Distributional patterns of Neotropical seasonally dry forest birds abiogeographical regionalization

David A Prieto-Torresab Octavio R Rojas-Sotoa Elisa Bonaccorsocd Diego Santiago-Alarcone and Adolfo G Navarro-Sigeurouenzab

aRed de Biologıa Evolutiva Laboratorio de Bioclimatologıa Instituto de Ecologıa AC carretera antigua a Coatepec No 351 El Haya 91070 Xalapa

Veracruz Mexico bMuseo de Zoologıa Departamento de Biologıa Evolutiva Facultad de Ciencias Universidad Nacional Autonoma de Mexico

Apartado Postal 70-399 Mexico City 04510 Mexico cLaboratorio de Biologıa Evolutiva Instituto BIOSFERA and Colegio de Ciencias Biologicas y

Ambientales Universidad San Francisco de Quito Diego de Robles y Vıa Interoceanica 17-1200-841 Quito Ecuador dBiodiversity Institute University

of Kansas Lawrence KS USA eRed de Biologıa y Conservacion de Vertebrados Instituto de Ecologıa AC El Haya Xalapa Veracruz Mexico

Accepted 12 November 2018

Abstract

Neotropical seasonally dry forests (NSDFs) are widely distributed and possess high levels of species richness and endemismhowever their biogeography remains only partially understood Using species distribution modelling and parsimony analysis ofendemicity we analysed the distributional patterns of the NSDF avifauna in order to identify their areas of endemism and pro-vide a better understanding of the historical relationships among those areas The strict consensus trees revealed 17 areas ofendemism for NSDFs which involve four large regions Baja California CaribbeanndashAntilles islands Mesoamerica and SouthAmerica These well-resolved clades are circumscribed by geographical and ecological barriers associated with the Gulf of Cali-fornia the leading edge of the Caribbean plate the Tehuantepec Isthmus the PolochicndashMotagua fault the Nicaragua Depres-sion the Choco forest the Amazon basin and the Andean Cordillera Relationships among groups of NSDFs found heresuggest that evolution of their avifauna involved a mixture of vicariance and dispersal events Our results support the idea ofindependent diversification patterns and biogeographical processes in each region including those previously associated with thePleistocene Arc Hypothesis for NSDFs of south-eastern South America This study provides a biogeographical framework toopen new lines of research related to the biotic diversification of NSDFscopy The Willi Hennig Society 2018

Introduction

Species heterogeneous distributions across Earthhave challenged biogeographers and ecologists tryingto unveil the evolutionary and ecological factors driv-ing patterns of biodiversity Although a variety of the-oretical and methodological approaches have beenapplied to diverse taxonomic groups the temporal andspatial diversification patterns of numerous biodiver-sity assemblages remain poorly understood (Wiens andDonoghue 2004 Weir and Hey 2006) This has been

the case for the avifauna associated with the Neotropi-cal seasonally dry forests (NSDFs) which involves highlevels of both species richness and endemism but hasreceived relatively little attention compared to the avi-fauna of other Neotropical ecosystems (Stotz et al1996 Herzog and Kessler 2002 Porzecanski andCracraft 2005 Rıos-Mu~noz andNavarro-Sigeurouenza 2012)Despite the vast accumulation of evidence (ie

palaeo-palynology climatology and genetic data) high-lighting the evolution and biological diversification ofplant taxa (Pennington et al 2006 2009 Wernecket al 2011 2012 Banda et al 2016) there are fewstudies addressing the biogeographical patterns of theavifauna associated with NSDFs at a continental scale(Ceballos 1995 Stotz et al 1996) Most studies have

Corresponding authorsE-mail addresses octaviorojasinecolmx (ORR-S) anddavidprietorresgmailcom (DAP-T)

CladisticsCladistics (2018) 1ndash15

101111cla12366

copy The Willi Hennig Society 2018

focused either on the taxonomy of species restricted toor associated with these forests or on analyses describ-ing distributional patterns of avifauna mainly withinisolated NSDF patches (eg Cracraft 1985 Herzogand Kessler 2002 Porzecanski and Cracraft 2005Rodrıguez-Ferraro and Blake 2008 Rıos-Mu~noz andNavarro-Sigeurouenza 2012 Oswald and Steadman 2015)Thus the evolution and biogeography of NSDFsbased on birds and other biological groups remainunclear exacerbating the precarious conservationstatus of these forests and their associated biota(Werneck et al 2011 Banda et al 2016) Under-standing the distribution and relationships amongNSDF birdsmdashincluding endemism patterns of speciesrichness and species turnovermdashwould provide relevantclues for identifying diversification centres and uniquebiodiversity areas which subsequently will aid in defin-ing priority conservation units at both regional andcontinental scales (Gordon and Ornelas 2000)Currently NSDFs show a discontinuous distribu-

tion from north-western Mexico to northern Argen-tina and from north-eastern to southern Brazil(Fig 1) The main NSDF patches are separated fromeach other by different ecosystems (eg humid mon-tane forests savannas and Amazonia rain forests)showing distinct plant species composition and phenol-ogy among patches in fact there are no two NSDFpatches that share more than half of their plant species(Pennington et al 2000 2006 2009 Banda et al

2016) Notwithstanding this Prado and Gibbs (1993)and Pennington et al (2000) highlighted a number ofunrelated NSDF tree species that are widespread inseveral of the disjunct NSDF areas These repeateddistribution patterns were considered as evidence of amore widespread and perhaps continuous forest for-mation with maximum extension during the Last Gla-cial Maximum (LGM) which may have spread atleast into Amazonia the Andean region and even tothe Caribbean coast (see Pennington et al 20002009)If any connections existed between some or all

NSDFs during recent geological time (ie the Pleis-tocene Arc Hypothesis) we might expect to find clearsimilarities in species composition (Prado and Gibbs1993 Pennington et al 2000 2009 Linares-Palominoet al 2011) Nevertheless diverse studies have foundthat the biota of NSDFs includes different speciesassociations and high levels of endemism inMesoamerica and South America (Ceballos 1995Herzog and Kessler 2002 Becerra 2005 Porzecanskiand Cracraft 2005 Rıos-Mu~noz and Navarro-Sigeurouenza 2012 Banda et al 2016 Oswald et al2017) suggesting that independent processes haveshaped species compositions and diversification pat-terns among NSDF biota Banda et al (2016) identi-fied 12 floristic groups with a clear northndashsouthdivision in NSDFs where the separation of a northerncluster (with four groups from Mexico to Colombia

Fig 1 Geographical distribution of Neotropical seasonally dry forests (NSDFs) showing the grid cells used for the parsimony analysis ofendemicity Numbers correspond to main NSDFs identified by Pennington et al (2000) and Banda et al (2016) Mexico (1) Central America(2) CaribbeanndashAntilles (3) Caribbean coast of Colombia and Venezuela (4) Inter-Andean valleys of Colombia (5) Pacific Equatorial (6) Inter-Andean valleys of southern Peru Apurimac-Mantaro (7) Sub-Andean Piedmont (8) the Chiquitano forests (9) Misiones Province (10) andCaatinga (11) Light grey represents the savanna and Chaco ecosystems Map modified from Pennington et al (2000)

2 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

and Venezuela including the Caribbean islands)reflects the effectiveness of the rain forests of Amazo-nia and the Choco as a barrier to migration In factPennington et al (2009) and Linares-Palomino et al(2011) suggested that widespread species in NSDFs arean exception only 143 of plant species are recordedin ~50 of the floristic nuclei defined across theregion This paucity of widespread NSDF speciesargues strongly against a widespread PleistoceneNSDF formation throughout the Neotropics rathersuggesting dispersal limitation in NSDF biota and amixed evolutionary history across their distributionwith ecological similarities arising from climatic con-vergence (Cracraft 1985 Mayle 2004 Becerra 2005Cortes et al 2015 Banda et al 2016 de Melo et al2016)More than 1000 bird species are known to inhabit the

NSDFs and this avifauna is composed of a mixture ofNSDF-restricted endemics lowland taxa and speciesthat are more common at higher or lower elevations(Stotz et al 1996) In addition several species consistof highly isolated populations often well differentiatedand with very limited or completely interrupted geneflow (Ribas et al 2009 Rodrıguez-Gomez et al 2013Navarro-Sigeurouenza et al 2017 OrsquoConnell et al 2017)Although much is known regarding the biology and dis-tribution of these birds no recent and thorough distri-butional analysis has been performed within a cladisticframework Cracraft (1985) produced the first analysisof distributional patterns and areas of endemism of theSouth American avifauna whereas Rıos-Mu~noz andNavarro-Sigeurouenza (2012) analysed species richness pat-terns in Mesoamerica suggesting a first biogeographicalregionalization for these forests in the region Howeverthese studies have only covered a fraction of the distri-bution of NSDFs and lack the general perspective nec-essary to explore the historical assemblage of NSDFavifaunas Only studying the distribution patterns ofbird species inhabiting the entire NSDFs could allowthe identification of particular regions (areas of ende-mism) with common spatial patterns for their associatedtaxa and facilitate the recognition of those processes(eg vicariance dispersal extinction) that promote theirevolution (eg Rosen and Smith 1988 Rojas-Sotoet al 2003 Sanchez-Gonzalez et al 2008 Morrone2014b)Methods using parsimony algorithms have been pro-

posed as valuable tools for detecting similaritiesamong biogeographical areas based on their associatedtaxa One of these methods is parsimony analysis ofendemicity or PAE (Rosen and Smith 1988 Morrone2014c) Despite criticisms in its application (eg itignores evolutionary relationships among species andis applicable only to closed ecosystems to which thelineages under analysis should be endemic see Brooksand van Veller 2003 Peterson 2008) diverse studies

have argued for its utility for establishing biogeo-graphical classifications and the relationships amongthese units (Morrone and Escalante 2002 Rojas-Sotoet al 2003 Nihei 2006 Morrone 2014c) particularlywhen phylogenies for all groups are unavailable Infact PAE has been increasingly used to understandbiological similarities between biogeographical areasbecause the results obtained consist of nested or hier-archical groups of biota where the congruent speciesrsquodistributional patterns demonstrate that taxa havebeen affected by common factors and terminaldichotomies are interpreted as areas sharing the mostrecent biogeographical history (Rosen and Smith1988 Morrone 2005 2014c)In this study we integrated species distribution mod-

els (SDMs) of ~1300 birds and PAE to identify areasof endemism for NSDF avifauna as well as their rela-tionships performing a regionalization of NSDF birdsacross the whole Neotropical region Given that pat-terns found in previous studies with NSDF plants(Linares-Palomino et al 2011 Banda et al 2016)have proved to be affected by dispersal to some extentwe hypothesized that birds might exhibit similar pat-terns where the potential areas of endemism must becircumscribed by geographical and ecological barriersthat have shaped the species composition and diversifi-cation patterns among NSDF avifauna Our resultsmay be used as a framework for comparison withindependent sets of evidence such as area cladogramsobtained through cladistic biogeographical and phylo-geographical studies (Rosen and Smith 1988 Morroneand Escalante 2002) Finally these results provide abetter understanding of the biogeographical relation-ships of NSDFs and their avifauna which alsoenhance the conceptual bases for future conservationdecisionsmdashtaking into account continental-level speciespatternsmdashto manage this highly threatened ecosystem

Methods

Definition of NSDFs

Given the difficulties in defining NSDF limits fromother ecosystemsmdashincluding savannas montane for-ests mangroves and more recently agricultural landmdashwe adopted a broad definition that includes all forestshaving a closed canopy typically dominated (gt 50)by semi-deciduous and deciduous trees that arepresent in frost-free areas with a mean annualtemperature gt 25 degC total annual precipitation of700ndash2000 mm and at least three or more dry months(precipitation lt 100 mm) per year (Murphy and Lugo1986 Sanchez-Azofeifa et al 2005 2013 Banda et al2016) This vulnerable ecosystem which encompasses42 ecoregions according to Olson et al (2001) is

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 3

discontinuously distributed in 18 countries (includingthe Caribbean islands) across the continent (Fig 1)

Species data gathering

We created a complete list of the bird speciesinhabiting NSDFs which was compiled from sourcesthat offer information on the habitat characteristicsfor each species (eg Stotz et al 1996 Gill and Dons-ker 2015) and from a database of presence records(see below) Then in a second selection we excludedall the species marginally inhabiting NSDFs and thoseof seasonal presence (ie inter-tropical migrants)because they may bias the identification of patterns ofendemicity (Rojas-Soto et al 2003 Sanchez-Gonzalezet al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza2012) Species marginally inhabiting NSDFs weredefined as those whose geographical ranges includedless than 10 of NSDFs (see Species distributionmodels section below) because their distribution corre-sponded to occasional or accidental records withinNSDF areas We followed the taxonomy of Gill andDonsker (2015) for the birds of Mexico and CentralAmerica as well as the South American ClassificationCommittee bird list (SACC Remsen et al 2017) andthe Clements Checklist (Clements et al 2015) for thebirds of South AmericaOccurrence records for each species were gathered

from (1) the Atlas of the Birds of Mexico (Navarro-Sigeurouenza et al 2002 2003) (2) Atlas de Registro deAves Brasileiras (ARA httparacemavegovbr) (3)specimen records directly obtained from ornithologicalcollections worldwide (see Appendix S1) (4) recordsobtained through the first authorrsquos fieldwork in Mex-ico and Venezuela and (5) online scientific collectiondatabases [ie Global Biodiversity Information Facility(GBIF) eBird and SiB-Colombia (sibcolombianet)]For all species records repeated in multiple sourceswere removed retaining only unique localities (within1 km2) In addition to identify problematic or impre-cise species occurrences we compared the spatial dis-tribution of obtained records with the ranges forspecies as defined in the websites Neotropical Birds(httpsneotropicalbirdscornelledu) and BirdLifeInternational (httpswwwbirdlifeorg) removing allthose mismatched records We omitted those recordswith geographical information that could not be veri-fied as well as those records without bioclimatic dataGeographical coordinates were transformed to decimaldegrees based on the WGS84 datum

Species distribution models

Traditionally PAE has been performed using datafrom individual specimens and localities Unfortu-nately for most species few observations andor

specimens are available and when they exist data aregenerally biased by site accessibility (Peterson 2001)Thus considering the immense efforts required todefine maps for the speciesrsquo distributional ranges theuse of computational algorithms to generate speciesdistribution models (SDMs) represent a good alterna-tive to obtain accurate speciesrsquo distribution maps(Peterson 2001 Soberon and Peterson 2005) UsingSDMs has the advantage of minimizing spatial gapsinherent to speciesrsquo distributional information improv-ing the resolution of biogeographical hypotheses inPAE by filling in poorly known andor non-surveyedareas (eg Rojas-Soto et al 2003 Sanchez-Gonzalezet al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza 2012)Although recent studies have shown that there are

uncertainties when forecasting species distributionsdepending on the algorithm used (Heikkinen et al2006) we decided to use the maximum entropymachine-learning algorithm (MaxEnt) given its provenperformance in calculating the most likely distributionof the focal species as a function of occurrence locali-ties and environmental variables (Elith et al 20062011 Elith and Leathwick 2007) For each species weobtained a potential SDM using MaxEnt 333k(Phillips et al 2006) To characterize the potential dis-tribution based on ecological niche modelling we usedthe environmental data from WorldClim 14 (Hijmanset al 2005) which includes a set of 19 climatic vari-ables summarizing aspects of precipitation and temper-ature at 30-arc second resolution (~1 km2 cell size)Although procedures for SDMs using the 19 climaticvariables have been discussed extensively elsewhere(eg Graham 2003b Peterson et al 2011 Dormannet al 2013) we used all 19 variables to simplify theprocess of modelling the distribution of the large num-ber of species selected for analyses (~1300) The Max-Ent algorithm compensates for co-linearity betweenvariables using a method for regularization that dealswith feature selection ranking the contribution of eachone throughout the analysis thus there is less need toremove correlated variables (Elith et al 2011)On the other hand given that SDMs must consider

historical factors affecting the species distributions weused specific areas for model calibration (ie accessiblearea or M) for each species (Soberon and Peterson2005 Barve et al 2011) These calibration areas wereestablished based on the intersection of occurrencerecords with the Terrestrial Ecoregions (Olson et al2001) and the Biogeographical Provinces of theNeotropical region (Morrone 2014a) Such considera-tion was based on the assumption that these regionsmay define the historical accessible areas for each spe-cies in geographical spaceBecause low sample size in occurrence records may

affect model performance (Pearson et al 2007 Owenset al 2013) we only modelled species with five or

4 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

and Venezuela including the Caribbean islands)reflects the effectiveness of the rain forests of Amazo-nia and the Choco as a barrier to migration In factPennington et al (2009) and Linares-Palomino et al(2011) suggested that widespread species in NSDFs arean exception only 143 of plant species are recordedin ~50 of the floristic nuclei defined across theregion This paucity of widespread NSDF speciesargues strongly against a widespread PleistoceneNSDF formation throughout the Neotropics rathersuggesting dispersal limitation in NSDF biota and amixed evolutionary history across their distributionwith ecological similarities arising from climatic con-vergence (Cracraft 1985 Mayle 2004 Becerra 2005Cortes et al 2015 Banda et al 2016 de Melo et al2016)More than 1000 bird species are known to inhabit the

NSDFs and this avifauna is composed of a mixture ofNSDF-restricted endemics lowland taxa and speciesthat are more common at higher or lower elevations(Stotz et al 1996) In addition several species consistof highly isolated populations often well differentiatedand with very limited or completely interrupted geneflow (Ribas et al 2009 Rodrıguez-Gomez et al 2013Navarro-Sigeurouenza et al 2017 OrsquoConnell et al 2017)Although much is known regarding the biology and dis-tribution of these birds no recent and thorough distri-butional analysis has been performed within a cladisticframework Cracraft (1985) produced the first analysisof distributional patterns and areas of endemism of theSouth American avifauna whereas Rıos-Mu~noz andNavarro-Sigeurouenza (2012) analysed species richness pat-terns in Mesoamerica suggesting a first biogeographicalregionalization for these forests in the region Howeverthese studies have only covered a fraction of the distri-bution of NSDFs and lack the general perspective nec-essary to explore the historical assemblage of NSDFavifaunas Only studying the distribution patterns ofbird species inhabiting the entire NSDFs could allowthe identification of particular regions (areas of ende-mism) with common spatial patterns for their associatedtaxa and facilitate the recognition of those processes(eg vicariance dispersal extinction) that promote theirevolution (eg Rosen and Smith 1988 Rojas-Sotoet al 2003 Sanchez-Gonzalez et al 2008 Morrone2014b)Methods using parsimony algorithms have been pro-

posed as valuable tools for detecting similaritiesamong biogeographical areas based on their associatedtaxa One of these methods is parsimony analysis ofendemicity or PAE (Rosen and Smith 1988 Morrone2014c) Despite criticisms in its application (eg itignores evolutionary relationships among species andis applicable only to closed ecosystems to which thelineages under analysis should be endemic see Brooksand van Veller 2003 Peterson 2008) diverse studies

have argued for its utility for establishing biogeo-graphical classifications and the relationships amongthese units (Morrone and Escalante 2002 Rojas-Sotoet al 2003 Nihei 2006 Morrone 2014c) particularlywhen phylogenies for all groups are unavailable Infact PAE has been increasingly used to understandbiological similarities between biogeographical areasbecause the results obtained consist of nested or hier-archical groups of biota where the congruent speciesrsquodistributional patterns demonstrate that taxa havebeen affected by common factors and terminaldichotomies are interpreted as areas sharing the mostrecent biogeographical history (Rosen and Smith1988 Morrone 2005 2014c)In this study we integrated species distribution mod-

els (SDMs) of ~1300 birds and PAE to identify areasof endemism for NSDF avifauna as well as their rela-tionships performing a regionalization of NSDF birdsacross the whole Neotropical region Given that pat-terns found in previous studies with NSDF plants(Linares-Palomino et al 2011 Banda et al 2016)have proved to be affected by dispersal to some extentwe hypothesized that birds might exhibit similar pat-terns where the potential areas of endemism must becircumscribed by geographical and ecological barriersthat have shaped the species composition and diversifi-cation patterns among NSDF avifauna Our resultsmay be used as a framework for comparison withindependent sets of evidence such as area cladogramsobtained through cladistic biogeographical and phylo-geographical studies (Rosen and Smith 1988 Morroneand Escalante 2002) Finally these results provide abetter understanding of the biogeographical relation-ships of NSDFs and their avifauna which alsoenhance the conceptual bases for future conservationdecisionsmdashtaking into account continental-level speciespatternsmdashto manage this highly threatened ecosystem

Methods

Definition of NSDFs

Given the difficulties in defining NSDF limits fromother ecosystemsmdashincluding savannas montane for-ests mangroves and more recently agricultural landmdashwe adopted a broad definition that includes all forestshaving a closed canopy typically dominated (gt 50)by semi-deciduous and deciduous trees that arepresent in frost-free areas with a mean annualtemperature gt 25 degC total annual precipitation of700ndash2000 mm and at least three or more dry months(precipitation lt 100 mm) per year (Murphy and Lugo1986 Sanchez-Azofeifa et al 2005 2013 Banda et al2016) This vulnerable ecosystem which encompasses42 ecoregions according to Olson et al (2001) is

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 3

discontinuously distributed in 18 countries (includingthe Caribbean islands) across the continent (Fig 1)

Species data gathering

We created a complete list of the bird speciesinhabiting NSDFs which was compiled from sourcesthat offer information on the habitat characteristicsfor each species (eg Stotz et al 1996 Gill and Dons-ker 2015) and from a database of presence records(see below) Then in a second selection we excludedall the species marginally inhabiting NSDFs and thoseof seasonal presence (ie inter-tropical migrants)because they may bias the identification of patterns ofendemicity (Rojas-Soto et al 2003 Sanchez-Gonzalezet al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza2012) Species marginally inhabiting NSDFs weredefined as those whose geographical ranges includedless than 10 of NSDFs (see Species distributionmodels section below) because their distribution corre-sponded to occasional or accidental records withinNSDF areas We followed the taxonomy of Gill andDonsker (2015) for the birds of Mexico and CentralAmerica as well as the South American ClassificationCommittee bird list (SACC Remsen et al 2017) andthe Clements Checklist (Clements et al 2015) for thebirds of South AmericaOccurrence records for each species were gathered

from (1) the Atlas of the Birds of Mexico (Navarro-Sigeurouenza et al 2002 2003) (2) Atlas de Registro deAves Brasileiras (ARA httparacemavegovbr) (3)specimen records directly obtained from ornithologicalcollections worldwide (see Appendix S1) (4) recordsobtained through the first authorrsquos fieldwork in Mex-ico and Venezuela and (5) online scientific collectiondatabases [ie Global Biodiversity Information Facility(GBIF) eBird and SiB-Colombia (sibcolombianet)]For all species records repeated in multiple sourceswere removed retaining only unique localities (within1 km2) In addition to identify problematic or impre-cise species occurrences we compared the spatial dis-tribution of obtained records with the ranges forspecies as defined in the websites Neotropical Birds(httpsneotropicalbirdscornelledu) and BirdLifeInternational (httpswwwbirdlifeorg) removing allthose mismatched records We omitted those recordswith geographical information that could not be veri-fied as well as those records without bioclimatic dataGeographical coordinates were transformed to decimaldegrees based on the WGS84 datum

Species distribution models

Traditionally PAE has been performed using datafrom individual specimens and localities Unfortu-nately for most species few observations andor

specimens are available and when they exist data aregenerally biased by site accessibility (Peterson 2001)Thus considering the immense efforts required todefine maps for the speciesrsquo distributional ranges theuse of computational algorithms to generate speciesdistribution models (SDMs) represent a good alterna-tive to obtain accurate speciesrsquo distribution maps(Peterson 2001 Soberon and Peterson 2005) UsingSDMs has the advantage of minimizing spatial gapsinherent to speciesrsquo distributional information improv-ing the resolution of biogeographical hypotheses inPAE by filling in poorly known andor non-surveyedareas (eg Rojas-Soto et al 2003 Sanchez-Gonzalezet al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza 2012)Although recent studies have shown that there are

uncertainties when forecasting species distributionsdepending on the algorithm used (Heikkinen et al2006) we decided to use the maximum entropymachine-learning algorithm (MaxEnt) given its provenperformance in calculating the most likely distributionof the focal species as a function of occurrence locali-ties and environmental variables (Elith et al 20062011 Elith and Leathwick 2007) For each species weobtained a potential SDM using MaxEnt 333k(Phillips et al 2006) To characterize the potential dis-tribution based on ecological niche modelling we usedthe environmental data from WorldClim 14 (Hijmanset al 2005) which includes a set of 19 climatic vari-ables summarizing aspects of precipitation and temper-ature at 30-arc second resolution (~1 km2 cell size)Although procedures for SDMs using the 19 climaticvariables have been discussed extensively elsewhere(eg Graham 2003b Peterson et al 2011 Dormannet al 2013) we used all 19 variables to simplify theprocess of modelling the distribution of the large num-ber of species selected for analyses (~1300) The Max-Ent algorithm compensates for co-linearity betweenvariables using a method for regularization that dealswith feature selection ranking the contribution of eachone throughout the analysis thus there is less need toremove correlated variables (Elith et al 2011)On the other hand given that SDMs must consider

historical factors affecting the species distributions weused specific areas for model calibration (ie accessiblearea or M) for each species (Soberon and Peterson2005 Barve et al 2011) These calibration areas wereestablished based on the intersection of occurrencerecords with the Terrestrial Ecoregions (Olson et al2001) and the Biogeographical Provinces of theNeotropical region (Morrone 2014a) Such considera-tion was based on the assumption that these regionsmay define the historical accessible areas for each spe-cies in geographical spaceBecause low sample size in occurrence records may

affect model performance (Pearson et al 2007 Owenset al 2013) we only modelled species with five or

4 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

discontinuously distributed in 18 countries (includingthe Caribbean islands) across the continent (Fig 1)

Species data gathering

We created a complete list of the bird speciesinhabiting NSDFs which was compiled from sourcesthat offer information on the habitat characteristicsfor each species (eg Stotz et al 1996 Gill and Dons-ker 2015) and from a database of presence records(see below) Then in a second selection we excludedall the species marginally inhabiting NSDFs and thoseof seasonal presence (ie inter-tropical migrants)because they may bias the identification of patterns ofendemicity (Rojas-Soto et al 2003 Sanchez-Gonzalezet al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza2012) Species marginally inhabiting NSDFs weredefined as those whose geographical ranges includedless than 10 of NSDFs (see Species distributionmodels section below) because their distribution corre-sponded to occasional or accidental records withinNSDF areas We followed the taxonomy of Gill andDonsker (2015) for the birds of Mexico and CentralAmerica as well as the South American ClassificationCommittee bird list (SACC Remsen et al 2017) andthe Clements Checklist (Clements et al 2015) for thebirds of South AmericaOccurrence records for each species were gathered

from (1) the Atlas of the Birds of Mexico (Navarro-Sigeurouenza et al 2002 2003) (2) Atlas de Registro deAves Brasileiras (ARA httparacemavegovbr) (3)specimen records directly obtained from ornithologicalcollections worldwide (see Appendix S1) (4) recordsobtained through the first authorrsquos fieldwork in Mex-ico and Venezuela and (5) online scientific collectiondatabases [ie Global Biodiversity Information Facility(GBIF) eBird and SiB-Colombia (sibcolombianet)]For all species records repeated in multiple sourceswere removed retaining only unique localities (within1 km2) In addition to identify problematic or impre-cise species occurrences we compared the spatial dis-tribution of obtained records with the ranges forspecies as defined in the websites Neotropical Birds(httpsneotropicalbirdscornelledu) and BirdLifeInternational (httpswwwbirdlifeorg) removing allthose mismatched records We omitted those recordswith geographical information that could not be veri-fied as well as those records without bioclimatic dataGeographical coordinates were transformed to decimaldegrees based on the WGS84 datum

Species distribution models

Traditionally PAE has been performed using datafrom individual specimens and localities Unfortu-nately for most species few observations andor

specimens are available and when they exist data aregenerally biased by site accessibility (Peterson 2001)Thus considering the immense efforts required todefine maps for the speciesrsquo distributional ranges theuse of computational algorithms to generate speciesdistribution models (SDMs) represent a good alterna-tive to obtain accurate speciesrsquo distribution maps(Peterson 2001 Soberon and Peterson 2005) UsingSDMs has the advantage of minimizing spatial gapsinherent to speciesrsquo distributional information improv-ing the resolution of biogeographical hypotheses inPAE by filling in poorly known andor non-surveyedareas (eg Rojas-Soto et al 2003 Sanchez-Gonzalezet al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza 2012)Although recent studies have shown that there are

uncertainties when forecasting species distributionsdepending on the algorithm used (Heikkinen et al2006) we decided to use the maximum entropymachine-learning algorithm (MaxEnt) given its provenperformance in calculating the most likely distributionof the focal species as a function of occurrence locali-ties and environmental variables (Elith et al 20062011 Elith and Leathwick 2007) For each species weobtained a potential SDM using MaxEnt 333k(Phillips et al 2006) To characterize the potential dis-tribution based on ecological niche modelling we usedthe environmental data from WorldClim 14 (Hijmanset al 2005) which includes a set of 19 climatic vari-ables summarizing aspects of precipitation and temper-ature at 30-arc second resolution (~1 km2 cell size)Although procedures for SDMs using the 19 climaticvariables have been discussed extensively elsewhere(eg Graham 2003b Peterson et al 2011 Dormannet al 2013) we used all 19 variables to simplify theprocess of modelling the distribution of the large num-ber of species selected for analyses (~1300) The Max-Ent algorithm compensates for co-linearity betweenvariables using a method for regularization that dealswith feature selection ranking the contribution of eachone throughout the analysis thus there is less need toremove correlated variables (Elith et al 2011)On the other hand given that SDMs must consider

historical factors affecting the species distributions weused specific areas for model calibration (ie accessiblearea or M) for each species (Soberon and Peterson2005 Barve et al 2011) These calibration areas wereestablished based on the intersection of occurrencerecords with the Terrestrial Ecoregions (Olson et al2001) and the Biogeographical Provinces of theNeotropical region (Morrone 2014a) Such considera-tion was based on the assumption that these regionsmay define the historical accessible areas for each spe-cies in geographical spaceBecause low sample size in occurrence records may

affect model performance (Pearson et al 2007 Owenset al 2013) we only modelled species with five or

4 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

more records for those that were geographicallyrestricted (Stotz et al 1996 Rıos-Mu~noz andNavarro-Sigeurouenza 2012 Gill and Donsker 2015) Forwidely distributed species we only considered thosewith at least 20 independent occurrence records SDMsfor species that had between five and 20 records weredeveloped using all presence data and assessed with aJackknife test (Pearson et al 2007) while SDMs forspecies with gt 20 records were generated using a ran-dom sampling of 90 of the locality records formodel training and the remaining 10 for internalmodel evaluation (ie testing data) In this last caseperformance of the MaxEnt models was evaluated bycalculating the commission and omission error values(Anderson et al 2003) and the Partial-ROC curve test(Peterson et al 2008) All models were run with noextrapolation to avoid artificial projections of extremevalues of ecological variables (Peterson 2008 Elithet al 2011) and all other MaxEnt parameters were setto defaultIn all cases we used the logistic response to obtain

digital maps containing the values for habitat suitabil-ity (continuous probability from 0 to 1 Phillips et al2006 Elith et al 2011) which were subsequently con-verted into binary presencendashabsence data based on thelsquotenth percentile training presencersquo (TPTP) as thresholdvalue (Liu et al 2013) This threshold represents a cri-terion that minimizes commission errorsmdashrejecting thelowest (10) suitability values of training recordsmdashinour final binary maps allowing for a better recoveryof species distributional areas (Escalante et al 2013Owens et al 2013) To generate only the ldquobesthypothesis maprdquo for each species to use in subsequentanalysis we compared our final maps against availabledistributional information of each species (egCracraft 1985 Gordon and Ornelas 2000 Herzogand Kessler 2002 Schulenberg et al 2010 Gill andDonsker 2015 Herzog et al 2016) discarding thosewith high commission errors andor those that werestatistically not significantConsidering these final maps we determined the spe-

ciesrsquo ecosystem specificity in order to identify thoseNSDF-restricted species This step was developedbased on two approaches (1) calculating the degree ofcoverage of the geographical distributions for eachspecies across the Neotropical ecosystems and (2)using the modification proposed by Sanchez-Gonzalezand Navarro-Sigeurouenza (2009) for published endemicityindices (Crisp et al 2001 Linder 2001) For the firstapproach we divided the speciesrsquo range area predictedin each ecosystem by the total speciesrsquo range area pre-dicted in the SDM For the second approach weobtained the index of restriction (IR) by substitutingthe number of quadrants (see Crisp et al 2001Linder 2001) for the number of ecosystems in which agiven species was reported as present by our models

This information was compared with the databases ofStotz et al (1996) which offer information about thehabitat for each species in order of preference Forboth analyses the area per ecosystem was delimited byselecting the terrestrial ecoregions of the world (Olsonet al 2001) which provides a framework for the iden-tification of and comparisons among representativehabitats along with the approximate boundaries ofnatural communities and ecosystems prior to majorland-use change Based on this idea we subdividedthe terrestrial Neotropical area into 14 major ecosys-temsmdashwhich involved 263 ecoregions (Olson et al2001)mdashand overlapped them with each SDM map

Parsimony analysis of endemicity (PAE)

To assess the relationships among NSDFs and toidentify areas of endemism we implemented a PAE(Rosen and Smith 1988 Morrone and Escalante2002) This method joins areas (analogous to terminaltaxa) based on their shared taxa (analogous to charac-ters) according to the most parsimonious explanation(Morrone 2014c) First we divided the study area into563 grid cells of size 1deg 9 1deg (Fig 1) and constructeda species presenceabsence binary matrix (coded as ldquo1rdquoif present or ldquo0rdquo if absent) on each cell (AppendicesS2 and S3) based on the individual SDM maps for allspecies (hereafter ldquocomplete species matrixrdquo) From acontinental (broad) scale this grid size could be con-sidered a fine-grained resolution which allows us toconnect small cells when locality data points are scat-tered and produce low data resolution (Gutierrez-Velazquez et al 2013)Then given that Pennington and Lavin (2016)

argued thatmdashat least for treesmdashNSDFs show phyloge-nies that are geographically structured (contrary to thepattern found in rain forests and savannas) we builtan additional binary matrix employing only those spe-cies defined as NSDF-restricted (hereafter ldquorestrictedspecies matrixrdquo see Appendix S2) This last step isimportant because it allows us to assess and comparethe distributional patterns of NSDF avifauna consider-ing different but parallel histories providing a betterunderstanding of the historical relationships amonggeographical areas (Ricklefs 1987 Wiens and Dono-ghue 2004 Weir and Hey 2006 OrsquoConnell et al2017) For both cases an extra hypothetical areacoded only with ldquo0rdquo was added to the matrices andused to root the cladogram which would be equivalentto a hypothetical ldquoancientrdquo area without taxa (Rosenand Smith 1988 Morrone and Escalante 2002Morrone 2014c)The data matrices were analysed with a heuristic

search based on the New Technology algorithm inTNT ver 15 software (Goloboff and Catalano 2016)finding the minimum length five times After

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 5

performing several tests surveying the weighting valuewe used a homoplasy penalization (k) equal to 3 asproposed by Goloboff (1993) which preserved theclades obtained with lowest values restricting homo-plasy severely and maximizing the number of synapo-morphies The obtained strict consensus trees wereanalysed in WinClada (Nixon 1999) and plotted(Appendices S4 and S5) in a geographical map usingArcMap 102 (ESRI 2010) Areas of endemism wererecovered based on at least two synapomorphies(Platnick 1991) also including taxa with consistencyindex (CI) = 05 (only if taxa showed reversions)because they may represent a distinctive feature ofoccurrence (autopomorphies) that is coded as uniquepresence to a given area (Escalante 2015) Nodal sup-port in cladograms was assessed by performing a boot-strap analysis with 1000 replicates in TNTBecause some authors have discussed the need to

increase the historical signal for finding relationshipsamong study areas and confer a ldquodynamic interpreta-tionrdquo to PAE (Rosen and Smith 1988 Cracraft 1991Nihei 2006 Vazquez-Miranda et al 2007) we per-formed two additional analyses taking into considera-tion the hierarchical information from genera andspecies together (Cracraft 1991 Porzecanski andCracraft 2005) In this sense our results from PAEcould be tested by a cladistic biogeographical analysisand compared with available phylogenies of taxainhabiting the areas investigated (Rosen and Smith1988 Morrone 2005 Sanchez-Gonzalez et al 2008)However we did not observe any difference in generalpatterns (ie NSDF regions identified) between theldquospeciesrdquo and ldquospecies-generardquo approaches (Appen-dices S4 and S5) The results shown herein are basedon strict consensus cladograms (ie complete speciesvs NSDF-restricted species) only from the analysesusing species informationFinally considering that patterns of relationships

among areas are easily retrieved within well-definedgeographical clades (Grismer 2000 Morrone 2014c)we performed two additional PAEs (ie complete andrestricted species approaches) based on the 17 NSDFgeographical groups or regions identified in the clado-grams for the 563 sampled grids (Appendix S4) Thislast step allowed us to increase the numbers of synapo-morphies grouping cell grids where data quantity islow and splitting quadrants that generate conflict(Morrone and Escalante 2002) determining the rela-tionships among areas (Morrone 2014c) The resultinggeographical groups were supported by taxa assynapomorphies (see Results Table 1) In all clado-grams we performed a bootstrap analysis with 1000replicates to assess nodal support The 17 NSDFregions identified herein were named based on a pro-posal by Pennington et al (2000) and Banda et al(2016)

Results

The occurrence database contained ~1 248 000records from 1298 bird species (from 78 families and511 genera) selected in this study For the 1242 speciescontaining more than 20 independent occurrencerecords all SDMs showed high AUC (Area UnderCurve for Receiver Operating Characteristic) values(ranking from 07 to 099) and Partial-ROC ratios(ranking from 115 to 199) For the remaining 56 spe-cies the Jackknife test showed that models were statis-tically significant (P lt 001) Thus performance valuesfor both modelling approaches indicated that modelsof the speciesrsquo potential distributions were accurateBased on the SDMs we observed that 364 of

bird species have at least 40 of their distributionwithin NSDFs while 304 were between 25 and40 and 331 had between 10 and 25 overlapwith NSDF areas Additionally the values of ecosys-temecological specificity showed that 409 of speciestended to be present in more than three ecosystems(IR lt 02) while only 137 are present between oneand two (IR gt 05) Most of the NSDF bird species(454) are present in at least three ecosystems(IR = 03) From these approaches we determinedthat ~43 (n = 557) of species initially considered arehighly associated andor endemic to NSDFs(Appendix S2) Final species selection was establishedbased on those with at least 33 of their distributionwithin the NSDF and an IR ge 03 where the NSDFrepresented the main ecosystem occupiedFor the complete species matrix approach (repre-

senting 563 sampled grids) we obtained a strict consen-sus cladogram of 16 367 steps [CI = 007 retentionindex (RI) = 086] while for the restricted speciesmatrix approach the consensus cladogram consisted of4172 steps (CI = 013 RI = 089) Both analysesrevealed 17 NSDF geographical groups showing~53 coincidence in the areas of endemism betweenthe two approaches (ie complete and restricted spe-cies) and containing almost the same synapomorphies(Table 1) Individual cladograms and their geographi-cal correspondence with the grid cell groupings forboth approaches are presented in Appendix S4Subsequently for PAE considering the 17 NSDF

regions in the complete species approach we obtained astrict consensus cladogram with 2040 steps (CI = 063RI = 075) while the restricted species approachshowed 705 steps (CI = 078 RI = 080) Both clado-grams were the base to propose our biotic regionaliza-tion of the NSDF (Fig 2) In this proposal weobserved that Baja California is the sister group (at thebase of the tree) of the rest of NSDF clades theCaribbeanndashAntilles islands are separated from the well-defined clade that includes the Mesoamerican and SouthAmerican areas The complete species cladogram

6 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

differed from the restricted one because the northern-most limit of NSDFs (SS group Fig 2a) appears in theformer as sister group of the Mesoamerican and SouthAmerican mainland grid cells whereas in the restrictedspecies cladogram these grid cells appear within north-western Mexico (Fig 2b)Following the nested pattern NSDFs showed a

clear division into two main well-defined clades onecontains the Mesoamerican region from north-westernMexico to Panama and another comprises all of theSouth American areas Focusing in the Mesoamericanregion we observed that the Yucatan Peninsula (YP)

is sister to all other NSDF groups located in CentralAmerica and north-western Mexico (Fig 2) Howeverit is noteworthy that the Panama region (Pa groupFig 2b) which represents the southernmost limit ofMesoamerican NSDF range is the most basal sub-clade for the restricted species approach whereas it isgeographically included within the South CentralAmerica group (CAmS in Fig 2a) for the completespecies cladogram However despite the slight differ-ences in topology between the two cladograms (com-plete and restricted species) for the main NSDF inMesoamerica where some are completely nested

Table 1Synapomorphic species for the clades (ie potential areas of endemism PAE) for avifauna of the Neotropical seasonally dry forests (NSDFs)considering both approaches (matrices of information) complete vs NSDF-restricted species Individual cladograms and the geographical corre-spondence obtained for the grid cells are presented in the supplementary material (Appendix S4)

NSDF region Species

Nodes in PAE

CompleteRestrictedspecies

CaribbeanndashAntillesIslands

Tachornis phoenicobia Quiscalus niger a ndash

Cuba (Cu) Coccyzus merlini Contopus caribaeus Corvus nasicus Icterus melanopsisMargarobyas lawrencii Melanerpes superciliaris Myiarchus sagrae Spindaliszena Tiaris canorus

c a

Haiti DominicanRepublic and PuertoRico (HP)

Amazona ventralis Chlorostilbon swainsonii Coccyzus longirostris Contopushispaniolensis Corvus leucognaphalus Dulus dominicus Euphonia musica Icterusdominicensis Melanerpes striatus Mellisuga minima Microligea palustris Myiarchusstolidus Nesoctites micromegas Patagioenas inornata Phaenicophilus palmarumPsittacara chloropterus Spinus dominicensis Todus angustirostris Todus subulatusTyto glaucops Vireo nanus

b ndash

Baja California (BC) Basilinna xantusii Geothlypis beldingi Toxostoma cinereum d ndashMesoamericanmainland

Amphispiza quinquestriata Camptostoma imberbe Megascops guatemalae Molothrusaeneus Pachyramphus aglaiae Stelgidopteryx serripennis Turdus rufopalliatus

e g

Sonora to Panama Arremonops rufivirgatus Geothlypis poliocephala Vireo hypochryseus f ndashNorthwestern Mexico(PS)

Amphispiza quinquestriata Polioptila nigriceps Turdus rufopalliatus ndash h

Northern Oaxaca (SP) Aimophila notosticta Calothorax pulcher ndash iYucatan Peninsula (YP) Cyanocorax yucatanicus Melanoptila glabrirostris Melanerpes pygmaeus Meleagris

ocellata Myiarchus yucatanensis Nyctiphrynus yucatanicusg j

Central America (CAm) Cantorchilus modestus Zimmerius vilissimus h ndashNorth Central America(CAmN)

Campylopterus rufus Icterus maculialatus i k

South Central America(CAmS)

Geothlypis chiriquensis Thamnophilus bridgesi j ndash

Northern SouthAmerica

Campylorhynchus griseus Icterus nigrogularis Myiarchus apicalis Picumnusgranadensis Picumnus squamulatus Sporophila intermedia

k e

Caribbean coast ofColombia andVenezuela (CCV)

Campylorhynchus nuchalis Crax daubentoni Crypturellus erythropus Euphoniatrinitatis Icterus icterus Inezia caudata Myiarchus venezuelensis Ortalisruficauda Saltator orenocensis Synallaxis cinnamomea Thamnophilusmelanonotus Thraupis glaucocolpa

l f

Pacific Equatorial (PE) Amazilia amazilia Arremon abeillei Basileuterus trifasciatus Chaetocercusbombus Columbina cruziana Contopus punensis Cyanocorax mystacalis Diveswarczewiczi Geothlypis auricularis Mecocerculus calopterus Melanopareiaelegans Mimus longicaudatus Myiodynastes bairdii Sturnella bellicosaThamnophilus bernardi

m c

Sub-Andean piedmont(SA-P)

Arremon torquatus Leptotila megalura Megascops hoyi Microstilbon burmeisteriPsilopsiagon aymara Rhynchotus maculicollis Sappho sparganurus

n b

Caatinga (CaB) Casiornis fuscus Herpsilochmus sellowi Pseudoseisura cristata Paroariadominicana Picumnus pygmaeus

o d

Synapomorhic species reported for both PAEs

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 7

(Fig 2a) and others showed some reciprocally mono-phyletic groups (Fig 2b) the division is clear betweenthe north-western Mexico [including the sections ofTamaulipas and Veracruz (TV) Pacific Mexican forest(PS) and Northern Oaxaca in southern Mexico (SP)]and the Central American (from southern Mexico toPanama) regions This last region is subsequentlydivided into two main subgroups (1) the northern(CAmN from southern Mexico to Nicaragua) and (2)the southern subgroup (CAmS from Nicaragua toPanama)Furthermore for both approaches we observed that

NSDFs in South America involved a well-resolvedclade divided into two reciprocally monophyleticgroups (Fig 2) One group comprises the grid cellsthat correspond to northern South America contain-ing all areas throughout the inter-Andean valleys(IAV from Colombia to the Caribbean coast) as wellas the north-western stretch of the Orinoco river (in-cluding the Cordillera de la Costa and the MaracaiboLake depression in Venezuela) and the dry forests

located in the western slope of the Sierra de Perija(from La Guajira to Santander-Cordoba Departmentsin Colombia CCV subgroup) The second group isformed by the forests located throughout the PacificEquatorial (PE) region (an area of endemism locatedfrom western Ecuador to north-western Peru includ-ing the central Andean Valleys) and a sister clade thatincludes the dry valleys located in southern Peru [ieApurımac-Mantaro and Tarapoto-Quillabamba forests(Ap-M)] as well as the south-eastern South Americaareas (ie NSDFs from western Bolivia and north-eastern Argentina to north-eastern Brazil) For thislast South American group two well-supported sub-groups are formed consecutively and separately onecontains the endemism area at the Sub-Andean pied-mont forests (SA-P) of north-western Argentina andwestern Bolivia and the other corresponds to theChiquitano and Misiones Province [ie south-easternBolivia north-eastern Argentina Paraguay and Uru-guay (C-MP)] and the endemism area of Caatinga(CaB located in north-eastern Brazil)

Fig 2 Regionalization of the Neotropical seasonally dry forests (NSDFs) based on the avifauna distribution and their relationships consideringcomplete (a) vs NSDF-restricted (b) species approaches (matrices of information) The cladograms revealed 17 NSDF geographical groupswhich involved four large regions The Baja California (A) CaribbeanndashAntilles islands (B) Mesoamerican (C) and South America (D) Acro-nyms in both cladograms and the map correspond to identified NSDF regions and areas of endemism [more than two synapomorphies (speciesrepresented with a black circle] BC = Baja California Cu = Cuba island HP = Haiti Dominican Republic and Puerto Rico islandsSS = Sonora and Sinaloa forests PS = Mexican Pacific slope lowlands SP = northern Oaxaca in southern Mexico TV = TamaulipasndashVeracruzstates YP = Yucatan Peninsula CAmN = northern Central America CAmS = nouthern Central America CAm = Central AmericaPa = Panama IAV = Inter-Andean Valleys of Colombia CCV = Caribbean ColombiandashVenezuela PE = Pacific Equatorial Ap-M = Apurimac-Mantaro SA-P = Sub-Andean Piedmont C-MP = Chiquitano forests and Misiones Province CaB = Brazilian Caatinga These well-resolvedclades are circumscribed by geographical and ecological barriers (dotted black lines) associated with the Gulf of California the leading edge ofthe Caribbean plate the Tehuantepec Isthmus the PolochicndashMotagua fault the Nicaragua Depression the Choco forest the Amazon basin andthe Andean Cordillera (including the Inter-Andean Valleys) Well-supported clades [more than two synapomorphies (species)] are shown with ablack circle in the acronyms of NSDF regions (see Table 1 for a complete description)

8 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

Discussion

The Baja Californiarsquos NSDF

In regional cladograms the Cape of Baja Californiaforms a distinct unit which supports the idea of asomewhat independent biogeographical history as hasbeen suggested in previous avian regionalizations Iso-lation of the NSDF in the Cape region from those inwestern Mexico (55ndash40 Mya) by the Gulf of Califor-nia has promoted the differentiation and a high degreeof endemism associated with the Cape of diversegroups including mammals birds snakes and insects(Grismer 2000 Rojas-Soto et al 2003 Rıos-Mu~nozand Navarro-Sigeurouenza 2012 OrsquoConnell et al 2017)However given the closest affinities of avian specieswith the nearby continent at the northern portion ofthe Peninsula (Rojas-Soto et al 2003) it seems clearthat the Cape avifauna originated through a terrestrialroute from the northern to the southern tip of the BajaCalifornia Peninsula (Rojas-Soto et al 2003OrsquoConnell et al 2017)

The CaribbeanndashAntilles avifauna

As found in previous biogeographical studies (egCrother and Guyer 1996 Zink et al 2000 Graham2003ab Vazquez-Miranda et al 2007) our analysessupport a monophyletic origin for groups of organismsinhabiting this region However a potential difficultyin explaining this monophyly is posited by the ques-tion of how birds reached those islands given that thecomplex geological history of the region included ldquoatleast eight events of fragmentation and seven ofhybridizationrdquo (Rosen 1985) In our study weobserved that several taxa acted as homoplasies whichcan be interpreted as the result of either dispersal orextinction events (Vazquez-Miranda et al 2007) Con-sidering that ~41 (n = 56 species) of NSDF avifaunais endemic and of these only eight species (two ofwhich are synapomorphies) were shared among thethree islands (Cuba Haiti Puerto Rico) the observedpattern is more probably explained by a mixture ofvicariance and dispersal events especially becausethose islands have had relatively recent bird exchanges(Vazquez-Miranda et al 2007) This idea is more par-simonious than assuming concordant dispersal pat-terns of individual taxa particularly considering thebiotarsquos affinities with both Central America andorSouth America (Crother and Guyer 1996 Zink et al2000 Graham 2003ab)Despite the existence of vicariance models explaining

affinities between the avifauna of this region with themainland (eg the temporal separation of the leadingedge of the Caribbean plate [beginning in the LateCretaceous and finishing in the MiocenendashMiddle

Pleiocene] Rosen 1985 Iturralde-Vinent and Mac-Phee 1999 Graham 2003ab Morrone 2014b) it isimportant to consider that dispersal events could havecontributed to defining the avifaunistic assemblages ofthis region as apparently occurred in the history ofthe woodpecker genus Melanerpes (Navarro-Sigeurouenzaet al 2017) It is difficult to compare the presentstudy with earlier studies because they used differentislands different taxa and different continental regionsas a reference Therefore further studies are needed toassess the relative age of these diversification centresand their biogeographical relationships

The Mesoamerican NSDFs

The division of this clade into sub-groups supports thelong isolation history of the identified Mesoamericanregions Our results highlight the importance of the Mex-ican Plateau and the Tehuantepec Isthmus in shapingthe distribution of Mesoamerican biotas in both high-lands and lowlands (Becerra 2005 Sanchez-Gonzalezet al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza 2012Banda et al 2016) The clear differentiation of thegroups on both sides of the Tehuantepec Isthmus(north-western Mexico Yucatan Peninsula and CentralAmerica Fig 2) as well as the subdivisions observedwithin the northern Mexican regions [Sonora-Sinaloa(SS) northern Oaxaca (SP) and TamaulipasndashVeracruz(TV) areas] might be explained by higher diversificationrates compared to other dry forest areas as detected forexample in the hummingbird Amazilia cyanocephala(Rodrıguez-Gomez et al 2013 OrsquoConnell et al 2017)Although the age of Mesoamerican NSDFs is notknown the fossil record and dated phylogenies suggest aMiocenendashPliocene age at least in north-western Mexico(Becerra 2005) These forests probably expanded toCentral America (where fossil evidence seems to suggestthat it invaded only after 25 Mya Graham and Dilcher1995) after the Sierra Madre Occidental and the Trans-mexican Volcanic belt were formed creating conditionsfor the establishment and persistence of NSDFsThe clear separation observed for the Sonora-Sina-

loa (SS) subgroup and the rest of the Mesoamericanmainland (Fig 2a) could be explained by the highaffinities with Nearctic biota of this northernmost area(Stotz et al 1996) In fact despite the high speciesturnover (n = 188) observed between the Sonora-Sina-loa subgroup and the rest of the MesoamericanNSDFs these northernmost forests shared 54 species(95) with those reported for the Baja Californiaregion Furthermore we propose a region formed byall areas west of the Tehuantepec Isthmus located inMexico [including NSDF in TamaulipasndashVeracruz(TV) and except those corresponding to Baja Califor-nia (BC)] which is supported by previous biogeo-graphical hypotheses and also suggested that the

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 9

north-eastern Gulf of Mexico is biogeographically dis-tinct from the southern and south-eastern part of theGulf coast (Liebherr 1994 Hernandez-Ba~nos et al1995 Marshall and Liebherr 2000 Rıos-Mu~noz andNavarro-Sigeurouenza 2012)On the other hand separation within the south-east-

ern Isthmus groups (light blues in Fig 2) supports thebiogeographical hypothesis of vicariant processes asso-ciated with the PolochicndashMotagua fault and the Nicar-agua Depression in the region (Morrone 2001Schuster et al 2003 Liede and Meve 2004) Here weidentified at least three distinct regions suggesting theexistence of bird assemblages with independent biogeo-graphical histories probably shaped by new immi-grants from dispersal events between South and NorthAmerica after the establishment of the Panama Isth-mus (Smith and Klicka 2010) For instance weobserved that ~73 of the NSDF avifauna is sharedbetween southern Central America and northern SouthAmerica (with ~72 of NSDF-restricted speciesshared between the NSDF from Panama and northernSouth America) confirming the existence of strongaffinities between the two regions (Linares-Palominoet al 2011 Rıos-Mu~noz and Navarro-Sigeurouenza 2012Banda et al 2016)In the Panama region differences in the affinities

between the cladograms for this region may be influ-enced by the inclusion (ie complete approach) or theexclusion (ie restricted approach) of species widelydistributed in the region and inhabiting other ecosys-tems surrounding NSDFs In this sense such differ-ences could be explained by different temporalpatterns of speciation documented for taxa associatedwith humid and dry habitats across the Isthmus ofPanama (Smith et al 2012) The Panama area hashad a history of expanding and contracting processesfor both the dry and the humid forests and even whenboth habitats consist of ldquoyoungrdquo and ldquooldrdquo speciesthe dry habitat species are geographically more distantfrom their nearest counterpart in Central America andnorthern South America compared with species fromthe humid habitats in these regions (Oswald andSteadman 2015 Oswald et al 2017) This phe-nomenon if generalized is probably an importantcontributor to biogeographical patterns in the south-ernmost section of the Mesoamerican region (Sanchez-Gonzalez et al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza 2012 Smith et al 2012)

South American NSDF areas and the Pleistocene ArcHypothesis

Previous studies of distributional patterns within theSouth American biota have accepted the hypothesisthat areas of endemism have arisen by isolation ofmore widespread biotas This scenario is commonly

associated with paleogeographical changes alterationsin river systems Pleistocene climatic fluctuations or amixture of all these processes (Herzog and Kessler2002 Werneck et al 2012) However we find littlesupport for widespread Pleistocene NSDF formationthroughout South America because few species(1033 n = 100 species) are widespread and sharedby more than 75 of South American areas Thisresult is concordant with NSDF plant patterns recentlyreported by Linares-Palomino et al (2011) and Bandaet al (2016) as well as with the restricted climatic cor-ridors found by Werneck et al (2011) which suggestthat LGM climate was probably too dry and cold tosupport large tracts of NSDFs across the continentSome authors suggest that long-distance ancient dis-

persal events of unrelated taxonomic groups are lesslikely than vicariant associations (Werneck et al2011 de Melo et al 2016) Likewise there is substan-tial support that NSDF is a dispersal-limited biomegiven the phylogenetic niche conservatism and stronggeographically structured phylogenies for woody plantclades (Pennington et al 2009 Pennington and Lavin2016) Thus an allopatric-vicariance explanationmdashde-lineated by major geographical or ecological barriersbefore the Pleistocenemdashseems more plausible toexplain the evolution and relationships among all theseforest areas which is supported by results based onendemic taxa For instance the biogeography of Bro-togeris parakeets implies a dynamic history for SouthAmerican biomes since the Pliocene corroboratingthat geological evolution of Amazonia has been impor-tant in shaping its biodiversity with taxa endemic toNSDFs and other biomes (Ribas et al 2009) In thissense our results suggest that the wide distributions ofa relatively low proportion of NSDF species and eco-logical similarity among them (including specialistsand generalists) may reflect only limited recent long-distance dispersal events and species exchange (Dicket al 2003 2004 Mayle 2004)The separation of a northern South American clade

(Fig 2) suggests an effect of the Amazonia rain foreststhe wet Choco forest and the final uplift of the north-ern Andes as barriers for the dispersal of dry forestspecies (Amorim 2001 Garzione et al 2008 Hoornet al 2010 Morrone 2014b Banda et al 2016) Clus-ters of the northern inter-Andean valleys (located alongthe Magdalena and Cauca Rivers) together with Carib-bean coastal areas reflect the heterogeneity of dryAndean regions (with important species turnover withMesoamerica and CaribbeanndashAntilles) where few spe-cies are distributed all the way from the northern tothe southern Andes (Nores 2004 Vazquez-Mirandaet al 2007) These results are also consistent withthose previously reported for other taxonomic groupsin the region that are associated with NSDFs (Linares-Palomino et al 2011 Morrone 2014b)

10 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

Although our analyses generally agree with thePleistocene Arc Hypothesis proposed originally byPrado and Gibbs (1993)mdashformation of a once exten-sive and largely contiguous seasonal woodland com-prising the Caatingas the Misiones and Sub-AndeanPiedmont nucleimdashthe clear separation among NSDFsfrom centralndashsouth-western and south-eastern SouthAmerica supports the idea that both uplift of theAndes and historical (Cenozoic and Quaternary) cli-matic regimes of north-western EcuadorndashPeru andwestern Bolivia-Argentina are probably responsible forthe high endemism and distributions of their NSDFbird communities (Hoorn et al 2010 Morrone2014b) These communities have been biogeographi-cally separated from neighbouring Bolivian lowlandsites (Herzog and Kessler 2002 Herzog et al 2016)In fact previous Andean NSDF legume phylogeniesshowed consistently high geographical structurethroughout their distribution suggesting that at leastthe NSDF flora has been assembled gradually over thepast ca 19 Myr with processes of in situ diversifica-tion largely undisturbed by new immigrants (Seuroarkinenet al 2012) From this perspective the NSDF on theslopes of centralndashsouth-western South America consti-tutes an important biogeographical area with ecologi-cal and geographical long-term stability and probablyoriginated by convergence scenarios resulting fromancient dispersal events that involved low niche differ-entiation and strong dispersal limitation from otherNSDF patches (Pennington et al 2009 Seuroarkinenet al 2012) This also suggests that NSDFs fromEcuador and northern Peru to western BoliviandashArgen-tina might constitute remnants of a previously widerexpansion according to the Pleistocene NSDF forma-tion (Seuroarkinen et al 2012 Cortes et al 2015 de Meloet al 2016) Passerine fossils support this hypothesisand suggest that some NSDF species had wider distri-butions in the past (Oswald and Steadman 2015) withthe Andean valleys (eg the Mara~non and ApurimacValleys) having direct influence on divergence andextinction processes by acting as semi-permeable barri-ers (Seuroarkinen et al 2011 Smith et al 2014 Wingerand Bates 2015)Finally we found a stretch of NSDF from north-

eastern Brazil south to ParaguayndashArgentina and northtowards eastern Bolivia (Fig 2) which is concordantwith previous studies showing a well-supported group-ing among these forests and suggesting the existence ofa historical stable unit known as the Open Dry Diago-nal (Mayle 2004 Porzecanski and Cracraft 2005Linares-Palomino et al 2011 Werneck et al 20112012) Today many dry habitat sister taxa are geo-graphically distant from one another suggesting thatthese species have either undergone long-distance dis-persal across large tracts of humid forests or that aridregions were once connected (Oswald et al 2017) The

absence of synapomorphies within the Chiquitano andMisiones Province forests (Table 1 Appendix S4) pro-vides some evidence for the recent formation of birdassemblages in this area (Herzog and Kessler 2002Smith et al 2014) as was found for woody plants(Linares-Palomino et al 2011 Banda et al 2016) Infact several authors suggest climatic and floristic char-acteristics closer to Atlantic rain forests than to NSDFsfor these areas (Olson et al 2001 Oliveira-Filho et al2006) which were probably colonized by dry forestspecies during drier and cooler periods of the Pleis-tocene These colonization processes were probablyfavoured by a potential narrow eastndashwest corridor con-necting NSDF in the extremes of north-eastern Braziland south-western South America during the LGM(Werneck et al 2011) This connection could serve asa biogeographical link sheltering floristic and faunisticdispersal routes resulting in partial expansions ofNSDFs in some areas of Amazonia (Werneck et al2011 2012) As a consequence of this last idea thelong-distance dispersal hypothesis of NSDFs in south-eastern South America proposed by Mayle (2004)remains a scenario that still needs independent testing

Final considerations

Our study suggests that evolution of the well-differ-entiated avifaunas in the four major geographicalgroups herein identified might have involved a mixtureof vicariance and dispersal events which provides newinsights into the diversification centres and historicalrelationships for the distributional patterns of thisecosystem and its associated avifauna Despite somespecies being shared among different clades (regions)and that this probably increased after closure of thePanama Isthmus (Morrone 2014b) and during themaximum extension of forests at the LGM (Bandaet al 2016) we observed a clear differentiation amongthe main areas with high degrees of endemism(Fig 2) This agrees with the idea of independent evo-lutionary histories among NSDF nuclei across theirdistribution (Becerra 2005 Cortes et al 2015 deMelo et al 2016)Differences observed with previous studies (eg

Porzecanski and Cracraft 2005 Rıos-Mu~noz andNavarro-Sigeurouenza 2012) are largely due to taxonomicuncertainties (ie species concepts and new species thathave been described or recognized since those studiesJetz et al 2002) as well as by differences in the defini-tion of sampling area sizes and algorithms used foranalyses For PAE there is a trend to decrease theabsolute number of steps and to increase the numberof synapomorphies as the size of the area increases(Morrone and Escalante 2002) Likewise it is impor-tant to note that this approach depends on detailed pri-mary biodiversity occurrence data In studies like this

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 11

with large data numbers it is worth noting thatobserved patterns of species distributional data couldbe influenced by diverse sources of error (eg taxo-nomic uncertainties identification errors or errors onlocality information associated with species occur-rences) Thus it is important to promote the constantupdate of the taxonomic status of specimens depositedin collections as well as the procedures for cleaningand revising biodiversity data from heterogeneoussources (Navarro-Sigeurouenza et al 2003 Bloom et al2017)Although our results mostly apply to the geographi-

cal scale and taxa used for this study the NSDF avi-fauna distribution patterns we found are non-randomand congruent with those for other organisms(Becerra 2005 Porzecanski and Cracraft 2005 Pen-nington et al 2009 Rıos-Mu~noz and Navarro-Sigeurouenza 2012 Oswald and Steadman 2015 Bandaet al 2016) The general pattern identified hereimplies the need to continue to study these areas ofendemism which certainly could have a much longerhistory than the temporal framework of the Pleis-tocene Thus from a conservation perspective mosthistorical areas of endemism must be considered aspriorities and incorporated into efforts of conserva-tions planning (eg Nori et al 2016) because theyreflect an unique history of the Earth and its biota Afailure to protect them would result in major losses ofunique species diversity for this highly threatenedecosystem The observed distributional patterns of theavifauna as well as their relationships across theNSDFs constitute evidence of the great complexity ofthe biogeography of this ecosystem in which each par-ticular pattern should be considered a stepping stonealong a true arc recognized as the biogeography ofthe Neotropics

Acknowledgments

We appreciate the efforts of museums (seeAppendix S1) and their curators as well as field collec-tors and institutions that made our study possibleDAP-T extends his gratitude to Consejo Nacional deCiencia y Tecnologıa (CONACyT Mexico) for a doc-toral scholarship (297538) and the Smithsonian-MasonSchool of Conservation for a scholarship during thecourse ldquoSpatial Ecology Geospatial Analysis andRemote Sensing for Conservation (MCCS 0500)rdquo thatprovided the tools and skills necessary for the spatialanalysis in this study The Rufford Foundation (DAP-T 16017-1 DAP-T 20284-2) Idea Wild (DAP-T) Con-sejo de Desarrollo Cientıfico Humanıstico y Tec-nologico at Universidad del Zulia (CONDES projectsCC-0247-13 and CC-0497-16 DAP-T) and the CONA-CyT project 152060 (AGNS and ORRS) provided

financial and logistical support J C Catarı P JCampos and Z Perez-Duran (Bolivia) as well as J ADelgado (Peru) helped during fieldwork A GordilloC Mota-Vargas and I Perozo helped in georeferencingbird locality data T Escalante helped with the PAEThe manuscript was improved by comments from HM Ortega-Andrade C Mota-Vargas A Lira-NoriegaJ L Parra L A Gonzalez and F Lopez

Conflicts of interest

The authors declare no competing interests in thisstudy

References

Amorim DS 2001 Dos Amazonias In Llorente J Morrone JJ(Eds) Introduccion a la Biogeografıa en LatinoamericaConceptos Teorıas Metodos y Aplicaciones Las Prensas deCiencias Universidad Nacional Autonoma de Mexico MexicoDF pp 245ndash255

Anderson RP Lew D Peterson AT 2003 Evaluatingpredictive models of speciesrsquo distributions criteria for selectingoptimal models Ecol Modell 162 211ndash232

Banda K Delgado-Salinas A Dexter KG Linares-Palomino ROliveira-Filho A Prado D Pullan M Quintana C Riina RRodrıguez M Weintritt J Acevedo-Rodrıguez P Adarve JAlvarez E Aranguren BA Arteaga JC Aymard G Casta~noA Ceballos-Mago N Cogollo A Cuadros H Delgado FDevia W Due~nas H Fajardo L Fernandez A FernandezMA Franklin J Freid EH Galetti LA Gonto RGonzalez-M R Graveson R Helmer EH Idarraga ALopez R Marcano-Vega H Martınez OG Maturo HMMcDonald M McLaren K Melo O Mijares F Mogni VMolina D Moreno NdP Nassar JM Neves DM OakleyLJ Oatham M Olvera-Luna AR Pezzini FF DominguezOJR Rıos ME Rivera O Rodrıguez N Rojas A SeuroarkinenT Sanchez R Smith M Vargas C Villanueva B PenningtonRT 2016 Plant diversity patterns in Neotropical dry forests andtheir conservation implications Science 353 1383ndash1387

Barve N Barve V Jimenez-Valverde A Lira-Noriega AMaher SP Peterson AT Soberon J Villalobos F 2011The crucial role of the accessible area in ecological nichemodeling and species distribution modeling Ecol Modell 2221810ndash1819

Becerra JX 2005 Timing the origin and expansion of the Mexicantropical dry forest Proc Natl Acad Sci USA 102 10919ndash10923

Bloom TDS Flower A DeChaine EG 2017 Why georeferencingmatters introducing a practical protocol to prepare speciesoccurrence records for spatial analysis Ecol Evol 8 765ndash777

Brooks DR van Veller MGP 2003 Critique of parsimonyanalysis of endemicity as a method of historical biogeography JBiogeogr 30 819ndash825

Ceballos G 1995 Vertebrate diversity ecology and conservation inNeotropical dry forests In Bullock S Medina E Mooney H(Eds) Seasonally Dry Tropical Forests Cambridge UniversityPress Cambridge UK pp 195ndash220

Clements J Schulenberg T Iliff M Roberson D FredericksT Sullivan B Wood C 2015 The eBirdClements checklist ofbirds of the world v2015 URL httpwww birds cornell educlementschecklistdownloadIOC [Verified 22 April 2017]

Cortes ALA Rapini A Daniel TF 2015 The Tetrameriumlineage (Acanthaceae Justicieae) does not support the PleistoceneArc hypothesis for South American seasonally dry forests AmJ Bot 102 992ndash1007

12 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

performing several tests surveying the weighting valuewe used a homoplasy penalization (k) equal to 3 asproposed by Goloboff (1993) which preserved theclades obtained with lowest values restricting homo-plasy severely and maximizing the number of synapo-morphies The obtained strict consensus trees wereanalysed in WinClada (Nixon 1999) and plotted(Appendices S4 and S5) in a geographical map usingArcMap 102 (ESRI 2010) Areas of endemism wererecovered based on at least two synapomorphies(Platnick 1991) also including taxa with consistencyindex (CI) = 05 (only if taxa showed reversions)because they may represent a distinctive feature ofoccurrence (autopomorphies) that is coded as uniquepresence to a given area (Escalante 2015) Nodal sup-port in cladograms was assessed by performing a boot-strap analysis with 1000 replicates in TNTBecause some authors have discussed the need to

increase the historical signal for finding relationshipsamong study areas and confer a ldquodynamic interpreta-tionrdquo to PAE (Rosen and Smith 1988 Cracraft 1991Nihei 2006 Vazquez-Miranda et al 2007) we per-formed two additional analyses taking into considera-tion the hierarchical information from genera andspecies together (Cracraft 1991 Porzecanski andCracraft 2005) In this sense our results from PAEcould be tested by a cladistic biogeographical analysisand compared with available phylogenies of taxainhabiting the areas investigated (Rosen and Smith1988 Morrone 2005 Sanchez-Gonzalez et al 2008)However we did not observe any difference in generalpatterns (ie NSDF regions identified) between theldquospeciesrdquo and ldquospecies-generardquo approaches (Appen-dices S4 and S5) The results shown herein are basedon strict consensus cladograms (ie complete speciesvs NSDF-restricted species) only from the analysesusing species informationFinally considering that patterns of relationships

among areas are easily retrieved within well-definedgeographical clades (Grismer 2000 Morrone 2014c)we performed two additional PAEs (ie complete andrestricted species approaches) based on the 17 NSDFgeographical groups or regions identified in the clado-grams for the 563 sampled grids (Appendix S4) Thislast step allowed us to increase the numbers of synapo-morphies grouping cell grids where data quantity islow and splitting quadrants that generate conflict(Morrone and Escalante 2002) determining the rela-tionships among areas (Morrone 2014c) The resultinggeographical groups were supported by taxa assynapomorphies (see Results Table 1) In all clado-grams we performed a bootstrap analysis with 1000replicates to assess nodal support The 17 NSDFregions identified herein were named based on a pro-posal by Pennington et al (2000) and Banda et al(2016)

Results

The occurrence database contained ~1 248 000records from 1298 bird species (from 78 families and511 genera) selected in this study For the 1242 speciescontaining more than 20 independent occurrencerecords all SDMs showed high AUC (Area UnderCurve for Receiver Operating Characteristic) values(ranking from 07 to 099) and Partial-ROC ratios(ranking from 115 to 199) For the remaining 56 spe-cies the Jackknife test showed that models were statis-tically significant (P lt 001) Thus performance valuesfor both modelling approaches indicated that modelsof the speciesrsquo potential distributions were accurateBased on the SDMs we observed that 364 of

bird species have at least 40 of their distributionwithin NSDFs while 304 were between 25 and40 and 331 had between 10 and 25 overlapwith NSDF areas Additionally the values of ecosys-temecological specificity showed that 409 of speciestended to be present in more than three ecosystems(IR lt 02) while only 137 are present between oneand two (IR gt 05) Most of the NSDF bird species(454) are present in at least three ecosystems(IR = 03) From these approaches we determinedthat ~43 (n = 557) of species initially considered arehighly associated andor endemic to NSDFs(Appendix S2) Final species selection was establishedbased on those with at least 33 of their distributionwithin the NSDF and an IR ge 03 where the NSDFrepresented the main ecosystem occupiedFor the complete species matrix approach (repre-

senting 563 sampled grids) we obtained a strict consen-sus cladogram of 16 367 steps [CI = 007 retentionindex (RI) = 086] while for the restricted speciesmatrix approach the consensus cladogram consisted of4172 steps (CI = 013 RI = 089) Both analysesrevealed 17 NSDF geographical groups showing~53 coincidence in the areas of endemism betweenthe two approaches (ie complete and restricted spe-cies) and containing almost the same synapomorphies(Table 1) Individual cladograms and their geographi-cal correspondence with the grid cell groupings forboth approaches are presented in Appendix S4Subsequently for PAE considering the 17 NSDF

regions in the complete species approach we obtained astrict consensus cladogram with 2040 steps (CI = 063RI = 075) while the restricted species approachshowed 705 steps (CI = 078 RI = 080) Both clado-grams were the base to propose our biotic regionaliza-tion of the NSDF (Fig 2) In this proposal weobserved that Baja California is the sister group (at thebase of the tree) of the rest of NSDF clades theCaribbeanndashAntilles islands are separated from the well-defined clade that includes the Mesoamerican and SouthAmerican areas The complete species cladogram

6 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

differed from the restricted one because the northern-most limit of NSDFs (SS group Fig 2a) appears in theformer as sister group of the Mesoamerican and SouthAmerican mainland grid cells whereas in the restrictedspecies cladogram these grid cells appear within north-western Mexico (Fig 2b)Following the nested pattern NSDFs showed a

clear division into two main well-defined clades onecontains the Mesoamerican region from north-westernMexico to Panama and another comprises all of theSouth American areas Focusing in the Mesoamericanregion we observed that the Yucatan Peninsula (YP)

is sister to all other NSDF groups located in CentralAmerica and north-western Mexico (Fig 2) Howeverit is noteworthy that the Panama region (Pa groupFig 2b) which represents the southernmost limit ofMesoamerican NSDF range is the most basal sub-clade for the restricted species approach whereas it isgeographically included within the South CentralAmerica group (CAmS in Fig 2a) for the completespecies cladogram However despite the slight differ-ences in topology between the two cladograms (com-plete and restricted species) for the main NSDF inMesoamerica where some are completely nested

Table 1Synapomorphic species for the clades (ie potential areas of endemism PAE) for avifauna of the Neotropical seasonally dry forests (NSDFs)considering both approaches (matrices of information) complete vs NSDF-restricted species Individual cladograms and the geographical corre-spondence obtained for the grid cells are presented in the supplementary material (Appendix S4)

NSDF region Species

Nodes in PAE

CompleteRestrictedspecies

CaribbeanndashAntillesIslands

Tachornis phoenicobia Quiscalus niger a ndash

Cuba (Cu) Coccyzus merlini Contopus caribaeus Corvus nasicus Icterus melanopsisMargarobyas lawrencii Melanerpes superciliaris Myiarchus sagrae Spindaliszena Tiaris canorus

c a

Haiti DominicanRepublic and PuertoRico (HP)

Amazona ventralis Chlorostilbon swainsonii Coccyzus longirostris Contopushispaniolensis Corvus leucognaphalus Dulus dominicus Euphonia musica Icterusdominicensis Melanerpes striatus Mellisuga minima Microligea palustris Myiarchusstolidus Nesoctites micromegas Patagioenas inornata Phaenicophilus palmarumPsittacara chloropterus Spinus dominicensis Todus angustirostris Todus subulatusTyto glaucops Vireo nanus

b ndash

Baja California (BC) Basilinna xantusii Geothlypis beldingi Toxostoma cinereum d ndashMesoamericanmainland

Amphispiza quinquestriata Camptostoma imberbe Megascops guatemalae Molothrusaeneus Pachyramphus aglaiae Stelgidopteryx serripennis Turdus rufopalliatus

e g

Sonora to Panama Arremonops rufivirgatus Geothlypis poliocephala Vireo hypochryseus f ndashNorthwestern Mexico(PS)

Amphispiza quinquestriata Polioptila nigriceps Turdus rufopalliatus ndash h

Northern Oaxaca (SP) Aimophila notosticta Calothorax pulcher ndash iYucatan Peninsula (YP) Cyanocorax yucatanicus Melanoptila glabrirostris Melanerpes pygmaeus Meleagris

ocellata Myiarchus yucatanensis Nyctiphrynus yucatanicusg j

Central America (CAm) Cantorchilus modestus Zimmerius vilissimus h ndashNorth Central America(CAmN)

Campylopterus rufus Icterus maculialatus i k

South Central America(CAmS)

Geothlypis chiriquensis Thamnophilus bridgesi j ndash

Northern SouthAmerica

Campylorhynchus griseus Icterus nigrogularis Myiarchus apicalis Picumnusgranadensis Picumnus squamulatus Sporophila intermedia

k e

Caribbean coast ofColombia andVenezuela (CCV)

Campylorhynchus nuchalis Crax daubentoni Crypturellus erythropus Euphoniatrinitatis Icterus icterus Inezia caudata Myiarchus venezuelensis Ortalisruficauda Saltator orenocensis Synallaxis cinnamomea Thamnophilusmelanonotus Thraupis glaucocolpa

l f

Pacific Equatorial (PE) Amazilia amazilia Arremon abeillei Basileuterus trifasciatus Chaetocercusbombus Columbina cruziana Contopus punensis Cyanocorax mystacalis Diveswarczewiczi Geothlypis auricularis Mecocerculus calopterus Melanopareiaelegans Mimus longicaudatus Myiodynastes bairdii Sturnella bellicosaThamnophilus bernardi

m c

Sub-Andean piedmont(SA-P)

Arremon torquatus Leptotila megalura Megascops hoyi Microstilbon burmeisteriPsilopsiagon aymara Rhynchotus maculicollis Sappho sparganurus

n b

Caatinga (CaB) Casiornis fuscus Herpsilochmus sellowi Pseudoseisura cristata Paroariadominicana Picumnus pygmaeus

o d

Synapomorhic species reported for both PAEs

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 7

(Fig 2a) and others showed some reciprocally mono-phyletic groups (Fig 2b) the division is clear betweenthe north-western Mexico [including the sections ofTamaulipas and Veracruz (TV) Pacific Mexican forest(PS) and Northern Oaxaca in southern Mexico (SP)]and the Central American (from southern Mexico toPanama) regions This last region is subsequentlydivided into two main subgroups (1) the northern(CAmN from southern Mexico to Nicaragua) and (2)the southern subgroup (CAmS from Nicaragua toPanama)Furthermore for both approaches we observed that

NSDFs in South America involved a well-resolvedclade divided into two reciprocally monophyleticgroups (Fig 2) One group comprises the grid cellsthat correspond to northern South America contain-ing all areas throughout the inter-Andean valleys(IAV from Colombia to the Caribbean coast) as wellas the north-western stretch of the Orinoco river (in-cluding the Cordillera de la Costa and the MaracaiboLake depression in Venezuela) and the dry forests

located in the western slope of the Sierra de Perija(from La Guajira to Santander-Cordoba Departmentsin Colombia CCV subgroup) The second group isformed by the forests located throughout the PacificEquatorial (PE) region (an area of endemism locatedfrom western Ecuador to north-western Peru includ-ing the central Andean Valleys) and a sister clade thatincludes the dry valleys located in southern Peru [ieApurımac-Mantaro and Tarapoto-Quillabamba forests(Ap-M)] as well as the south-eastern South Americaareas (ie NSDFs from western Bolivia and north-eastern Argentina to north-eastern Brazil) For thislast South American group two well-supported sub-groups are formed consecutively and separately onecontains the endemism area at the Sub-Andean pied-mont forests (SA-P) of north-western Argentina andwestern Bolivia and the other corresponds to theChiquitano and Misiones Province [ie south-easternBolivia north-eastern Argentina Paraguay and Uru-guay (C-MP)] and the endemism area of Caatinga(CaB located in north-eastern Brazil)

Fig 2 Regionalization of the Neotropical seasonally dry forests (NSDFs) based on the avifauna distribution and their relationships consideringcomplete (a) vs NSDF-restricted (b) species approaches (matrices of information) The cladograms revealed 17 NSDF geographical groupswhich involved four large regions The Baja California (A) CaribbeanndashAntilles islands (B) Mesoamerican (C) and South America (D) Acro-nyms in both cladograms and the map correspond to identified NSDF regions and areas of endemism [more than two synapomorphies (speciesrepresented with a black circle] BC = Baja California Cu = Cuba island HP = Haiti Dominican Republic and Puerto Rico islandsSS = Sonora and Sinaloa forests PS = Mexican Pacific slope lowlands SP = northern Oaxaca in southern Mexico TV = TamaulipasndashVeracruzstates YP = Yucatan Peninsula CAmN = northern Central America CAmS = nouthern Central America CAm = Central AmericaPa = Panama IAV = Inter-Andean Valleys of Colombia CCV = Caribbean ColombiandashVenezuela PE = Pacific Equatorial Ap-M = Apurimac-Mantaro SA-P = Sub-Andean Piedmont C-MP = Chiquitano forests and Misiones Province CaB = Brazilian Caatinga These well-resolvedclades are circumscribed by geographical and ecological barriers (dotted black lines) associated with the Gulf of California the leading edge ofthe Caribbean plate the Tehuantepec Isthmus the PolochicndashMotagua fault the Nicaragua Depression the Choco forest the Amazon basin andthe Andean Cordillera (including the Inter-Andean Valleys) Well-supported clades [more than two synapomorphies (species)] are shown with ablack circle in the acronyms of NSDF regions (see Table 1 for a complete description)

8 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

Discussion

The Baja Californiarsquos NSDF

In regional cladograms the Cape of Baja Californiaforms a distinct unit which supports the idea of asomewhat independent biogeographical history as hasbeen suggested in previous avian regionalizations Iso-lation of the NSDF in the Cape region from those inwestern Mexico (55ndash40 Mya) by the Gulf of Califor-nia has promoted the differentiation and a high degreeof endemism associated with the Cape of diversegroups including mammals birds snakes and insects(Grismer 2000 Rojas-Soto et al 2003 Rıos-Mu~nozand Navarro-Sigeurouenza 2012 OrsquoConnell et al 2017)However given the closest affinities of avian specieswith the nearby continent at the northern portion ofthe Peninsula (Rojas-Soto et al 2003) it seems clearthat the Cape avifauna originated through a terrestrialroute from the northern to the southern tip of the BajaCalifornia Peninsula (Rojas-Soto et al 2003OrsquoConnell et al 2017)

The CaribbeanndashAntilles avifauna

As found in previous biogeographical studies (egCrother and Guyer 1996 Zink et al 2000 Graham2003ab Vazquez-Miranda et al 2007) our analysessupport a monophyletic origin for groups of organismsinhabiting this region However a potential difficultyin explaining this monophyly is posited by the ques-tion of how birds reached those islands given that thecomplex geological history of the region included ldquoatleast eight events of fragmentation and seven ofhybridizationrdquo (Rosen 1985) In our study weobserved that several taxa acted as homoplasies whichcan be interpreted as the result of either dispersal orextinction events (Vazquez-Miranda et al 2007) Con-sidering that ~41 (n = 56 species) of NSDF avifaunais endemic and of these only eight species (two ofwhich are synapomorphies) were shared among thethree islands (Cuba Haiti Puerto Rico) the observedpattern is more probably explained by a mixture ofvicariance and dispersal events especially becausethose islands have had relatively recent bird exchanges(Vazquez-Miranda et al 2007) This idea is more par-simonious than assuming concordant dispersal pat-terns of individual taxa particularly considering thebiotarsquos affinities with both Central America andorSouth America (Crother and Guyer 1996 Zink et al2000 Graham 2003ab)Despite the existence of vicariance models explaining

affinities between the avifauna of this region with themainland (eg the temporal separation of the leadingedge of the Caribbean plate [beginning in the LateCretaceous and finishing in the MiocenendashMiddle

Pleiocene] Rosen 1985 Iturralde-Vinent and Mac-Phee 1999 Graham 2003ab Morrone 2014b) it isimportant to consider that dispersal events could havecontributed to defining the avifaunistic assemblages ofthis region as apparently occurred in the history ofthe woodpecker genus Melanerpes (Navarro-Sigeurouenzaet al 2017) It is difficult to compare the presentstudy with earlier studies because they used differentislands different taxa and different continental regionsas a reference Therefore further studies are needed toassess the relative age of these diversification centresand their biogeographical relationships

The Mesoamerican NSDFs

The division of this clade into sub-groups supports thelong isolation history of the identified Mesoamericanregions Our results highlight the importance of the Mex-ican Plateau and the Tehuantepec Isthmus in shapingthe distribution of Mesoamerican biotas in both high-lands and lowlands (Becerra 2005 Sanchez-Gonzalezet al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza 2012Banda et al 2016) The clear differentiation of thegroups on both sides of the Tehuantepec Isthmus(north-western Mexico Yucatan Peninsula and CentralAmerica Fig 2) as well as the subdivisions observedwithin the northern Mexican regions [Sonora-Sinaloa(SS) northern Oaxaca (SP) and TamaulipasndashVeracruz(TV) areas] might be explained by higher diversificationrates compared to other dry forest areas as detected forexample in the hummingbird Amazilia cyanocephala(Rodrıguez-Gomez et al 2013 OrsquoConnell et al 2017)Although the age of Mesoamerican NSDFs is notknown the fossil record and dated phylogenies suggest aMiocenendashPliocene age at least in north-western Mexico(Becerra 2005) These forests probably expanded toCentral America (where fossil evidence seems to suggestthat it invaded only after 25 Mya Graham and Dilcher1995) after the Sierra Madre Occidental and the Trans-mexican Volcanic belt were formed creating conditionsfor the establishment and persistence of NSDFsThe clear separation observed for the Sonora-Sina-

loa (SS) subgroup and the rest of the Mesoamericanmainland (Fig 2a) could be explained by the highaffinities with Nearctic biota of this northernmost area(Stotz et al 1996) In fact despite the high speciesturnover (n = 188) observed between the Sonora-Sina-loa subgroup and the rest of the MesoamericanNSDFs these northernmost forests shared 54 species(95) with those reported for the Baja Californiaregion Furthermore we propose a region formed byall areas west of the Tehuantepec Isthmus located inMexico [including NSDF in TamaulipasndashVeracruz(TV) and except those corresponding to Baja Califor-nia (BC)] which is supported by previous biogeo-graphical hypotheses and also suggested that the

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 9

north-eastern Gulf of Mexico is biogeographically dis-tinct from the southern and south-eastern part of theGulf coast (Liebherr 1994 Hernandez-Ba~nos et al1995 Marshall and Liebherr 2000 Rıos-Mu~noz andNavarro-Sigeurouenza 2012)On the other hand separation within the south-east-

ern Isthmus groups (light blues in Fig 2) supports thebiogeographical hypothesis of vicariant processes asso-ciated with the PolochicndashMotagua fault and the Nicar-agua Depression in the region (Morrone 2001Schuster et al 2003 Liede and Meve 2004) Here weidentified at least three distinct regions suggesting theexistence of bird assemblages with independent biogeo-graphical histories probably shaped by new immi-grants from dispersal events between South and NorthAmerica after the establishment of the Panama Isth-mus (Smith and Klicka 2010) For instance weobserved that ~73 of the NSDF avifauna is sharedbetween southern Central America and northern SouthAmerica (with ~72 of NSDF-restricted speciesshared between the NSDF from Panama and northernSouth America) confirming the existence of strongaffinities between the two regions (Linares-Palominoet al 2011 Rıos-Mu~noz and Navarro-Sigeurouenza 2012Banda et al 2016)In the Panama region differences in the affinities

between the cladograms for this region may be influ-enced by the inclusion (ie complete approach) or theexclusion (ie restricted approach) of species widelydistributed in the region and inhabiting other ecosys-tems surrounding NSDFs In this sense such differ-ences could be explained by different temporalpatterns of speciation documented for taxa associatedwith humid and dry habitats across the Isthmus ofPanama (Smith et al 2012) The Panama area hashad a history of expanding and contracting processesfor both the dry and the humid forests and even whenboth habitats consist of ldquoyoungrdquo and ldquooldrdquo speciesthe dry habitat species are geographically more distantfrom their nearest counterpart in Central America andnorthern South America compared with species fromthe humid habitats in these regions (Oswald andSteadman 2015 Oswald et al 2017) This phe-nomenon if generalized is probably an importantcontributor to biogeographical patterns in the south-ernmost section of the Mesoamerican region (Sanchez-Gonzalez et al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza 2012 Smith et al 2012)

South American NSDF areas and the Pleistocene ArcHypothesis

Previous studies of distributional patterns within theSouth American biota have accepted the hypothesisthat areas of endemism have arisen by isolation ofmore widespread biotas This scenario is commonly

associated with paleogeographical changes alterationsin river systems Pleistocene climatic fluctuations or amixture of all these processes (Herzog and Kessler2002 Werneck et al 2012) However we find littlesupport for widespread Pleistocene NSDF formationthroughout South America because few species(1033 n = 100 species) are widespread and sharedby more than 75 of South American areas Thisresult is concordant with NSDF plant patterns recentlyreported by Linares-Palomino et al (2011) and Bandaet al (2016) as well as with the restricted climatic cor-ridors found by Werneck et al (2011) which suggestthat LGM climate was probably too dry and cold tosupport large tracts of NSDFs across the continentSome authors suggest that long-distance ancient dis-

persal events of unrelated taxonomic groups are lesslikely than vicariant associations (Werneck et al2011 de Melo et al 2016) Likewise there is substan-tial support that NSDF is a dispersal-limited biomegiven the phylogenetic niche conservatism and stronggeographically structured phylogenies for woody plantclades (Pennington et al 2009 Pennington and Lavin2016) Thus an allopatric-vicariance explanationmdashde-lineated by major geographical or ecological barriersbefore the Pleistocenemdashseems more plausible toexplain the evolution and relationships among all theseforest areas which is supported by results based onendemic taxa For instance the biogeography of Bro-togeris parakeets implies a dynamic history for SouthAmerican biomes since the Pliocene corroboratingthat geological evolution of Amazonia has been impor-tant in shaping its biodiversity with taxa endemic toNSDFs and other biomes (Ribas et al 2009) In thissense our results suggest that the wide distributions ofa relatively low proportion of NSDF species and eco-logical similarity among them (including specialistsand generalists) may reflect only limited recent long-distance dispersal events and species exchange (Dicket al 2003 2004 Mayle 2004)The separation of a northern South American clade

(Fig 2) suggests an effect of the Amazonia rain foreststhe wet Choco forest and the final uplift of the north-ern Andes as barriers for the dispersal of dry forestspecies (Amorim 2001 Garzione et al 2008 Hoornet al 2010 Morrone 2014b Banda et al 2016) Clus-ters of the northern inter-Andean valleys (located alongthe Magdalena and Cauca Rivers) together with Carib-bean coastal areas reflect the heterogeneity of dryAndean regions (with important species turnover withMesoamerica and CaribbeanndashAntilles) where few spe-cies are distributed all the way from the northern tothe southern Andes (Nores 2004 Vazquez-Mirandaet al 2007) These results are also consistent withthose previously reported for other taxonomic groupsin the region that are associated with NSDFs (Linares-Palomino et al 2011 Morrone 2014b)

10 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

Although our analyses generally agree with thePleistocene Arc Hypothesis proposed originally byPrado and Gibbs (1993)mdashformation of a once exten-sive and largely contiguous seasonal woodland com-prising the Caatingas the Misiones and Sub-AndeanPiedmont nucleimdashthe clear separation among NSDFsfrom centralndashsouth-western and south-eastern SouthAmerica supports the idea that both uplift of theAndes and historical (Cenozoic and Quaternary) cli-matic regimes of north-western EcuadorndashPeru andwestern Bolivia-Argentina are probably responsible forthe high endemism and distributions of their NSDFbird communities (Hoorn et al 2010 Morrone2014b) These communities have been biogeographi-cally separated from neighbouring Bolivian lowlandsites (Herzog and Kessler 2002 Herzog et al 2016)In fact previous Andean NSDF legume phylogeniesshowed consistently high geographical structurethroughout their distribution suggesting that at leastthe NSDF flora has been assembled gradually over thepast ca 19 Myr with processes of in situ diversifica-tion largely undisturbed by new immigrants (Seuroarkinenet al 2012) From this perspective the NSDF on theslopes of centralndashsouth-western South America consti-tutes an important biogeographical area with ecologi-cal and geographical long-term stability and probablyoriginated by convergence scenarios resulting fromancient dispersal events that involved low niche differ-entiation and strong dispersal limitation from otherNSDF patches (Pennington et al 2009 Seuroarkinenet al 2012) This also suggests that NSDFs fromEcuador and northern Peru to western BoliviandashArgen-tina might constitute remnants of a previously widerexpansion according to the Pleistocene NSDF forma-tion (Seuroarkinen et al 2012 Cortes et al 2015 de Meloet al 2016) Passerine fossils support this hypothesisand suggest that some NSDF species had wider distri-butions in the past (Oswald and Steadman 2015) withthe Andean valleys (eg the Mara~non and ApurimacValleys) having direct influence on divergence andextinction processes by acting as semi-permeable barri-ers (Seuroarkinen et al 2011 Smith et al 2014 Wingerand Bates 2015)Finally we found a stretch of NSDF from north-

eastern Brazil south to ParaguayndashArgentina and northtowards eastern Bolivia (Fig 2) which is concordantwith previous studies showing a well-supported group-ing among these forests and suggesting the existence ofa historical stable unit known as the Open Dry Diago-nal (Mayle 2004 Porzecanski and Cracraft 2005Linares-Palomino et al 2011 Werneck et al 20112012) Today many dry habitat sister taxa are geo-graphically distant from one another suggesting thatthese species have either undergone long-distance dis-persal across large tracts of humid forests or that aridregions were once connected (Oswald et al 2017) The

absence of synapomorphies within the Chiquitano andMisiones Province forests (Table 1 Appendix S4) pro-vides some evidence for the recent formation of birdassemblages in this area (Herzog and Kessler 2002Smith et al 2014) as was found for woody plants(Linares-Palomino et al 2011 Banda et al 2016) Infact several authors suggest climatic and floristic char-acteristics closer to Atlantic rain forests than to NSDFsfor these areas (Olson et al 2001 Oliveira-Filho et al2006) which were probably colonized by dry forestspecies during drier and cooler periods of the Pleis-tocene These colonization processes were probablyfavoured by a potential narrow eastndashwest corridor con-necting NSDF in the extremes of north-eastern Braziland south-western South America during the LGM(Werneck et al 2011) This connection could serve asa biogeographical link sheltering floristic and faunisticdispersal routes resulting in partial expansions ofNSDFs in some areas of Amazonia (Werneck et al2011 2012) As a consequence of this last idea thelong-distance dispersal hypothesis of NSDFs in south-eastern South America proposed by Mayle (2004)remains a scenario that still needs independent testing

Final considerations

Our study suggests that evolution of the well-differ-entiated avifaunas in the four major geographicalgroups herein identified might have involved a mixtureof vicariance and dispersal events which provides newinsights into the diversification centres and historicalrelationships for the distributional patterns of thisecosystem and its associated avifauna Despite somespecies being shared among different clades (regions)and that this probably increased after closure of thePanama Isthmus (Morrone 2014b) and during themaximum extension of forests at the LGM (Bandaet al 2016) we observed a clear differentiation amongthe main areas with high degrees of endemism(Fig 2) This agrees with the idea of independent evo-lutionary histories among NSDF nuclei across theirdistribution (Becerra 2005 Cortes et al 2015 deMelo et al 2016)Differences observed with previous studies (eg

Porzecanski and Cracraft 2005 Rıos-Mu~noz andNavarro-Sigeurouenza 2012) are largely due to taxonomicuncertainties (ie species concepts and new species thathave been described or recognized since those studiesJetz et al 2002) as well as by differences in the defini-tion of sampling area sizes and algorithms used foranalyses For PAE there is a trend to decrease theabsolute number of steps and to increase the numberof synapomorphies as the size of the area increases(Morrone and Escalante 2002) Likewise it is impor-tant to note that this approach depends on detailed pri-mary biodiversity occurrence data In studies like this

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 11

with large data numbers it is worth noting thatobserved patterns of species distributional data couldbe influenced by diverse sources of error (eg taxo-nomic uncertainties identification errors or errors onlocality information associated with species occur-rences) Thus it is important to promote the constantupdate of the taxonomic status of specimens depositedin collections as well as the procedures for cleaningand revising biodiversity data from heterogeneoussources (Navarro-Sigeurouenza et al 2003 Bloom et al2017)Although our results mostly apply to the geographi-

cal scale and taxa used for this study the NSDF avi-fauna distribution patterns we found are non-randomand congruent with those for other organisms(Becerra 2005 Porzecanski and Cracraft 2005 Pen-nington et al 2009 Rıos-Mu~noz and Navarro-Sigeurouenza 2012 Oswald and Steadman 2015 Bandaet al 2016) The general pattern identified hereimplies the need to continue to study these areas ofendemism which certainly could have a much longerhistory than the temporal framework of the Pleis-tocene Thus from a conservation perspective mosthistorical areas of endemism must be considered aspriorities and incorporated into efforts of conserva-tions planning (eg Nori et al 2016) because theyreflect an unique history of the Earth and its biota Afailure to protect them would result in major losses ofunique species diversity for this highly threatenedecosystem The observed distributional patterns of theavifauna as well as their relationships across theNSDFs constitute evidence of the great complexity ofthe biogeography of this ecosystem in which each par-ticular pattern should be considered a stepping stonealong a true arc recognized as the biogeography ofthe Neotropics

Acknowledgments

We appreciate the efforts of museums (seeAppendix S1) and their curators as well as field collec-tors and institutions that made our study possibleDAP-T extends his gratitude to Consejo Nacional deCiencia y Tecnologıa (CONACyT Mexico) for a doc-toral scholarship (297538) and the Smithsonian-MasonSchool of Conservation for a scholarship during thecourse ldquoSpatial Ecology Geospatial Analysis andRemote Sensing for Conservation (MCCS 0500)rdquo thatprovided the tools and skills necessary for the spatialanalysis in this study The Rufford Foundation (DAP-T 16017-1 DAP-T 20284-2) Idea Wild (DAP-T) Con-sejo de Desarrollo Cientıfico Humanıstico y Tec-nologico at Universidad del Zulia (CONDES projectsCC-0247-13 and CC-0497-16 DAP-T) and the CONA-CyT project 152060 (AGNS and ORRS) provided

financial and logistical support J C Catarı P JCampos and Z Perez-Duran (Bolivia) as well as J ADelgado (Peru) helped during fieldwork A GordilloC Mota-Vargas and I Perozo helped in georeferencingbird locality data T Escalante helped with the PAEThe manuscript was improved by comments from HM Ortega-Andrade C Mota-Vargas A Lira-NoriegaJ L Parra L A Gonzalez and F Lopez

Conflicts of interest

The authors declare no competing interests in thisstudy

References

Amorim DS 2001 Dos Amazonias In Llorente J Morrone JJ(Eds) Introduccion a la Biogeografıa en LatinoamericaConceptos Teorıas Metodos y Aplicaciones Las Prensas deCiencias Universidad Nacional Autonoma de Mexico MexicoDF pp 245ndash255

Anderson RP Lew D Peterson AT 2003 Evaluatingpredictive models of speciesrsquo distributions criteria for selectingoptimal models Ecol Modell 162 211ndash232

Banda K Delgado-Salinas A Dexter KG Linares-Palomino ROliveira-Filho A Prado D Pullan M Quintana C Riina RRodrıguez M Weintritt J Acevedo-Rodrıguez P Adarve JAlvarez E Aranguren BA Arteaga JC Aymard G Casta~noA Ceballos-Mago N Cogollo A Cuadros H Delgado FDevia W Due~nas H Fajardo L Fernandez A FernandezMA Franklin J Freid EH Galetti LA Gonto RGonzalez-M R Graveson R Helmer EH Idarraga ALopez R Marcano-Vega H Martınez OG Maturo HMMcDonald M McLaren K Melo O Mijares F Mogni VMolina D Moreno NdP Nassar JM Neves DM OakleyLJ Oatham M Olvera-Luna AR Pezzini FF DominguezOJR Rıos ME Rivera O Rodrıguez N Rojas A SeuroarkinenT Sanchez R Smith M Vargas C Villanueva B PenningtonRT 2016 Plant diversity patterns in Neotropical dry forests andtheir conservation implications Science 353 1383ndash1387

Barve N Barve V Jimenez-Valverde A Lira-Noriega AMaher SP Peterson AT Soberon J Villalobos F 2011The crucial role of the accessible area in ecological nichemodeling and species distribution modeling Ecol Modell 2221810ndash1819

Becerra JX 2005 Timing the origin and expansion of the Mexicantropical dry forest Proc Natl Acad Sci USA 102 10919ndash10923

Bloom TDS Flower A DeChaine EG 2017 Why georeferencingmatters introducing a practical protocol to prepare speciesoccurrence records for spatial analysis Ecol Evol 8 765ndash777

Brooks DR van Veller MGP 2003 Critique of parsimonyanalysis of endemicity as a method of historical biogeography JBiogeogr 30 819ndash825

Ceballos G 1995 Vertebrate diversity ecology and conservation inNeotropical dry forests In Bullock S Medina E Mooney H(Eds) Seasonally Dry Tropical Forests Cambridge UniversityPress Cambridge UK pp 195ndash220

Clements J Schulenberg T Iliff M Roberson D FredericksT Sullivan B Wood C 2015 The eBirdClements checklist ofbirds of the world v2015 URL httpwww birds cornell educlementschecklistdownloadIOC [Verified 22 April 2017]

Cortes ALA Rapini A Daniel TF 2015 The Tetrameriumlineage (Acanthaceae Justicieae) does not support the PleistoceneArc hypothesis for South American seasonally dry forests AmJ Bot 102 992ndash1007

12 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

differed from the restricted one because the northern-most limit of NSDFs (SS group Fig 2a) appears in theformer as sister group of the Mesoamerican and SouthAmerican mainland grid cells whereas in the restrictedspecies cladogram these grid cells appear within north-western Mexico (Fig 2b)Following the nested pattern NSDFs showed a

clear division into two main well-defined clades onecontains the Mesoamerican region from north-westernMexico to Panama and another comprises all of theSouth American areas Focusing in the Mesoamericanregion we observed that the Yucatan Peninsula (YP)

is sister to all other NSDF groups located in CentralAmerica and north-western Mexico (Fig 2) Howeverit is noteworthy that the Panama region (Pa groupFig 2b) which represents the southernmost limit ofMesoamerican NSDF range is the most basal sub-clade for the restricted species approach whereas it isgeographically included within the South CentralAmerica group (CAmS in Fig 2a) for the completespecies cladogram However despite the slight differ-ences in topology between the two cladograms (com-plete and restricted species) for the main NSDF inMesoamerica where some are completely nested

Table 1Synapomorphic species for the clades (ie potential areas of endemism PAE) for avifauna of the Neotropical seasonally dry forests (NSDFs)considering both approaches (matrices of information) complete vs NSDF-restricted species Individual cladograms and the geographical corre-spondence obtained for the grid cells are presented in the supplementary material (Appendix S4)

NSDF region Species

Nodes in PAE

CompleteRestrictedspecies

CaribbeanndashAntillesIslands

Tachornis phoenicobia Quiscalus niger a ndash

Cuba (Cu) Coccyzus merlini Contopus caribaeus Corvus nasicus Icterus melanopsisMargarobyas lawrencii Melanerpes superciliaris Myiarchus sagrae Spindaliszena Tiaris canorus

c a

Haiti DominicanRepublic and PuertoRico (HP)

Amazona ventralis Chlorostilbon swainsonii Coccyzus longirostris Contopushispaniolensis Corvus leucognaphalus Dulus dominicus Euphonia musica Icterusdominicensis Melanerpes striatus Mellisuga minima Microligea palustris Myiarchusstolidus Nesoctites micromegas Patagioenas inornata Phaenicophilus palmarumPsittacara chloropterus Spinus dominicensis Todus angustirostris Todus subulatusTyto glaucops Vireo nanus

b ndash

Baja California (BC) Basilinna xantusii Geothlypis beldingi Toxostoma cinereum d ndashMesoamericanmainland

Amphispiza quinquestriata Camptostoma imberbe Megascops guatemalae Molothrusaeneus Pachyramphus aglaiae Stelgidopteryx serripennis Turdus rufopalliatus

e g

Sonora to Panama Arremonops rufivirgatus Geothlypis poliocephala Vireo hypochryseus f ndashNorthwestern Mexico(PS)

Amphispiza quinquestriata Polioptila nigriceps Turdus rufopalliatus ndash h

Northern Oaxaca (SP) Aimophila notosticta Calothorax pulcher ndash iYucatan Peninsula (YP) Cyanocorax yucatanicus Melanoptila glabrirostris Melanerpes pygmaeus Meleagris

ocellata Myiarchus yucatanensis Nyctiphrynus yucatanicusg j

Central America (CAm) Cantorchilus modestus Zimmerius vilissimus h ndashNorth Central America(CAmN)

Campylopterus rufus Icterus maculialatus i k

South Central America(CAmS)

Geothlypis chiriquensis Thamnophilus bridgesi j ndash

Northern SouthAmerica

Campylorhynchus griseus Icterus nigrogularis Myiarchus apicalis Picumnusgranadensis Picumnus squamulatus Sporophila intermedia

k e

Caribbean coast ofColombia andVenezuela (CCV)

Campylorhynchus nuchalis Crax daubentoni Crypturellus erythropus Euphoniatrinitatis Icterus icterus Inezia caudata Myiarchus venezuelensis Ortalisruficauda Saltator orenocensis Synallaxis cinnamomea Thamnophilusmelanonotus Thraupis glaucocolpa

l f

Pacific Equatorial (PE) Amazilia amazilia Arremon abeillei Basileuterus trifasciatus Chaetocercusbombus Columbina cruziana Contopus punensis Cyanocorax mystacalis Diveswarczewiczi Geothlypis auricularis Mecocerculus calopterus Melanopareiaelegans Mimus longicaudatus Myiodynastes bairdii Sturnella bellicosaThamnophilus bernardi

m c

Sub-Andean piedmont(SA-P)

Arremon torquatus Leptotila megalura Megascops hoyi Microstilbon burmeisteriPsilopsiagon aymara Rhynchotus maculicollis Sappho sparganurus

n b

Caatinga (CaB) Casiornis fuscus Herpsilochmus sellowi Pseudoseisura cristata Paroariadominicana Picumnus pygmaeus

o d

Synapomorhic species reported for both PAEs

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 7

(Fig 2a) and others showed some reciprocally mono-phyletic groups (Fig 2b) the division is clear betweenthe north-western Mexico [including the sections ofTamaulipas and Veracruz (TV) Pacific Mexican forest(PS) and Northern Oaxaca in southern Mexico (SP)]and the Central American (from southern Mexico toPanama) regions This last region is subsequentlydivided into two main subgroups (1) the northern(CAmN from southern Mexico to Nicaragua) and (2)the southern subgroup (CAmS from Nicaragua toPanama)Furthermore for both approaches we observed that

NSDFs in South America involved a well-resolvedclade divided into two reciprocally monophyleticgroups (Fig 2) One group comprises the grid cellsthat correspond to northern South America contain-ing all areas throughout the inter-Andean valleys(IAV from Colombia to the Caribbean coast) as wellas the north-western stretch of the Orinoco river (in-cluding the Cordillera de la Costa and the MaracaiboLake depression in Venezuela) and the dry forests

located in the western slope of the Sierra de Perija(from La Guajira to Santander-Cordoba Departmentsin Colombia CCV subgroup) The second group isformed by the forests located throughout the PacificEquatorial (PE) region (an area of endemism locatedfrom western Ecuador to north-western Peru includ-ing the central Andean Valleys) and a sister clade thatincludes the dry valleys located in southern Peru [ieApurımac-Mantaro and Tarapoto-Quillabamba forests(Ap-M)] as well as the south-eastern South Americaareas (ie NSDFs from western Bolivia and north-eastern Argentina to north-eastern Brazil) For thislast South American group two well-supported sub-groups are formed consecutively and separately onecontains the endemism area at the Sub-Andean pied-mont forests (SA-P) of north-western Argentina andwestern Bolivia and the other corresponds to theChiquitano and Misiones Province [ie south-easternBolivia north-eastern Argentina Paraguay and Uru-guay (C-MP)] and the endemism area of Caatinga(CaB located in north-eastern Brazil)

Fig 2 Regionalization of the Neotropical seasonally dry forests (NSDFs) based on the avifauna distribution and their relationships consideringcomplete (a) vs NSDF-restricted (b) species approaches (matrices of information) The cladograms revealed 17 NSDF geographical groupswhich involved four large regions The Baja California (A) CaribbeanndashAntilles islands (B) Mesoamerican (C) and South America (D) Acro-nyms in both cladograms and the map correspond to identified NSDF regions and areas of endemism [more than two synapomorphies (speciesrepresented with a black circle] BC = Baja California Cu = Cuba island HP = Haiti Dominican Republic and Puerto Rico islandsSS = Sonora and Sinaloa forests PS = Mexican Pacific slope lowlands SP = northern Oaxaca in southern Mexico TV = TamaulipasndashVeracruzstates YP = Yucatan Peninsula CAmN = northern Central America CAmS = nouthern Central America CAm = Central AmericaPa = Panama IAV = Inter-Andean Valleys of Colombia CCV = Caribbean ColombiandashVenezuela PE = Pacific Equatorial Ap-M = Apurimac-Mantaro SA-P = Sub-Andean Piedmont C-MP = Chiquitano forests and Misiones Province CaB = Brazilian Caatinga These well-resolvedclades are circumscribed by geographical and ecological barriers (dotted black lines) associated with the Gulf of California the leading edge ofthe Caribbean plate the Tehuantepec Isthmus the PolochicndashMotagua fault the Nicaragua Depression the Choco forest the Amazon basin andthe Andean Cordillera (including the Inter-Andean Valleys) Well-supported clades [more than two synapomorphies (species)] are shown with ablack circle in the acronyms of NSDF regions (see Table 1 for a complete description)

8 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

Discussion

The Baja Californiarsquos NSDF

In regional cladograms the Cape of Baja Californiaforms a distinct unit which supports the idea of asomewhat independent biogeographical history as hasbeen suggested in previous avian regionalizations Iso-lation of the NSDF in the Cape region from those inwestern Mexico (55ndash40 Mya) by the Gulf of Califor-nia has promoted the differentiation and a high degreeof endemism associated with the Cape of diversegroups including mammals birds snakes and insects(Grismer 2000 Rojas-Soto et al 2003 Rıos-Mu~nozand Navarro-Sigeurouenza 2012 OrsquoConnell et al 2017)However given the closest affinities of avian specieswith the nearby continent at the northern portion ofthe Peninsula (Rojas-Soto et al 2003) it seems clearthat the Cape avifauna originated through a terrestrialroute from the northern to the southern tip of the BajaCalifornia Peninsula (Rojas-Soto et al 2003OrsquoConnell et al 2017)

The CaribbeanndashAntilles avifauna

As found in previous biogeographical studies (egCrother and Guyer 1996 Zink et al 2000 Graham2003ab Vazquez-Miranda et al 2007) our analysessupport a monophyletic origin for groups of organismsinhabiting this region However a potential difficultyin explaining this monophyly is posited by the ques-tion of how birds reached those islands given that thecomplex geological history of the region included ldquoatleast eight events of fragmentation and seven ofhybridizationrdquo (Rosen 1985) In our study weobserved that several taxa acted as homoplasies whichcan be interpreted as the result of either dispersal orextinction events (Vazquez-Miranda et al 2007) Con-sidering that ~41 (n = 56 species) of NSDF avifaunais endemic and of these only eight species (two ofwhich are synapomorphies) were shared among thethree islands (Cuba Haiti Puerto Rico) the observedpattern is more probably explained by a mixture ofvicariance and dispersal events especially becausethose islands have had relatively recent bird exchanges(Vazquez-Miranda et al 2007) This idea is more par-simonious than assuming concordant dispersal pat-terns of individual taxa particularly considering thebiotarsquos affinities with both Central America andorSouth America (Crother and Guyer 1996 Zink et al2000 Graham 2003ab)Despite the existence of vicariance models explaining

affinities between the avifauna of this region with themainland (eg the temporal separation of the leadingedge of the Caribbean plate [beginning in the LateCretaceous and finishing in the MiocenendashMiddle

Pleiocene] Rosen 1985 Iturralde-Vinent and Mac-Phee 1999 Graham 2003ab Morrone 2014b) it isimportant to consider that dispersal events could havecontributed to defining the avifaunistic assemblages ofthis region as apparently occurred in the history ofthe woodpecker genus Melanerpes (Navarro-Sigeurouenzaet al 2017) It is difficult to compare the presentstudy with earlier studies because they used differentislands different taxa and different continental regionsas a reference Therefore further studies are needed toassess the relative age of these diversification centresand their biogeographical relationships

The Mesoamerican NSDFs

The division of this clade into sub-groups supports thelong isolation history of the identified Mesoamericanregions Our results highlight the importance of the Mex-ican Plateau and the Tehuantepec Isthmus in shapingthe distribution of Mesoamerican biotas in both high-lands and lowlands (Becerra 2005 Sanchez-Gonzalezet al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza 2012Banda et al 2016) The clear differentiation of thegroups on both sides of the Tehuantepec Isthmus(north-western Mexico Yucatan Peninsula and CentralAmerica Fig 2) as well as the subdivisions observedwithin the northern Mexican regions [Sonora-Sinaloa(SS) northern Oaxaca (SP) and TamaulipasndashVeracruz(TV) areas] might be explained by higher diversificationrates compared to other dry forest areas as detected forexample in the hummingbird Amazilia cyanocephala(Rodrıguez-Gomez et al 2013 OrsquoConnell et al 2017)Although the age of Mesoamerican NSDFs is notknown the fossil record and dated phylogenies suggest aMiocenendashPliocene age at least in north-western Mexico(Becerra 2005) These forests probably expanded toCentral America (where fossil evidence seems to suggestthat it invaded only after 25 Mya Graham and Dilcher1995) after the Sierra Madre Occidental and the Trans-mexican Volcanic belt were formed creating conditionsfor the establishment and persistence of NSDFsThe clear separation observed for the Sonora-Sina-

loa (SS) subgroup and the rest of the Mesoamericanmainland (Fig 2a) could be explained by the highaffinities with Nearctic biota of this northernmost area(Stotz et al 1996) In fact despite the high speciesturnover (n = 188) observed between the Sonora-Sina-loa subgroup and the rest of the MesoamericanNSDFs these northernmost forests shared 54 species(95) with those reported for the Baja Californiaregion Furthermore we propose a region formed byall areas west of the Tehuantepec Isthmus located inMexico [including NSDF in TamaulipasndashVeracruz(TV) and except those corresponding to Baja Califor-nia (BC)] which is supported by previous biogeo-graphical hypotheses and also suggested that the

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 9

north-eastern Gulf of Mexico is biogeographically dis-tinct from the southern and south-eastern part of theGulf coast (Liebherr 1994 Hernandez-Ba~nos et al1995 Marshall and Liebherr 2000 Rıos-Mu~noz andNavarro-Sigeurouenza 2012)On the other hand separation within the south-east-

ern Isthmus groups (light blues in Fig 2) supports thebiogeographical hypothesis of vicariant processes asso-ciated with the PolochicndashMotagua fault and the Nicar-agua Depression in the region (Morrone 2001Schuster et al 2003 Liede and Meve 2004) Here weidentified at least three distinct regions suggesting theexistence of bird assemblages with independent biogeo-graphical histories probably shaped by new immi-grants from dispersal events between South and NorthAmerica after the establishment of the Panama Isth-mus (Smith and Klicka 2010) For instance weobserved that ~73 of the NSDF avifauna is sharedbetween southern Central America and northern SouthAmerica (with ~72 of NSDF-restricted speciesshared between the NSDF from Panama and northernSouth America) confirming the existence of strongaffinities between the two regions (Linares-Palominoet al 2011 Rıos-Mu~noz and Navarro-Sigeurouenza 2012Banda et al 2016)In the Panama region differences in the affinities

between the cladograms for this region may be influ-enced by the inclusion (ie complete approach) or theexclusion (ie restricted approach) of species widelydistributed in the region and inhabiting other ecosys-tems surrounding NSDFs In this sense such differ-ences could be explained by different temporalpatterns of speciation documented for taxa associatedwith humid and dry habitats across the Isthmus ofPanama (Smith et al 2012) The Panama area hashad a history of expanding and contracting processesfor both the dry and the humid forests and even whenboth habitats consist of ldquoyoungrdquo and ldquooldrdquo speciesthe dry habitat species are geographically more distantfrom their nearest counterpart in Central America andnorthern South America compared with species fromthe humid habitats in these regions (Oswald andSteadman 2015 Oswald et al 2017) This phe-nomenon if generalized is probably an importantcontributor to biogeographical patterns in the south-ernmost section of the Mesoamerican region (Sanchez-Gonzalez et al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza 2012 Smith et al 2012)

South American NSDF areas and the Pleistocene ArcHypothesis

Previous studies of distributional patterns within theSouth American biota have accepted the hypothesisthat areas of endemism have arisen by isolation ofmore widespread biotas This scenario is commonly

associated with paleogeographical changes alterationsin river systems Pleistocene climatic fluctuations or amixture of all these processes (Herzog and Kessler2002 Werneck et al 2012) However we find littlesupport for widespread Pleistocene NSDF formationthroughout South America because few species(1033 n = 100 species) are widespread and sharedby more than 75 of South American areas Thisresult is concordant with NSDF plant patterns recentlyreported by Linares-Palomino et al (2011) and Bandaet al (2016) as well as with the restricted climatic cor-ridors found by Werneck et al (2011) which suggestthat LGM climate was probably too dry and cold tosupport large tracts of NSDFs across the continentSome authors suggest that long-distance ancient dis-

persal events of unrelated taxonomic groups are lesslikely than vicariant associations (Werneck et al2011 de Melo et al 2016) Likewise there is substan-tial support that NSDF is a dispersal-limited biomegiven the phylogenetic niche conservatism and stronggeographically structured phylogenies for woody plantclades (Pennington et al 2009 Pennington and Lavin2016) Thus an allopatric-vicariance explanationmdashde-lineated by major geographical or ecological barriersbefore the Pleistocenemdashseems more plausible toexplain the evolution and relationships among all theseforest areas which is supported by results based onendemic taxa For instance the biogeography of Bro-togeris parakeets implies a dynamic history for SouthAmerican biomes since the Pliocene corroboratingthat geological evolution of Amazonia has been impor-tant in shaping its biodiversity with taxa endemic toNSDFs and other biomes (Ribas et al 2009) In thissense our results suggest that the wide distributions ofa relatively low proportion of NSDF species and eco-logical similarity among them (including specialistsand generalists) may reflect only limited recent long-distance dispersal events and species exchange (Dicket al 2003 2004 Mayle 2004)The separation of a northern South American clade

(Fig 2) suggests an effect of the Amazonia rain foreststhe wet Choco forest and the final uplift of the north-ern Andes as barriers for the dispersal of dry forestspecies (Amorim 2001 Garzione et al 2008 Hoornet al 2010 Morrone 2014b Banda et al 2016) Clus-ters of the northern inter-Andean valleys (located alongthe Magdalena and Cauca Rivers) together with Carib-bean coastal areas reflect the heterogeneity of dryAndean regions (with important species turnover withMesoamerica and CaribbeanndashAntilles) where few spe-cies are distributed all the way from the northern tothe southern Andes (Nores 2004 Vazquez-Mirandaet al 2007) These results are also consistent withthose previously reported for other taxonomic groupsin the region that are associated with NSDFs (Linares-Palomino et al 2011 Morrone 2014b)

10 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

Although our analyses generally agree with thePleistocene Arc Hypothesis proposed originally byPrado and Gibbs (1993)mdashformation of a once exten-sive and largely contiguous seasonal woodland com-prising the Caatingas the Misiones and Sub-AndeanPiedmont nucleimdashthe clear separation among NSDFsfrom centralndashsouth-western and south-eastern SouthAmerica supports the idea that both uplift of theAndes and historical (Cenozoic and Quaternary) cli-matic regimes of north-western EcuadorndashPeru andwestern Bolivia-Argentina are probably responsible forthe high endemism and distributions of their NSDFbird communities (Hoorn et al 2010 Morrone2014b) These communities have been biogeographi-cally separated from neighbouring Bolivian lowlandsites (Herzog and Kessler 2002 Herzog et al 2016)In fact previous Andean NSDF legume phylogeniesshowed consistently high geographical structurethroughout their distribution suggesting that at leastthe NSDF flora has been assembled gradually over thepast ca 19 Myr with processes of in situ diversifica-tion largely undisturbed by new immigrants (Seuroarkinenet al 2012) From this perspective the NSDF on theslopes of centralndashsouth-western South America consti-tutes an important biogeographical area with ecologi-cal and geographical long-term stability and probablyoriginated by convergence scenarios resulting fromancient dispersal events that involved low niche differ-entiation and strong dispersal limitation from otherNSDF patches (Pennington et al 2009 Seuroarkinenet al 2012) This also suggests that NSDFs fromEcuador and northern Peru to western BoliviandashArgen-tina might constitute remnants of a previously widerexpansion according to the Pleistocene NSDF forma-tion (Seuroarkinen et al 2012 Cortes et al 2015 de Meloet al 2016) Passerine fossils support this hypothesisand suggest that some NSDF species had wider distri-butions in the past (Oswald and Steadman 2015) withthe Andean valleys (eg the Mara~non and ApurimacValleys) having direct influence on divergence andextinction processes by acting as semi-permeable barri-ers (Seuroarkinen et al 2011 Smith et al 2014 Wingerand Bates 2015)Finally we found a stretch of NSDF from north-

eastern Brazil south to ParaguayndashArgentina and northtowards eastern Bolivia (Fig 2) which is concordantwith previous studies showing a well-supported group-ing among these forests and suggesting the existence ofa historical stable unit known as the Open Dry Diago-nal (Mayle 2004 Porzecanski and Cracraft 2005Linares-Palomino et al 2011 Werneck et al 20112012) Today many dry habitat sister taxa are geo-graphically distant from one another suggesting thatthese species have either undergone long-distance dis-persal across large tracts of humid forests or that aridregions were once connected (Oswald et al 2017) The

absence of synapomorphies within the Chiquitano andMisiones Province forests (Table 1 Appendix S4) pro-vides some evidence for the recent formation of birdassemblages in this area (Herzog and Kessler 2002Smith et al 2014) as was found for woody plants(Linares-Palomino et al 2011 Banda et al 2016) Infact several authors suggest climatic and floristic char-acteristics closer to Atlantic rain forests than to NSDFsfor these areas (Olson et al 2001 Oliveira-Filho et al2006) which were probably colonized by dry forestspecies during drier and cooler periods of the Pleis-tocene These colonization processes were probablyfavoured by a potential narrow eastndashwest corridor con-necting NSDF in the extremes of north-eastern Braziland south-western South America during the LGM(Werneck et al 2011) This connection could serve asa biogeographical link sheltering floristic and faunisticdispersal routes resulting in partial expansions ofNSDFs in some areas of Amazonia (Werneck et al2011 2012) As a consequence of this last idea thelong-distance dispersal hypothesis of NSDFs in south-eastern South America proposed by Mayle (2004)remains a scenario that still needs independent testing

Final considerations

Our study suggests that evolution of the well-differ-entiated avifaunas in the four major geographicalgroups herein identified might have involved a mixtureof vicariance and dispersal events which provides newinsights into the diversification centres and historicalrelationships for the distributional patterns of thisecosystem and its associated avifauna Despite somespecies being shared among different clades (regions)and that this probably increased after closure of thePanama Isthmus (Morrone 2014b) and during themaximum extension of forests at the LGM (Bandaet al 2016) we observed a clear differentiation amongthe main areas with high degrees of endemism(Fig 2) This agrees with the idea of independent evo-lutionary histories among NSDF nuclei across theirdistribution (Becerra 2005 Cortes et al 2015 deMelo et al 2016)Differences observed with previous studies (eg

Porzecanski and Cracraft 2005 Rıos-Mu~noz andNavarro-Sigeurouenza 2012) are largely due to taxonomicuncertainties (ie species concepts and new species thathave been described or recognized since those studiesJetz et al 2002) as well as by differences in the defini-tion of sampling area sizes and algorithms used foranalyses For PAE there is a trend to decrease theabsolute number of steps and to increase the numberof synapomorphies as the size of the area increases(Morrone and Escalante 2002) Likewise it is impor-tant to note that this approach depends on detailed pri-mary biodiversity occurrence data In studies like this

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 11

with large data numbers it is worth noting thatobserved patterns of species distributional data couldbe influenced by diverse sources of error (eg taxo-nomic uncertainties identification errors or errors onlocality information associated with species occur-rences) Thus it is important to promote the constantupdate of the taxonomic status of specimens depositedin collections as well as the procedures for cleaningand revising biodiversity data from heterogeneoussources (Navarro-Sigeurouenza et al 2003 Bloom et al2017)Although our results mostly apply to the geographi-

cal scale and taxa used for this study the NSDF avi-fauna distribution patterns we found are non-randomand congruent with those for other organisms(Becerra 2005 Porzecanski and Cracraft 2005 Pen-nington et al 2009 Rıos-Mu~noz and Navarro-Sigeurouenza 2012 Oswald and Steadman 2015 Bandaet al 2016) The general pattern identified hereimplies the need to continue to study these areas ofendemism which certainly could have a much longerhistory than the temporal framework of the Pleis-tocene Thus from a conservation perspective mosthistorical areas of endemism must be considered aspriorities and incorporated into efforts of conserva-tions planning (eg Nori et al 2016) because theyreflect an unique history of the Earth and its biota Afailure to protect them would result in major losses ofunique species diversity for this highly threatenedecosystem The observed distributional patterns of theavifauna as well as their relationships across theNSDFs constitute evidence of the great complexity ofthe biogeography of this ecosystem in which each par-ticular pattern should be considered a stepping stonealong a true arc recognized as the biogeography ofthe Neotropics

Acknowledgments

We appreciate the efforts of museums (seeAppendix S1) and their curators as well as field collec-tors and institutions that made our study possibleDAP-T extends his gratitude to Consejo Nacional deCiencia y Tecnologıa (CONACyT Mexico) for a doc-toral scholarship (297538) and the Smithsonian-MasonSchool of Conservation for a scholarship during thecourse ldquoSpatial Ecology Geospatial Analysis andRemote Sensing for Conservation (MCCS 0500)rdquo thatprovided the tools and skills necessary for the spatialanalysis in this study The Rufford Foundation (DAP-T 16017-1 DAP-T 20284-2) Idea Wild (DAP-T) Con-sejo de Desarrollo Cientıfico Humanıstico y Tec-nologico at Universidad del Zulia (CONDES projectsCC-0247-13 and CC-0497-16 DAP-T) and the CONA-CyT project 152060 (AGNS and ORRS) provided

financial and logistical support J C Catarı P JCampos and Z Perez-Duran (Bolivia) as well as J ADelgado (Peru) helped during fieldwork A GordilloC Mota-Vargas and I Perozo helped in georeferencingbird locality data T Escalante helped with the PAEThe manuscript was improved by comments from HM Ortega-Andrade C Mota-Vargas A Lira-NoriegaJ L Parra L A Gonzalez and F Lopez

Conflicts of interest

The authors declare no competing interests in thisstudy

References

Amorim DS 2001 Dos Amazonias In Llorente J Morrone JJ(Eds) Introduccion a la Biogeografıa en LatinoamericaConceptos Teorıas Metodos y Aplicaciones Las Prensas deCiencias Universidad Nacional Autonoma de Mexico MexicoDF pp 245ndash255

Anderson RP Lew D Peterson AT 2003 Evaluatingpredictive models of speciesrsquo distributions criteria for selectingoptimal models Ecol Modell 162 211ndash232

Banda K Delgado-Salinas A Dexter KG Linares-Palomino ROliveira-Filho A Prado D Pullan M Quintana C Riina RRodrıguez M Weintritt J Acevedo-Rodrıguez P Adarve JAlvarez E Aranguren BA Arteaga JC Aymard G Casta~noA Ceballos-Mago N Cogollo A Cuadros H Delgado FDevia W Due~nas H Fajardo L Fernandez A FernandezMA Franklin J Freid EH Galetti LA Gonto RGonzalez-M R Graveson R Helmer EH Idarraga ALopez R Marcano-Vega H Martınez OG Maturo HMMcDonald M McLaren K Melo O Mijares F Mogni VMolina D Moreno NdP Nassar JM Neves DM OakleyLJ Oatham M Olvera-Luna AR Pezzini FF DominguezOJR Rıos ME Rivera O Rodrıguez N Rojas A SeuroarkinenT Sanchez R Smith M Vargas C Villanueva B PenningtonRT 2016 Plant diversity patterns in Neotropical dry forests andtheir conservation implications Science 353 1383ndash1387

Barve N Barve V Jimenez-Valverde A Lira-Noriega AMaher SP Peterson AT Soberon J Villalobos F 2011The crucial role of the accessible area in ecological nichemodeling and species distribution modeling Ecol Modell 2221810ndash1819

Becerra JX 2005 Timing the origin and expansion of the Mexicantropical dry forest Proc Natl Acad Sci USA 102 10919ndash10923

Bloom TDS Flower A DeChaine EG 2017 Why georeferencingmatters introducing a practical protocol to prepare speciesoccurrence records for spatial analysis Ecol Evol 8 765ndash777

Brooks DR van Veller MGP 2003 Critique of parsimonyanalysis of endemicity as a method of historical biogeography JBiogeogr 30 819ndash825

Ceballos G 1995 Vertebrate diversity ecology and conservation inNeotropical dry forests In Bullock S Medina E Mooney H(Eds) Seasonally Dry Tropical Forests Cambridge UniversityPress Cambridge UK pp 195ndash220

Clements J Schulenberg T Iliff M Roberson D FredericksT Sullivan B Wood C 2015 The eBirdClements checklist ofbirds of the world v2015 URL httpwww birds cornell educlementschecklistdownloadIOC [Verified 22 April 2017]

Cortes ALA Rapini A Daniel TF 2015 The Tetrameriumlineage (Acanthaceae Justicieae) does not support the PleistoceneArc hypothesis for South American seasonally dry forests AmJ Bot 102 992ndash1007

12 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

(Fig 2a) and others showed some reciprocally mono-phyletic groups (Fig 2b) the division is clear betweenthe north-western Mexico [including the sections ofTamaulipas and Veracruz (TV) Pacific Mexican forest(PS) and Northern Oaxaca in southern Mexico (SP)]and the Central American (from southern Mexico toPanama) regions This last region is subsequentlydivided into two main subgroups (1) the northern(CAmN from southern Mexico to Nicaragua) and (2)the southern subgroup (CAmS from Nicaragua toPanama)Furthermore for both approaches we observed that

NSDFs in South America involved a well-resolvedclade divided into two reciprocally monophyleticgroups (Fig 2) One group comprises the grid cellsthat correspond to northern South America contain-ing all areas throughout the inter-Andean valleys(IAV from Colombia to the Caribbean coast) as wellas the north-western stretch of the Orinoco river (in-cluding the Cordillera de la Costa and the MaracaiboLake depression in Venezuela) and the dry forests

located in the western slope of the Sierra de Perija(from La Guajira to Santander-Cordoba Departmentsin Colombia CCV subgroup) The second group isformed by the forests located throughout the PacificEquatorial (PE) region (an area of endemism locatedfrom western Ecuador to north-western Peru includ-ing the central Andean Valleys) and a sister clade thatincludes the dry valleys located in southern Peru [ieApurımac-Mantaro and Tarapoto-Quillabamba forests(Ap-M)] as well as the south-eastern South Americaareas (ie NSDFs from western Bolivia and north-eastern Argentina to north-eastern Brazil) For thislast South American group two well-supported sub-groups are formed consecutively and separately onecontains the endemism area at the Sub-Andean pied-mont forests (SA-P) of north-western Argentina andwestern Bolivia and the other corresponds to theChiquitano and Misiones Province [ie south-easternBolivia north-eastern Argentina Paraguay and Uru-guay (C-MP)] and the endemism area of Caatinga(CaB located in north-eastern Brazil)

Fig 2 Regionalization of the Neotropical seasonally dry forests (NSDFs) based on the avifauna distribution and their relationships consideringcomplete (a) vs NSDF-restricted (b) species approaches (matrices of information) The cladograms revealed 17 NSDF geographical groupswhich involved four large regions The Baja California (A) CaribbeanndashAntilles islands (B) Mesoamerican (C) and South America (D) Acro-nyms in both cladograms and the map correspond to identified NSDF regions and areas of endemism [more than two synapomorphies (speciesrepresented with a black circle] BC = Baja California Cu = Cuba island HP = Haiti Dominican Republic and Puerto Rico islandsSS = Sonora and Sinaloa forests PS = Mexican Pacific slope lowlands SP = northern Oaxaca in southern Mexico TV = TamaulipasndashVeracruzstates YP = Yucatan Peninsula CAmN = northern Central America CAmS = nouthern Central America CAm = Central AmericaPa = Panama IAV = Inter-Andean Valleys of Colombia CCV = Caribbean ColombiandashVenezuela PE = Pacific Equatorial Ap-M = Apurimac-Mantaro SA-P = Sub-Andean Piedmont C-MP = Chiquitano forests and Misiones Province CaB = Brazilian Caatinga These well-resolvedclades are circumscribed by geographical and ecological barriers (dotted black lines) associated with the Gulf of California the leading edge ofthe Caribbean plate the Tehuantepec Isthmus the PolochicndashMotagua fault the Nicaragua Depression the Choco forest the Amazon basin andthe Andean Cordillera (including the Inter-Andean Valleys) Well-supported clades [more than two synapomorphies (species)] are shown with ablack circle in the acronyms of NSDF regions (see Table 1 for a complete description)

8 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

Discussion

The Baja Californiarsquos NSDF

In regional cladograms the Cape of Baja Californiaforms a distinct unit which supports the idea of asomewhat independent biogeographical history as hasbeen suggested in previous avian regionalizations Iso-lation of the NSDF in the Cape region from those inwestern Mexico (55ndash40 Mya) by the Gulf of Califor-nia has promoted the differentiation and a high degreeof endemism associated with the Cape of diversegroups including mammals birds snakes and insects(Grismer 2000 Rojas-Soto et al 2003 Rıos-Mu~nozand Navarro-Sigeurouenza 2012 OrsquoConnell et al 2017)However given the closest affinities of avian specieswith the nearby continent at the northern portion ofthe Peninsula (Rojas-Soto et al 2003) it seems clearthat the Cape avifauna originated through a terrestrialroute from the northern to the southern tip of the BajaCalifornia Peninsula (Rojas-Soto et al 2003OrsquoConnell et al 2017)

The CaribbeanndashAntilles avifauna

As found in previous biogeographical studies (egCrother and Guyer 1996 Zink et al 2000 Graham2003ab Vazquez-Miranda et al 2007) our analysessupport a monophyletic origin for groups of organismsinhabiting this region However a potential difficultyin explaining this monophyly is posited by the ques-tion of how birds reached those islands given that thecomplex geological history of the region included ldquoatleast eight events of fragmentation and seven ofhybridizationrdquo (Rosen 1985) In our study weobserved that several taxa acted as homoplasies whichcan be interpreted as the result of either dispersal orextinction events (Vazquez-Miranda et al 2007) Con-sidering that ~41 (n = 56 species) of NSDF avifaunais endemic and of these only eight species (two ofwhich are synapomorphies) were shared among thethree islands (Cuba Haiti Puerto Rico) the observedpattern is more probably explained by a mixture ofvicariance and dispersal events especially becausethose islands have had relatively recent bird exchanges(Vazquez-Miranda et al 2007) This idea is more par-simonious than assuming concordant dispersal pat-terns of individual taxa particularly considering thebiotarsquos affinities with both Central America andorSouth America (Crother and Guyer 1996 Zink et al2000 Graham 2003ab)Despite the existence of vicariance models explaining

affinities between the avifauna of this region with themainland (eg the temporal separation of the leadingedge of the Caribbean plate [beginning in the LateCretaceous and finishing in the MiocenendashMiddle

Pleiocene] Rosen 1985 Iturralde-Vinent and Mac-Phee 1999 Graham 2003ab Morrone 2014b) it isimportant to consider that dispersal events could havecontributed to defining the avifaunistic assemblages ofthis region as apparently occurred in the history ofthe woodpecker genus Melanerpes (Navarro-Sigeurouenzaet al 2017) It is difficult to compare the presentstudy with earlier studies because they used differentislands different taxa and different continental regionsas a reference Therefore further studies are needed toassess the relative age of these diversification centresand their biogeographical relationships

The Mesoamerican NSDFs

The division of this clade into sub-groups supports thelong isolation history of the identified Mesoamericanregions Our results highlight the importance of the Mex-ican Plateau and the Tehuantepec Isthmus in shapingthe distribution of Mesoamerican biotas in both high-lands and lowlands (Becerra 2005 Sanchez-Gonzalezet al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza 2012Banda et al 2016) The clear differentiation of thegroups on both sides of the Tehuantepec Isthmus(north-western Mexico Yucatan Peninsula and CentralAmerica Fig 2) as well as the subdivisions observedwithin the northern Mexican regions [Sonora-Sinaloa(SS) northern Oaxaca (SP) and TamaulipasndashVeracruz(TV) areas] might be explained by higher diversificationrates compared to other dry forest areas as detected forexample in the hummingbird Amazilia cyanocephala(Rodrıguez-Gomez et al 2013 OrsquoConnell et al 2017)Although the age of Mesoamerican NSDFs is notknown the fossil record and dated phylogenies suggest aMiocenendashPliocene age at least in north-western Mexico(Becerra 2005) These forests probably expanded toCentral America (where fossil evidence seems to suggestthat it invaded only after 25 Mya Graham and Dilcher1995) after the Sierra Madre Occidental and the Trans-mexican Volcanic belt were formed creating conditionsfor the establishment and persistence of NSDFsThe clear separation observed for the Sonora-Sina-

loa (SS) subgroup and the rest of the Mesoamericanmainland (Fig 2a) could be explained by the highaffinities with Nearctic biota of this northernmost area(Stotz et al 1996) In fact despite the high speciesturnover (n = 188) observed between the Sonora-Sina-loa subgroup and the rest of the MesoamericanNSDFs these northernmost forests shared 54 species(95) with those reported for the Baja Californiaregion Furthermore we propose a region formed byall areas west of the Tehuantepec Isthmus located inMexico [including NSDF in TamaulipasndashVeracruz(TV) and except those corresponding to Baja Califor-nia (BC)] which is supported by previous biogeo-graphical hypotheses and also suggested that the

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 9

north-eastern Gulf of Mexico is biogeographically dis-tinct from the southern and south-eastern part of theGulf coast (Liebherr 1994 Hernandez-Ba~nos et al1995 Marshall and Liebherr 2000 Rıos-Mu~noz andNavarro-Sigeurouenza 2012)On the other hand separation within the south-east-

ern Isthmus groups (light blues in Fig 2) supports thebiogeographical hypothesis of vicariant processes asso-ciated with the PolochicndashMotagua fault and the Nicar-agua Depression in the region (Morrone 2001Schuster et al 2003 Liede and Meve 2004) Here weidentified at least three distinct regions suggesting theexistence of bird assemblages with independent biogeo-graphical histories probably shaped by new immi-grants from dispersal events between South and NorthAmerica after the establishment of the Panama Isth-mus (Smith and Klicka 2010) For instance weobserved that ~73 of the NSDF avifauna is sharedbetween southern Central America and northern SouthAmerica (with ~72 of NSDF-restricted speciesshared between the NSDF from Panama and northernSouth America) confirming the existence of strongaffinities between the two regions (Linares-Palominoet al 2011 Rıos-Mu~noz and Navarro-Sigeurouenza 2012Banda et al 2016)In the Panama region differences in the affinities

between the cladograms for this region may be influ-enced by the inclusion (ie complete approach) or theexclusion (ie restricted approach) of species widelydistributed in the region and inhabiting other ecosys-tems surrounding NSDFs In this sense such differ-ences could be explained by different temporalpatterns of speciation documented for taxa associatedwith humid and dry habitats across the Isthmus ofPanama (Smith et al 2012) The Panama area hashad a history of expanding and contracting processesfor both the dry and the humid forests and even whenboth habitats consist of ldquoyoungrdquo and ldquooldrdquo speciesthe dry habitat species are geographically more distantfrom their nearest counterpart in Central America andnorthern South America compared with species fromthe humid habitats in these regions (Oswald andSteadman 2015 Oswald et al 2017) This phe-nomenon if generalized is probably an importantcontributor to biogeographical patterns in the south-ernmost section of the Mesoamerican region (Sanchez-Gonzalez et al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza 2012 Smith et al 2012)

South American NSDF areas and the Pleistocene ArcHypothesis

Previous studies of distributional patterns within theSouth American biota have accepted the hypothesisthat areas of endemism have arisen by isolation ofmore widespread biotas This scenario is commonly

associated with paleogeographical changes alterationsin river systems Pleistocene climatic fluctuations or amixture of all these processes (Herzog and Kessler2002 Werneck et al 2012) However we find littlesupport for widespread Pleistocene NSDF formationthroughout South America because few species(1033 n = 100 species) are widespread and sharedby more than 75 of South American areas Thisresult is concordant with NSDF plant patterns recentlyreported by Linares-Palomino et al (2011) and Bandaet al (2016) as well as with the restricted climatic cor-ridors found by Werneck et al (2011) which suggestthat LGM climate was probably too dry and cold tosupport large tracts of NSDFs across the continentSome authors suggest that long-distance ancient dis-

persal events of unrelated taxonomic groups are lesslikely than vicariant associations (Werneck et al2011 de Melo et al 2016) Likewise there is substan-tial support that NSDF is a dispersal-limited biomegiven the phylogenetic niche conservatism and stronggeographically structured phylogenies for woody plantclades (Pennington et al 2009 Pennington and Lavin2016) Thus an allopatric-vicariance explanationmdashde-lineated by major geographical or ecological barriersbefore the Pleistocenemdashseems more plausible toexplain the evolution and relationships among all theseforest areas which is supported by results based onendemic taxa For instance the biogeography of Bro-togeris parakeets implies a dynamic history for SouthAmerican biomes since the Pliocene corroboratingthat geological evolution of Amazonia has been impor-tant in shaping its biodiversity with taxa endemic toNSDFs and other biomes (Ribas et al 2009) In thissense our results suggest that the wide distributions ofa relatively low proportion of NSDF species and eco-logical similarity among them (including specialistsand generalists) may reflect only limited recent long-distance dispersal events and species exchange (Dicket al 2003 2004 Mayle 2004)The separation of a northern South American clade

(Fig 2) suggests an effect of the Amazonia rain foreststhe wet Choco forest and the final uplift of the north-ern Andes as barriers for the dispersal of dry forestspecies (Amorim 2001 Garzione et al 2008 Hoornet al 2010 Morrone 2014b Banda et al 2016) Clus-ters of the northern inter-Andean valleys (located alongthe Magdalena and Cauca Rivers) together with Carib-bean coastal areas reflect the heterogeneity of dryAndean regions (with important species turnover withMesoamerica and CaribbeanndashAntilles) where few spe-cies are distributed all the way from the northern tothe southern Andes (Nores 2004 Vazquez-Mirandaet al 2007) These results are also consistent withthose previously reported for other taxonomic groupsin the region that are associated with NSDFs (Linares-Palomino et al 2011 Morrone 2014b)

10 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

Although our analyses generally agree with thePleistocene Arc Hypothesis proposed originally byPrado and Gibbs (1993)mdashformation of a once exten-sive and largely contiguous seasonal woodland com-prising the Caatingas the Misiones and Sub-AndeanPiedmont nucleimdashthe clear separation among NSDFsfrom centralndashsouth-western and south-eastern SouthAmerica supports the idea that both uplift of theAndes and historical (Cenozoic and Quaternary) cli-matic regimes of north-western EcuadorndashPeru andwestern Bolivia-Argentina are probably responsible forthe high endemism and distributions of their NSDFbird communities (Hoorn et al 2010 Morrone2014b) These communities have been biogeographi-cally separated from neighbouring Bolivian lowlandsites (Herzog and Kessler 2002 Herzog et al 2016)In fact previous Andean NSDF legume phylogeniesshowed consistently high geographical structurethroughout their distribution suggesting that at leastthe NSDF flora has been assembled gradually over thepast ca 19 Myr with processes of in situ diversifica-tion largely undisturbed by new immigrants (Seuroarkinenet al 2012) From this perspective the NSDF on theslopes of centralndashsouth-western South America consti-tutes an important biogeographical area with ecologi-cal and geographical long-term stability and probablyoriginated by convergence scenarios resulting fromancient dispersal events that involved low niche differ-entiation and strong dispersal limitation from otherNSDF patches (Pennington et al 2009 Seuroarkinenet al 2012) This also suggests that NSDFs fromEcuador and northern Peru to western BoliviandashArgen-tina might constitute remnants of a previously widerexpansion according to the Pleistocene NSDF forma-tion (Seuroarkinen et al 2012 Cortes et al 2015 de Meloet al 2016) Passerine fossils support this hypothesisand suggest that some NSDF species had wider distri-butions in the past (Oswald and Steadman 2015) withthe Andean valleys (eg the Mara~non and ApurimacValleys) having direct influence on divergence andextinction processes by acting as semi-permeable barri-ers (Seuroarkinen et al 2011 Smith et al 2014 Wingerand Bates 2015)Finally we found a stretch of NSDF from north-

eastern Brazil south to ParaguayndashArgentina and northtowards eastern Bolivia (Fig 2) which is concordantwith previous studies showing a well-supported group-ing among these forests and suggesting the existence ofa historical stable unit known as the Open Dry Diago-nal (Mayle 2004 Porzecanski and Cracraft 2005Linares-Palomino et al 2011 Werneck et al 20112012) Today many dry habitat sister taxa are geo-graphically distant from one another suggesting thatthese species have either undergone long-distance dis-persal across large tracts of humid forests or that aridregions were once connected (Oswald et al 2017) The

absence of synapomorphies within the Chiquitano andMisiones Province forests (Table 1 Appendix S4) pro-vides some evidence for the recent formation of birdassemblages in this area (Herzog and Kessler 2002Smith et al 2014) as was found for woody plants(Linares-Palomino et al 2011 Banda et al 2016) Infact several authors suggest climatic and floristic char-acteristics closer to Atlantic rain forests than to NSDFsfor these areas (Olson et al 2001 Oliveira-Filho et al2006) which were probably colonized by dry forestspecies during drier and cooler periods of the Pleis-tocene These colonization processes were probablyfavoured by a potential narrow eastndashwest corridor con-necting NSDF in the extremes of north-eastern Braziland south-western South America during the LGM(Werneck et al 2011) This connection could serve asa biogeographical link sheltering floristic and faunisticdispersal routes resulting in partial expansions ofNSDFs in some areas of Amazonia (Werneck et al2011 2012) As a consequence of this last idea thelong-distance dispersal hypothesis of NSDFs in south-eastern South America proposed by Mayle (2004)remains a scenario that still needs independent testing

Final considerations

Our study suggests that evolution of the well-differ-entiated avifaunas in the four major geographicalgroups herein identified might have involved a mixtureof vicariance and dispersal events which provides newinsights into the diversification centres and historicalrelationships for the distributional patterns of thisecosystem and its associated avifauna Despite somespecies being shared among different clades (regions)and that this probably increased after closure of thePanama Isthmus (Morrone 2014b) and during themaximum extension of forests at the LGM (Bandaet al 2016) we observed a clear differentiation amongthe main areas with high degrees of endemism(Fig 2) This agrees with the idea of independent evo-lutionary histories among NSDF nuclei across theirdistribution (Becerra 2005 Cortes et al 2015 deMelo et al 2016)Differences observed with previous studies (eg

Porzecanski and Cracraft 2005 Rıos-Mu~noz andNavarro-Sigeurouenza 2012) are largely due to taxonomicuncertainties (ie species concepts and new species thathave been described or recognized since those studiesJetz et al 2002) as well as by differences in the defini-tion of sampling area sizes and algorithms used foranalyses For PAE there is a trend to decrease theabsolute number of steps and to increase the numberof synapomorphies as the size of the area increases(Morrone and Escalante 2002) Likewise it is impor-tant to note that this approach depends on detailed pri-mary biodiversity occurrence data In studies like this

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 11

with large data numbers it is worth noting thatobserved patterns of species distributional data couldbe influenced by diverse sources of error (eg taxo-nomic uncertainties identification errors or errors onlocality information associated with species occur-rences) Thus it is important to promote the constantupdate of the taxonomic status of specimens depositedin collections as well as the procedures for cleaningand revising biodiversity data from heterogeneoussources (Navarro-Sigeurouenza et al 2003 Bloom et al2017)Although our results mostly apply to the geographi-

cal scale and taxa used for this study the NSDF avi-fauna distribution patterns we found are non-randomand congruent with those for other organisms(Becerra 2005 Porzecanski and Cracraft 2005 Pen-nington et al 2009 Rıos-Mu~noz and Navarro-Sigeurouenza 2012 Oswald and Steadman 2015 Bandaet al 2016) The general pattern identified hereimplies the need to continue to study these areas ofendemism which certainly could have a much longerhistory than the temporal framework of the Pleis-tocene Thus from a conservation perspective mosthistorical areas of endemism must be considered aspriorities and incorporated into efforts of conserva-tions planning (eg Nori et al 2016) because theyreflect an unique history of the Earth and its biota Afailure to protect them would result in major losses ofunique species diversity for this highly threatenedecosystem The observed distributional patterns of theavifauna as well as their relationships across theNSDFs constitute evidence of the great complexity ofthe biogeography of this ecosystem in which each par-ticular pattern should be considered a stepping stonealong a true arc recognized as the biogeography ofthe Neotropics

Acknowledgments

We appreciate the efforts of museums (seeAppendix S1) and their curators as well as field collec-tors and institutions that made our study possibleDAP-T extends his gratitude to Consejo Nacional deCiencia y Tecnologıa (CONACyT Mexico) for a doc-toral scholarship (297538) and the Smithsonian-MasonSchool of Conservation for a scholarship during thecourse ldquoSpatial Ecology Geospatial Analysis andRemote Sensing for Conservation (MCCS 0500)rdquo thatprovided the tools and skills necessary for the spatialanalysis in this study The Rufford Foundation (DAP-T 16017-1 DAP-T 20284-2) Idea Wild (DAP-T) Con-sejo de Desarrollo Cientıfico Humanıstico y Tec-nologico at Universidad del Zulia (CONDES projectsCC-0247-13 and CC-0497-16 DAP-T) and the CONA-CyT project 152060 (AGNS and ORRS) provided

financial and logistical support J C Catarı P JCampos and Z Perez-Duran (Bolivia) as well as J ADelgado (Peru) helped during fieldwork A GordilloC Mota-Vargas and I Perozo helped in georeferencingbird locality data T Escalante helped with the PAEThe manuscript was improved by comments from HM Ortega-Andrade C Mota-Vargas A Lira-NoriegaJ L Parra L A Gonzalez and F Lopez

Conflicts of interest

The authors declare no competing interests in thisstudy

References

Amorim DS 2001 Dos Amazonias In Llorente J Morrone JJ(Eds) Introduccion a la Biogeografıa en LatinoamericaConceptos Teorıas Metodos y Aplicaciones Las Prensas deCiencias Universidad Nacional Autonoma de Mexico MexicoDF pp 245ndash255

Anderson RP Lew D Peterson AT 2003 Evaluatingpredictive models of speciesrsquo distributions criteria for selectingoptimal models Ecol Modell 162 211ndash232

Banda K Delgado-Salinas A Dexter KG Linares-Palomino ROliveira-Filho A Prado D Pullan M Quintana C Riina RRodrıguez M Weintritt J Acevedo-Rodrıguez P Adarve JAlvarez E Aranguren BA Arteaga JC Aymard G Casta~noA Ceballos-Mago N Cogollo A Cuadros H Delgado FDevia W Due~nas H Fajardo L Fernandez A FernandezMA Franklin J Freid EH Galetti LA Gonto RGonzalez-M R Graveson R Helmer EH Idarraga ALopez R Marcano-Vega H Martınez OG Maturo HMMcDonald M McLaren K Melo O Mijares F Mogni VMolina D Moreno NdP Nassar JM Neves DM OakleyLJ Oatham M Olvera-Luna AR Pezzini FF DominguezOJR Rıos ME Rivera O Rodrıguez N Rojas A SeuroarkinenT Sanchez R Smith M Vargas C Villanueva B PenningtonRT 2016 Plant diversity patterns in Neotropical dry forests andtheir conservation implications Science 353 1383ndash1387

Barve N Barve V Jimenez-Valverde A Lira-Noriega AMaher SP Peterson AT Soberon J Villalobos F 2011The crucial role of the accessible area in ecological nichemodeling and species distribution modeling Ecol Modell 2221810ndash1819

Becerra JX 2005 Timing the origin and expansion of the Mexicantropical dry forest Proc Natl Acad Sci USA 102 10919ndash10923

Bloom TDS Flower A DeChaine EG 2017 Why georeferencingmatters introducing a practical protocol to prepare speciesoccurrence records for spatial analysis Ecol Evol 8 765ndash777

Brooks DR van Veller MGP 2003 Critique of parsimonyanalysis of endemicity as a method of historical biogeography JBiogeogr 30 819ndash825

Ceballos G 1995 Vertebrate diversity ecology and conservation inNeotropical dry forests In Bullock S Medina E Mooney H(Eds) Seasonally Dry Tropical Forests Cambridge UniversityPress Cambridge UK pp 195ndash220

Clements J Schulenberg T Iliff M Roberson D FredericksT Sullivan B Wood C 2015 The eBirdClements checklist ofbirds of the world v2015 URL httpwww birds cornell educlementschecklistdownloadIOC [Verified 22 April 2017]

Cortes ALA Rapini A Daniel TF 2015 The Tetrameriumlineage (Acanthaceae Justicieae) does not support the PleistoceneArc hypothesis for South American seasonally dry forests AmJ Bot 102 992ndash1007

12 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

Discussion

The Baja Californiarsquos NSDF

In regional cladograms the Cape of Baja Californiaforms a distinct unit which supports the idea of asomewhat independent biogeographical history as hasbeen suggested in previous avian regionalizations Iso-lation of the NSDF in the Cape region from those inwestern Mexico (55ndash40 Mya) by the Gulf of Califor-nia has promoted the differentiation and a high degreeof endemism associated with the Cape of diversegroups including mammals birds snakes and insects(Grismer 2000 Rojas-Soto et al 2003 Rıos-Mu~nozand Navarro-Sigeurouenza 2012 OrsquoConnell et al 2017)However given the closest affinities of avian specieswith the nearby continent at the northern portion ofthe Peninsula (Rojas-Soto et al 2003) it seems clearthat the Cape avifauna originated through a terrestrialroute from the northern to the southern tip of the BajaCalifornia Peninsula (Rojas-Soto et al 2003OrsquoConnell et al 2017)

The CaribbeanndashAntilles avifauna

As found in previous biogeographical studies (egCrother and Guyer 1996 Zink et al 2000 Graham2003ab Vazquez-Miranda et al 2007) our analysessupport a monophyletic origin for groups of organismsinhabiting this region However a potential difficultyin explaining this monophyly is posited by the ques-tion of how birds reached those islands given that thecomplex geological history of the region included ldquoatleast eight events of fragmentation and seven ofhybridizationrdquo (Rosen 1985) In our study weobserved that several taxa acted as homoplasies whichcan be interpreted as the result of either dispersal orextinction events (Vazquez-Miranda et al 2007) Con-sidering that ~41 (n = 56 species) of NSDF avifaunais endemic and of these only eight species (two ofwhich are synapomorphies) were shared among thethree islands (Cuba Haiti Puerto Rico) the observedpattern is more probably explained by a mixture ofvicariance and dispersal events especially becausethose islands have had relatively recent bird exchanges(Vazquez-Miranda et al 2007) This idea is more par-simonious than assuming concordant dispersal pat-terns of individual taxa particularly considering thebiotarsquos affinities with both Central America andorSouth America (Crother and Guyer 1996 Zink et al2000 Graham 2003ab)Despite the existence of vicariance models explaining

affinities between the avifauna of this region with themainland (eg the temporal separation of the leadingedge of the Caribbean plate [beginning in the LateCretaceous and finishing in the MiocenendashMiddle

Pleiocene] Rosen 1985 Iturralde-Vinent and Mac-Phee 1999 Graham 2003ab Morrone 2014b) it isimportant to consider that dispersal events could havecontributed to defining the avifaunistic assemblages ofthis region as apparently occurred in the history ofthe woodpecker genus Melanerpes (Navarro-Sigeurouenzaet al 2017) It is difficult to compare the presentstudy with earlier studies because they used differentislands different taxa and different continental regionsas a reference Therefore further studies are needed toassess the relative age of these diversification centresand their biogeographical relationships

The Mesoamerican NSDFs

The division of this clade into sub-groups supports thelong isolation history of the identified Mesoamericanregions Our results highlight the importance of the Mex-ican Plateau and the Tehuantepec Isthmus in shapingthe distribution of Mesoamerican biotas in both high-lands and lowlands (Becerra 2005 Sanchez-Gonzalezet al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza 2012Banda et al 2016) The clear differentiation of thegroups on both sides of the Tehuantepec Isthmus(north-western Mexico Yucatan Peninsula and CentralAmerica Fig 2) as well as the subdivisions observedwithin the northern Mexican regions [Sonora-Sinaloa(SS) northern Oaxaca (SP) and TamaulipasndashVeracruz(TV) areas] might be explained by higher diversificationrates compared to other dry forest areas as detected forexample in the hummingbird Amazilia cyanocephala(Rodrıguez-Gomez et al 2013 OrsquoConnell et al 2017)Although the age of Mesoamerican NSDFs is notknown the fossil record and dated phylogenies suggest aMiocenendashPliocene age at least in north-western Mexico(Becerra 2005) These forests probably expanded toCentral America (where fossil evidence seems to suggestthat it invaded only after 25 Mya Graham and Dilcher1995) after the Sierra Madre Occidental and the Trans-mexican Volcanic belt were formed creating conditionsfor the establishment and persistence of NSDFsThe clear separation observed for the Sonora-Sina-

loa (SS) subgroup and the rest of the Mesoamericanmainland (Fig 2a) could be explained by the highaffinities with Nearctic biota of this northernmost area(Stotz et al 1996) In fact despite the high speciesturnover (n = 188) observed between the Sonora-Sina-loa subgroup and the rest of the MesoamericanNSDFs these northernmost forests shared 54 species(95) with those reported for the Baja Californiaregion Furthermore we propose a region formed byall areas west of the Tehuantepec Isthmus located inMexico [including NSDF in TamaulipasndashVeracruz(TV) and except those corresponding to Baja Califor-nia (BC)] which is supported by previous biogeo-graphical hypotheses and also suggested that the

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 9

north-eastern Gulf of Mexico is biogeographically dis-tinct from the southern and south-eastern part of theGulf coast (Liebherr 1994 Hernandez-Ba~nos et al1995 Marshall and Liebherr 2000 Rıos-Mu~noz andNavarro-Sigeurouenza 2012)On the other hand separation within the south-east-

ern Isthmus groups (light blues in Fig 2) supports thebiogeographical hypothesis of vicariant processes asso-ciated with the PolochicndashMotagua fault and the Nicar-agua Depression in the region (Morrone 2001Schuster et al 2003 Liede and Meve 2004) Here weidentified at least three distinct regions suggesting theexistence of bird assemblages with independent biogeo-graphical histories probably shaped by new immi-grants from dispersal events between South and NorthAmerica after the establishment of the Panama Isth-mus (Smith and Klicka 2010) For instance weobserved that ~73 of the NSDF avifauna is sharedbetween southern Central America and northern SouthAmerica (with ~72 of NSDF-restricted speciesshared between the NSDF from Panama and northernSouth America) confirming the existence of strongaffinities between the two regions (Linares-Palominoet al 2011 Rıos-Mu~noz and Navarro-Sigeurouenza 2012Banda et al 2016)In the Panama region differences in the affinities

between the cladograms for this region may be influ-enced by the inclusion (ie complete approach) or theexclusion (ie restricted approach) of species widelydistributed in the region and inhabiting other ecosys-tems surrounding NSDFs In this sense such differ-ences could be explained by different temporalpatterns of speciation documented for taxa associatedwith humid and dry habitats across the Isthmus ofPanama (Smith et al 2012) The Panama area hashad a history of expanding and contracting processesfor both the dry and the humid forests and even whenboth habitats consist of ldquoyoungrdquo and ldquooldrdquo speciesthe dry habitat species are geographically more distantfrom their nearest counterpart in Central America andnorthern South America compared with species fromthe humid habitats in these regions (Oswald andSteadman 2015 Oswald et al 2017) This phe-nomenon if generalized is probably an importantcontributor to biogeographical patterns in the south-ernmost section of the Mesoamerican region (Sanchez-Gonzalez et al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza 2012 Smith et al 2012)

South American NSDF areas and the Pleistocene ArcHypothesis

Previous studies of distributional patterns within theSouth American biota have accepted the hypothesisthat areas of endemism have arisen by isolation ofmore widespread biotas This scenario is commonly

associated with paleogeographical changes alterationsin river systems Pleistocene climatic fluctuations or amixture of all these processes (Herzog and Kessler2002 Werneck et al 2012) However we find littlesupport for widespread Pleistocene NSDF formationthroughout South America because few species(1033 n = 100 species) are widespread and sharedby more than 75 of South American areas Thisresult is concordant with NSDF plant patterns recentlyreported by Linares-Palomino et al (2011) and Bandaet al (2016) as well as with the restricted climatic cor-ridors found by Werneck et al (2011) which suggestthat LGM climate was probably too dry and cold tosupport large tracts of NSDFs across the continentSome authors suggest that long-distance ancient dis-

persal events of unrelated taxonomic groups are lesslikely than vicariant associations (Werneck et al2011 de Melo et al 2016) Likewise there is substan-tial support that NSDF is a dispersal-limited biomegiven the phylogenetic niche conservatism and stronggeographically structured phylogenies for woody plantclades (Pennington et al 2009 Pennington and Lavin2016) Thus an allopatric-vicariance explanationmdashde-lineated by major geographical or ecological barriersbefore the Pleistocenemdashseems more plausible toexplain the evolution and relationships among all theseforest areas which is supported by results based onendemic taxa For instance the biogeography of Bro-togeris parakeets implies a dynamic history for SouthAmerican biomes since the Pliocene corroboratingthat geological evolution of Amazonia has been impor-tant in shaping its biodiversity with taxa endemic toNSDFs and other biomes (Ribas et al 2009) In thissense our results suggest that the wide distributions ofa relatively low proportion of NSDF species and eco-logical similarity among them (including specialistsand generalists) may reflect only limited recent long-distance dispersal events and species exchange (Dicket al 2003 2004 Mayle 2004)The separation of a northern South American clade

(Fig 2) suggests an effect of the Amazonia rain foreststhe wet Choco forest and the final uplift of the north-ern Andes as barriers for the dispersal of dry forestspecies (Amorim 2001 Garzione et al 2008 Hoornet al 2010 Morrone 2014b Banda et al 2016) Clus-ters of the northern inter-Andean valleys (located alongthe Magdalena and Cauca Rivers) together with Carib-bean coastal areas reflect the heterogeneity of dryAndean regions (with important species turnover withMesoamerica and CaribbeanndashAntilles) where few spe-cies are distributed all the way from the northern tothe southern Andes (Nores 2004 Vazquez-Mirandaet al 2007) These results are also consistent withthose previously reported for other taxonomic groupsin the region that are associated with NSDFs (Linares-Palomino et al 2011 Morrone 2014b)

10 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

Although our analyses generally agree with thePleistocene Arc Hypothesis proposed originally byPrado and Gibbs (1993)mdashformation of a once exten-sive and largely contiguous seasonal woodland com-prising the Caatingas the Misiones and Sub-AndeanPiedmont nucleimdashthe clear separation among NSDFsfrom centralndashsouth-western and south-eastern SouthAmerica supports the idea that both uplift of theAndes and historical (Cenozoic and Quaternary) cli-matic regimes of north-western EcuadorndashPeru andwestern Bolivia-Argentina are probably responsible forthe high endemism and distributions of their NSDFbird communities (Hoorn et al 2010 Morrone2014b) These communities have been biogeographi-cally separated from neighbouring Bolivian lowlandsites (Herzog and Kessler 2002 Herzog et al 2016)In fact previous Andean NSDF legume phylogeniesshowed consistently high geographical structurethroughout their distribution suggesting that at leastthe NSDF flora has been assembled gradually over thepast ca 19 Myr with processes of in situ diversifica-tion largely undisturbed by new immigrants (Seuroarkinenet al 2012) From this perspective the NSDF on theslopes of centralndashsouth-western South America consti-tutes an important biogeographical area with ecologi-cal and geographical long-term stability and probablyoriginated by convergence scenarios resulting fromancient dispersal events that involved low niche differ-entiation and strong dispersal limitation from otherNSDF patches (Pennington et al 2009 Seuroarkinenet al 2012) This also suggests that NSDFs fromEcuador and northern Peru to western BoliviandashArgen-tina might constitute remnants of a previously widerexpansion according to the Pleistocene NSDF forma-tion (Seuroarkinen et al 2012 Cortes et al 2015 de Meloet al 2016) Passerine fossils support this hypothesisand suggest that some NSDF species had wider distri-butions in the past (Oswald and Steadman 2015) withthe Andean valleys (eg the Mara~non and ApurimacValleys) having direct influence on divergence andextinction processes by acting as semi-permeable barri-ers (Seuroarkinen et al 2011 Smith et al 2014 Wingerand Bates 2015)Finally we found a stretch of NSDF from north-

eastern Brazil south to ParaguayndashArgentina and northtowards eastern Bolivia (Fig 2) which is concordantwith previous studies showing a well-supported group-ing among these forests and suggesting the existence ofa historical stable unit known as the Open Dry Diago-nal (Mayle 2004 Porzecanski and Cracraft 2005Linares-Palomino et al 2011 Werneck et al 20112012) Today many dry habitat sister taxa are geo-graphically distant from one another suggesting thatthese species have either undergone long-distance dis-persal across large tracts of humid forests or that aridregions were once connected (Oswald et al 2017) The

absence of synapomorphies within the Chiquitano andMisiones Province forests (Table 1 Appendix S4) pro-vides some evidence for the recent formation of birdassemblages in this area (Herzog and Kessler 2002Smith et al 2014) as was found for woody plants(Linares-Palomino et al 2011 Banda et al 2016) Infact several authors suggest climatic and floristic char-acteristics closer to Atlantic rain forests than to NSDFsfor these areas (Olson et al 2001 Oliveira-Filho et al2006) which were probably colonized by dry forestspecies during drier and cooler periods of the Pleis-tocene These colonization processes were probablyfavoured by a potential narrow eastndashwest corridor con-necting NSDF in the extremes of north-eastern Braziland south-western South America during the LGM(Werneck et al 2011) This connection could serve asa biogeographical link sheltering floristic and faunisticdispersal routes resulting in partial expansions ofNSDFs in some areas of Amazonia (Werneck et al2011 2012) As a consequence of this last idea thelong-distance dispersal hypothesis of NSDFs in south-eastern South America proposed by Mayle (2004)remains a scenario that still needs independent testing

Final considerations

Our study suggests that evolution of the well-differ-entiated avifaunas in the four major geographicalgroups herein identified might have involved a mixtureof vicariance and dispersal events which provides newinsights into the diversification centres and historicalrelationships for the distributional patterns of thisecosystem and its associated avifauna Despite somespecies being shared among different clades (regions)and that this probably increased after closure of thePanama Isthmus (Morrone 2014b) and during themaximum extension of forests at the LGM (Bandaet al 2016) we observed a clear differentiation amongthe main areas with high degrees of endemism(Fig 2) This agrees with the idea of independent evo-lutionary histories among NSDF nuclei across theirdistribution (Becerra 2005 Cortes et al 2015 deMelo et al 2016)Differences observed with previous studies (eg

Porzecanski and Cracraft 2005 Rıos-Mu~noz andNavarro-Sigeurouenza 2012) are largely due to taxonomicuncertainties (ie species concepts and new species thathave been described or recognized since those studiesJetz et al 2002) as well as by differences in the defini-tion of sampling area sizes and algorithms used foranalyses For PAE there is a trend to decrease theabsolute number of steps and to increase the numberof synapomorphies as the size of the area increases(Morrone and Escalante 2002) Likewise it is impor-tant to note that this approach depends on detailed pri-mary biodiversity occurrence data In studies like this

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 11

with large data numbers it is worth noting thatobserved patterns of species distributional data couldbe influenced by diverse sources of error (eg taxo-nomic uncertainties identification errors or errors onlocality information associated with species occur-rences) Thus it is important to promote the constantupdate of the taxonomic status of specimens depositedin collections as well as the procedures for cleaningand revising biodiversity data from heterogeneoussources (Navarro-Sigeurouenza et al 2003 Bloom et al2017)Although our results mostly apply to the geographi-

cal scale and taxa used for this study the NSDF avi-fauna distribution patterns we found are non-randomand congruent with those for other organisms(Becerra 2005 Porzecanski and Cracraft 2005 Pen-nington et al 2009 Rıos-Mu~noz and Navarro-Sigeurouenza 2012 Oswald and Steadman 2015 Bandaet al 2016) The general pattern identified hereimplies the need to continue to study these areas ofendemism which certainly could have a much longerhistory than the temporal framework of the Pleis-tocene Thus from a conservation perspective mosthistorical areas of endemism must be considered aspriorities and incorporated into efforts of conserva-tions planning (eg Nori et al 2016) because theyreflect an unique history of the Earth and its biota Afailure to protect them would result in major losses ofunique species diversity for this highly threatenedecosystem The observed distributional patterns of theavifauna as well as their relationships across theNSDFs constitute evidence of the great complexity ofthe biogeography of this ecosystem in which each par-ticular pattern should be considered a stepping stonealong a true arc recognized as the biogeography ofthe Neotropics

Acknowledgments

We appreciate the efforts of museums (seeAppendix S1) and their curators as well as field collec-tors and institutions that made our study possibleDAP-T extends his gratitude to Consejo Nacional deCiencia y Tecnologıa (CONACyT Mexico) for a doc-toral scholarship (297538) and the Smithsonian-MasonSchool of Conservation for a scholarship during thecourse ldquoSpatial Ecology Geospatial Analysis andRemote Sensing for Conservation (MCCS 0500)rdquo thatprovided the tools and skills necessary for the spatialanalysis in this study The Rufford Foundation (DAP-T 16017-1 DAP-T 20284-2) Idea Wild (DAP-T) Con-sejo de Desarrollo Cientıfico Humanıstico y Tec-nologico at Universidad del Zulia (CONDES projectsCC-0247-13 and CC-0497-16 DAP-T) and the CONA-CyT project 152060 (AGNS and ORRS) provided

financial and logistical support J C Catarı P JCampos and Z Perez-Duran (Bolivia) as well as J ADelgado (Peru) helped during fieldwork A GordilloC Mota-Vargas and I Perozo helped in georeferencingbird locality data T Escalante helped with the PAEThe manuscript was improved by comments from HM Ortega-Andrade C Mota-Vargas A Lira-NoriegaJ L Parra L A Gonzalez and F Lopez

Conflicts of interest

The authors declare no competing interests in thisstudy

References

Amorim DS 2001 Dos Amazonias In Llorente J Morrone JJ(Eds) Introduccion a la Biogeografıa en LatinoamericaConceptos Teorıas Metodos y Aplicaciones Las Prensas deCiencias Universidad Nacional Autonoma de Mexico MexicoDF pp 245ndash255

Anderson RP Lew D Peterson AT 2003 Evaluatingpredictive models of speciesrsquo distributions criteria for selectingoptimal models Ecol Modell 162 211ndash232

Banda K Delgado-Salinas A Dexter KG Linares-Palomino ROliveira-Filho A Prado D Pullan M Quintana C Riina RRodrıguez M Weintritt J Acevedo-Rodrıguez P Adarve JAlvarez E Aranguren BA Arteaga JC Aymard G Casta~noA Ceballos-Mago N Cogollo A Cuadros H Delgado FDevia W Due~nas H Fajardo L Fernandez A FernandezMA Franklin J Freid EH Galetti LA Gonto RGonzalez-M R Graveson R Helmer EH Idarraga ALopez R Marcano-Vega H Martınez OG Maturo HMMcDonald M McLaren K Melo O Mijares F Mogni VMolina D Moreno NdP Nassar JM Neves DM OakleyLJ Oatham M Olvera-Luna AR Pezzini FF DominguezOJR Rıos ME Rivera O Rodrıguez N Rojas A SeuroarkinenT Sanchez R Smith M Vargas C Villanueva B PenningtonRT 2016 Plant diversity patterns in Neotropical dry forests andtheir conservation implications Science 353 1383ndash1387

Barve N Barve V Jimenez-Valverde A Lira-Noriega AMaher SP Peterson AT Soberon J Villalobos F 2011The crucial role of the accessible area in ecological nichemodeling and species distribution modeling Ecol Modell 2221810ndash1819

Becerra JX 2005 Timing the origin and expansion of the Mexicantropical dry forest Proc Natl Acad Sci USA 102 10919ndash10923

Bloom TDS Flower A DeChaine EG 2017 Why georeferencingmatters introducing a practical protocol to prepare speciesoccurrence records for spatial analysis Ecol Evol 8 765ndash777

Brooks DR van Veller MGP 2003 Critique of parsimonyanalysis of endemicity as a method of historical biogeography JBiogeogr 30 819ndash825

Ceballos G 1995 Vertebrate diversity ecology and conservation inNeotropical dry forests In Bullock S Medina E Mooney H(Eds) Seasonally Dry Tropical Forests Cambridge UniversityPress Cambridge UK pp 195ndash220

Clements J Schulenberg T Iliff M Roberson D FredericksT Sullivan B Wood C 2015 The eBirdClements checklist ofbirds of the world v2015 URL httpwww birds cornell educlementschecklistdownloadIOC [Verified 22 April 2017]

Cortes ALA Rapini A Daniel TF 2015 The Tetrameriumlineage (Acanthaceae Justicieae) does not support the PleistoceneArc hypothesis for South American seasonally dry forests AmJ Bot 102 992ndash1007

12 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

north-eastern Gulf of Mexico is biogeographically dis-tinct from the southern and south-eastern part of theGulf coast (Liebherr 1994 Hernandez-Ba~nos et al1995 Marshall and Liebherr 2000 Rıos-Mu~noz andNavarro-Sigeurouenza 2012)On the other hand separation within the south-east-

ern Isthmus groups (light blues in Fig 2) supports thebiogeographical hypothesis of vicariant processes asso-ciated with the PolochicndashMotagua fault and the Nicar-agua Depression in the region (Morrone 2001Schuster et al 2003 Liede and Meve 2004) Here weidentified at least three distinct regions suggesting theexistence of bird assemblages with independent biogeo-graphical histories probably shaped by new immi-grants from dispersal events between South and NorthAmerica after the establishment of the Panama Isth-mus (Smith and Klicka 2010) For instance weobserved that ~73 of the NSDF avifauna is sharedbetween southern Central America and northern SouthAmerica (with ~72 of NSDF-restricted speciesshared between the NSDF from Panama and northernSouth America) confirming the existence of strongaffinities between the two regions (Linares-Palominoet al 2011 Rıos-Mu~noz and Navarro-Sigeurouenza 2012Banda et al 2016)In the Panama region differences in the affinities

between the cladograms for this region may be influ-enced by the inclusion (ie complete approach) or theexclusion (ie restricted approach) of species widelydistributed in the region and inhabiting other ecosys-tems surrounding NSDFs In this sense such differ-ences could be explained by different temporalpatterns of speciation documented for taxa associatedwith humid and dry habitats across the Isthmus ofPanama (Smith et al 2012) The Panama area hashad a history of expanding and contracting processesfor both the dry and the humid forests and even whenboth habitats consist of ldquoyoungrdquo and ldquooldrdquo speciesthe dry habitat species are geographically more distantfrom their nearest counterpart in Central America andnorthern South America compared with species fromthe humid habitats in these regions (Oswald andSteadman 2015 Oswald et al 2017) This phe-nomenon if generalized is probably an importantcontributor to biogeographical patterns in the south-ernmost section of the Mesoamerican region (Sanchez-Gonzalez et al 2008 Rıos-Mu~noz and Navarro-Sigeurouenza 2012 Smith et al 2012)

South American NSDF areas and the Pleistocene ArcHypothesis

Previous studies of distributional patterns within theSouth American biota have accepted the hypothesisthat areas of endemism have arisen by isolation ofmore widespread biotas This scenario is commonly

associated with paleogeographical changes alterationsin river systems Pleistocene climatic fluctuations or amixture of all these processes (Herzog and Kessler2002 Werneck et al 2012) However we find littlesupport for widespread Pleistocene NSDF formationthroughout South America because few species(1033 n = 100 species) are widespread and sharedby more than 75 of South American areas Thisresult is concordant with NSDF plant patterns recentlyreported by Linares-Palomino et al (2011) and Bandaet al (2016) as well as with the restricted climatic cor-ridors found by Werneck et al (2011) which suggestthat LGM climate was probably too dry and cold tosupport large tracts of NSDFs across the continentSome authors suggest that long-distance ancient dis-

persal events of unrelated taxonomic groups are lesslikely than vicariant associations (Werneck et al2011 de Melo et al 2016) Likewise there is substan-tial support that NSDF is a dispersal-limited biomegiven the phylogenetic niche conservatism and stronggeographically structured phylogenies for woody plantclades (Pennington et al 2009 Pennington and Lavin2016) Thus an allopatric-vicariance explanationmdashde-lineated by major geographical or ecological barriersbefore the Pleistocenemdashseems more plausible toexplain the evolution and relationships among all theseforest areas which is supported by results based onendemic taxa For instance the biogeography of Bro-togeris parakeets implies a dynamic history for SouthAmerican biomes since the Pliocene corroboratingthat geological evolution of Amazonia has been impor-tant in shaping its biodiversity with taxa endemic toNSDFs and other biomes (Ribas et al 2009) In thissense our results suggest that the wide distributions ofa relatively low proportion of NSDF species and eco-logical similarity among them (including specialistsand generalists) may reflect only limited recent long-distance dispersal events and species exchange (Dicket al 2003 2004 Mayle 2004)The separation of a northern South American clade

(Fig 2) suggests an effect of the Amazonia rain foreststhe wet Choco forest and the final uplift of the north-ern Andes as barriers for the dispersal of dry forestspecies (Amorim 2001 Garzione et al 2008 Hoornet al 2010 Morrone 2014b Banda et al 2016) Clus-ters of the northern inter-Andean valleys (located alongthe Magdalena and Cauca Rivers) together with Carib-bean coastal areas reflect the heterogeneity of dryAndean regions (with important species turnover withMesoamerica and CaribbeanndashAntilles) where few spe-cies are distributed all the way from the northern tothe southern Andes (Nores 2004 Vazquez-Mirandaet al 2007) These results are also consistent withthose previously reported for other taxonomic groupsin the region that are associated with NSDFs (Linares-Palomino et al 2011 Morrone 2014b)

10 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

Although our analyses generally agree with thePleistocene Arc Hypothesis proposed originally byPrado and Gibbs (1993)mdashformation of a once exten-sive and largely contiguous seasonal woodland com-prising the Caatingas the Misiones and Sub-AndeanPiedmont nucleimdashthe clear separation among NSDFsfrom centralndashsouth-western and south-eastern SouthAmerica supports the idea that both uplift of theAndes and historical (Cenozoic and Quaternary) cli-matic regimes of north-western EcuadorndashPeru andwestern Bolivia-Argentina are probably responsible forthe high endemism and distributions of their NSDFbird communities (Hoorn et al 2010 Morrone2014b) These communities have been biogeographi-cally separated from neighbouring Bolivian lowlandsites (Herzog and Kessler 2002 Herzog et al 2016)In fact previous Andean NSDF legume phylogeniesshowed consistently high geographical structurethroughout their distribution suggesting that at leastthe NSDF flora has been assembled gradually over thepast ca 19 Myr with processes of in situ diversifica-tion largely undisturbed by new immigrants (Seuroarkinenet al 2012) From this perspective the NSDF on theslopes of centralndashsouth-western South America consti-tutes an important biogeographical area with ecologi-cal and geographical long-term stability and probablyoriginated by convergence scenarios resulting fromancient dispersal events that involved low niche differ-entiation and strong dispersal limitation from otherNSDF patches (Pennington et al 2009 Seuroarkinenet al 2012) This also suggests that NSDFs fromEcuador and northern Peru to western BoliviandashArgen-tina might constitute remnants of a previously widerexpansion according to the Pleistocene NSDF forma-tion (Seuroarkinen et al 2012 Cortes et al 2015 de Meloet al 2016) Passerine fossils support this hypothesisand suggest that some NSDF species had wider distri-butions in the past (Oswald and Steadman 2015) withthe Andean valleys (eg the Mara~non and ApurimacValleys) having direct influence on divergence andextinction processes by acting as semi-permeable barri-ers (Seuroarkinen et al 2011 Smith et al 2014 Wingerand Bates 2015)Finally we found a stretch of NSDF from north-

eastern Brazil south to ParaguayndashArgentina and northtowards eastern Bolivia (Fig 2) which is concordantwith previous studies showing a well-supported group-ing among these forests and suggesting the existence ofa historical stable unit known as the Open Dry Diago-nal (Mayle 2004 Porzecanski and Cracraft 2005Linares-Palomino et al 2011 Werneck et al 20112012) Today many dry habitat sister taxa are geo-graphically distant from one another suggesting thatthese species have either undergone long-distance dis-persal across large tracts of humid forests or that aridregions were once connected (Oswald et al 2017) The

absence of synapomorphies within the Chiquitano andMisiones Province forests (Table 1 Appendix S4) pro-vides some evidence for the recent formation of birdassemblages in this area (Herzog and Kessler 2002Smith et al 2014) as was found for woody plants(Linares-Palomino et al 2011 Banda et al 2016) Infact several authors suggest climatic and floristic char-acteristics closer to Atlantic rain forests than to NSDFsfor these areas (Olson et al 2001 Oliveira-Filho et al2006) which were probably colonized by dry forestspecies during drier and cooler periods of the Pleis-tocene These colonization processes were probablyfavoured by a potential narrow eastndashwest corridor con-necting NSDF in the extremes of north-eastern Braziland south-western South America during the LGM(Werneck et al 2011) This connection could serve asa biogeographical link sheltering floristic and faunisticdispersal routes resulting in partial expansions ofNSDFs in some areas of Amazonia (Werneck et al2011 2012) As a consequence of this last idea thelong-distance dispersal hypothesis of NSDFs in south-eastern South America proposed by Mayle (2004)remains a scenario that still needs independent testing

Final considerations

Our study suggests that evolution of the well-differ-entiated avifaunas in the four major geographicalgroups herein identified might have involved a mixtureof vicariance and dispersal events which provides newinsights into the diversification centres and historicalrelationships for the distributional patterns of thisecosystem and its associated avifauna Despite somespecies being shared among different clades (regions)and that this probably increased after closure of thePanama Isthmus (Morrone 2014b) and during themaximum extension of forests at the LGM (Bandaet al 2016) we observed a clear differentiation amongthe main areas with high degrees of endemism(Fig 2) This agrees with the idea of independent evo-lutionary histories among NSDF nuclei across theirdistribution (Becerra 2005 Cortes et al 2015 deMelo et al 2016)Differences observed with previous studies (eg

Porzecanski and Cracraft 2005 Rıos-Mu~noz andNavarro-Sigeurouenza 2012) are largely due to taxonomicuncertainties (ie species concepts and new species thathave been described or recognized since those studiesJetz et al 2002) as well as by differences in the defini-tion of sampling area sizes and algorithms used foranalyses For PAE there is a trend to decrease theabsolute number of steps and to increase the numberof synapomorphies as the size of the area increases(Morrone and Escalante 2002) Likewise it is impor-tant to note that this approach depends on detailed pri-mary biodiversity occurrence data In studies like this

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 11

with large data numbers it is worth noting thatobserved patterns of species distributional data couldbe influenced by diverse sources of error (eg taxo-nomic uncertainties identification errors or errors onlocality information associated with species occur-rences) Thus it is important to promote the constantupdate of the taxonomic status of specimens depositedin collections as well as the procedures for cleaningand revising biodiversity data from heterogeneoussources (Navarro-Sigeurouenza et al 2003 Bloom et al2017)Although our results mostly apply to the geographi-

cal scale and taxa used for this study the NSDF avi-fauna distribution patterns we found are non-randomand congruent with those for other organisms(Becerra 2005 Porzecanski and Cracraft 2005 Pen-nington et al 2009 Rıos-Mu~noz and Navarro-Sigeurouenza 2012 Oswald and Steadman 2015 Bandaet al 2016) The general pattern identified hereimplies the need to continue to study these areas ofendemism which certainly could have a much longerhistory than the temporal framework of the Pleis-tocene Thus from a conservation perspective mosthistorical areas of endemism must be considered aspriorities and incorporated into efforts of conserva-tions planning (eg Nori et al 2016) because theyreflect an unique history of the Earth and its biota Afailure to protect them would result in major losses ofunique species diversity for this highly threatenedecosystem The observed distributional patterns of theavifauna as well as their relationships across theNSDFs constitute evidence of the great complexity ofthe biogeography of this ecosystem in which each par-ticular pattern should be considered a stepping stonealong a true arc recognized as the biogeography ofthe Neotropics

Acknowledgments

We appreciate the efforts of museums (seeAppendix S1) and their curators as well as field collec-tors and institutions that made our study possibleDAP-T extends his gratitude to Consejo Nacional deCiencia y Tecnologıa (CONACyT Mexico) for a doc-toral scholarship (297538) and the Smithsonian-MasonSchool of Conservation for a scholarship during thecourse ldquoSpatial Ecology Geospatial Analysis andRemote Sensing for Conservation (MCCS 0500)rdquo thatprovided the tools and skills necessary for the spatialanalysis in this study The Rufford Foundation (DAP-T 16017-1 DAP-T 20284-2) Idea Wild (DAP-T) Con-sejo de Desarrollo Cientıfico Humanıstico y Tec-nologico at Universidad del Zulia (CONDES projectsCC-0247-13 and CC-0497-16 DAP-T) and the CONA-CyT project 152060 (AGNS and ORRS) provided

financial and logistical support J C Catarı P JCampos and Z Perez-Duran (Bolivia) as well as J ADelgado (Peru) helped during fieldwork A GordilloC Mota-Vargas and I Perozo helped in georeferencingbird locality data T Escalante helped with the PAEThe manuscript was improved by comments from HM Ortega-Andrade C Mota-Vargas A Lira-NoriegaJ L Parra L A Gonzalez and F Lopez

Conflicts of interest

The authors declare no competing interests in thisstudy

References

Amorim DS 2001 Dos Amazonias In Llorente J Morrone JJ(Eds) Introduccion a la Biogeografıa en LatinoamericaConceptos Teorıas Metodos y Aplicaciones Las Prensas deCiencias Universidad Nacional Autonoma de Mexico MexicoDF pp 245ndash255

Anderson RP Lew D Peterson AT 2003 Evaluatingpredictive models of speciesrsquo distributions criteria for selectingoptimal models Ecol Modell 162 211ndash232

Banda K Delgado-Salinas A Dexter KG Linares-Palomino ROliveira-Filho A Prado D Pullan M Quintana C Riina RRodrıguez M Weintritt J Acevedo-Rodrıguez P Adarve JAlvarez E Aranguren BA Arteaga JC Aymard G Casta~noA Ceballos-Mago N Cogollo A Cuadros H Delgado FDevia W Due~nas H Fajardo L Fernandez A FernandezMA Franklin J Freid EH Galetti LA Gonto RGonzalez-M R Graveson R Helmer EH Idarraga ALopez R Marcano-Vega H Martınez OG Maturo HMMcDonald M McLaren K Melo O Mijares F Mogni VMolina D Moreno NdP Nassar JM Neves DM OakleyLJ Oatham M Olvera-Luna AR Pezzini FF DominguezOJR Rıos ME Rivera O Rodrıguez N Rojas A SeuroarkinenT Sanchez R Smith M Vargas C Villanueva B PenningtonRT 2016 Plant diversity patterns in Neotropical dry forests andtheir conservation implications Science 353 1383ndash1387

Barve N Barve V Jimenez-Valverde A Lira-Noriega AMaher SP Peterson AT Soberon J Villalobos F 2011The crucial role of the accessible area in ecological nichemodeling and species distribution modeling Ecol Modell 2221810ndash1819

Becerra JX 2005 Timing the origin and expansion of the Mexicantropical dry forest Proc Natl Acad Sci USA 102 10919ndash10923

Bloom TDS Flower A DeChaine EG 2017 Why georeferencingmatters introducing a practical protocol to prepare speciesoccurrence records for spatial analysis Ecol Evol 8 765ndash777

Brooks DR van Veller MGP 2003 Critique of parsimonyanalysis of endemicity as a method of historical biogeography JBiogeogr 30 819ndash825

Ceballos G 1995 Vertebrate diversity ecology and conservation inNeotropical dry forests In Bullock S Medina E Mooney H(Eds) Seasonally Dry Tropical Forests Cambridge UniversityPress Cambridge UK pp 195ndash220

Clements J Schulenberg T Iliff M Roberson D FredericksT Sullivan B Wood C 2015 The eBirdClements checklist ofbirds of the world v2015 URL httpwww birds cornell educlementschecklistdownloadIOC [Verified 22 April 2017]

Cortes ALA Rapini A Daniel TF 2015 The Tetrameriumlineage (Acanthaceae Justicieae) does not support the PleistoceneArc hypothesis for South American seasonally dry forests AmJ Bot 102 992ndash1007

12 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

Although our analyses generally agree with thePleistocene Arc Hypothesis proposed originally byPrado and Gibbs (1993)mdashformation of a once exten-sive and largely contiguous seasonal woodland com-prising the Caatingas the Misiones and Sub-AndeanPiedmont nucleimdashthe clear separation among NSDFsfrom centralndashsouth-western and south-eastern SouthAmerica supports the idea that both uplift of theAndes and historical (Cenozoic and Quaternary) cli-matic regimes of north-western EcuadorndashPeru andwestern Bolivia-Argentina are probably responsible forthe high endemism and distributions of their NSDFbird communities (Hoorn et al 2010 Morrone2014b) These communities have been biogeographi-cally separated from neighbouring Bolivian lowlandsites (Herzog and Kessler 2002 Herzog et al 2016)In fact previous Andean NSDF legume phylogeniesshowed consistently high geographical structurethroughout their distribution suggesting that at leastthe NSDF flora has been assembled gradually over thepast ca 19 Myr with processes of in situ diversifica-tion largely undisturbed by new immigrants (Seuroarkinenet al 2012) From this perspective the NSDF on theslopes of centralndashsouth-western South America consti-tutes an important biogeographical area with ecologi-cal and geographical long-term stability and probablyoriginated by convergence scenarios resulting fromancient dispersal events that involved low niche differ-entiation and strong dispersal limitation from otherNSDF patches (Pennington et al 2009 Seuroarkinenet al 2012) This also suggests that NSDFs fromEcuador and northern Peru to western BoliviandashArgen-tina might constitute remnants of a previously widerexpansion according to the Pleistocene NSDF forma-tion (Seuroarkinen et al 2012 Cortes et al 2015 de Meloet al 2016) Passerine fossils support this hypothesisand suggest that some NSDF species had wider distri-butions in the past (Oswald and Steadman 2015) withthe Andean valleys (eg the Mara~non and ApurimacValleys) having direct influence on divergence andextinction processes by acting as semi-permeable barri-ers (Seuroarkinen et al 2011 Smith et al 2014 Wingerand Bates 2015)Finally we found a stretch of NSDF from north-

eastern Brazil south to ParaguayndashArgentina and northtowards eastern Bolivia (Fig 2) which is concordantwith previous studies showing a well-supported group-ing among these forests and suggesting the existence ofa historical stable unit known as the Open Dry Diago-nal (Mayle 2004 Porzecanski and Cracraft 2005Linares-Palomino et al 2011 Werneck et al 20112012) Today many dry habitat sister taxa are geo-graphically distant from one another suggesting thatthese species have either undergone long-distance dis-persal across large tracts of humid forests or that aridregions were once connected (Oswald et al 2017) The

absence of synapomorphies within the Chiquitano andMisiones Province forests (Table 1 Appendix S4) pro-vides some evidence for the recent formation of birdassemblages in this area (Herzog and Kessler 2002Smith et al 2014) as was found for woody plants(Linares-Palomino et al 2011 Banda et al 2016) Infact several authors suggest climatic and floristic char-acteristics closer to Atlantic rain forests than to NSDFsfor these areas (Olson et al 2001 Oliveira-Filho et al2006) which were probably colonized by dry forestspecies during drier and cooler periods of the Pleis-tocene These colonization processes were probablyfavoured by a potential narrow eastndashwest corridor con-necting NSDF in the extremes of north-eastern Braziland south-western South America during the LGM(Werneck et al 2011) This connection could serve asa biogeographical link sheltering floristic and faunisticdispersal routes resulting in partial expansions ofNSDFs in some areas of Amazonia (Werneck et al2011 2012) As a consequence of this last idea thelong-distance dispersal hypothesis of NSDFs in south-eastern South America proposed by Mayle (2004)remains a scenario that still needs independent testing

Final considerations

Our study suggests that evolution of the well-differ-entiated avifaunas in the four major geographicalgroups herein identified might have involved a mixtureof vicariance and dispersal events which provides newinsights into the diversification centres and historicalrelationships for the distributional patterns of thisecosystem and its associated avifauna Despite somespecies being shared among different clades (regions)and that this probably increased after closure of thePanama Isthmus (Morrone 2014b) and during themaximum extension of forests at the LGM (Bandaet al 2016) we observed a clear differentiation amongthe main areas with high degrees of endemism(Fig 2) This agrees with the idea of independent evo-lutionary histories among NSDF nuclei across theirdistribution (Becerra 2005 Cortes et al 2015 deMelo et al 2016)Differences observed with previous studies (eg

Porzecanski and Cracraft 2005 Rıos-Mu~noz andNavarro-Sigeurouenza 2012) are largely due to taxonomicuncertainties (ie species concepts and new species thathave been described or recognized since those studiesJetz et al 2002) as well as by differences in the defini-tion of sampling area sizes and algorithms used foranalyses For PAE there is a trend to decrease theabsolute number of steps and to increase the numberof synapomorphies as the size of the area increases(Morrone and Escalante 2002) Likewise it is impor-tant to note that this approach depends on detailed pri-mary biodiversity occurrence data In studies like this

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 11

with large data numbers it is worth noting thatobserved patterns of species distributional data couldbe influenced by diverse sources of error (eg taxo-nomic uncertainties identification errors or errors onlocality information associated with species occur-rences) Thus it is important to promote the constantupdate of the taxonomic status of specimens depositedin collections as well as the procedures for cleaningand revising biodiversity data from heterogeneoussources (Navarro-Sigeurouenza et al 2003 Bloom et al2017)Although our results mostly apply to the geographi-

cal scale and taxa used for this study the NSDF avi-fauna distribution patterns we found are non-randomand congruent with those for other organisms(Becerra 2005 Porzecanski and Cracraft 2005 Pen-nington et al 2009 Rıos-Mu~noz and Navarro-Sigeurouenza 2012 Oswald and Steadman 2015 Bandaet al 2016) The general pattern identified hereimplies the need to continue to study these areas ofendemism which certainly could have a much longerhistory than the temporal framework of the Pleis-tocene Thus from a conservation perspective mosthistorical areas of endemism must be considered aspriorities and incorporated into efforts of conserva-tions planning (eg Nori et al 2016) because theyreflect an unique history of the Earth and its biota Afailure to protect them would result in major losses ofunique species diversity for this highly threatenedecosystem The observed distributional patterns of theavifauna as well as their relationships across theNSDFs constitute evidence of the great complexity ofthe biogeography of this ecosystem in which each par-ticular pattern should be considered a stepping stonealong a true arc recognized as the biogeography ofthe Neotropics

Acknowledgments

We appreciate the efforts of museums (seeAppendix S1) and their curators as well as field collec-tors and institutions that made our study possibleDAP-T extends his gratitude to Consejo Nacional deCiencia y Tecnologıa (CONACyT Mexico) for a doc-toral scholarship (297538) and the Smithsonian-MasonSchool of Conservation for a scholarship during thecourse ldquoSpatial Ecology Geospatial Analysis andRemote Sensing for Conservation (MCCS 0500)rdquo thatprovided the tools and skills necessary for the spatialanalysis in this study The Rufford Foundation (DAP-T 16017-1 DAP-T 20284-2) Idea Wild (DAP-T) Con-sejo de Desarrollo Cientıfico Humanıstico y Tec-nologico at Universidad del Zulia (CONDES projectsCC-0247-13 and CC-0497-16 DAP-T) and the CONA-CyT project 152060 (AGNS and ORRS) provided

financial and logistical support J C Catarı P JCampos and Z Perez-Duran (Bolivia) as well as J ADelgado (Peru) helped during fieldwork A GordilloC Mota-Vargas and I Perozo helped in georeferencingbird locality data T Escalante helped with the PAEThe manuscript was improved by comments from HM Ortega-Andrade C Mota-Vargas A Lira-NoriegaJ L Parra L A Gonzalez and F Lopez

Conflicts of interest

The authors declare no competing interests in thisstudy

References

Amorim DS 2001 Dos Amazonias In Llorente J Morrone JJ(Eds) Introduccion a la Biogeografıa en LatinoamericaConceptos Teorıas Metodos y Aplicaciones Las Prensas deCiencias Universidad Nacional Autonoma de Mexico MexicoDF pp 245ndash255

Anderson RP Lew D Peterson AT 2003 Evaluatingpredictive models of speciesrsquo distributions criteria for selectingoptimal models Ecol Modell 162 211ndash232

Banda K Delgado-Salinas A Dexter KG Linares-Palomino ROliveira-Filho A Prado D Pullan M Quintana C Riina RRodrıguez M Weintritt J Acevedo-Rodrıguez P Adarve JAlvarez E Aranguren BA Arteaga JC Aymard G Casta~noA Ceballos-Mago N Cogollo A Cuadros H Delgado FDevia W Due~nas H Fajardo L Fernandez A FernandezMA Franklin J Freid EH Galetti LA Gonto RGonzalez-M R Graveson R Helmer EH Idarraga ALopez R Marcano-Vega H Martınez OG Maturo HMMcDonald M McLaren K Melo O Mijares F Mogni VMolina D Moreno NdP Nassar JM Neves DM OakleyLJ Oatham M Olvera-Luna AR Pezzini FF DominguezOJR Rıos ME Rivera O Rodrıguez N Rojas A SeuroarkinenT Sanchez R Smith M Vargas C Villanueva B PenningtonRT 2016 Plant diversity patterns in Neotropical dry forests andtheir conservation implications Science 353 1383ndash1387

Barve N Barve V Jimenez-Valverde A Lira-Noriega AMaher SP Peterson AT Soberon J Villalobos F 2011The crucial role of the accessible area in ecological nichemodeling and species distribution modeling Ecol Modell 2221810ndash1819

Becerra JX 2005 Timing the origin and expansion of the Mexicantropical dry forest Proc Natl Acad Sci USA 102 10919ndash10923

Bloom TDS Flower A DeChaine EG 2017 Why georeferencingmatters introducing a practical protocol to prepare speciesoccurrence records for spatial analysis Ecol Evol 8 765ndash777

Brooks DR van Veller MGP 2003 Critique of parsimonyanalysis of endemicity as a method of historical biogeography JBiogeogr 30 819ndash825

Ceballos G 1995 Vertebrate diversity ecology and conservation inNeotropical dry forests In Bullock S Medina E Mooney H(Eds) Seasonally Dry Tropical Forests Cambridge UniversityPress Cambridge UK pp 195ndash220

Clements J Schulenberg T Iliff M Roberson D FredericksT Sullivan B Wood C 2015 The eBirdClements checklist ofbirds of the world v2015 URL httpwww birds cornell educlementschecklistdownloadIOC [Verified 22 April 2017]

Cortes ALA Rapini A Daniel TF 2015 The Tetrameriumlineage (Acanthaceae Justicieae) does not support the PleistoceneArc hypothesis for South American seasonally dry forests AmJ Bot 102 992ndash1007

12 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

with large data numbers it is worth noting thatobserved patterns of species distributional data couldbe influenced by diverse sources of error (eg taxo-nomic uncertainties identification errors or errors onlocality information associated with species occur-rences) Thus it is important to promote the constantupdate of the taxonomic status of specimens depositedin collections as well as the procedures for cleaningand revising biodiversity data from heterogeneoussources (Navarro-Sigeurouenza et al 2003 Bloom et al2017)Although our results mostly apply to the geographi-

cal scale and taxa used for this study the NSDF avi-fauna distribution patterns we found are non-randomand congruent with those for other organisms(Becerra 2005 Porzecanski and Cracraft 2005 Pen-nington et al 2009 Rıos-Mu~noz and Navarro-Sigeurouenza 2012 Oswald and Steadman 2015 Bandaet al 2016) The general pattern identified hereimplies the need to continue to study these areas ofendemism which certainly could have a much longerhistory than the temporal framework of the Pleis-tocene Thus from a conservation perspective mosthistorical areas of endemism must be considered aspriorities and incorporated into efforts of conserva-tions planning (eg Nori et al 2016) because theyreflect an unique history of the Earth and its biota Afailure to protect them would result in major losses ofunique species diversity for this highly threatenedecosystem The observed distributional patterns of theavifauna as well as their relationships across theNSDFs constitute evidence of the great complexity ofthe biogeography of this ecosystem in which each par-ticular pattern should be considered a stepping stonealong a true arc recognized as the biogeography ofthe Neotropics

Acknowledgments

We appreciate the efforts of museums (seeAppendix S1) and their curators as well as field collec-tors and institutions that made our study possibleDAP-T extends his gratitude to Consejo Nacional deCiencia y Tecnologıa (CONACyT Mexico) for a doc-toral scholarship (297538) and the Smithsonian-MasonSchool of Conservation for a scholarship during thecourse ldquoSpatial Ecology Geospatial Analysis andRemote Sensing for Conservation (MCCS 0500)rdquo thatprovided the tools and skills necessary for the spatialanalysis in this study The Rufford Foundation (DAP-T 16017-1 DAP-T 20284-2) Idea Wild (DAP-T) Con-sejo de Desarrollo Cientıfico Humanıstico y Tec-nologico at Universidad del Zulia (CONDES projectsCC-0247-13 and CC-0497-16 DAP-T) and the CONA-CyT project 152060 (AGNS and ORRS) provided

financial and logistical support J C Catarı P JCampos and Z Perez-Duran (Bolivia) as well as J ADelgado (Peru) helped during fieldwork A GordilloC Mota-Vargas and I Perozo helped in georeferencingbird locality data T Escalante helped with the PAEThe manuscript was improved by comments from HM Ortega-Andrade C Mota-Vargas A Lira-NoriegaJ L Parra L A Gonzalez and F Lopez

Conflicts of interest

The authors declare no competing interests in thisstudy

References

Amorim DS 2001 Dos Amazonias In Llorente J Morrone JJ(Eds) Introduccion a la Biogeografıa en LatinoamericaConceptos Teorıas Metodos y Aplicaciones Las Prensas deCiencias Universidad Nacional Autonoma de Mexico MexicoDF pp 245ndash255

Anderson RP Lew D Peterson AT 2003 Evaluatingpredictive models of speciesrsquo distributions criteria for selectingoptimal models Ecol Modell 162 211ndash232

Banda K Delgado-Salinas A Dexter KG Linares-Palomino ROliveira-Filho A Prado D Pullan M Quintana C Riina RRodrıguez M Weintritt J Acevedo-Rodrıguez P Adarve JAlvarez E Aranguren BA Arteaga JC Aymard G Casta~noA Ceballos-Mago N Cogollo A Cuadros H Delgado FDevia W Due~nas H Fajardo L Fernandez A FernandezMA Franklin J Freid EH Galetti LA Gonto RGonzalez-M R Graveson R Helmer EH Idarraga ALopez R Marcano-Vega H Martınez OG Maturo HMMcDonald M McLaren K Melo O Mijares F Mogni VMolina D Moreno NdP Nassar JM Neves DM OakleyLJ Oatham M Olvera-Luna AR Pezzini FF DominguezOJR Rıos ME Rivera O Rodrıguez N Rojas A SeuroarkinenT Sanchez R Smith M Vargas C Villanueva B PenningtonRT 2016 Plant diversity patterns in Neotropical dry forests andtheir conservation implications Science 353 1383ndash1387

Barve N Barve V Jimenez-Valverde A Lira-Noriega AMaher SP Peterson AT Soberon J Villalobos F 2011The crucial role of the accessible area in ecological nichemodeling and species distribution modeling Ecol Modell 2221810ndash1819

Becerra JX 2005 Timing the origin and expansion of the Mexicantropical dry forest Proc Natl Acad Sci USA 102 10919ndash10923

Bloom TDS Flower A DeChaine EG 2017 Why georeferencingmatters introducing a practical protocol to prepare speciesoccurrence records for spatial analysis Ecol Evol 8 765ndash777

Brooks DR van Veller MGP 2003 Critique of parsimonyanalysis of endemicity as a method of historical biogeography JBiogeogr 30 819ndash825

Ceballos G 1995 Vertebrate diversity ecology and conservation inNeotropical dry forests In Bullock S Medina E Mooney H(Eds) Seasonally Dry Tropical Forests Cambridge UniversityPress Cambridge UK pp 195ndash220

Clements J Schulenberg T Iliff M Roberson D FredericksT Sullivan B Wood C 2015 The eBirdClements checklist ofbirds of the world v2015 URL httpwww birds cornell educlementschecklistdownloadIOC [Verified 22 April 2017]

Cortes ALA Rapini A Daniel TF 2015 The Tetrameriumlineage (Acanthaceae Justicieae) does not support the PleistoceneArc hypothesis for South American seasonally dry forests AmJ Bot 102 992ndash1007

12 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

Cracraft J 1985 Historical biogeography and patterns ofdifferentiation within the South American avifauna areas ofendemism Ornithol Monogr 36 49ndash84

Cracraft J 1991 Patterns of diversification within continentalbiotas hierarchical congruence among the areas of endemism ofAustralian vertebrates Aust Syst Bot 4 211ndash227

Crisp MD Laffan S Linder HP Monro A 2001 Endemismin the Australian Flora J Biogeogr 28 183ndash198

Crother BI Guyer C 1996 Caribbean historical biogeographywas the dispersal-vicariance debate eliminated by anextraterrestrial bolide Herpetologica 52 440ndash465

Dick CW Abdul-Salim K Bermingham E 2003 Molecularsystematic analysis reveals cryptic Tertiary diversification of awidespread tropical rain forest tree Am Nat 162 691ndash703

Dick CW Roubik DW Gruber KF Bermingham E 2004Long-distance gene flow and cross-Andean dispersal of lowlandrainforest bees (Apidae Euglossini) revealed by comparativemitochondrial DNA phylogeography Mol Ecol 13 3775ndash3785

Dormann CF Elith J Bacher S Buchmann C Carl GCarre G Marquez JRG Gruber B Lafourcade B Leit~aoPJ 2013 Collinearity a review of methods to deal with it and asimulation study evaluating their performance Ecography 3627ndash46

Elith J Leathwick J 2007 Predicting species distributions frommuseum and herbarium records using multiresponse models fittedwith multivariate adaptive regression splines Divers Distrib 13265ndash275

Elith J Graham CH Anderson RP Dudık M Ferrier SGuisan A Hijmans RJ Huettmann F Leathwick JRLehmann A Li J Lohmann L Loiselle BA Manion GMoritz C Nakamura M Nakazawa Y Overton JMMPeterson AT Phillips SJ Richardson K Scachetti-Pereira RSchapire RE Soberon J Williams S Wisz MSZimmermann NE 2006 Novel methods improve prediction ofspeciesrsquo distributions from occurrence data Ecography 29 129ndash151

Elith J Phillips SJ Hastie T Dudık M Chee YE YatesCJ 2011 A statistical explanation of MaxEnt for ecologistsDivers Distrib 17 43ndash57

Escalante T 2015 Parsimony analysis of endemicity and analysisof endemicity a fair comparison Syst Biodivers 13 413ndash418

Escalante T Rodrıguez-Tapia G Linaje M Illoldi-Rangel PGonzalez-Lopez R 2013 Identification of areas of endemismfrom species distribution models threshold selection andNearctic mammals TIP 16 5ndash17

ESRI 2010 ArcMap 100 Environmental System ResearchInstitute Inc California USA

Garzione CN Hoke GD Libarkin JC Withers SMacFadden B Eiler J Ghosh P Mulch A 2008 Rise ofthe Andes Science 320 1304ndash1307

Gill F Donsker D 2015 IOC World Bird List (v 53) URLhttpwwwworldbirdnamesorgDOI-5master_ioc_list_v53xls[Verified 22 April 2017]

Goloboff PA 1993 Estimating character weights during treesearch Cladistics 91 83ndash91

Goloboff PA Catalano SA 2016 TNT version 15 including afull implementation of phylogenetic morphometrics Cladistics32 221ndash238

Gordon CE Ornelas JF 2000 Comparing endemism andhabitat restriction in Mesoamerican tropical deciduous forestbirds implications for biodiversity conservation planning BirdConserv Int 10 289ndash303

Graham A 2003a Geohistory models and Cenozoicpaleoenvironments of the Caribbean region Syst Bot 28 378ndash386

Graham MH 2003b Confronting multicollinearity in ecologicalmultiple regression Ecology 84 2809ndash2815

Graham A Dilcher D 1995 The Cenozoic record of tropical dryforest in northern Latin America and the southern United StatesIn Bullock S Medina E Mooney H (Eds) Seasonally DryTropical Forests Cambridge University Press Cambridge UKpp 124ndash145

Grismer LL 2000 Evolutionary biogeography on Mexicorsquos BajaCalifornia peninsula a synthesis of molecules and historicalgeology Proc Natl Acad Sci USA 97 14017ndash14018

Gutierrez-Velazquez A Rojas-Soto O Reyes-Castillo P HalffterG 2013 The classic theory of Mexican transition zone undermodern biogeographic approaches the distributional congruencepatterns of Passalidae (Coleoptera) Invertebr Syst 27 282ndash293

Heikkinen RK Luoto M Araujo MB Virkkala R ThuillerW Sykes MT 2006 Methods and uncertainties in bioclimaticenvelope modelling under climate change Prog Phys Geogr 30751ndash777

Hernandez-Ba~nos BE Peterson AT Navarro-Sigeurouenza AGEscalante-Pliego BP 1995 Bird faunas of the humid montaneforests of Mesoamerica biogeographic patterns and priorities forconservation Bird Conserv Int 5 251ndash277

Herzog SK Kessler M 2002 Biogeography and composition ofdry forest bird communities in Bolivia J Ornithol 143 171ndash204

Herzog S Terrill R Jahn A Remsen J Maillard O Garcıa-elSolıs V MacLeod R Maccormick A Vidoz J 2016 Birds ofBolivia Field Guide ARMONIA Santa Cruz de la Sierra Bolivia

Hijmans RJ Cameron SE Parra JL Jones PG Jarvis A2005 Very high resolution interpolated climate surfaces forglobal land areas Int J Climatol 25 1965ndash1978

Hoorn C Wesselingh F Ter Steege H Bermudez M Mora ASevink J Sanmartın I Sanchez-Meseguer A Anderson CFigueiredo J 2010 Amazonia through time Andean upliftclimate change landscape evolution and biodiversity Science330 927ndash931

Iturralde-Vinent MA MacPhee RD 1999 Paleogeography ofthe Caribbean region implications for Cenozoic biogeographyBull AMNH 238 1ndash95

Jetz W Thomas G Joy J Hartmann K Mooers A 2002 Theglobal diversity of birds in space and time Nature 491 444ndash448

Liebherr JK 1994 Biogeographic patterns of montane Mexicanand Central American Carabidae (Coleoptera) Can Entomol126 841ndash860

Liede S Meve U 2004 Revision of Metastelma (Apocynaceae-Asclepiadoideae) in southwestern North America and CentralAmerica Ann Mo Bot Gard 91 31ndash86

Linares-Palomino R Olivaeira-Filho AT Pennington RT 2011Neotropical seasonally dry forests diversity endemism andbiogeography of woody plants In Dirzo R Young HSMooney HA Ceballos G (Eds) Seasonally Dry TropicalForests Ecology and Conservation Island Press WashingtonDC USA pp 1ndash23

Linder HP 2001 Plant diversity and endemism in sub-Saharantropical Africa J Biogeogr 28 169ndash182

Liu C White M Newell G 2013 Selecting thresholds for theprediction of species occurrence with presence-only data JBiogeogr 40 778ndash789

Marshall C Liebherr J 2000 Cladistic biogeography of theMexican transition zone J Biogeogr 27 203ndash216

Mayle FE 2004 Assessment of the Neotropical dry forest refugiahypothesis in the light of palaeoecological data and vegetationmodel simulations J Quat Sci 19 713ndash720

de Melo WA Lima-Ribeiro MS Terribile LC CollevattiRG 2016 Coalescent simulation and paleodistributionmodeling for Tabebuia rosealba do not support South Americandry forest refugia hypothesis PLoS ONE 11 e0159314

Morrone JJ 2001 Biogeografıa de America Latina y el CaribeVolumen 3 Manuales y tesis de la Sociedad EntomologicaAragonesa Zaragoza Spain

Morrone JJ 2005 Hacia una sıntesis biogeografica de MexicoRev Mex Biodivers 76 207ndash252

Morrone JJ 2014a Biogeographical regionalisation of theNeotropical region Zootaxa 3782 1ndash110

Morrone JJ 2014b Cladistic biogeography of the Neotropicalregion identifying the main events in the diversification of theterrestrial biota Cladistics 30 202ndash214

Morrone JJ 2014c Parsimony analysis of endemicity (PAE)revisited J Biogeogr 41 842ndash854

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 13

Morrone JJ Escalante T 2002 Parsimony analysis of endemicity(PAE) of Mexican terrestrial mammals at different area unitswhen size matters J Biogeogr 29 1095ndash1104

Murphy PG Lugo AE 1986 Ecology of tropical dry forestAnnu Rev Ecol Syst 17 67ndash88

Navarro-Sigeurouenza AG Peterson AT Gordillo-Martınez A2002 A Mexican case study on a centralised database fromworld natural history museums Data Sci J 1 45ndash53

Navarro-Sigeurouenza AG Peterson AT Gordillo-Martınez A2003 Museums working together the atlas of the birds ofMexico Bull Br Orn Club Suppl 123A 207ndash225

Navarro-Sigeurouenza AG Vazquez-Miranda H Hernandez-AlonsoG Garcıa-Trejo EA Sanchez-Gonzalez LA 2017 Complexbiogeographic scenarios revealed in the diversification of thelargest woodpecker radiation in the New World MolPhylogenet Evol 112 53ndash67

Nihei SS 2006 Misconceptions about parsimony analysis ofendemicity J Biogeogr 33 2099ndash2106

Nixon K 1999 WINCLADA para IBM PC Ver 09 99UNAM21 Version Beta URL httpwwwdiversityoflifeorgwinclada [Verified 01 March 2018]

Nores M 2004 The implications of Tertiary and Quaternary sealevel rise events for avian distribution patterns in the lowlands ofnorthern South America Glob Ecol Biogeogr 13 149ndash161

Nori J Torres R Lescano JN Cordier JM Periago MEBaldo D 2016 Protected areas and spatial conservationpriorities for endemic vertebrates of the Gran Chaco one of themost threatened ecoregions of the world Divers Distrib 221212ndash1219

OrsquoConnell KA Streicher JW Smith EN Fujita MK 2017Geographical features are the predominant driver of moleculardiversification in widely distributed North American whipsnakesMol Ecol 26 5729ndash5751

Oliveira-Filho AT Jarenkow JA Rodal MN 2006 Floristicrelationships of seasonally dry forests of eastern South Americabased on tree species distribution patterns SystematicsAssociation Special Volume 69 159

Olson DM Dinerstein E Wikramanayake ED Burgess NDPowell GV Underwood EC Drsquoamico JA Itoua I StrandHE Morrison JC 2001 Terrestrial ecoregions of the world anew map of life on earth a new global map of terrestrialecoregions provides an innovative tool for conservingbiodiversity Bioscience 51 933ndash938

Oswald JA Steadman DW 2015 The changing diversity anddistribution of dry forests passerine birds in northwestern sincethe last ice age Auk 132 836ndash862

Oswald JA Overcast I Mauck WM Andersen MJ SmithBT 2017 Isolation with asymmetric gene flow during thenonsynchronous divergence of dry forest birds Mol Ecol 261386ndash1400

Owens HL Campbell LP Dornak LL Saupe EE Barve NSoberon J Ingenloff K Lira-Noriega A Hensz CMMyers CE 2013 Constraints on interpretation of ecologicalniche models by limited environmental ranges on calibrationareas Ecol Modell 263 10ndash18

Pearson RG Raxworthy CJ Nakamura M Peterson AT2007 Predicting species distributions from small numbers ofoccurrence records a test case using cryptic geckos inMadagascar J Biogeogr 34 102ndash117

Pennington RT Lavin M 2016 The contrasting nature of woodyplant species in different Neotropical forest biomes reflectsdifferences in ecological stability New Phytol 210 25ndash37

Pennington RT Prado DE Pendry CA 2000 Neotropicalseasonally dry forests and quaternary vegetation changes JBiogeogr 27 261ndash273

Pennington RT Ratter JA Lewis GP 2006 An overview ofthe plant diversity biogeography and conservation ofNeotropical savannas and seasonally dry forests In PenningtonRT Ratter JA Lewis GP (Eds) Neotropical Savannas andSeasonally Dry Forests Plant Biodiversity Biogeography andConservation CRC Press Boca Raton FL USA pp 1ndash29

Pennington RT Lavin M Oliveira-Filho A 2009 Woody plantdiversity evolution and ecology in the tropics perspectives fromseasonally dry tropical forests Annu Rev Ecol Evol Syst 40437ndash457

Peterson AT 2001 Predicting speciesrsquo geographic distributionsbased on ecological niche modeling Condor 103 599ndash605

Peterson AT 2008 Parsimony analysis of endemism (PAE) andstudies of Mexican biogeography Rev Mex Biodivers 79 541ndash542

Peterson AT Papes M Soberon J 2008 Rethinking receiveroperating characteristic analysis applications in ecological nichemodeling Ecol Modell 213 63ndash72

Peterson AT Soberon J Pearson RG Anderson RPMartınez-Meyer E Nakamura M Araujo MB 2011Ecological Niches and Geographic Distributions PrincetonUniversity Press Princeton NJ USA

Phillips SJ Anderson RP Schapire RE 2006 Maximumentropy modeling of species geographic distributions EcolModell 190 231ndash259

Platnick N 1991 On areas of endemism Aust Syst Bot 4 11ndash12Porzecanski AL Cracraft J 2005 Cladistic analysis of

distributions and endemism (CADE) using raw distributions ofbirds to unravel the biogeography of the South Americanaridlands J Biogeogr 32 261ndash275

Prado DE Gibbs PE 1993 Patterns of species distributions inthe dry seasonal forests of South America Ann Mo Bot Gard80 902ndash927

Remsen J Jr Cadena C Jaramillo A Nores M Pacheco JPerez-Eman J Robbins M Stiles F Stotz D Zimmer K2017 A classification of the bird species of South AmericaSouth American Classification Committee AmericanOrnithologistsrsquo Union Available at httpwwwmuseumlsuedu~RemsenSACCBaselinehtm [verified 22 April 2017]

Ribas CC Miyaki CY Cracraft J 2009 Phylogeneticrelationships diversification and biogeography in NeotropicalBrotogeris parakeets J Biogeogr 36 1712ndash1729

Ricklefs RE 1987 Community diversity relative roles of andregional processes testing predictions of local-process theoriesScience 235 167ndash171

Rıos-Mu~noz CA Navarro-Sigeurouenza AG 2012 Patterns ofspecies richness and biogeographic regionalization of theavifaunas of the seasonally dry tropical forest in MesoamericaStud Neotrop Fauna Environ 47 171ndash182

Rodrıguez-Ferraro A Blake JG 2008 Diversity patterns of birdassemblages in arid zones of Northern Venezuela Condor 110405ndash420

Rodrıguez-Gomez F Gutierrez-Rodrıguez C Ornelas JFCarine M 2013 Genetic phenotypic and ecological divergencewith gene flow at the Isthmus of Tehuantepec the case of theazure-crowned hummingbird (Amazilia cyanocephala) JBiogeogr 40 1360ndash1373

Rojas-Soto OR Alcantara-Ayala O Navarro-Sigeurouenza AG2003 Regionalization of the avifauna of the Baja CaliforniaPeninsula Mexico a parsimony analysis of endemicity anddistributional modeling approach J Biogeogr 30 449ndash461

Rosen DE 1985 Geological hierarchies and biogeographiccongruence in the Caribbean Ann Mo Bot Gard 72 636ndash659

Rosen BR Smith AB 1988 Tectonics from fossils Analysis ofreef-coral and sea-urchin distributions from late Cretaceous toRecent using a new method J Geol Soc Lond 37 275ndash306

Sanchez-Azofeifa GA Quesada M Rodrıguez JP Nassar JMStoner KE Castillo A Garvin T Zent EL Calvo-Alvarado JC Kalacska MER Fajardo L Gamon JACuevas-Reyes P 2005 Research priorities for Neotropical dryforests Biotropica 37 477ndash485

Sanchez-Azofeifa GA Powers JS Fernandes GW Quesada M2013 Tropical Dry Forests in the Americas Ecology Conservationand Management CRC Press Boca Raton FL USA

Sanchez-Gonzalez LA Navarro-Sigeurouenza AG 2009 Historymeets ecology a geographical analysis of ecological restriction inthe Neotropical humid montane forests avifaunas DiversDistrib 15 1ndash11

14 David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15

Sanchez-Gonzalez LA Morrone JJ Navarro-Sigeurouenza AG2008 Distributional patterns of the Neotropical humid montaneforest avifaunas Biol J Linn Soc Lond 94 175ndash194

Seuroarkinen TE Marcelo-Pe~na JL Yomona AD Simon MFPennington RT Hughes CE 2011 Underestimated endemicspecies diversity in the dry inter-Andean valley of the RıoMara~non northern Peru an example from Mimosa(Leguminosae Mimosoideae) Taxon 60 139ndash150

Seuroarkinen T Pennington RT Lavin M Simon MF HughesCE 2012 Evolutionary islands in the Andes persistence andisolation explain high endemism in Andean dry tropical forestsJ Biogeogr 39 884ndash900

Schulenberg TS Stotz DF Lane DF OrsquoNeill JP ParkerTA III 2010 Birds of Peru Revised and Updated EditionPrinceton University Press Princeton NJ USA

Schuster JC Cano EB Reyes-Castillo P 2003 Proculus giantLatin-American passalids revision phylogeny and biogeographyActa Zool Mex 90 281ndash306

Smith BT Klicka J 2010 The profound influence of the LatePliocene Panamanian uplift on the exchange diversification anddistribution of New World birds Ecography 33 333ndash342

Smith BT Amei A Klicka J 2012 Evaluating the role ofcontracting and expanding rainforest in initiating cycles ofspeciation across the Isthmus of Panama Proc R Soc LondSer B Biol Sci 279 3520ndash3526

Smith BT McCormack JE Cuervo AM Hickerson MJAleixo A Cadena CD Perez-Eman J Burney CW XieX Harvey MG Faircloth BC Glenn TC DerryberryEP Prejean J Fields S Brumfield RT 2014 The drivers oftropical speciation Nature 515 406ndash409

Soberon J Peterson AT 2005 Interpretation of models offundamental ecological niches and speciesrsquo distributional areasBiodivers Inf 2 1ndash10

Stotz DF Fitzpatrick JW Parker TAIII Moskovits DKSnow D 1996 Neotropical Birds Ecology and ConservationUniversity of Chicago Press Chicago IL USA

Vazquez-Miranda H Navarro-Sigeurouenza AG Morrone JJ 2007Biogeographical patterns of the avifaunas of the Caribbean BasinIslands a parsimony perspective Cladistics 23 180ndash200

Weir JT Hey J 2006 Divergent timing and patterns of speciesaccumulation in lowland and highland Neotropical birdsEvolution 60 842ndash855

Werneck FP Costa GC Colli GR Prado DE Sites JW Jr2011 Revisiting the historical distribution of Seasonally DryTropical Forests new insights based on palaeodistribution modellingand palynological evidence Glob Ecol Biogeogr 20 272ndash288

Werneck FP Gamble T Colli GR Rodrigues MT Sites JW2012 Deep diversification and long-term persistence in the SouthAmerican lsquodry diagonalrsquo integrating continent-wide phylogeographyand distribution modeling of geckos Evolution 66 3014ndash3034

Wiens JJ Donoghue MJ 2004 Historical biogeography ecologyand species richness Trends Ecol Evol 19 639ndash644

Winger BM Bates JM 2015 The tempo of trait divergence ingeographic isolation avian speciation across the Mara~non Valleyof Peru Evolution 69 772ndash787

Zink RM Blackwell-Rago RC Ronquist F 2000 The shiftingroles of dispersal and vicariance in biogeography Proc R SocLond Ser B Biol Sci 267 497ndash503

Supporting Information

Additional supporting information may be foundonline in the Supporting Information section at theend of the articleAppendix S1 Ornithological collections of interna-

tional museums that kindly provided data analysed inthis study

Appendix S2 List of the 1298 bird species ofNeotropical seasonally dry forests considered in theparsimony analysis of endemicityAppendix S3 Matrix of all species and cells used in

this studyAppendix S4 Cladogram and geographical corre-

spondence of the groupings of grid cells obtainedfrom the potential distributional data in the parsi-mony analysis of endemicity considering the com-plete (a) vs Neotropical seasonally dry forest(NDSF)-restricted (b) species matrices from ldquospeciesinformationrdquo analyses Letters in cladograms showthe groups supported by synapomorphies included inTable 1 Well-supported clades [more than twosynapomorphies (species)] are shown with a blackcircle Black square in the cladogram corresponds tothe outgroup Acronyms in maps correspond toidentified NDFS regions and areas of endemismBC = Baja California Cu = Cuba island HP = HaitiDominican Republic and Puerto Rico islandsSS = Sonora and Sinaloa forests PS = Mexican Paci-fic slope lowlands SP = northern Oaxaca in southernMexico TV = TamaulipasndashVeracruz YP = YucatanPeninsula CAmN = northern Central AmericaCAmS = Southern Central America CAm = CentralAmerica Pa = Panama IAV = Inter-Andean Valleysof Colombia CCV = Caribbean ColombiandashVene-zuela PE = Pacific Equatorial Ap-M = Apurimac-Mantaro SA-P = Sub-Andean Piedmont C-MP =Chiquitano forests and Misiones ProvinceCaB = Brazilian CaatingaAppendix S5 Cladogram and geographical correspon-

dence of the groupings of grid cells obtained from thepotential distributional data in the parsimony analysis ofendemicity considering the complete (a) vs Neotropicalseasonally dry forest (NDSF)-restricted (b) species matri-ces from ldquospecies-genera informationrdquo analyses Letters incladograms show the groups supported by synapomor-phies Well-supported clades [more than two synapomor-phies (species)] are shown with a black circle Blacksquare in the cladogram corresponds to the outgroupAcronyms in maps correspond to identified NDFSregions and areas of endemism BC = Baja CaliforniaCu = Cuba island HP = Haiti Dominican Republic andPuerto Rico islands SS = Sonora and Sinaloa forestsPS = Mexican Pacific slope lowlands SP = northernOaxaca in southern Mexico TV = TamaulipasndashVeracruzYP = Yucatan Peninsula CAmN = northern CentralAmerica CAmS = southern Central AmericaCAm = Central America Pa = Panama IAV = Inter-Andean Valleys of Colombia CCV = CaribbeanColombiandashVenezuela PE = Pacific Equatorial Ap-M = Apurimac-Mantaro SA-P = Sub-Andean Pied-mont C-MP = Chiquitano forests and MisionesProvince CaB = Brazilian Caatinga

David A Prieto-Torres et al Cladistics 0 (2018) 1ndash15 15