, R.,… Scott, G. R. (2019). Control of breathing and respiratory gasexchange in ducks native to high altitude in the Andes. Journal of Experimental Biology, 222(Pt 7), jeb198622. https://doi.org/ 10.1242/jeb.198622

Johnsgard,P.A.(1966).Thebiologyandrelationshipsofthetorrentduck.Wildfowl,17,66–74.

Johnsgard,P.A. (2010).Ducks, geese, and swans of the world: Tribe mer‐ganettini (Torrent Duck). Lincoln, NE: The University of Nebraska‐LincolnLibraries.

Johnson,A.W.(1963).Notesonthedistribution,reproductionanddis‐playoftheAndeanTorrentDuck,Merganetta armata. Ibis,105,114–116.https://doi.org/10.1111/j.1474‐919X.1963.tb02483.x

Jombart, T. (2008).Adegenet:ARpackage for themultivariate analy‐sis of genetic markers. Bioinformatics, 24, 1403–1405. https://doi.org/10.1093/bioinformatics/btn129

Kawecki, T. J., & Ebert, D. (2004). Conceptual issues in localadaptation. Ecology Letters, 7, 1225–1241. https://doi.org/10.1111/j.1461‐0248.2004.00684.x

Kawecki,T.J.,Lenski,R.E.,Ebert,D.,Hollis,B.,Olivieri,I.,&Whitlock,M.C.(2012).Experimentalevolution.Trends in Ecology and Evolution,27(10),547–560.https://doi.org/10.1016/j.tree.2012.06.001

Lavretsky, P., DaCosta, J. M., Hernández‐Baños, B. E., Engilis, A. Jr,Sorensen,M.D.,&Peters, J. L. (2015). SpeciationgenomicsandarolefortheZchromosomeintheearlystagesofdivergencebetweenMexicanducksandmallards.Molecular Ecology,24(21),5364–5378.https://doi.org/10.1111/mec.13402

Lavretsky,P.,Peters, J.L.,Winker,K.,Bahn,V.,Kulikova, I.,Zhuravlev,Y.N.,…McCracken,K.G. (2016).Becomingpure: Identifyinggen‐erational classes of admixed individuals within lesser and greaterscaup populations.Molecular Ecology, 25(3), 661–674. https://doi.org/10.1111/mec.13487

Lenormand, T. (2002). Gene flow and the limits to natural selection.Trends in Ecology and Evolution,17,183–189.https://doi.org/10.1016/S0169‐5347(02)02497‐7

Levins,R.(1968).Evolution in changing environments: Some theoretical ex‐plorations.Princeton,NJ:PrincetonUniversityPress.

Ma, L., Ji, Y. J., & Zhang,D. X. (2015). Statisticalmeasures of geneticdifferentiationofpopulations:Rationales,historyandcurrentstates.Current Zoology, 61(5), 886–896. https://doi.org/10.1093/czoolo/61.5.886

Marin,J.C.,Casey,C.S.,Kadwell,M.,Yaya,K.,Hoces,D.,Olazabal,J.,…Wheeler,J.C. (2007).Mitochondrialphylogeographyanddemo‐graphichistoryofthevicuña:Implicationsforconservation.Heredity,99,70–80.https://doi.org/10.1038/sj.hdy.6800966

Martins,A.C.,Scherz,M.D.,&Renner,S.(2014).Severaloriginsofflo‐raloil intheAngelonieae,aSouthernhemispheredisjunctcladeofPlantaginaceae. American Journal of Botany, 101(12), 2113–2120.https://doi.org/10.3732/ajb.1400470

Mayr,E. (1963).Animal species and evolution.Cambridge,MA:HarvardUniversityPress.

McCauley,D. E. (1991).Genetic consequences of local population ex‐tinctionandrecolonization.Trends in Ecology and Evolution,6(1),5–8.https://doi.org/10.1016/0169‐5347(91)90139‐O

McCracken,K.G.,Barger,C.P.,Bulgarella,M., Johnson,K.P.,Kuhner,M.K.,Moore,A.V.,…Wilson,R.E. (2009).Signaturesofhigh‐alti‐tudeadaptationinthemajorhemoglobinoffivespeciesofAndeandabblingducks.The American Naturalist,174,631–650.https://doi.org/10.1086/606020

McCracken, K.G., Johnson,W. P., & Sheldon, F.H. (2001).Molecularpopulation genetics, phylogeography, and conservation biology oftheMottledDuck(Anas fulvigula).Conservation Genetics,2,87–102.

Moffet,G.M.(1970).AstudyofnestingTorrentDucksintheAndes.The Living Bird,9,5–27.

Murillo‐A, J. C., Stuessy, T. F., & Ruiz, E. (2016). Explaining disjunctdistributions in the flora of southern SouthAmerica: Evolutionaryhistory and biogeography of Myrceugenia (Myrtaceae). Journal of Biogeography,43,979–990.

Naiman,R.J.,Magnuson,J.J.,McKnight,D.M.,Stanford,J.A.,&Karr,J.R.(1995).Freshwaterecosystemsandtheirmanagement:anationalinitiative. Science, 270, 584–585. https://doi.org/10.1126/science.270.5236.584

Naranjo, L. G., & Ávila, V. J. (2003). Distribución habitacional y dietadelpatodelostorrentes(Merganetta armata)enelParqueRegionalNatural Ucumari en la cordillera central de Colombia. Ornitología Colombiana,1,22–28.

Nei,M.(2013).Mutation‐driven evolution.Oxford,UK:OxfordUniversityPress.

Neigel, J.E. (1997).Acomparisonofalternativestrategies forestimat‐ing gene flow from geneticmarkers.Annual Review of Ecology and Systematics, 28, 105–128. https://doi.org/10.1146/annurev.ecolsys.28.1.105

Nielsen,R.,&Slatkin,M. (2013).An introduction to population genetics. Sunderland,MA:SinauerAssociates.

Nielsen,R.,&Wakeley,J.(2001).Distinguishingmigrationfromisolation:Amarkovchainmontecarloapproach.Genetics,158,885–896.

Nores, M. (1995). Insular biogeography of birds on mountain‐tops inNorth Western Argentina. Journal of Biogeography, 22(1), 61–70.https://doi.org/10.2307/2846073

Omernik,J.M.,&Bailey,R.G.(1997).Distinguishingbetweenwatershedsandecoregions.Journal of the American Water Resources Association,33(5), 935–949. https://doi.org/10.1111/j.1752‐1688.1997.tb04115.x

Orr,H.A. (2005). Thegenetic basis of reproductive isolation: InsightsfromDrosophila.InJ.Hey,W.M.Fitch,&F.Ayala(Eds.),Systematics and the origin of species: On Ernst Mayr’s 100th anniversary(pp.9–23).Washington,DC:TheNationalAcademiesPress.

Paradis,E.(2010).pegas:AnRpackageforpopulationgeneticswithanintegrated‐modular approach.Bioinformatics,26, 419–420. https://doi.org/10.1093/bioinformatics/btp696

Paradis,E.,Baillie,S.R.,Sutherland,W.J.,&Gregory,R.D.(1998).Patternsofnatalandbreedingdispersalinbirds.Journal of Animal Ecology,67,518–536.https://doi.org/10.1046/j.1365‐2656.1998.00215.x

Pernollet,C.A.,Estades,C.F.,&Pavez,E.F. (2012).Estructura socialdelpatocortacorrientesMerganetta armata armata,enChilecentral.Boletin Chileno De Ornitología,18(2),23–34.

16  |     ALZA et AL.

Pernollet,C.A.,Pavez,E.F.,&Estades,C.F. (2013).Habitatselectionby Torrent Ducks (Merganetta armata armata) in Central Chile:Conservation implications of hydropower production.Waterbirds,36(3),287–299.

Peters,J.L.,Gretes,W.,&Omland,K.E.(2005).LatePleistocenediver‐gencebetweeneasternandwesternpopulationsofwoodducks(Aix sponsa) inferredbythe ‘isolationwithmigration’coalescentmodel.Molecular Ecology,14,3407–3418.

Peters, J. L., Lavretsky,P.,DaCosta, J.M.,Biefeld,R.R., Feddersen,J. C., & Sorenson, M. D. (2016). Population genomic data de‐lineate conservation units in mottled ducks (Anas fulvigula).Biological Conservation, 203, 272–281. https://doi.org/10.1016/j.biocon.2016.10.003

Peters,J.L.,Zhuravlev,Y.,Fefelov,I.,Humphries,E.M.,&Omland,K.E.(2008).MultilocusphylogeographyofaHolarcticduck:ColonizationofNorthAmericafromEurasiabygadwalls(Anas strepera).Evolution,62,1469–1483.

Pfeifer,B.,Wittelsbürger,U.,Ramos‐Onsins,S.E.,&Lercher,M.J.(2014).PopGenome:AnefficientSwissarmyknifeforpopulationgenomicanalysesinR.Molecular Biology and Evolution,31,1929–1936.https://doi.org/10.1093/molbev/msu136

Phillimore,A.B.,&Owens,I.P.F.(2006).Aresubspeciesusefulinevolu‐tionaryandconservationbiology?Proceedings of the Royal Society B: Biological Sciences,273(1590),1049–1053.https://doi.org/10.1098/rspb.2005.3425

Posada,D.,&Crandall,K.A.(2001).Intraspecificgenegenealogies:Treesgrafting into networks. Trends in Ecology and Evolution,16, 37–45.https://doi.org/10.1016/S0169‐5347(00)02026‐7

Purcell, S., Neale, B., Todd‐Brown, K., Thomas, L., Ferreira, M. A. R.,Bender,D.,…Sham,P.C. (2007). PLINK:A tool set forwhole‐ge‐nome association and population‐based linkage analyses. The American Journal of Human Genetics, 81, 559–575. https://doi.org/10.1086/519795

Quiroga,M.P.(2010).Comparisonbetweentheaustraltemperateforestandthetucumano‐bolivianoforestthroughgenerawithspeciesofarborealhabits.Gayana Botánica,67(2),176–187.

Rabassa, J.,Coronato,A.,&Martínez,O. (2011).LateCenozoicglacia‐tions in the Patagonia and Tierra del Fuego: An updated review.Biological Journal of the Linnean Society,103,315–335.

Rabassa,J.,Coronato,A.M.,&Salemme,M.(2005).ChronologyofthelateCenozoicPatagonianglaciationsandtheircorrelationwithbio‐stratigraphic units of the pampean region (Argentina). Journal of South American Earth Sciences,20,81–103.https://doi.org/10.1016/j.jsames.2005.07.004

Rambaut,A.(2007).Se‐Al 2.0a11.Oxford,UK:UniversityofOxford.Ramírez,L.M.,Botero,A.,&Kattan,G.(2014).Distribuciónyabundancia

delpatode los torrentesMerganetta armata (Aves:Anatidae)enelRíoQuindio,Colombia.Boletín Científico Museo de Historia Natural,18(2),172–180.

Ridgely,R.S.,&Tudor,G.(2009).The field guide to the songbirds of South America: The passerines.Austin,TX:UniversityofTexasPress.

Rivera,J.A.,Penalba,O.C.,Villalba,R.,&Araneo,D.C.(2017).Spatio‐Temporalpatternsofthe2010–2015extremehydrologicaldroughtacross the central Andes, Argentina. Water, 9, 652. https://doi.org/10.3390/w9090652

Roig‐Juñent,S.,Flores,G.,Claver,S.,Debandi,G.,&Marvaldi,A.(2001).Monte Desert (Argentina): Insect biodiversity and natural areas.Journal of Arid Environments, 47, 77–94. https://doi.org/10.1006/jare.2000.0688

Sæther, B.‐E., Lande, R., Engren, S., Weimerskirch, H., Magnar, L.,Altwegg, L.,…Visser,M.E. (2005).Generation timeand temporalscalingofbirdpopulationdynamics.Nature,436,99–102.https://doi.org/10.1038/nature03666

Seltzer, G. O. (1990). Recent glacial history and paleoclimate of thePeruvian‐Bolivian Andes. Quaternary Science Reviews, 9, 137–152.https://doi.org/10.1016/0277‐3791(90)90015‐3

Shaffer,M.L.(1981).Minimumpopulationsizesforspeciesconservation.BioScience,31,131–134.https://doi.org/10.2307/1308256

Slatkin,M. (1987).Gene flow and the geographic structure of naturalpopulations. Science, 236, 787–792. https://doi.org/10.1126/science.3576198

Slatkin,M.(1993).Isolationbydistanceequilibriumandnon‐equilibriumpopulations.Evolution,47,264–279.

Soulé,M.E.,&Mills,L.S.(1998).Populationsgenetics–Noneedtoiso‐lategenetics.Science,282,1658–1659.

Sullivan, S.M.P.,Watzin,M.C.,&Keeton,W.S. (2007).A riverscapeperspectiveonhabitatassociationamongriverinebirdassemblagesintheLakeChamplainBasin,USA.Landscape Ecology,22,1169–1186.

Tamura,K.,&Nei,M.(1993).Estimationofthenumberofnucleotidesub‐stitutionsinthecontrolregionofmitochondrialDNAinhumansandchimpanzees.Molecular Biology and Evolution,10,512–526.

Ubeda,C.,Cerón,G.,&Trejo,A.(2007).DescripcióndeunNuevocom‐portamientoenhembradepatocortacorrientes(Merganetta armata,ANATIDAE).Boletín Chileno de Ornitología,13,47–49.

Valqui,T.(2008).Phylogeography of Nothoprocta Tinamous and the phylog‐eny of the Tinamidae.LSUDoctoralDissertations.2805.

Voelker, G. (2002). Molecular phylogenetics and the historical bio‐geography of dippers (Cinclus). Ibis, 144, 577–584. https://doi.org/10.1046/j.1474‐919X.2002.00084.x

Vuilleumier,F.(1998).AquantitativesurveyofspeciationphenomenainPatagonianbirds.Ornitologia Neotropical,2,5–28.

Weller,M.W. (1975).Habitat selectionbywaterfowlofArgentine IslaGrande.The Wilson Bulletin,87(1),83–90.

Wilson,R.E.,Peters,J.L.,&McCracken,K.G.(2012).Geneticandphe‐notypic divergencebetween low‐ andhigh‐altitudepopulationsoftworecentlydivergedcinnamontealsubspecies.Evolution,67,170–184.https://doi.org/10.1111/j.1558‐5646.2012.01740.x

Wright,S.(1943).Isolationbydistance.Genetics,28,114–138.Wright,S.(1969).Evolution and genetics of populations, vol. 2: The theory of

gene frequencies.Chicago,IL:UniversityofChicagoPress.Zachos,J.,Pagani,M.,Sloan,L.,Thomas,E.,&Billups,K.(2001).Trends,

rhythms,andaberrationsinglobalclimate65Matopresent.Science,292,686–693.https://doi.org/10.1126/science.1059412

Zhou, H., Alexander, D., & Lange, K. (2011). A quasi‐Newton ac‐celeration for high‐dimensional optimization algorithms.Statistics and Computing, 21, 261–273. https://doi.org/10.1007/s11222‐009‐9166‐3

Zwick,P.(1992).Streamhabitatfragmentation,athreattobiodiversity.Biodiversity and Conservation, 1, 80–97. https://doi.org/10.1007/BF00731036

SUPPORTING INFORMATION

Additional supporting information may be found online in theSupportingInformationsectionattheendofthearticle.

How to cite this article:AlzaL,LavretskyP,PetersJL,etal.Olddivergenceandrestrictedgeneflowbetweentorrentduck(Merganetta armata)subspeciesintheCentralandSouthernAndes.Ecol Evol. 2019;00:1–16. https://doi.org/10.1002/ece3.5538