<<

Universidade Federal de Goiás Instituto de Ciências Biológicas Programa de Pós-graduação em Ecologia e Evolução

Uma Perspectiva Macroecológica sobre o Risco de Extinção em Mamíferos

VINÍCIUS SILVA REIS

Goiânia 2019

VINÍCIUS SILVA REIS

Uma Perspectiva Macroecológica sobre o Risco de Extinção em Mamíferos

Tese apresentada ao Programa de Pós-graduação em Ecologia e Evolução do Departamento de Ecologia do Instituto de Ciências Biológicas da Universidade Federal de Goiás como requisito parcial para a obtenção do título de Doutor em Ecologia e Evolução. Orientador: Profº Drº Matheus de Souza Lima- Ribeiro Co-orientadora: Profª Drª Levi Carina Terribile

Goiânia 2019

DEDI CATÓRIA

Ao meu pai Wilson e à minha mãe Iris por sempre acreditarem em mim.

“Esper o que próxima vez que eu te veja, você s eja um novo homem com uma vasta gama de novas experiências e aventuras. Não he site, nem se permita dar desculpas. Apenas vá e faça. Vá e faça. Você ficará muito, muito feliz por ter feito”.

Trecho da carta escrita por Christopher McCandless a Ron Franz contida em Into the Wild de Jon Krakauer (Livre tradução) .

AGRADECIMENTOS

Eis que a aventura do doutoramento esteve bem longe de ser um caminho solitário. Não poderia ter sido um caminho tão feliz se eu não tivesse encontrado pessoas que me ensinaram desde método científico até como se bebe cerveja de verdade. São aos que estiveram comigo desde sempre, aos que permaneceram comigo e às novas amizades que eu construí quando me mudei pra Goiás que quero agradecer por terem me apoiado no nascimento desta tese:

À minha família, em especial meu pai Wilson, minha mãe Iris e minha irmã Flora, por me apoiarem e me incentivarem em cada conquista diária.

Aos meus orientadores Matheus e Carina, por terem me recebido desde 2013 no Laboratório de Macroecologia da Universidade Federal de Jataí, pela paciência e dedicação em me orientar ao longo desses anos.

Aos queridos amigos que fiz, em especial às mulheres-cientistas que abrilhantaram meus dias na capital goiana: Dani, Lara, Flávia, Bárbbara, Olívia, Raísa, Naty e aos colegas do Lets – Laboratório de Ecologia Teórica de Síntese pelos momentos de descontração, debate e cooperação, em especial à Leila, Flávio, Danilo e Lucas. Aos queridos Sara Villén e Diego Llusias pelo carinho no tempo em que partilhamos o mesmo teto da Casa-Horta.

Aos amigos que fiz em Jataí, em especial minha família adotiva Jataiense: Tia Marisa, Tracy e Táric por terem me acolhido como novo morador na república na qual passamos tantos momentos engraçados e felizes.

A Justin Travis, Greta Bocedi e Stephen Palmer por terem me recebido por seis meses no grupo de pesquisa deles no Zoology Building da Universidade de Aberdeen – Escócia. A todos os colegas da Sala 422 do Zoology Building por terem feito a experiência do intercâmbio acadêmico mais rica e interessante a cada dia. Aos amigos brasileiros que ganhei na Escócia, em especial Priscila, Elias, Ramon, Mariana, Juliana e Paulo.

A Sidney Gouveia, da Universidade Federal de Sergipe por ter me acolhido em seu laboratório para que eu pudesse finalizar esta tese e a Davi Crescente pela ajuda e disponibilidade para tirar minhas dúvidas.

O presente trabalho foi realizado com apoio da Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Código de Financiamento 001, mediante a concessão de um bolsa de doutorado e uma bolsa de doutorado sanduíche edital nº 19/2016, processo nº 88881.135886/2016-01.

SUMÁRIO

Resumo ...... 12 Abstract ...... 13 Introdução Geral ...... 14 Literatura citada ...... 16 Capítulo 1 - Historical, Ecological and Geographical Correlates of Risk ………………………………………………………………… 17 Introduction ...... 18 Material and Methods ...... 20 Results ...... 27 Discussion ...... 28 Literature Cited ...... 30 Supplementary Information ...... 42 Capítulo 2 - Landmasses prevent habitat tracking and challenge terrestrial with extinction risk …………..………………………………..…… 52 Introduction ...... 53 Material and Methods ...... 55 Results ...... 62 Discussion ...... 65 Literature Cited ...... 66 Supplementary Information ...... 74 Capítulo 3 - A Macroecological Perspective on Late Extinction …………………………………………………………. 76 Introduction ...... 77 Material and Methods ...... 79 Results ...... 83 Discussion ...... 85 Literature Cited ...... 86 Supplementary Information ...... 92 Considerações finais ...... 99

RESUMO

A dinâmica da diversidade no planeta acarreta o aparecimento e desaparecimento de espécies ao longo do tempo. O registro fóssil atualmente forma um corpo consistente de evidências sobre os eventos de extinção e esse fenômeno levanta uma questão fundamental no estudo da biodiversidade: Porque algumas espécies se extinguem, e desaparecem dos sistemas, e outras não? Nesta tese adotei uma abordagem macroecológica para atingir o objetivo geral de elucidar causas globais para o risco de extinção em mamíferos. No primeiro capítulo, particionei a importância relativa da idade filogenética, do tamanho corporal e da mudança do tamanho da distribuição geográfica das espécies na determinação do risco de extinção global de mamíferos. Levei em consideração a incerteza na estimativa dos preditores, tanto em espécies atuais, como em extintas. Mostrei que o tamanho do corpo é o melhor preditor único do risco de extinção em mamíferos, seguido do efeito combinado entre tamanho corporal e idade filogenética. Tanto a inserção de espécies extintas quanto a consideração da incerteza dos preditores se mostraram importantes na análise do risco de extinção. No segundo capítulo, analisei se o tamanho das massas de terra pode influenciar o risco de extinção de mamíferos via restrição da mudança do tamanho da distribuição geográfica e/ou limitação da dispersão em busca de habitats adequados à sobrevivência. Encontrei que os limites físicos continentais influenciam, em escala global, o risco de extinção dos mamíferos. Os limites continentais impedem as espécies de mamíferos a seguirem ambientes adequados à sobrevivência em um contexto de mudança do clima. Mostrei ainda que essa limitação funciona como uma restrição espacial horizontal para a dispersão dos mamíferos. Assim, mamíferos mais limitados pelo tamanho continental possuem maior probabilidade de serem extintos ao longo do tempo com o curso da mudança do clima. No terceiro capítulo, avaliei se este mesmo efeito restritivo das massas de terra sobre a capacidade das espécies de seguirem ambientes adequados à sobrevivência influenciou a extinção da Megafauna no final do Quaternário. Nesse contexto, comparei ainda dois métodos de análise dos dados obtidos, o de quadrados mínimos generalizados e o da regressão quantílica. Concluí que a extinção que dizimou boa parte da Megafauna foi um evento complexo e não linear. A Megafauna foi limitada em sua capacidade de se dispersar para novas regiões nas quais o clima permitiria a sobrevivência por mais tempo. Essa limitação foi imposta pela borda dos continentes, o que fez com que massas de terra de menor tamanho perdessem mais espécies de megafauna.

Palavras-chave: modelagem de nicho ecológico; habitat tracking; distribuição geográfica; massas de terra; megafauna.

12

ABSTRACT

The dynamics of diversity on the planet lead to the appearance and disappearance of over time. The fossil record currently forms a consistent body of evidence about extinction events and this phenomenon raises a fundamental question in the study of : Why do some species become extinct and disappear from systems and others not? In this thesis I took a macroecological approach to achieve the overall goal of elucidating global causes for mammalian extinction risk. In the first chapter, I partitioned the relative importance of phylogenetic age, body size and the change of the geographical distribution of species in determining the risk of global extinction. I accounted for the uncertainty in estimating predictors in both current and extinct species. I showed that body size is the single best predictor of extinction risk in mammals, followed by the combined effect of body size and phylogenetic age. Both the insertion of extinct species and the consideration of the uncertainty in predictors were important in extinction risk analysis. In the second chapter, I examined whether landmass size have influenced mammalian extinction risk by restricting the change in the size of geographic distribution and/or limiting dispersal in search of suitable habitats for survival. I found that continental physical limits globally influence the extinction risk of mammals. Continental boundaries prevent mammalian species from following environments suitable for survival in a context of climate change. I further showed that this limitation functions as a horizontal spatial constraint for mammalian dispersal. Thus, mammals that are more limited by landmass size are more likely to become extinct over time in the course of climate change. In the third chapter, I assessed whether this same restrictive effect of landmasses on the ability of species to follow suitable environments for survival influenced the extinction of the Megafauna at the end of the Quaternary. In this context, I also compared two methods of analysis of the obtained data, the generalized least squares and the quantile regression. I concluded that the extinction that wiped out much of Megafauna was a complex, nonlinear event. Megafauna has been limited on its ability to disperse to new regions where the climate would allow it to survive longer. This limitation was imposed by the edge of the continents, causing smaller landmasses to lose more species of megafauna.

Key-words: ecological niche modeling; habitat tracking; geographical distribution; landmasses; megafauna.

13

INTRODUÇÃO GERAL

A Macroecologia compõe um programa de pesquisa relativamente novo. Segundo Brown (1995), a Macroecologia é definida como o estudo da relação dos organismos e o ambiente de modo a caracterizar e explicar padrões estatísticos de abundância, distribuição e diversidade. O que identifica a Macroecologia como sendo um novo olhar sobre o estudo da biodiversidade é justamente a análise em grandes escalas espaciais e/ou temporais. Ainda, adquire caráter sintético quando tem por meta a inclusão de avanços em outras áreas como a paleobiologia, biogeografia e ciências da Terra (Brown, 1995). Em uma definição mais simplificada, a Macroecologia pode ser definida como o estudo da distribuição e abundância dos organismos em amplas escalas de tempo e espaço (Gaston; Blackburn, 1999, 2000).

No que pode ser considerado o marco fundador da Macroecologia, o trabalho de Brown e Maurer (1989), podemos observar características fundamentais da abordagem macroecológica: Utilização de análises computacionais; compilação de grandes quantidades de informação sobre características relevantes de uma biota taxonomicamente definida e a análise da distribuição estatística de tais variáveis entre as espécies. Neste caso, os autores utilizaram informações sobre tamanho corporal, densidade populacional e distribuição geográfica de aves. Ainda segundo os mesmos autores, esses pontos permitem a elucidação de processos ecológicos que afetam a diversificação de um táxon. O rápido desenvolvimento deste campo (veja Smith et al., 2008; Beck et al., 2012), permitiu também o avanço da compilação e digitalização de grandes bancos de dados de traços biológicos importantes para a macroecologia como de tamanho corporal e traços de história de vida [ex. Smith et al. (2003); PanTheria (Jones et al, 2009); EltonTraits (Wilman et al., 2014); Myhrvold et al. (2015)], de pontos de ocorrência [ex. GBIF (2017); PaleoBioDB (2017)] e super-árvores filogenéticas de grupos como, por exemplo, os mamíferos (Bininda-Emonds et al., 2007; Kuhn et al., 2011; Faurby; Svenning, 2015).

A extinção, como um dos componentes reguladores da biodiversidade (Purvis et al., 2000), também pode ser alvo de estudo da Macroecologia (ex. Kitzes; Harte, 2014; Sandom et al., 2014; Saupe et al., 2015; Congreve et al., 2018; Spooner et al., 2018). Nesse contexto, a relação entre o tamanho corporal e o tamanho da distribuição

14

geográfica de vários grupos animais vem sendo salientada como importante na explicação dos padrões de extinção de espécies globalmente (Diniz-Filho, 2004; Lima- Ribeiro et al., 2014). O argumento geral é que uma espécie deve possuir um tamanho de distribuição geográfica que propicie suficiente aquisição de energia para a sobrevivência das populações a longo prazo. Caso a distribuição geográfica venha a diminuir e atingir níveis insustentáveis em termos de aquisição de recursos, a espécie pode se extinguir, de início localmente e em seguida globalmente (Lima-Ribeiro et al., 2014). O tamanho do corpo por si só vem sendo amplamente citado como sendo um indicador do risco de extinção de uma espécie (Gaston; Blackburn, 1995; Purvis et al., 2000; Cardillo; Bromham, 2001; Cardillo, 2003; Cardillo et al., 2005; Davidson et al., 2009; Fritz et al., 2009; Ripple et al., 2017).

Assim, o tamanho de corpo pode ser considerado uma característica biológica chave relacionada ao risco de extinção em animais (Fisher et al., 2003; Cardillo et al., 2008). Outros traços biológicos citados por influenciar o risco de extinção são o grau de especialização (Harcourt et al., 2002), tamanho da área de vida (Améca y Juárez et al., 2014) e características de história de vida como número de filhotes, idade de maturação e tempo de gestação (Johnson, 2002; Botha-Brink et al., 2016). Dada a estrutura filogenética sobre a qual os caracteres biológicos estão distribuídos ao longo da árvore da vida, podemos esperar que espécies mais próximas tenham valores mais semelhantes de um determinado traço do que esperado ao acaso (veja Felsenstein, 1985; Purvis, 2008, Davies; Yessoufou, 2013). Como espécies mais próximas possuem características mais parecidas e tais características podem estar ligadas ao risco de extinção, podemos esperar ainda, que o próprio risco de extinção não esteja distribuído aleatoriamente em uma filogenia (Bennet; Owens, 1997; Purvis, 2008). De fato, alguns autores tem elencado características da própria história evolutiva das espécies que as levam a uma maior probabilidade de desaparecimento (ex. Gaston; Blackburn, 1997; Meijaard et al., 2008).

Nesta tese, utilizo dados de tamanho corporal, filogenias e dados de distribuição geográfica para inferir causas relacionadas ao risco de extinção em mamíferos. Empreguei amplamente, ao longo dos capítulos, técnicas de modelagem de nicho ecológico (veja Peterson et al., 2011) para obter informações sobre a dinâmica da distribuição geográfica de mamíferos ao longo do tempo. Os modelos de nicho

15

ecológico (MNE) têm representado grande avanço na investigação de questões ecológicas que necessitam de uma abordagem em amplas escalas espaciais ou temporais (ex. Terribile et al., 2012). Os MNE’s têm sido utilizados também, para estudos sobre risco de extinção (Lima-Ribeiro et al., 2012; Lima-Ribeiro et al., 2013; Lima-Ribeiro et al., 2017). Escolhi os mamíferos como modelo para este trabalho por se tratar de um grupo relativamente bem estudado, com filogenias e informações de distribuição geográfica disponíveis, além de serem vastamente distribuídos ao redor do globo (Bininda-Emonds et al., 2007; Kuhn et al., 2011; Faurby; Svenning, 2015).

No decorrer de três capítulos escritos em forma de artigo e em inglês, questiono quais fatores levam algumas espécies de mamíferos a terem maior risco de extinção do que outras. Mais especificamente, no primeiro capítulo relaciono tamanho corporal, idade filogenética e mudança no tamanho da distribuição geográfica ao longo do Quaternário com os riscos de extinção das espécies de mamíferos atuais conforme a avaliação da União Internacional para a Conservação da Natureza (IUCN, 2018). Também utilizei, espécies fósseis para esta análise, além de considerar a incerteza na estimação dos preditores na análise dos resultados. No segundo capítulo, relaciono o risco de extinção ao tamanho das massas de terra que as espécies habitam para avaliar se as barreiras continentais impediram as espécies de mudarem o tamanho da distribuição geográfica ou de seguirem ambientes adequados à sobrevivência ao longo do Quaternário. No terceiro capítulo investigo se essa mesma limitação imposta pela barreira continental às espécies terrestres influenciou a onda de extinção em massa que extirpou grande parte da Megafauna de mamíferos que proliferou pela Terra ao longo da última era glacial. Ao final deste documento, evidencio aspectos gerais que pude concluir ao longo do trabalho.

Literatura Citada

AMECA Y JUÁREZ, E. I. et al. Identifying species' characteristics associated with natural population die‐offs in mammals. conservation, v. 17, n. 1, p. 35- 43, 2014.

BECK, Jan et al. What's on the horizon for macroecology? Ecography, v. 35, n. 8, p. 673-683, 2012.

16

BENNETT, Peter M.; OWENS, Ian PF. Variation in extinction risk among birds: chance or evolutionary predisposition? Proceedings of the Royal Society of London. Series B: Biological Sciences, v. 264, n. 1380, p. 401-408, 1997.

BININDA-EMONDS, Olaf R. P. et al. The delayed rise of present-day mammals. Nature, v. 446, n. 7135, p. 507, 2007.

BOTHA-BRINK, Jennifer et al. Breeding young as a survival strategy during Earth’s greatest mass extinction. Scientific Reports, v. 6, p. 24053, 2016.

BROWN, James H. Macroecology. University of Press, 1995.

BROWN, James H.; MAURER, Brian A. Macroecology: the division of food and space among species on continents. Science, v. 243, n. 4895, p. 1145-1150, 1989.

CARDILLO, Marcel. Biological determinants of extinction risk: why are smaller species less vulnerable? In: Animal Conservation forum. Cambridge University Press, p. 63-69, 2003.

CARDILLO, Marcel; BROMHAM, Lindell. Body size and risk of extinction in Australian mammals. Conservation biology, v. 15, n. 5, p. 1435-1440, 2001.

CARDILLO, Marcel et al. Multiple causes of high extinction risk in large mammal species. Science, v. 309, n. 5738, p. 1239-1241, 2005.

CARDILLO, Marcel et al. The predictability of extinction: biological and external correlates of decline in mammals. Proceedings of the Royal Society B: Biological Sciences, v. 275, n. 1641, p. 1441-1448, 2008.

CONGREVE, Curtis R.; FALK, Amanda R.; LAMSDELL, James C. Biological hierarchies and the nature of extinction. Biological Reviews, v. 93, n. 2, p. 811- 826, 2018.

DAVIDSON, Ana D. et al. Multiple ecological pathways to extinction in mammals. Proceedings of the National Academy of Sciences, v. 106, n. 26, p. 10702-10705, 2009.

DAVIES, T. Jonathan; YESSOUFOU, Kowiyou. Revisiting the impacts of non-random extinction on the tree-of-life. Biology letters, v. 9, n. 4, p. 20130343, 2013.

DINIZ‐FILHO, José Alexandre Felizola. Macroecology and the hierarchical expansion of evolutionary theory. Global Ecology and , v. 13, n. 1, p. 1-5, 2004.

FAURBY, Søren; SVENNING, Jens-Christian. A species-level phylogeny of all extant and late Quaternary extinct mammals using a novel heuristic-hierarchical Bayesian approach. Molecular phylogenetics and evolution, v. 84, p. 14-26, 2015.

17

FELSENSTEIN, Joseph. Phylogenies and the comparative method. The American Naturalist, v. 125, n. 1, p. 1-15, 1985.

FISHER, Diana O.; BLOMBERG, Simon P.; OWENS, Ian PF. Extrinsic versus intrinsic factors in the decline and extinction of Australian . Proceedings of the Royal Society of London. Series B: Biological Sciences, v. 270, n. 1526, p. 1801-1808, 2003.

FRITZ, Susanne A.; BININDA‐EMONDS, Olaf RP; PURVIS, Andy. Geographical variation in predictors of mammalian extinction risk: big is bad, but only in the tropics. Ecology letters, v. 12, n. 6, p. 538-549, 2009.

GASTON, Kevin J.; BLACKBURN, Tim M. Birds, body size and the threat of extinction. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, v. 347, n. 1320, p. 205-212, 1995.

GASTON, Kevin J.; BLACKBURN, Tim. Evolutionary age and risk of extinction in the global avifauna. Evolutionary Ecology, v. 11, n. 5, p. 557-565, 1997.

GASTON, Kevin J.; BLACKBURN, Tim M. A critique for macroecology. Oikos, p. 353-368, 1999.

GASTON, Kevin; BLACKBURN, Tim. Pattern and process in macroecology. John Wiley & Sons, 2000.

GBIF. Global Biodiversity Information Facility, 2017. Available from: https://www.gbif.org/.

HARCOURT, A. H., COPPETO, S. A., PARKS, S. A. Rarity, specialization and extinction in . Journal of Biogeography, v. 29, n. 4, pp. 445-456, 2002.

IUCN. International Union of Conservation of Nature Red List of Threatened Species. Version 2018-1. Available from: http://www.iucnredlist.org/. Accessed in: 19th July 2018.

JOHNSON, Chrisopher N. Determinants of loss of mammal species during the Late Quaternary ‘megafauna’ : life history and ecology, but not body size. Proceedings of the Royal Society of London. Series B: Biological Sciences, v. 269, n. 1506, p. 2221-2227, 2002.

JONES, Kate E. et al. PanTHERIA: a species‐level database of life history, ecology, and geography of extant and recently extinct mammals. Ecology, v. 90, n. 9, p. 2648-2648, 2009.

KITZES, Justin; HARTE, John. Beyond the species–area relationship: improving macroecological extinction estimates. Methods in Ecology and Evolution, v. 5, n. 1, p. 1-8, 2014.

18

KUHN, Tyler S.; MOOERS, Arne Ø.; THOMAS, Gavin H. A simple polytomy resolver for dated phylogenies. Methods in Ecology and Evolution, v. 2, n. 5, p. 427-436, 2011.

LIMA-RIBEIRO, Matheus Souza et al. Potential suitable areas of giant ground dropped before its extinction in : the evidences from bioclimatic envelope modeling. Natureza & Conservação, v. 10, n. 2, p.145-151, 2012.

LIMA-RIBEIRO, Matheus Souza et al. Climate and set the place and time of Proboscidean extinction in late Quaternary of South America. Palaeogeography, Palaeoclimatology, Palaeoecology, v. 392, p. 546-556, 2013.

LIMA-RIBEIRO, Matheus S. et al. Constraint envelope analyses of macroecological patterns reveal climatic effects on mammal extinctions. Quaternary research, v. 82, n. 1, p. 260-269, 2014.

LIMA‐RIBEIRO, Matheus S. et al. Fossil record improves biodiversity risk assessment under future climate change scenarios. Diversity and Distributions, v. 23, n. 8, p. 922-933, 2017.

MEIJAARD, Erik et al. Phylogenetic age is positively correlated with sensitivity to timber harvest in Bornean mammals. Biotropica, v. 40, n. 1, p. 76-85, 2008.

MYHRVOLD, Nathan P. et al. An amniote life‐history database to perform comparative analyses with birds, mammals, and reptiles: Ecological Archives E096‐269. Ecology, v. 96, n. 11, p. 3109-3109, 2015.

PBDB. The Paleobiology database, 2017. Available from: https://paleobiodb.org/.

PETERSON, A. Townsend et al. Ecological niches and geographic distributions (MPB-49). Princeton University Press, 2011.

PURVIS, Andy et al. Predicting extinction risk in declining species. Proceedings of the royal society of London. Series B: Biological Sciences, v. 267, n. 1456, p. 1947- 1952, 2000.

PURVIS, Andy. Phylogenetic approaches to the study of extinction. Annual Review of Ecology, Evolution, and Systematics, v. 39, p. 301-319, 2008.

RIPPLE, William J. et al. Extinction risk is most acute for the world’s largest and smallest vertebrates. Proceedings of the National Academy of Sciences, v. 114, n. 40, p. 10678-10683, 2017.

SANDOM, Christopher et al. Global late Quaternary megafauna extinctions linked to humans, not climate change. Proceedings of the Royal Society B: Biological Sciences, v. 281, n. 1787, p. 20133254, 2014.

SAUPE, Erin E. et al. Niche breadth and geographic range size as determinants of species survival on geological time scales. Global Ecology and Biogeography, v. 24, n. 10, p. 1159-1169, 2015. 19

SMITH, Felisa A. et al. Body mass of late Quaternary mammals. Ecology, v. 84, n. 12, p. 3403-3403, 2003.

SMITH, Felisa A. et al. Macroecology: more than the division of food and space among species on continents. Progress in Physical Geography, v. 32, n. 2, p. 115-138, 2008.

SPOONER, Fiona E. B.; PEARSON, Richard G.; FREEMAN, Robin. Rapid warming is associated with population decline among terrestrial birds and mammals globally. Global change biology, v. 24, n. 10, p. 4521-4531, 2018.

TERRIBILE, Levi Carina et al. Areas of climate stability of species ranges in the Brazilian Cerrado: disentangling uncertainties through time. Natureza & Conservação, v. 10, n. 2, p. 152-159, 2012.

WILMAN, Hamish et al. EltonTraits 1.0: Species‐level foraging attributes of the world's birds and mammals. Ecology, v. 95, n. 7, p. 2027-2027, 2014.

20

CAPÍTULO 1

21

Historical, Ecological and Geographical Correlates of Extinction Risk in Mammals

Vinicius Silva Reis1,2, Flávia Machado1,2, Lucas Jardim2, Wellington Hannibal3, Levi Carina Terribile1, Matheus Souza Lima-Ribeiro1

1 – Laboratório de Macroecologia, Universidade Federal de Jataí, Jataí, Brasil; 2 – Instituto de Ciências Biológicas V, Laboratório de Ecologia Teórica e Síntese, Universidade Federal de Goiás, Goiânia, Brasil; 3 – Laboratório de Ecologia e Biogeografia de Mamíferos, Universidade Estadual de Goiás, Quirinópolis, Brasil.

Abstract

Extinction risk has been correlated in literature with geographic (range size, range shift), ecological (body size, life history traits) and historical (clade age, phylogenetic distinctiveness) characteristics, but the relative contribution of each one of these factors remains unknown. Here, we disentangled the contribution of historical, ecological and geographical correlates on mammals’ extinction risk taking into account the uncertainty in the estimation of such predictors using both extant and extinct species. We computed the phylogenetic branch length as a proxy for evolutionary age; gathered body mass from online datasets and through phylogenetic imputation methods as a proxy for life history traits and built ecological niche models to estimate range shift over the last for each species. To account for uncertainty in the estimation of each predictor, we built 1000 phylogenies by randomly inserting extinct species (phylogenetically uncertain taxa) at the most derivate clade and built 1000 ENMs ensembles by randomly combining methods and paleoclimatic simulations. We developed general linear models with binary logit link for the response variable (not threatened vs. threatened/extinct) and generated models for individual and four possible combinations of the predictors. We partitioned the estimated deviance for each model into individual and shared deviances. Body mass emerged as the single best predictor of mammals’ extinction risk regardless of the combinations of variables and extant/extinct species used here. When all species were considered together, the uncertainty on estimation of evolutionary age most affected the influence of this predictor. Future studies should benefit from regional and extrinsic aspects determining extinction risk.

1) Introduction

The dynamics of biodiversity entails the appearance of new species as well the extirpation of non-adapted clades (Raup, 1994). Species extinction is part of the diversification processes and evolutionary history of all clades (Jablonsky, 1999; Quental; Marshall, 2010). The fossil record provides a consistent body of evidence corroborating extinction phenomenon and raises a fundamental issue in the study of biodiversity: why do some species become extinct and disappear from , while others keep surviving? The extinction process does not develop randomly

22

(Bennett; Owens, 1997; Fritz; Purvis, 2010) and this suggests that some species might possess biological traits that influence survival or extinction through time (Purvis, 2008).

The reduction of the geographic range size is used by the International Union of Conservation of Nature (IUCN, 2016) to classify species regarding extinction risk. Taking into account that if a geographical range size decreases in such a way to reduce the species into few individuals, geographical range size becomes intuitively a good predictor of extinction risk. Then, species limited to small geographic distributions are more probable to face extinction (McKinney, 1997; Purvis et al., 2000a). In this case, the vulnerability of the species relates to its capacity of population recovery, more specifically to reproductive life history traits (Hutchings et al., 2012). In turn, life history traits linked to reproductive capacity keep allometric relations with body size of (Peters, 1983; Hilbers et al., 2016). Large-bodied animals are more vulnerable to extinction than small-bodied ones (Cardillo et al., 2008), since large body might implicate a relative slow reproductive cycle and undermost capacity of population recovery.

Although body size by itself is not cause of many ecological and evolutionary patterns (Koslowski; Wiener, 1997), it is useful as surrogate of causal biological traits hard to obtain (Gaston; Blackburn, 2000). Thus, body size has been extensively linked to species extinction risk (see eg. Cardillo; Bromhan, 2000; Purvis et al., 2000a, 2000b; Cardillo et al., 2005, 2008; Davidson et al., 2009). Given that both body size and geographical range size are traits prone to be structured throughout the phylogenetic history of a clade [see Diniz-Filho; Tôrres (2002); Ashton (2004); Blackburn et al. (2004); Freckleton; Jetz (2009) and Böhning-Gaese et al. (2006); Zacaï et al. (2017) for a contrasting opinion regarding range size], phylogenetic history can be also connected to extinction risk probability. In general, it has been affirmed that phylogenetically older species are more vulnerable to extinction [Gaston; Blackburn (1997); Johnson et al. (2002); Meijaard et al. (2008), but see Arregoitia et al. (2013) for opposite evidences]. Body size, geographic range and phylogenetic age have already been singly tested against extinction probability, encompassing limited groups of organisms and excluding already extinct species, such the late-Quaternary extinct mammalian megafauna. To our knowledge there is still no evaluation of the relative importance of body size,

23

geographic range and phylogenetic age in the generation of extinction risk patterns in wide taxonomic scales.

Here, we partitioned the relative influence of species age, body size and range shift determining the global pattern of mammals’ extinction risk, taking into account the intrinsic uncertainties to estimate these predictors for both extant and extinct species. Mammals are a diverse clade, with an old evolutionary history and possesses both extinct and extant representants with varying body sizes and reproductive characteristics (see Bininda-Emonds et al., 2007 and Faurby; Svenning, 2015 for phylogenies). The late and relatively well-preserved fossil record (Koch; Barnosky, 2006; Lima-Ribeiro; Diniz-Filho, 2013a) make mammals an interesting biological model to disentangle the contribution of historical, ecological and geographical correlates from extinction risk analyses including not only living species but also extinct ones. We test specifically which factors, body size, range shift or phylogenetic age, better predict the mammals’ extinction risk. To avoid biased results, we also considered uncertainty in predictor estimation to evaluate the relative impact of historical, ecological and geographical correlates on mammals’ extinction risk.

2) Material and Methods

Historical correlates

As a historical correlate of extinction risk, we estimated the phylogenetic age of each species. We randomly selected 100 phylogenies from Kuhn et al. (2011), which represent different combinations of possible resolutions of polytomies contained in Bininda-Emonds et al. (2007) super-tree. Since Kuhn’s et al. (2011) phylogenies only comprehended extant species, we considered missing extinct species (n=154) as phylogenetically uncertain taxon (hereafter PUT; Rangel et al., 2015). The PUTs were randomly added into the phylogenies at branches subsequent to the most derived consensus clade (MDCC). Branch length probabilistically weighted the position of PUTs; larger branches, representing longer evolutionary histories, were more likely to receive a PUT than shorter branches. The MDCCs were determined from the taxonomic classification of extinct species available in Fossilworks (Alroy, 2018, Table S1), and represent the most derived taxonomic level (class, order, family or ) an extinct 24

species shares with its extant relatives. In phylogenetic terms, the MDCCs reflect the most common ancestors from which PUTs unequivocally descend (Martins et al., 2013; Rangel et al., 2015).

The PUT procedure was repeated 10 times across every phylogeny, such that 1,000 phylogenetic trees (100 phylogenies * 10 repetitions) were obtained. This set of trees represent both the uncertainty of polytomies contained initially in the Bininda- Emonds’ et al. (2007) super-tree and the uncertainty of the phylogenetic position of newly added extinct species (PUTs). We considered the length of the individual terminal branch connected to each tip as the phylogenetic age of its respective species, which was used as the historical predictor of mammals’ extinction risk. We also used this same set of 1000 phylogenies for the generation of the ecological correlates described below.

Ecological correlates

Given allometric relationships with key life history traits (Damuth, 1981), we considered body size as the ecological correlate of species extinction risk. For extant species, we gathered information about body mass in grams from on-line databases and scientific literature, filling the missing data in this order: PanTheria (Jones et al. 2009); EltonTraits 1.0 (Wilman et al., 2014); Encyclopedia of Life (Parr et al., 2014) and literature regarding specific taxonomic groups: such Mittermeier et al. (2013) for primates and Paglia et al. (2012), Voss et al. (2004, 2005), Pavan and Voss (2016), Pine and Hardley-Jr (2008), Lew et al. (2006), Flores et al. (2008) for small mammals. For extinct species, we accessed the body mass database for late Quaternary mammals from Smith et al. (2003), Turvey and Fritz (2011) and Dantas et al. (2017).

Approximately 1.4% (56 out of 4020) of the species still missed empirical body mass estimation, all of them extinct species. To fill these gaps, we estimated the lacking body masses from phylogenetic imputation by using the random method as implemented in the package “missForest” (version 1.4, Stekhoven; Bühlmann, 2012). We used two sets of predictors in the imputation process: phylogenetic eigenvectors and life history traits. We extracted the phylogenetic eigenvectors from the set of 1,000 phylogenetic trees (see previous section “Historical correlates”) using the phylogenetic eigenvector maps (Guénard et al., 2013) as implemented in the package “MPSEM”

25

(version 0.3-3, Guénard, 2017). We selected the first 250 eigenvectors for imputation minimal errors of imputed body mass (See evaluation metrics for the imputation process in Fig. S1). The life history traits were gathered from PanTheria (Jones et al., 2009) and Myhrvold et al. (2015) databases, namely: age at first birth, gestation length, home range area, interbirth interval, litter size, number of litters per , body mass of the neonate, population density, age at sexual maturity, generation length and time for female maturity. Extinct species also missed all life history information, meanwhile the extant ones had information for at least one of the mentioned life history characteristics.

Geographical correlates

We considered the species range shift – i.e. increase or decrease in geographical range – since the (LGM, ~21 ky BP) to the mid- (~6 ky BP) as the geographical correlate of extinction risk. We used ecological niche modeling (ENM) to estimate the species distribution throughout the last ice age.

We gathered georeferenced occurrence records for extant species from the Global Biodiversity Information Facility (GBIF, 2017), using the package “rgbif”, (version 0.9.9, Chamberlain, 2017). We also gathered occurrence information for both extinct and extant species from fossil datasets that provide dated species records, such as: The Paleobiology Database (PBDB, 2017), a global dataset, by the package “paleobioDB” (version 0.5.0, Varela et al. 2015a); The FosSahul Database (Rodríguez- Rey et al., 2016) for Australian fossil mammal occurrences; the New and Old Worlds Database (Fortelius, 2015) and the faunal database from The Stage Three Project (The Stage Three Project, 2017) for Eurasian mammals; the georeferenced information contained in Faith (2014) for African fossil mammal occurrences; the FaunMap (Graham; Lundelius, 2014) and Databases for , Card (2011) and the for South America, Lima-Ribeiro et al. (2013b).

We carefully reviewed the occurrence points from all databases for taxonomic synonymies and inconsistences related to fossil aging. Only dated occurrences identified at species level remained. We calibrated C14 dated records for calendrical aging with the CALIB Radiocarbon Calibration software (version 7.1, Stuiver et al., 2018) with the IntCal13 calibration curve (Reimer et al., 2013) or the SHCal13 curve (Hogg et al., 26

2013) for north or south hemisphere records, respectively. Yet, we eliminated duplicated records overlapping a same grid cell of 0.5° (~55 Km) spatial resolution (lat/long) and with the same age at 1 ky of temporal resolution. At the end, we obtained 241,911 occurrences for 2,210 living and 75 extinct species.

We obtained 19 bioclimatic layers for LGM and mid-Holocene from the ecoClimate database (www.ecoClimate.org; Lima-Ribeiro et al., 2018). We based on Varela et al. (2015b) to select the AOGCMs that maximize uncertainty among climate simulations at LGM, namely: CCSM and MIROC (Table S2). As bioclimatic variables are correlated, we carried out a factorial analysis to choose the bioclimatic variables that best keep the variance of all climatic characteristics with minimal collinearity. Factorial analyses were performed for each biogeographical realm and the number of factors was determined from scree plot (see more details about factorial analysis in Hair et al. 2009). Then we selected the bioclimatic variables with the greater loadings within each factor (Table S3). Before the modeling process, we standardized the bioclimatic variables by ranging between 0 and 1 to ensure that predictors are comparable throughout ENM process. For this we used the function “decostand” from package “vegan” version 2.5-1 (Dixon, 2003; Oksanen et al., 2018).

To acquire climatic information for those occurrences whose dating did not fit within the time bins of LGM nor mid-Holocene, we performed a temporal interpolation using the δ18 Oxygen curve from Lisiecki and Raymo (2005) at a 1 ky resolution. The δ18 Oxygen curve is the ratio curve between the quantity of the 18O and 16O stable isotopes present in biogenic components such invertebrate marine shells and can be used as a proxy of temperature dynamics throughout the geological time (Mulitza et al., 2003). Because interpolation accuracy decreases at longer time ago, we eliminated all fossil records aged older than 700 ky BP.

To consider uncertainties in ENM predictions, we used four methods based on different data and statistical characteristics: Bioclim, Domain (Gower distance), Maximum entropy (Maxent) and Support Vector Machine (SVM, see Table S4). Additionally, we used the simplest euclidean environmental distance method alone to build ENMs for species with less than five occurrence points. We combined climate conditions from both fossil and modern occurrences in a multitemporal calibration approach (see Maiorano et al., 2013; Lima-Ribeiro et al., 2017) to predict the potential 27

distributions of living and extinct mammals onto LGM and mid-Holocene. We delimited the biogeographical realms in which every species occurs as the background area to build its respective ENMs. From background areas, we randomly sampled 1000 points to fit the presence-background methods (i.e. Maxent and SVM). We fitted machine learning presence-background methods with the simplest configuration for classification tasks: we considered a linear kernel function with probabilistic output for SVM and only the linear feature to combine climatic predictors in Maxent, with logistic output. According to Varela et al. (2011), simple choices for methods and parametrization procedures are part of a set of good practices when modeling ecological niches from fossil records. We did not consider presence-absence modeling methods because most species lack absence data, as well as due to the impossibility of generating absence information for extinct species.

To avoid unreliable evaluations based on pseudo-absences (Golicher et al., 2012), the ENM predictions were evaluated using only presence records. Predictions for extinct species and species with less than 5 occurrences were also not evaluated. For species with occurrences ranging from 5 to 25 points, ENMs were evaluated by using the jackknife procedure known as leave-one-out (LOO, Pearson et al., 2007). For species presenting more than 25 occurrences, records were randomly split into 75% for training and 25% for testing. We repeated this procedure 20 times to minimize effects from spurious partitions. In both cases, we computed the “D” statistics proposed by Pearson et al. (2007), as follows:

D = ∑ Xi (1 - Pi)

where Pi represents the proportion of study area (background) predicted as species presence and Xi represents the rate of success in predicting the testing occurrences. For species with occurrences ranging from 5 to 25 records, Xi follows the original elaboration by Pearson et al. (2007) and represents the binary variable informing success (Xi = 1) or failure (Xi = 0) of ENMs to predict the omitted occurrences in the

LOO procedure. For species with more than 25 occurrences, Xi varies continuously from 0 to 1 and represents the True Positive Rate; the proportion of testing occurrences correctly predicted by ENMs (see Fielding; Bell, 1997). During the evaluation process,

28

we used the lowest presence threshold (LPT) criterion to obtain binary maps and confront ENM predictions against testing presences.

We developed consensus maps for LGM and mid-Holocene from distinct methods and AOGCMs in an ensemble approach (Araújo; New, 2007). After standardizing the ENM predictions to suitability vary between 0 and 1, the consensus maps were obtained by a weighted mean of the 8 predictions (4 methods * 2 AOGCMs) of each species in each period. We used the D-statistics as weights, thus predictions better evaluated had more importance on the final consensus maps. For species whose ENMs were not evaluated, the consensus maps were generated by arithmetic mean among predictions.

In accordance with evolutionary age and body mass, we also considered uncertainty on range shift across 1,000 consensus maps for each species by randomly combining the 8 predictions (4 methods * 2 AOGCMs). To transform these continuous consensus maps into binary occurrence maps, we used a generic approach that searches for the threshold that maintain species distribution over 50% of the background area at the LGM. This threshold was then used to binarize both LGM and mid-Holocene continuous maps for each species. Considering we are interested in estimating the magnitude of range shift since the LGM and not the absolute distribution areas themselves, this generic approach of binarization proves efficient once it guarantees equal area (50% of background) for species either reduce or increase geographical range through time. With this criterion, it is also possible to avoid spurious extreme values of range area by getting extreme thresholds. We estimated geographical range size at the LGM and mid-Holocene by counting the number of grid cells each species was predicted to occupy throughout the background. The range shift, an increase or decrease in geographical distribution over time, was accessed by the difference in geographical range sizes between mid-Holocene and LGM and expressed by the number of cells. Thus, positive range shifts are intuitively connected with range increment and negative ones with range loss. Each species has then 1,000 values of range shift.

Data analyses

To disentangle the relative role of historical, ecological and geographical correlates determining the extinction risk in mammals, we started obtaining the threat

29

status of living species from the International Union for Conservation of Nature red list (IUCN, 2016). Then we converted the IUCN status in a binary response variable indicating the mammals’ extinction risk, as follows: 1) threatened species (T, n=533), including vulnerable, endangered, critically endangered and extinct species (including ); 2) non-threatened species (NT, n=1,752), including least concern and near threatened species.

We then fitted a generalized linear models (GLM) using the binary extinction risk (T vs. NT) as response variable, and evolutionary age, body size and range shift as predictors. The GLM was built with the binary family and logit link from “glm” function of the “stats” package (R Core Team 2017). Similar to partial regression, we used the function “Dsquared” from “modEvA” package (version 1.3.2, Barbosa et al., 2016) to partition the amount of adjusted deviance (D-squared) accounted for each predictor alone (individual contributions) and all combinations among them (shared contributions) (Figure 1). According to Guisan and Zimmerman (2000), the D-squared from GLMs is equivalent to the R-squared from linear models and reports the proportion of the deviance in response variable that predictors account for. The adjusted version of the D-squared enables comparing models with different number of predictors, such as we aim here across individual and shared contributions. By taking into account the uncertainty into predictors, we calculated 1,000 replicates of D-squared for individual and shared contributions among evolutionary age, body mass and range shift (Figure 1). Finally, we compared the replicates set of deviances with Kruskal- Wallis test in order to check if they were statistically different among each other.

B P BP B P A Frequency BS PS

1000 replicates of unique 1000 replicates of unique contribution of body size to S contribution of phylogenetic extinction risk age to extinction risk

BP – Body size and S R

phylogenetic age PS – Phylogenetic y age and range shift nc BS – Body size and range shift A – Deviance shared Freque by all predictors 30 1000 replicates of non- Portions of shared 1000 replicates of unique contributions among two contribution of range shift to explained deviance in or all predictors. extinction risk extinction risk

Figure 1. Deviance partition of extinction risk among historical, ecological and geographical correlates. Unique deviances are: B – body mass (grams); P – phylogenetic age (length of the terminal branch) and S – range shift (number of grid cells). Shared deviances are: BP – body mass and phylogenetic age; PS – phylogenetic age and range shift; BS – body mass and range shift and A – deviance shared by all predictors. R – Residual deviance; i.e. not explained by any predictor. As we had 1000 replicates of each predictor for each species, the same number of values are contained in each histogram.

3) Results

Overall, the historical, ecological and geographical correlates accounted for a relatively small deviance of mammals’ extinction risk, regardless considering extinct Quaternary megafauna or omitting them (D-squared < 15%). However, the explanatory power of the model that has taken fossil species into account (Figures 2 and 3, Kruskal- Wallis test, H=6667.4, df=7, p-value<0.001) showed an improvement compared to the model without fossil species (Figures S2 and S3, Kruskal-Wallis test, H=6025.06, df=7, p-value<0.001).

The partitioned D-squared showed body mass as the strongest correlate of mammals’ extinction risk. The inclusion of extinct species increased at least four times the explanatory power of body mass as the unique best predictor (D-squared). The shared contribution between body mass and phylogenetic age was the second best predictor of extinction risk in mammals. Similarly, none of other combinations between predictors explained better the deviance in extinction risk in mammals than body mass and phylogenetic age together (Figure 3). The part of the deviance not explained by any of the predictors alone nor their combination depicted more than 85% of the deviance in all models (Figures 3 and S3).

b p s bp ps bs a R 31

Figure 2. Median percentage of deviance for unique and shared predictors of extinction risk. Unique deviances are: b – body mass (grams); p – phylogenetic age (length of the terminal branch) and s – range shift (number of grid cells). Shared deviances are: bp – body mass and phylogenetic age; ps – phylogenetic age and range shift; bs – body mass and range shift and a – deviance shared by all predictors. R – Residual deviance, i.e. not explained by any predictor. 1000 replicates for each portion of deviance. Extinct species included.

P B

0.15

10.3 -0.003 equency

r F 0.0005 -0.009 0.0008

-0.03 Body Mass Phylogenetic Age

Body size and S R phylogenetic age (m=0.15)

Phylogenetic age and ncy range shift (m=0.0008) Body size and range

shift (m=-0.009) Freque Deviance shared by all predictors (m=0.0005)

Shared Deviances Range Shift Residual

Figure 3. Percentage of unique and shared deviances obtained by the generalized linear models considering extinct species. Unique deviances are: B – body mass (grams); P – phylogenetic age (length of the terminal branch) and S – range shift (number of grid cells). The portion of deviance not explained by any variable or set of variables represents 89.6±0.19% in average of the total deviance of the model. The unique best predictor of extinction risk in mammals is body size (10.3±0.21), followed by the shared deviance between body size and phylogenetic age (0.15±0.10, both values in bold). 1000 replicates for each portion of deviance.

4) Discussion

Our findings highlight that extinct species are indispensable in extinction risk analyses and indicate body size as the best single correlate of mammals’ extinction risk. It is paradoxical that extinct species are often neglected from extinction risk analyses, especially in the modern world whose extinctions have been happening faster than earlier (Nott et al., 1995; Pimm et al., 1995). Omitting extinct species seems historically

32

justifiable behind the guise of paleontological shortfall, but this routine should be discarded nowadays with the increasing availability of fossil databases worldwide (Brewer et al., 2012). Understand the genuine role of extinction correlates is crucial for planning better conservation policies in the modern world and foremost prevent that further species perish soon. Our findings undoubtedly show that including extinct species most truly highlight the relative role of extinction correlates and provide more realistic assessments of the global extinction picture.

Body size has already been pointed as a key correlate of extinction risk across many taxa and biogeographical regions (Purvis et al., 2000a, 2000b; Cardillo; Brohman, 2001; Cardillo et al., 2005, 2008; Davidson et al., 2009). As argued by Johnson et al. (2002), body size is a robust surrogate of life history traits associated to reproductive rates, which actually underlie the mammals’ extinction risk. The velocity at which populations recover after disturbances dictate the fate of species (Audzijonyte; Kuparinen, 2016), such that large-bodied mammals often presenting slow reproductive rates become more prone to extinctions in a changing world (Cardillo, 2003; Cardillo et al. 2005). However, body size also scales with the opposite reasoning by focusing on dispersal capability of species (Angert et al., 2011; Bowmann et al., 2002; Schloss et al., 2012). Species able to migrate for longer distances, also large-bodied ones, most likely safe from extinctions by tracking suitable habitats in a changing world. Because of the contrasting allometric relationships with multiple life history traits, body size represents an ecological correlate of extinction risk with dubious effect. Depending of focal surrogates (whether eg. dispersal ability), body size may increase or decrease species extinction risk. Being the double player the stronger correlate of extinction risk, body size deserves special attention in extinction analyses and further studies should disentangle its contrasting effects on species survival.

The common influence of body mass and evolutionary history together emerged as the second best predictor of extinction risk in mammals. We interpret such shared influence as being an artifact of the phylogenetic structure that body mass itself possesses (eg. Diniz-Filho; Tôrres, 2002). As we did not opt for the usage of phylogenetic comparative methods (Felsenstein, 1985), the effect of the phylogenetic autocorrelation accrued from the common evolutionary history of mammal species was not separated from body mass data. Regarding phylogenetic age itself, despite many

33

works claim phylogenetically older species to be in higher extinction risk (see McDowall, 1969; Gaston; Blackburn, 1997 for birds and Meijaard et al., 2008; Johnson et al., 2002 for mammals), we observed little influence of the evolutionary history alone on extinction risk. Arregoitia et al. (2013) found that, at a global scale, there is no evidence to support that mammals in older lineages are at greater risk, which can be related to the cancellation of regional and clade effects among themselves. The same happens when we observe geographical causes of extinction: the range shift dynamic between the LGM and Holocene had also little influence on extinction risk. Although geographical range loss is used to assess species extinction risk (see IUCN, 2016, for example), historical range loss due to climate change had sparing influence on extinction when ecological correlates such as body size are accounted for.

We found that ecological factors, mediated by the allometric relations between body size and life history traits have the strongest influence in shaping extinction risk patterns of mammal species, both for extant and already extinct mammal taxa. Since we have only analyzed intrinsic putative factors of extinction risk, future works should benefit from the perspective of the interaction among body size and extrinsic factors such as biotic interactions. Global analysis such ours can mask regional effects of putative predictors of extinction risk as affirmed by Arregoitia et al (2013), so works in biogeographical and scales might be useful for guiding more specific actions for conservation.

Literature Cited

ALROY, John. Fossilworks: gateway to the paleobiology database. 2018. Available in: http//: fossilworks.org. Acessed in 2018.

ANGERT, A. L. et al. Do species’ traits predict recent shifts at expanding range edges?. Ecology Letters, v. 14, n. 7, pp. 677-689, 2011.

ARAÚJO, Miguel B.; NEW, Mark. Ensemble forecasting of species distributions. Trends in ecology & evolution, v. 22, n. 1, p. 42-47, 2007.

ARREGOITIA, Luis D. V.; BLOMBERG, Simon P.; FISHER, Diana O. Phylogenetic correlates of extinction risk in mammals: species in older lineages are not at greater risk. Proceedings of the Royal Society of London B: Biological Sciences, v. 280, n. 1765, p. 20131092, 2013.

34

ASHTON, K. G. Comparing phylogenetic signal in intraspecific and interspecific body size datasets. Journal of evolutionary biology, v. 17, n. 5, p. 1157-1161, 2004.

AUDZIJONYTE, Asta; KUPARINEN, Anna. The role of life histories and trophic interactions in population recovery. Conservation Biology, v. 30, n. 4, p. 734- 743, 2016.

BARBOSA, A. M., BROWN, J. A., JIMENEZ-VALVERDE, A., REAL, R. modEvA: Model Evaluation and Analysis. R package version 1.3.2, 2016. Available at: https://CRAN.R-project.org/package=modEvA

BENNETT, P. M.; OWENS, I. P. Variation in extinction risk among birds: chance or evolutionary predisposition?. Proceedings of the Royal Society of London B: Biological Sciences, v. 264, n. 1380, p. 401-408, 1997.

BININDA-EMONDS, Olaf R. P. et al. The delayed rise of present-day mammals. Nature, v. 446, n. 7135, p. 507, 2007.

BLACKBURN, Tim M. et al. The influence of spatial resolution on macroecological patterns of range size variation: a case study using (Aves: Psittaciformes) of the world. Journal of Biogeography, v. 31, n. 2, p. 285-293, 2004.

BÖHNING-GAESE, Katrin et al. Range size: disentangling current traits and phylogenetic and biogeographic factors. The American Naturalist, v. 167, n. 4, p. 555-567, 2006.

BOWMAN, J., JAEGER, J. A., FAHRIG, L. Dispersal distance of mammals is proportional to home range size. Ecology, v. 83, n. 7, pp. 2049-2055, 2002.

BREWER, Simon; JACKSON, Stephen T.; WILLIAMS, John W. Paleoecoinformatics: applying geohistorical data to ecological questions. Trends in Ecology & Evolution, v. 27, n. 2, p. 104-112, 2012.

CARD – The Canadian Archaeological Radiocarbon Database. Available from: http://www.canadianarchaeology.ca/. Accessed in: 15th January 2011.

CARDILLO, Marcel. Biological determinants of extinction risk: why are smaller species less vulnerable? In: Animal Conservation forum. Cambridge University Press, p. 63-69, 2003.

CARDILLO, Marcel; BROMHAM, Lindell. Body size and risk of extinction in Australian mammals. Conservation biology, v. 15, n. 5, p. 1435-1440, 2001.

CARDILLO, Marcel; MACE, Georgina; GITTLEMAN, John L.; JONES, Kate E.; BIEBLY, Jon; PURVIS, Andy. The predictability of extinction: biological and external correlates of decline in mammals. Proceedings of the Royal Society of London B: Biological Sciences, v. 275, n. 1641, p. 1441-1448, 2008.

CARDILLO, Marcel; MACE, Georgina; KATE, Jones; BININDA-EMONDS, Olaf L. P.; SECHREST, Wes; ORME, C. David L.; PURVIS, Andy. Multiple causes of 35

high extinction risk in large mammal species. Science, v. 309, n. 5738, p. 1239- 1241, 2005.

CHAMBERLAIN, Scott. Rgbif: Interface to the Global “Biodiversity” Information Facility API. R package version 0.9.9, 2017. Available from: https:// www. CRAN. R-project.org/ package=rgbif.

DAMUTH, John. Population density and body size in mammals. Nature, v. 290, n. 5808, p. 699, 1981.

DANTAS, Mário A. T.; SANTOS, Luciano V.; COZZUOL, Mario A. Quanto pesa um megamamífero extinto? Estimando o peso da megafauna do Quaternário da região intertropical brasileira. In: LANGER, Max C.; ALHALABI, Wafa A.; FERREIRA, Gabriel de S.; HERMANSON, Guilherme; MARSOLA, Júlio C. de A.; ONARY-ALVES, Sílvio Y (ed.). Paleontologia em Destaque: boletim informativo da Sociedade Brasileira de Paleontologia, ano 32, 2017.

DAVIDSON, Ana D.; HAMILTON, Marcus J.; BOYER, Alison G.; BROWN, James H.; CEBALLOS, Gerardo. Multiple ecological pathways to extinction in mammals. Proceedings of the National Academy of Sciences, v. 106, n. 26, p. 10702-10705, 2009.

DINIZ-FILHO, José Alexandre F.; TÔRRES, Natália M. Phylogenetic comparative methods and the geographic range size–body size relationship in new world terrestrial . Evolutionary Ecology, v. 16, n. 4, p. 351-367, 2002.

DIXON, Philip. VEGAN, a package of R functions for community ecology. Journal of Vegetation Science, v. 14, n. 6, p. 927-930, 2003.

FAITH, J. Tyler. and Holocene mammal extinctions on continental . Earth-Science Reviews, v. 128, p. 105-121, 2014.

FAURBY, Søren; SVENNING, Jens-Christian. A species-level phylogeny of all extant and late Quaternary extinct mammals using a novel heuristic-hierarchical Bayesian approach. Molecular phylogenetics and evolution, v. 84, p. 14-26, 2015.

FELSENSTEIN, Joseph. Phylogenies and the comparative method. The American Naturalist, v. 125, n. 1, p. 1-15, 1985.

FIELDING, Allan H.; BELL, John F. A review of methods for the assessment of predictions errors in conservation presence/absence models. Environmental Conservation, v. 24, n. 1, p. 38-49, 1997.

FLORES, David A.; BARQUEZ, Rubén M.; DÍAZ, María M. A new species of Philander Brisson, 1762 (Didelphimorphia, Didelphidae). Mammalian Biology- Zeitschrift für Säugetierkunde, v. 73, n. 1, p. 14-24, 2008.

36

FORTELIUS, M. New and Old Worlds database of fossil mammals (NOW), 2015. Available from: http://www.helsinki.fi/science/now/. Accessed in: 4th October 2017.

FRECKLETON, Robert P.; JETZ, Walter. Space versus phylogeny: disentangling phylogenetic and spatial signals in comparative data. Proceedings of the Royal Society B: Biological Sciences, v. 276, n. 1654, p. 21-30, 2008.

FRITZ, Susanne A.; PURVIS, Andy. Selectivity in mammalian extinction risk and threat types: a new measure of phylogenetic signal strength in binary traits. Conservation Biology, v. 24, n. 4, p. 1042-1051, 2010.

GASTON, Kevin; BLACKBURN, Tim. Evolutionary age and risk of extinction in the global avifauna. Evolutionary Ecology, v. 11, n. 5, p. 557-565, 1997.

GASTON, Kevin; BLACKBURN, Tim. Pattern and process in macroecology. John Wiley & Sons, 2000.

GBIF. Global Biodiversity Information Facility, 2017. Available from: https://www.gbif.org/. Data downloaded in: March 2016.

GOLICHER, Duncan et al. Pseudo-absences, pseudo-models and pseudo-niches: pitfalls of model selection based on the area under the curve. International Journal of Geographical Information Science, v. 26, n. 11, p. 2049-2063, 2012.

GRAHAM, R. W.; LUNDELIUS-Jr, E. L. FAUNMAP II: New data for North America with a temporal extension for the Blancan, Irvingtonian and early Rancholabrean. FAUNMAP II Database, version 1.0, 2010. Available from: http://www.ucmp.berkeley.edu/faunmap/. Acessed in: 15th August 2014.

GUÉNARD, Guillaume; LEGENDRE, Pierre; PERES‐NETO, Pedro. Phylogenetic eigenvector maps: a framework to model and predict species traits. Methods in Ecology and Evolution, v. 4, n. 12, p. 1120-1131, 2013.

GUÉNARD, Guillaume. MPSEM: Modeling phylogenetic signals using eigenvector maps. R package version 0.3-3. https://CRAN.R-project.org/package=MPSEM, 2017.

GUISAN, Antoine; ZIMMERMANN, Niklaus E. Predictive habitat distribution models in ecology. Ecological modelling, v. 135, n. 2-3, p. 147-186, 2000.

HAIR, Joseph F. et al. Análise multivariada de dados. Bookman Editora, 2009.

HILBERS, J. P. et al. An allometric approach to quantify the extinction vulnerability of birds and mammals. Ecology, v. 97, n. 3, p. 615-626, 2016.

HOGG, Alan G.; QUAN, Hua; BLACKWELL, Paul G.; NIU, Mu; BUCK, Caitlin E.; GUILDERSON, Thomas P.; HEATON, Timothy J.; PALMER, Johnathan G.; REIMER, Paula J.; REIMER, Ron W.; TURNEY, Christian S. M.;

37

ZIMMERMAN, Susan R. H. SHCal13 Southern Hemisphere calibration, 0– 50,000 cal BP. Radiocarbon, v. 55, n. 4, p. 1889-1903, 2013.

HUTCHINGS, Jeffrey A. et al. Life‐history correlates of extinction risk and recovery potential. Ecological Applications, v. 22, n. 4, p. 1061-1067, 2012.

IUCN. International Union of Conservation of Nature Red List of Threatened Species. Version 2017-3. Available from: http://www.iucnredlist.org/. Accessed in: 2nd December 2016.

JABLONSKI, David. The future of the fossil record. Science, v. 284, n. 5423, p. 2114- 2116, 1999.

JOHNSON, Christopher N. Determinants of loss of mammal species during the Late Quaternary ‘megafauna’ extinctions: life history and ecology, but not body size. Proceedings of the Royal Society of London B: Biological Sciences, v. 269, n. 1506, p. 2221-2227, 2002.

JOHNSON, Christopher N.; DELEAN, Steven; BALMFORD, Andrew. Phylogeny and the selectivity of extinction in Australian marsupials. Animal Conservation, v. 5, n. 2, p. 135-142, 2002.

JONES, Kate E. et al. PanTHERIA: a species‐level database of life history, ecology, and geography of extant and recently extinct mammals. Ecology, v. 90, n. 9, p. 2648-2648, 2009.

KOCH, Paul L.; BARNOSKY, Anthony D. Late Quaternary extinctions: state of the debate. Annual Review of Ecology, Evolution, and Systematics, v. 37, 2006.

KOZIOWSKI, J.; WEINER, J. Interspecific allometries are by-products of body size optimization. American Naturalist, v. 149, n. 2, p. 352-380, 1997.

KUHN, Tyler S.; MOOERS, Arne Ø.; THOMAS, Gavin H. A simple polytomy resolver for dated phylogenies. Methods in Ecology and Evolution, v. 2, n. 5, p. 427-436, 2011.

LEW, Daniel; PÉREZ-HERNÁNDEZ, Roger; VENTURA, Jacint. Two new species of Philander (Didelphimorphia, Didelphidae) from northern South America. Journal of Mammalogy, v. 87, n. 2, p. 224-237, 2006.

LIMA-RIBEIRO, Matheus S., DINIZ-FILHO, José A. F. Modelos ecológicos e a extinção da megafauna: clima e homem na América do Sul. São Carlos: Cubo, 2013a.

LIMA-RIBEIRO, Matheus S.; DINIZ-FILHO, José. A. F. American megafaunal extinctions and arrival: improved evaluation using a meta-analytical approach. Quaternary International, v. 299, p. 38-52, 2013b.

38

LIMA‐RIBEIRO, Matheus S. et al. Fossil record improves biodiversity risk assessment under future climate change scenarios. Diversity and Distributions, v. 23, n. 8, p. 922-933, 2017.

LIMA-RIBEIRO, Matheus S.; VARELA, Sara; GONZÁLEZ-HERÁNDEZ, Javier; OLIVEIRA, Guilherme de; DINIZ-FILHO, José A. F.; PETERSON, A. Townsend; TERRIBILE, Levi C. The ecoClimate Database, Available from: http://ecoclimate.org. Accessed in: 20th March 2018.

LISIECKI, Lorraine E.; RAYMO, Maureen E. A ‐Pleistocene stack of 57 globally distributed benthic δ18O records. Paleoceanography, v. 20, n. 1, 2005.

MAIORANO, Luigi et al. Building the niche through time: using 13,000 years of data to predict the effects of climate change on three tree species in . Global Ecology and Biogeography, v. 22, n. 3, p. 302-317, 2013. MARTINS, Wellington S. et al. SUNPLIN: simulation with uncertainty for phylogenetic investigations. BMC bioinformatics, v. 14, n. 1, p. 324, 2013.

MCDOWALL, R. M. Extinction and endemism in land birds. Tuatara, v. 17, n. 1, p. 1-12, 1969.

MCKINNEY, Michael L. Extinction vulnerability and selectivity: combining ecological and paleontological views. Annual Review of Ecology and Systematics, v. 28, n. 1, p. 495-516, 1997.

MEIJAARD, Erik; SHEIL, Douglas; MARSHALL, Andrew J.; NASI, Robert. Phylogenetic age is positively correlated with sensitivity to timber harvest in Bornean mammals. Biotropica, v. 40, n. 1, p. 76-85, 2008.

MITTERMEIER, Russell A.; WILSON, Don E.; RYLANDS, Anthony B. (Ed.). Handbook of the mammals of the world: primates. Edicions, 2013.

MULITZA, Stefan; BOLTOVSKOY, Demetrio; DONNER, Barbara; MEGGERS, Helge; PAUL, André; WEFER, Gerold. Temperature: δ 18 O relationships of planktonic foraminifera collected from surface waters. Palaeogeography, Palaeoclimatology, Palaeoecology, v. 202, n. 1, p. 143-152, 2003.

MYHRVOLD, Nathan P. et al. An amniote life‐history database to perform comparative analyses with birds, mammals, and reptiles: Ecological Archives E096‐269. Ecology, v. 96, n. 11, p. 3109-3109, 2015.

NOTT, M. P.; ROGERS, E.; PIMM, S. Extinction Rates: Modern extinctions in the kilo-death range. Current Biology, v. 5, n. 1, p. 14-17, 1995.

OKSANEN, Jari; BLANCHET, F. Guillaume; FRIENDLY, Michael; KINDT, Roeland; LEGENDRE, Pierre; MCGLINN, Dan; MINCHIN, Peter R.; O’HARA, R. B.; SIMPSON, Gavin L.; SOLYMOS, Peter; STVENS, M. Henry H.; SZOECS, Eduard; WAGNER, Helene. vegan: Community Ecology Package. R package version 2.5-1. https://CRAN.R-project.org/package=vegan. 2018.

39

PAGLIA, Adriano P.; FONSECA, Gustavo A. B. da; RYLANDS, Anthony B.; HERRMANN, Gisela; AGUIAR, Ludmilla R. S.; CHIARELLO, Adriano G.; LEITE, Yuri L. R.; COSTA, Leonora P.; SICILIANO, Salvatore; KIERULFF, Maria Cecilia M.; MENDES, Sérgio L.; TAVARES, Valéria da C.; MITTERMEIER, Russel A.; PATTON, James L. Lista Anotada dos Mamíferos do Brasil, 2ª edição. Occasional papers in conservation biology, v. 6, p. 76, 2012.

PARR, Cynthia S. et al. The encyclopedia of life v2: providing global access to knowledge about life on earth. Biodiversity Data Journal, n. 2, 2014.

PAVAN, Silvia E.; VOSS, Robert S. A revised subgeneric classification of short-tailed (Didelphidae: Monodelphis). American Museum Novitates, n. 3868, p. 1-44, 2016.

PBDB. The Paleobiology database, 2017. Available from: https://paleobiodb.org/.

PEARSON, Richard G; RAXWORTHY, Christopher J; NAKAMURA, Miguel; PETERSON, A. Townsend. Predicting species distributions from small numbers of occurrence records: a test case using cryptic geckos in . Journal of biogeography, v. 34, n. 1, p. 102-117, 2007.

PETERS, R. H. Ecological implications of body mass. Cambridge: Cambridge University Press, 1983.

PIMM, Stuart L. et al. The future of biodiversity. Science, v. 269, n. 5222, p. 347-350, 1995.

PINE, R. H.; HANDLEY-Jr, O. Genus Monodelphis (Burnett, 1830), p. 82-107, 2008. In: GARDNER, Alfred L. (ed.). Mammals of South America, volume 1: marsupials, xenarthrans, shrews, and bats. University of Chicago Press, 2008.

PURVIS, Andy; GITTLEMAN, John L.; COWLISHAW, Guy; MACE, Georgina M. Predicting extinction risk in declining species. Proceedings of the Royal Society of London B: Biological Sciences, v. 267, n. 1456, p. 1947-1952, 2000a.

PURVIS, Andy; JONES, Kate E.; MACE, Georgina M. Extinction. BioEssays, v. 22, n. 12, p. 1123-1133, 2000b.

PURVIS, Andy. Phylogenetic approaches to the study of extinction. Annual Review of Ecology, Evolution, and Systematics, v. 39, 2008.

QUENTAL, Tiago B.; MARSHALL, Charles R. Diversity dynamics: molecular phylogenies need the fossil record. Trends in Ecology & Evolution, v. 25, n. 8, p. 434-441, 2010.

R CORE TEAM. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Viena, Austria, 2016. Available from: https:// www.R-project.org/. Downloaded in 7th July 2017.

40

RANGEL, Thiago F. et al. Phylogenetic uncertainty revisited: Implications for ecological analyses. Evolution, v. 69, n. 5, p. 1301-1312, 2015.

RAUP, David M. The role of extinction in evolution. Proceedings of the National Academy of Sciences, v. 91, n. 15, p. 6758-6763, 1994.

REIMER, Paula J.; et al. IntCal13 and Marine13 radiocarbon age calibration curves 0– 50,000 years cal BP. Radiocarbon, v. 55, n. 4, p. 1869-1887, 2013.

RODRÍGUEZ-REY, Marta; PÉREZ-HERRANDO, Salvador; BROOK, Barry W.; SALTRÉ, Frédérik; ALROY, John; BEETON, Nicholas; BIRD, Michael I.; COOPER, Alan; GILLESPIE, Richard; JACOBS, Zenobia; JOHNSON, Christopher N.; MILLER, Gifford H.; PRIDEAUX, Gavin J.; ROBERTS, Richard G.; TURNEY, Chris S. M.; BRADSHAW, Corey J. A. A comprehensive database of quality-rated fossil ages for Sahul’s Quaternary vertebrates. Scientific data, v. 3, p. 160053, 2016.

SCHLOSS, C. A., NUÑEZ, T. A., LAWLER, J. J. Dispersal will limit ability of mammals to track climate change in the Western Hemisphere. Proceedings of the National Academy of Sciences, v. 109, n. 22, pp. 8606-8611, 2012.

SMITH, Felisa A.; LYONS, Kathleen S.; ERNEST, S. K. Morgan; JONES, Kate E.; KAUFMAN, Dawn; DAYAN, Tamar; MARQUET, Pablo A.; BROWN, James, H.; HASKELL, John P. Body mass of late Quaternary mammals. Ecology, v. 84, n. 12, p. 3403-3403, 2003.

STEKHOVEN, Daniel J.; BÜHLMANN, Peter. MissForest – non-parametric missing value imputation for mixed-type data. Bioinformatics, v. 28, n. 1, p. 112-118, 2011.

STUIVER, M.; REIMER, P. J.; REIMER, R. W. CALIB 7.1, 2018. Available from: http://calib.org/. Accessed in: 12th March 2018.

THE STAGE THREE PROJECT. Institute for Quaternary Research at Cambridge University, 2017. Available from: https://www.esc.cam.ac.uk/research/research- groups/research-projects/stage-three-project/stage-three-project-database- downloads. Accessed in: 5th October 2017.

TURVEY, Samuel T.; FRITZ, Susanne A. The ghosts of mammal’s past: biological and geographical patterns of global mammalian extinction across the Holocene. Philosophical Transactions of the Royal Society B: Biological Sciences, v. 366, n. 1577, p. 2564-2576, 2011.

VARELA, Sara; GONZÁLEZ-HERNÁNDEZ, Javier; SGARBI, Luciano F.; MARSHAL, Charles; UHEN, Mark D.; PETERS, Shanan; MCCLENNEN, Michael. paleobioDB: an R package for downloading, visualizing and processing data from the Paleobiology Database. Ecography, v. 38, n. 4, p. 419-425, 2015a.

41

VARELA, Sara; LIMA-RIBEIRO, Matheus S.; TERRIBILE, Levi Carina. A short guide to the climatic variables of the last glacial maximum for biogeographers. PloS one, v. 10, n. 6, p. e0129037, 2015b.

VARELA, Sara; LOBO, Jorge M.; HORTAL, Joaquín. Using species distribution models in paleobiogeography: a matter of data, predictors and concepts. Palaeogeography, Palaeoclimatology, Palaeoecology, v. 310, n. 3-4, p. 451-463, 2011.

VOSS, Robert S.; LUNDE, Darrin P.; JANSA, Sharon A. On the contents of Gracilinanus, with the description of a previously unrecognized clade of small Didelphid Marsupials. American Museum Novitates, p. 1-36, 2005.

VOSS, Robert S.; TARIFA, Teresa; YENSEN, Eric. An introduction to Marmosops (Marsupialia: Didelphidae), with the description of a new species from Bolivia and notes on the and distribution of other Bolivian forms. American Museum Novitates, p. 1-40, 2004.

WILMAN, Hamish; BELMAKER, Jonathan; SIMPSON, Jennifer; ROSA, Carolina de la; RIVADENEIRA, Marcelo M.; WALTER, Jetz. EltonTraits 1.0: Species‐level foraging attributes of the world's birds and mammals. Ecology, v. 95, n. 7, p. 2027-2027, 2014.

ZACAÏ, Axelle et al. Phylogenetic conservatism of species range size is the combined outcome of phylogeny and environmental stability. Journal of Biogeography, v. 44, n. 11, p. 2451-2462, 2017.

42

Supplementary Information – Chapter 01

A B

y

Frequenc

Normalized mean root squared error Diferences in mean

C D

y

equenc

Fr

Diferences in median Diferences in standard deviation

Figure S1. Evaluation metrics for the performance of the imputation process. (A) Normalized mean root squared error. (B) Differences between the mean of body mass in grams of the 1000 datasets with imputed values and the mean of original dataset without these values. (C) Differences between the median of body mass in grams of the 1000 datasets with imputed values and the median of original dataset without these values. (D) Differences between the standard deviation of body mass in grams of the 1000 datasets with imputed values and the standard deviation of original dataset without these values. All species without body mass were extinct and there is evidence that extinction phenomenon is concentrated in some branches of the phylogeny [i.e. non-random, see Purvis et al. (2000); Purvis (2008); Fritz and Purvis (2010) for a review]. Given that, we argue that this relative high value of standardized error could not be avoided since we only had phylogeny and life history as predictors of body mass in mammals and we have no information about life history characteristics for fossil species.

43

b p s bp ps bs a R

Figure S2. Median percentage of deviance for unique and shared predictors of extinction risk. Unique deviances are: b – body mass (grams); p – phylogenetic age (length of the terminal branch) and s – range shift (number of grid cells). Shared deviances are: bp – body mass and phylogenetic age; ps – phylogenetic age and range shift; bs – body mass and range shift and a – deviance shared by all predictors. R – Residual deviance, i.e. not explained by any predictor. 1000 replicates for each portion of deviance. Extinct species not considered. 1000 replicates for each portion of deviance.

44

B P

0.014 2.83 -0.01

Frequency 0.0002 0.003 0.0003

Body Mass -0.032 Phylogenetic Age

Body size and S R phylogenetic age (m=0.014)

Phylogenetic age and ncy

range shift (m=0.0003) ue Body size and range

shift (m=-0.003) Freq Deviance shared by all predictors (m=0.0002)

Shared Deviances Range Shift Residual

Figure S3. Percentage of unique and shared deviances obtained by the generalized linear models when extinct species are removed. Unique deviances are: B – body mass (grams); P – phylogenetic age (length of the terminal branch) and S – range shift (number of grid cells). The portion of deviance not explained by any variable or set of variables represents 97.2±0.051% in average of the total deviance of the model. The unique best predictor of extinction risk in mammals is body size (2.83±0.015%), followed by the shared deviance between body size and phylogenetic age (0.15±0.10%, both values in bold). 1000 replicates for each portion of deviance.

45

Table S1. Extinct species inserted as phylogenetic uncertain taxa in Kuhn’s et al. (2011) one hundred-phylogeny block and the respective most recent common ancestor (Alroy, 2018), in diverse taxonomic resolutions. The insertion generated 1000 phylogenies used in this work.

Phylogenetically Uncertain Taxon Most Derived Consensus Clade (PUT) (MCDA) Plesiorycteropus madagascariensis Mammalia patachonica Mammalia bahiense Mammalia Neolicaphrium recens Mammalia larensis Mammalia platensis Mammalia Cynotherium sardous Canidae Canidae Protocyon troglodytes Canidae serum Homotherium latidens Felidae Miracinonyx trumani Felidae fatalis Felidae Brachyprotoma obtusata Arctodus simus Ursidae tarijense Ursidae Arctotherium wingei Ursidae Capromeryx minor Stockoceros conklingi Antilocapridae bombifrons priscus Bovidae Megalovis guangxiensis Bovidae antiquus Bovidae Soergelia minor Bovidae Spirocerus kiakhtensis Bovidae hesternus macrocephala Camelidae Hemiauchenia paradoxa Camelidae major Camelidae Palaeolama mirifica Camelidae Agalmaceros blicki Cervidae ultra Cervidae Cervidae giganteus Cervidae brachyceros Cervidae Morenelaphus lujanensis Cervidae Navahoceros fricki Cervidae Paraceros fragilis Cervidae

46

Praemegaceros cazioti Cervidae Sangamona fugitiva Cervidae Kolpochoerus majus Suidae Metridiochoerus compactus Suidae seguini Propraopus sulcatus Dasypodidae Propraopus punctatus Dasypodidae hopei Protemnodon nombe Macropodidae Protemnodon tumbuna Macropodidae atlas Macropodidae Sthenurus stirlingi Macropodidae Sthenurus tindalei Macropodidae Troposodon minor Macropodidae Borungaboodie hatcheri Caloprymnus campestris Potoroidae oscillans Potoroidae gigas Vombatidae Nesiotites similis Soricidae Aztlanolagus agilis Leporidae Megalibgwilia ramsayi Tachyglossidae principale Hippidion devillei Equidae Coelodonta antiquitatis Rhinicerontidae hemitoechus Rhinicerontidae Stephanorhinus kirchbergensis Rhinicerontidae Diabolotherium nordenskioldi rodens Megalonychidae Neocnus comes Megalonychidae Parocnus browni Megalonychidae bambuiorum brasiliensis Atelidae floresiensis Pachylemur insignis Lemuridae Mammuthus exilis hyodon waringi Proboscidea trigonocephalus Proboscidea Stegodon orientalis Proboscidea Stegodon florensis Proboscidea ohioensis Castoridae Neochoerus aesopi Caviidae Pliomys lenki Cricetidae

47

Microgale macpheei Microgale dirus Canis Cryptoprocta Leopardus amnicola Leopardus Tremarctos floridanus Tremarctos Antidorcas australis Antidorcas Antidorcas bondi Antidorcas Bos primigenius Bos Bubalus palaeokerabau Bubalus Damaliscus niro Damaliscus Damaliscus hypsodon Damaliscus collinum Bovidae Gazella atlantica Gazella Gazella tingitana Gazella leucophaeus Hippotragus Alces scotti Alces sivalensis Hexaprotodon lemerlei Hippopotamus Hippopotamus madagascariensis Hippopotamus Phanourios minutes Hippopotamus Catagonus stenocephalus Catagonus Dasypus ferragus Macropus Macropus greyi Macropus Onychogalea lunata Onychogalea Thylogale christenseni Thylogale Vombatus hacketti Vombatus Ochotona whartoni Ochotona Zaglossus hacketti Zaglossus Perameles eremiana Perameles hydruntinus Equus Equus semiplicatus Equus Tapirus augustus Tapirus Tapirus copei Tapirus antiquus Elephas Elephas cypriotes Elephas Elephas iolensis Elephas Elephas mnaidriensis Elephas Hystrix refossa Hystrix Conilurus albipes Conilurus Pseudomys gouldii Pseudomys Hypogeomys australis Hypogeomys clavicaudatus

48

Glyptodon clavipes Cingulata reticulatus Cingulata floridanum Cingulata Glyptotherium cylindricum Cingulata euphractus Cingulata tuberculatus Cingulata occidentalis Cingulata Holmesina paulacoutoi Cingulata Pampatherium humboldti Cingulata Pampatherium typum Cingulata optatum Maokopia ronaldi Diprotodontia Prolagus sardus Lagomorpha Chaeropus ecaudatus Peramelemorphia laurillardi tarijense Pilosa cuvieri Pilosa robustum Pilosa armatus Pilosa darwinii Pilosa chiliensis Pilosa leptocephalum Pilosa deformis Pilosa maquinense Pilosa majori Primates Archaeolemur edwardsi Primates stenognathus Primates edwardsi Primates Megaladapis madagascariensis Primates ingens Primates

49

Table S2. Atmosphere-ocean general circulation models (Aogcm’s) with references of their respective modeling group(s) and institute ID’s.

Model Name Institute ID Modeling Group(s)

CCSM RSMAS University of Miami - RSMAS Agency for Marine- Earth Science and Technology, Atmosphere and Ocean Research MIROC MIROC Institute (The University of Tokyo), and National Institute for Environmental Studies

50

Table S3. Bioclimatic variables selected to represent the climatic characteristics of Wallacean biogeographical realms. bio1 – Annual mean temperature; bio2 – Mean diurnal range (Mean of monthly (maximal temperature – minimal temperature)); bio3 – Isothermality ((bio2 / bio7) * 100); bio4 – Temperature seasonality (standard deviation * 100); bio5 – Maximum temperature of warmest month; bio7 – Temperature annual range (bio5 - bio6); bio13 – Precipitation of wettest month; bio14 – Precipitation of driest month; bio15 – Precipitation seasonality (coefficient of variation); bio16 – Precipitation of wettest quarter; bio17 – Precipitation of driest quarter.

Biogeographical Realms Bioclimatic variables

African bio1, bio2, bio16, bio17 Australian bio1, bio2, bio16, bio17 Indo-malayan bio1, bio4, bio16, bio17 Neartic bio5, bio7, bio13, bio15 Neotropic bio1, bio2, bio3, bio16, bio17 Paleartic bio2, bio4, bio5, bio13, bio14

51

Table S4. Algorithms, modeling methods and the type of data used to build the ENM’s.

Method Statistical feature Data type R package

Euclidean distance Environmental Presence-only stats1 distance Bioclim Climatic envelope Presence-only dismo2 Domain Environmental Presence-only dismo2 (Gower distance) distance Maximum entropy Machine learning Presence/background dismo2 (Maxent) Support vector Machine learning Presence/background kernlab3 machine (SVM)

1R Core Team (2017); 2Hijmans et al. (2017); 3Karatzoglou et al. (2004)

Literature Cited

ALROY, John. Fossilworks: gateway to the paleobiology database. 2018. Available in: http//: fossilworks.org. Acessed in 2018.

HIJMANS, Robert J.; PHILLIPS, Steven; LEATWICK, John; ELITH, Jane. dismo: Species distribution modeling. R package version 1.1-4. https://CRAN.R- project.org/package=dismo, 2017.

KARATZOGLOU, Alexandros et al. kernlab-an S4 package for kernel methods in R. Journal of statistical software, v. 11, n. 9, p. 1-20, 2004.

KUHN, Tyler S.; MOOERS, Arne Ø.; THOMAS, Gavin H. A simple polytomy resolver for dated phylogenies. Methods in Ecology and Evolution, v. 2, n. 5, p. 427-436, 2011.

R CORE TEAM. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Viena, Austria, 2016. Available from: https:// www.R-project.org/. Downloaded in 7th July 2017.

52

CAPÍTULO 2

53

Landmasses prevent habitat tracking and challenge terrestrial mammals with extinction risk

Vinícius Silva Reis1,2, Levi Carina Terribile1, Matheus Souza Lima-Ribeiro1

1 – Laboratório de Macroecologia, Universidade Federal de Jataí, Jataí, Brasil; 2 – Instituto de Ciências Biológicas V, Laboratório de Ecologia Teórica e Síntese, Universidade Federal de Goiás, Goiânia, Brasil.

Abstract

Given the change on climate worldwide, species might be able to change the geographical distribution in order to cope with these changes. However terrestrial species are limited by the physical space, namely the landmasses, available for them to spread. Here we test the hypothesis that the landmasses act as a physical barrier for terrestrial mammal species to shift their geographical ranges as well as to track suitable habitats for survival. We used species distribution modeling techniques to generate information for species range shift and displacement in a potential scenario (without landmasses) and a realized scenario (with landmasses). We combined this information with the threat status of IUCN, shared in three categories: Not-threatened, threatened and extinct species. We found that despite the landmasses constraint species to shift the size of the distribution, it is very unlikely that this process influence extinction risk. In the other hand, landmasses limit the capacity of species to track suitable habitats for survival and this limitation influences extinction risk. Threatened and extinctic species have greater disparities between potential and realized habitat tracking compared with not-threatened ones. The spatially limited nature of the landmasses act as a horizontal large-scale driver of extinction risk via the restriction of species geographic distribution displacement.

Introduction

Species extinction risk is linked to a set of non-exclusive and interactive factors such as reproductive cycles, body size, level of exploitation and geographical distribution (Cardillo et al., 2005; Davidson et al., 2009; Purvis et al., 2000). The geographical ranges greatly vary among and within biogeographical regions across the tree of life. The species range size is primarily determined by intrinsic factors limiting energetic profit (Brown; Maurer, 1987, 1989) and can be restricted by secondary factors such as climate conditions (Coristine; Kerr, 2015), biotic interactions (Pigot; Tobias, 2013), niche breadth (Slatyer et al., 2013) and dispersal capability (Böhning-Gaese et al., 2006) of the species (see review in Brown et al., 1996 and Gaston, 2009). Together, such limiting factors entails a clear pattern of range sizes, with most species distributing across narrow geographical ranges (Brown, 1995; Gaston; Blackburn, 2006).

54

Although not representing a linear relationship (Gaston, 2003), the species extinction risk is generally negatively related with their range sizes (Jablonski, 2005; Payne; Finnegan, 2007). Larger geographical distributions might encompass higher number of individuals and the probability of extinction may be reduced by the presence of conspecifics (Allee effect, Stephens and Sutherland, 1999). Conversely, the species might not be able to maintain minimum viable populations in reduced geographic ranges and then extinguish.

According to Brown and Maurer (1987, 1989), the minimum viable population size is the main factor regulating the relation between the range size and body size of species, which constrain the species in a triangular-shaped envelope across that bivariate space. Larger-bodied species needs broader geographical ranges to acquire enough energy for the maintenance of viable populations, whereas smaller-bodied species can occupy both large or small geographical distribution. Similarly, Lima-Ribeiro et al. (2014) also highlighted the role of the range shift in shaping species constraints through the last ice age. Larger-bodied species were challenged to track suitable habitats for long distances, whereas smaller-bodied species tracked their habitat for both short and long distances. These macroecological envelopes suggest a process of species-level selection acting upon functional constraints, such that species closer to the inferior boundary of the triangular-shaped constraint envelope are more prone to extinction (Diniz-Filho, 2004; Lima-Ribeiro et al., 2014).

Considering a climate change-scenario, species should be able to move the geographical distribution over space and track more favorable climate conditions for survival, avoiding extinction (Gould; Eldredge, 1993; Chen et al., 2011; Pecl et. al., 2017). For terrestrial species, the space available for shifts on geographical distribution is ultimately the size of the landmass inhabited by the species. Then, the size of the landmasses might represent a physical barrier for terrestrial species to dispersal towards suitable conditions for survival. If suitable climatic conditions extrapolate continental boundaries, terrestrial species will be unable to follow such conditions due to the land- ocean limit. Inside landmasses, species dispersal might be limited by the internal margins of the biogeographical realms, that are based on mountain chains and tectonic disruptions (Ficetola et al., 2017).

55

Here, we analyzed the role of the landmasses as dispersal barriers in determining the terrestrial mammal’s extinction risk at global scale. Our hypothesis is that the landmasses act as a physical barrier for terrestrial species to shift their geographical ranges as well as to track suitable habitats for survival. If the landmasses limited species to shift the range and track suitable habitats for survival, with a consequent geographical range reduction, we expect that threatened species might have been more constrained by the landmasses than the not-threatened ones. We define here range shift to be the change in the size of the geographical distribution. Habitat tracking sensu Eldredge and Gould (1972) is the distance that the geographical range displaces as a consequence of following suitable climatic conditions for survival. We aim to shed light on large scale determinants of extinction risk of the species accounting for macroecological traits such geographical range size. We used terrestrial non-volant mammals as a model group. Mammals are relatively well known, with extant and extinct species across all continents encompassing a variety of body and geographical range sizes (Lyons; Smith, 2013).

Material and Methods

Data sampling

We gathered georeferenced occurrence records for all mammal species worldwide from multiple online datasets and timeframes (Table 1). All occurrence records were extensively revised and only those with accepted species level taxonomy were maintained. For every species, we spatially filtered duplicated records into the same pixel with 0.5o resolution (lat/long), and maintained only aged fossil records. To obtain the calendrical ages, the fossil occurrences dated by radiocarbon were calibrated using the CALIB Radiocarbon Calibration software (version 7.1, Stuiver et al., 2018) with the IntCal13 calibration curve (Reimer et al., 2013) or the SHCal13 curve (Hogg et al., 2013) for records from northern or southern hemispheres, respectively. After taxonomic, spatial and age filters, we obtained occurrence records for 2880 mammals, including 154 late-Quaternary extinct species.

56

Table 1. Datasets used to assemble modern and fossil occurrence records of mammal species for ecological niche modeling.

Source Species Spatial Extent Reference

The Global Extant Global GBIF (2017); Biodiversity Package “rgbif”, Information Facility version 0.9.9 (Chamberlain, 2017). The Paleobiology Extinct/Extant Global PBDB (2017); Database Package “paleobioDB”, version 0.5.0 (Varela et al. 2015a) The FosSahul Extinct/Extant Rodríguez-Rey et Database al. (2016) The New and Old Extinct/Extant and Fortelius (2015) Worlds Database North America The Stage Three Extinct/Extant Eurasia The Stage Three Project Project (2017) Faith (2014) Extinct Africa Faith (2014) FaunMap Database Extinct/Extant North-America Graham and Lundelius (2014) Card Database Extinct/Extant North-America Card (2011) Lima-Ribeiro et al. Extinct/Extant South-America Lima-Ribeiro et al. (2013) (2013)

To build the ecological niche models, we obtained a set of 19 bioclimatic variables from ecoClimate database (www.ecoclimate.org; Lima-Ribeiro et al. 2015) for present, mid-Holocene (6 ky) and Last Glacial Maximum (LGM, 21 ky). We performed a cluster analysis and used K-means criterion to select two atmosphere-ocean coupled global circulation models (AOGCMs) that maximize uncertainty among climate simulations at LGM (see details in Varela et al. 2015), namely: CCSM and MIROC (Table S1). Together, these climatic layers represent key aspects of climate dynamics over the last glacial cycle across both oceans and landmasses (Lima-Ribeiro et al. 2015).

57

To avoid multicollinearity among predictors, we performed factorial analyses with varimax rotation and selected the bioclimatic variables with the greatest loadings within each factor. The number of factors was determined from scree plot and variables were independently selected for each Wallacean biogeographical realm (see Wallace 1876) (see the set of selected variables in Table S3). The set of selected variables kept the variance from all climatic characteristics across biogeographical realms, with minimal collinearity among predictors. To ensure comparable predictors when modeling species distribution, we standardized the bioclimatic variables to range between 0 and 1 using the function “decostand” from R-package “vegan” (Dixon, 2003; Oksanen et al., 2018).

Given that the age of fossil records extrapolates the specific periods of LGM and mid-Holocene, we interpolated the climate conditions at every 1 ky (1,000 years) by using the δ18 Oxygen curve from Lisiecki and Raymo (2005) as a covariate. The δ18 Oxygen curve represents the ratio between the 18O and 16O stable isotopes present in biogenic components such as invertebrate marine shells and is useful to detect temperature dynamics over geological time (Mulitza et al., 2003). Because the accuracy of interpolated climates decreases in earlier periods, we eliminated all fossil records aged older than 700 ky BP.

Paleodistribution modeling

We used ecological niche models (ENM) to map geographical distributions of 2880 living and extinct mammals throughout the last ice age. To consider methodological uncertainty across ENMs, we used four modelling methods based on distinct data sets and statistical features: Bioclim, Domain (Gower distance), Maximum entropy (Maxent) and Support Vector Machine (SVM) (see details in Table S2). Because absence data are conceptually intangible for extinct species, and we aimed assessing range dynamics based on potential distributions instead of realized ones, presence-absence methods were not considered here (Jiménez-Valverde et al. 2008). Yet, the machine learning methods based on presence-background data were fitted using the simplest configuration for classification tasks, including linear kernel function with probabilistic output for SVM and only the linear feature to combine climatic predictors in Maxent, with logistic output. Besides attain our specific goals, using simple methods

58

and parameterizations is part of good practices when modelling ecological niche using fossil records (Varela et al., 2011).

To build ENMs, we combined climate conditions at both fossil and modern occurrences of species, and predicted the potential distributions for living and extinct mammals onto LGM and mid-Holocene. Since predicting climate spaces closer to species fundamental niches, the multi-temporal calibration approach has been recommended to access range dynamics over time (Maiorano et al., 2013; Lima-Ribeiro et al., 2017). To minimize computational effort, we considered all biogeographical realms from which each species occurs as the background area to predict its potential distribution. From background, we sampled 1000 random points to fit presence- background methods (Maxent and SVM).

We evaluated the ENMs by using only modern occurrences (fossil records were omitted from model evaluation and ENMs for extinct species were not evaluated) in two categories: 1) species with number of occurrences ranging from 5 to 25 had their ENMs evaluated with the jackknife procedure known as leave-one-out (see Pearson et al. 2007); 2) ENMs for species with samples larger than 25 occurrences were evaluated by splitting the modern occurrences into 75% for training and 25% for testing, with 20 repetitions to minimize effects from spurious partitions. In both categories, we computed the “D” statistics proposed by Pearson et al. (2007), as follows:

D = ∑ Xi (1 - Pi)

where Pi is the proportion of the background area predicted as species presence, and Xi is the success rate in correctly predicting the testing occurrences. For species in the first category (5-25 occurrences), Xi represents a binary rate informing success (Xi = 1) or failure (Xi = 0) of ENMs to predict the species occurrences in the leave-one-out procedure. For species in the second category (> 25 occurrences), Xi represents the True Positive Rate (TPR), the proportion of testing occurrences correctly predicted by ENMs (TPR continually varies from 0 to 1; see Fielding and Bell, 1997). In both cases, D provides a useful measure of predictive ability (see details in Pearson et al. 2007). To compute success rates, we used the lowest presence threshold (LPT) criterion, the lowest predicted value of climate suitability among the training occurrences, to obtain binary maps and confront ENM predictions against testing presences.

59

Finally, we combined ENMs from distinct methods and AOGCMs in an ensemble approach (Araújo; New, 2007) to map the consensual distribution of species at LGM and mid-Holocene. By using D-statistic as weights, consensus maps were obtained by weighted averaging the 12 predictions (4 methods * 3 AOGCMs) for each species in each period. As ENMs from extinct species were not evaluated, their ensembles were computed by arithmetic mean among predictions. Before averaging, we standardized the predicted suitabilities of all ENMs to vary between 0 and 1. To avoid information loss, we yet considered the species with small number of occurrences (< 5 records) in our analyses. For these species, we used only Euclidean distance method to predict its potential distributions, and ENMs were not evaluated. These ENMs were combined by arithmetically averaging the predictions from distinct AOGCMs.

Range shift and displacement

To evaluate if landmasses acted as dispersal barriers over the last ice age determining the terrestrial mammal’s extinction risk, we analyzed the two main geographical responses a species exhibit when facing climate change: the change in the size of the geographic distribution, the range shift, and spatial the displacement of the distribution, the so-called habitat tracking. We measured both species range shift and displacement from LGM to mid-Holocene under two scenarios: 1) restricting range dynamics to the interior of landmasses, likely as species realized in the past (hereafter, realized scenario), and 2) enabling unlimited range dynamics throughout the Earth’s surface with no continental barrier (hereafter, potential scenario). Considering terrestrial species, the potential scenario simulates the geographical range dynamics that could be achieved without any physical barrier imposed by the geographical limits of landmasses.

To compute the species range shift and displacement over the last ice age, we first transformed the consensus maps into binary maps to estimate the species range size in each period. We applied a general approach to find the threshold that maintain species distributed over 50% of background (the occupied biogeographical realms) at LGM, which was used to binarize all consensus maps (LGM and mid-Holocene, for both realized and potential scenarios). Because we aim evaluating range shift rather than realized range sizes in the past, this threshold criterion is advantageous, as it guarantees equal area (50% of background) to species either increase or reduce geographical ranges 60

over time, and avoid spurious errors such as extinguishing living species by getting extreme thresholds. From binary maps, we measured the species range size as the number of occupied grid cells, and then estimated the range shift for both realized and potential scenarios as the difference between range sizes at mid-Holocene and LGM. Range shift measures the magnitude of species geographical range expansion or reduction over time; positive range shifts indicate that species expanded its range, whereas negative values indicate range contraction.

To estimate range displacement, we applied the bioclimatic velocity reasoning from Serra-Díaz et al. (2014) to compute the distance by which the species had tracked its suitable habitats across space over the last ice age. The bioclimatic velocity measures the species exposure to climate change and benefits from inclusion of species-specific climatic tolerance to deal with habitat tracking via ecological niche models (see Carroll et al., 2015, for a comparison between biotic and climate change velocity metrics). With binary maps from each species, we calculated the distance of every occupied grid-cell in the LGM to the nearest occupied grid-cell in the mid-Holocene, and then added the bioclimatic distances cell-by-cell weighted by its respective climatic suitability throughout the background, as follows:

RD = ∑ di Si

th where di and Si are the bioclimatic distance and climatic suitability for i grid-cell from the background. Thus, species were more exposed to climate change across the more suitable habitats (grid-cells) that displaced for longer distances across space and time.

Data analyses

We explored the role of landmasses as geographical barriers increasing species extinction risk by comparing range shift and displacement between realized and potential scenarios. We estimated the magnitude of dispersal limitation imposed by continental barriers as the difference of range shift and displacement for each species between realized and potential scenarios. Then, we tested if threatened species had higher limitations than non-threatened species. To categorize the species in threat status, we considered the IUCN Red List of Threatened Species assessment (IUCN, 2018) and converted them into three categories: 1) Not Threatened (NT), including species

61

originally classified as least concern and near threatened; 2) Threatened (T), including vulnerable and endangered species and 3) Extinct (EX), including species critically endangered, extinct in the wild and extinct (mainly ). If the landmasses exert a restriction on changes of geographical distribution, we expect that the potential range shift and the potential habitat tracking are in average greater than the realized ones. If the limitation imposed by the landmass boundaries influences extinction risk of species via range shift and habitat tracking, we also expect that the difference between potential and realized scenarios of range shift and habitat tracking is greater for threatened and extinct species than for not-threatened species. Furthermore, we expect negative correlations between potential and realized scenarios of range shift and habitat tracking, with a higher negative slope for threatened and extinct species than for not-threatened ones. We carried out all the statistical analysis in R program (R Core Team, 2017).

When considering species as sample units, it is worth considering the usage of phylogenetic methods for hypothesis testing (see Garamszegi, 2014). Phylogenetic comparative methods are useful to avoid high probability of Type 1 error due to sample phylogenetic non-independence (Blomberg et al., 2003, Revell et al., 2008). Although we used basically geographic range characteristics to test whether and how continents affect extinction risk of mammals, we opted not to use such set of comparative methods, for there is not a cohesive body of evidence affirming that phylogenetic close species have more similar geographic ranges than expected by chance: Freckleton and Jetz (2009) found influence of phylogenetic dependence on range size among species of three mammal orders and Blackburn et al. (2004) found the same dependence for one bird order. In the other hand, Böhning-Gaese et al. (2006) and Gaston and Blackburn (1997) demonstrated the range size to have a labile behavior among species, with weak evidence for phylogenetic structure in such trait. More recently, Zacaï et al. (2017) affirmed that a pair of phylogenetically related species would resemble range sizes depending on the environmental stability of the regions species occur. They argue that phylogenetic structure of range size is an outcome of intrinsic and extrinsic factors, including environmental stability.

62

Results

All range dynamic metrics, range size, range shift and habitat tracking were higher for the potential than the realized scenario (paired t-test - Range size: t=70.37, df=2879, p-value<0.001; Range shift: t=67.85, df=2879, p-value<0.001; Habitat tracking: t=146.09, df=2879, p-value<0.001; Figure 1).

(A) (B) (C)

ge size ge

ran

habitat tracking habitat

shift

mof th

Logarithm of Logarithm

Logarithm of absolute range range absolute of Logarithm Logari Realized Potential Realized Potential Realized Potential

Figure 1 – Geographic range size (A), range shift (B) and habitat tracking (C) from realized and potential scenarios.

The difference between potential and realized range shift did not differ among not-threatened, threatened and extinct mammals (F=1.8, df=2, p-value=0.15, Figure 2). Furthermore, although 61% of the species presenting loss of geographic distribution from the LGM to mid-Holocene were classified as extinct, this association was not significant (χ2=1.84, df=2, p-value=0.40, Table 6).

nge shift nge

a

r

in Difference

NT T EX 63 Threat categories

Figure 2 – Differences between potential and realized range shift across the threat categories. Not threatened (NT,), Threatened (T) and Extinct (EX).

Table 6. Number of species that presented range expansion (positive range shift) or range contraction (negative range shift) in both realized and potential scenarios. Range Expansion Range Contraction

With Without With Without Landmasses Landmasses Landmasses Landmasses Not threatened 778 2124 1369 23 Threatened 129 384 258 3 Extinct 72 186 114 0

Species classified in distinct threat categories differed regarding the mean difference between potential and realized habitat tracking (F=12.7, df=2, p-value<0.001, Figure 3). Threatened and extinct species were significantly more limited to track their habitats displacing over oceans across the last ice age than not-threatened ones (planned comparison NT vs. T/EX: F=24.27, df=1, p<0.001).

t tracking t

ce in habita in ce

Differen

NT T EX Threat categories

Figure 3 - Differences between potential and realized habitat tracking for species classified as not-threatened (NT), threatened (T) and extinct (EX).

64

By considering a quantile regression approach, the relation between potential and realized range shift did not present any negative slope for not threatened species (β=0.17, p-value=0.02), threatened (β=0.42, p-value<0.001) and nor for extinct ones (β=0.37, p-value<0.001, Figure 4). Conversely, realized habitat tracking decreased with growing potential habitat tracking for extinct (β=-3.95, p-valor<0.001), threatened (β=- 2.57, p-value=0.02) and not-threatened species (β=-1.86, p-valor<0.0001, Figure 5).

shift range

alized alized

Logarithm ofre Logarithm

Logarithm of potential range shift Figure 4 - Relationship between realized range shift and potential range shift. Blue circles represent not-threatened (NT), yellow circles represent threatened (T) and the filled red circles represent extinct species (EX).

tracking habitat

zed

reali

of ithm Logar

Logarithm of potential habitat tracking 65

Figure 5 - Relationship between realized habitat tracking and potential tracking. Blue circles represent not threatened species, yellow circles represent threatened species and the filled red circles represent critically endangered and extinct species. (β=-1.98, p-valor<0.001).

Discussion

Our findings highlight that landmasses limit both geographical range size and range shift, although it is unlikely that such processes influence extinction risk of mammal species. However, landmasses influence extinction risk by constraining species to track suitable habitats for survival. Extinct and threatened species moved shorter distances in relation to the distances they could have coursed through in order to survive from climate change during Quaternary due to the physical barrier imposed by the landmasses.

Despite the average loss of geographic range size imposed by the landmasses since the LGM, this may not be claimed as a mechanism that influence extinction risk of species. Processes occurring within the landmasses might have more importance on triggering range loss and influencing extinction probability. It is possible to list regional environmental characteristics (Collen et al., 2011) such as peculiarities in landscape and land-use scenarios (Crooks et al., 2017; Powers; Jetz, 2019), landscape features and its interaction with animal behavior (Fordham et al., 2014) and the synergetic effect among them (González-Suárez; Revilla, 2014) as being responsible for changes in range size within landmasses. Besides that, DiMarco and Santini (2015) affirm that human activity has a major power of shaping range size of mammals. Thus, it is reasonable to envisage that such set of factors are prone to change from one landmass to another, inflating the complexity of interactive factors shaping range shift phenomenon.

Extinct and threatened mammals could follow only partially the distances necessary to track suitable habitats due to physical limitation imposed by continents. The relation we describe here between lack of space to spread and extinction risk in a is analogous to what has been related for some mountaintop biotas: As climate changes, adequate conditions for survival and reproduction shift upwards, forcing communities to follow such displacement in order to survive (Forero- Medina et al., 2011; Rumpf et al., 2019). At some point of climate change, it is likely that no more land area with suitable environmental conditions might be available for

66

species to colonize, enhancing probability of extinction (Freeman et al., 2018). This represents a vertical space constraint for species habitat tracking. In the case of our study, we document a horizontal space constraint for habitat tracking, in which species that are more spatially constrained by the landmasses are more prone to get extinct.

Our study shed light on large-scale determinants of extinction risk. The influence of landmasses on shaping range shift and habitat tracking of species should be seen as a mechanism of species level selection in which species capable to follow suitable habitats for survival despite the landmass-constraints are more prone to survive to climate change. Future developments should focus on the power of biological traits (eg. phylogenetic history) to predict which species are going to have more probability of being extirpated given the landmass-constraint for dispersal.

Literature Cited

ARAÚJO, Miguel B.; NEW, Mark. Ensemble forecasting of species distributions. Trends in ecology & evolution, v. 22, n. 1, p. 42-47, 2007.

BLACKBURN, Tim M. et al. The influence of spatial resolution on macroecological patterns of range size variation: a case study using parrots (Aves: Psittaciformes) of the world. Journal of Biogeography, v. 31, n. 2, p. 285-293, 2004.

BÖHNING-GAESE, Katrin et al. Range size: disentangling current traits and phylogenetic and biogeographic factors. The American Naturalist, v. 167, n. 4, p. 555-567, 2006.

BROWN, James H.; MAURER, Brian A. Evolution of species assemblages: effects of energetic constraints and species dynamics on the diversification of the North American avifauna. The American Naturalist, v. 130, n. 1, p. 1-17, 1987.

BROWN, James H.; MAURER, Brian A. Macroecology: the division of food and space among species on continents. Science, v. 243, n. 4895, p. 1145-1150, 1989.

BROWN, James H. et al. Macroecology. University of Chicago Press, 1995.

BROWN, James H.; STEVENS, George C.; KAUFMAN, Dawn M. The geographic range: size, shape, boundaries, and internal structure. Annual review of ecology and systematics, v. 27, n. 1, p. 597-623, 1996.

CARD – The Canadian Archaeological Radiocarbon Database. Available from: http://www.canadianarchaeology.ca/. Accessed in: 15th January 2011.

67

CARDILLO, Marcel et al. Multiple causes of high extinction risk in large mammal species. Science, v. 309, n. 5738, p. 1239-1241, 2005.

CARROLL, Carlos et al. Biotic and climatic velocity identify contrasting areas of vulnerability to climate change. PLoS One, v. 10, n. 10, p. e0140486, 2015.

CHAMBERLAIN, Scott. Rgbif: Interface to the Global “Biodiversity” Information Facility API. R package version 0.9.9, 2017. Available from: https:// www. CRAN. R-project.org/ package=rgbif.

CHEN, I.-Ching et al. Rapid range shifts of species associated with high levels of climate warming. Science, v. 333, n. 6045, p. 1024-1026, 2011.

COLLEN, Ben et al. Predicting how populations decline to extinction. Philosophical Transactions of the Royal Society B: Biological Sciences, v. 366, n. 1577, p. 2577-2586, 2011.

CORISTINE, Laura E.; KERR, Jeremy T. Temperature‐related geographical shifts among passerines: contrasting processes along poleward and equatorward range margins. Ecology and Evolution, v. 5, n. 22, p. 5162-5176, 2015.

CROOKS, Kevin R. et al. Quantification of habitat fragmentation reveals extinction risk in terrestrial mammals. Proceedings of the National Academy of Sciences, v. 114, n. 29, p. 7635-7640, 2017.

DAVIDSON, Ana D. et al. Multiple ecological pathways to extinction in mammals. Proceedings of the National Academy of Sciences, v. 106, n. 26, p. 10702-10705, 2009.

DIXON, Philip. VEGAN, a package of R functions for community ecology. Journal of Vegetation Science, v. 14, n. 6, p. 927-930, 2003.

DI MARCO, Moreno; SANTINI, Luca. Human pressures predict species’ geographic range size better than biological traits. Global Change Biology, v. 21, n. 6, p. 2169-2178, 2015.

DINIZ‐FILHO, José Alexandre Felizola. Macroecology and the hierarchical expansion of evolutionary theory. Global Ecology and Biogeography, v. 13, n. 1, p. 1-5, 2004.

ELDREDGE, Niles; GOULD, Stephen Jay. Punctuated equilibria: an alternative to phyletic gradualism. In: Schopf, T. J. M. (eds) Models in Paleobiology. Freeman, Cooper, Sna Francisco, p. 82-115, 1972.

FAITH, J. Tyler. Late Pleistocene and Holocene mammal extinctions on continental Africa. Earth-Science Reviews, v. 128, p. 105-121, 2014.

FICETOLA, Gentile Francesco; MAZEL, Florent; THUILLER, Wilfried. Global determinants of zoogeographical boundaries. Nature Ecology & Evolution, v. 1, n. 4, p. 0089, 2017. 68

FIELDING, Allan H.; BELL, John F. A review of methods for the assessment of predictions errors in conservation presence/absence models. Environmental Conservation, v. 24, n. 1, p. 38-49, 1997.

FORDHAM, Damien A. et al. How interactions between animal movement and landscape processes modify local range dynamics and extinction risk. Biology letters, v. 10, n. 5, p. 20140198, 2014.

FORERO‐MEDINA, GERMAN; JOPPA, Lucas; PIMM, Stuart L. Constraints to species’ elevational range shifts as climate changes. Conservation Biology, v. 25, n. 1, p. 163-171, 2011.

FORTELIUS, M. New and Old Worlds database of fossil mammals (NOW), 2015. Available from: http://www.helsinki.fi/science/now/. Accessed in: 4th October 2017.

FRECKLETON, Robert P.; JETZ, Walter. Space versus phylogeny: disentangling phylogenetic and spatial signals in comparative data. Proceedings of the Royal Society B: Biological Sciences, v. 276, n. 1654, p. 21-30, 2008.

FREEMAN, Benjamin G. et al. Climate change causes upslope shifts and mountaintop extirpations in a tropical bird community. Proceedings of the National Academy of Sciences, v. 115, n. 47, p. 11982-11987, 2018.

GARAMSZEGI, László Zsolt. Modern phylogenetic comparative methods and their application in evolutionary biology. Springer Berlin Heidelberg, Berlin, Heidelberg, v. 10, 2014.

GASTON, Kevin J. The structure and dynamics of geographic ranges. Oxford University Press on Demand, 2003.

GASTON, Kevin J. Geographic range limits: achieving synthesis. Proceedings of the Royal Society B: Biological Sciences, v. 276, n. 1661, p. 1395-1406, 2009.

GASTON, Kevin J.; BLACKBURN, Tim M. Age, area and avian diversification. Biological Journal of the Linnean Society, v. 62, n. 2, p. 239- 253, 1997.

GASTON, Kevin; BLACKBURN, Tim. Pattern and process in macroecology. John Wiley & Sons, 2006.

GBIF. Global Biodiversity Information Facility, 2017. Available from: https://www.gbif.org/.

GRAHAM, R. W.; LUNDELIUS-Jr, E. L. FAUNMAP II: New data for North America with a temporal extension for the Blancan, Irvingtonian and early Rancholabrean. FAUNMAP II Database, version 1.0, 2010. Available from: http://www.ucmp.berkeley.edu/faunmap/. Acessed in: 15th August 2014.

69

GONZALEZ-SUAREZ, Manuela; REVILLA, Eloy. Generalized drivers in the mammalian endangerment process. PLoS One, v. 9, n. 2, p. e90292, 2014.

GOULD, Stephan; ELDREDGE, Niles. Punctuated equilibrium comes of age. Nature, v. 366, n. 6452, p. 223, 1993.

HOGG, Alan G. et al. SHCal13 Southern Hemisphere calibration, 0–50,000 years cal BP. Radiocarbon, v. 55, n. 4, p. 1889-1903, 2013.

IUCN. International Union of Conservation of Nature Red List of Threatened Species. Version 2018-1. Available from: http://www.iucnredlist.org/. Accessed in: 19th July 2018.

JABLONSKI, David. Mass extinctions and macroevolution. Paleobiology, v. 31, n. S2, p. 192-210, 2005.

JIMÉNEZ‐VALVERDE, Alberto; LOBO, Jorge M.; HORTAL, Joaquín. Not as good as they seem: the importance of concepts in species distribution modelling. Diversity and distributions, v. 14, n. 6, p. 885-890, 2008.

LIMA-RIBEIRO, Matheus S.; DINIZ-FILHO, José. A. F. American megafaunal extinctions and human arrival: improved evaluation using a meta-analytical approach. Quaternary International, v. 299, p. 38-52, 2013.

LIMA-RIBEIRO, Matheus S. et al. Constraint envelope analyses of macroecological patterns reveal climatic effects on Pleistocene mammal extinctions. Quaternary Research, v. 82, n. 1, p. 260-269, 2014.

LIMA-RIBEIRO, Matheus Souza et al. EcoClimate: a database of climate data from multiple models for past, present, and future for macroecologists and biogeographers. Biodiversity Informatics, v. 10, 2015.

LIMA‐RIBEIRO, Matheus S. et al. Fossil record improves biodiversity risk assessment under future climate change scenarios. Diversity and Distributions, v. 23, n. 8, p. 922-933, 2017.

LISIECKI, Lorraine E.; RAYMO, Maureen E. A Pliocene‐Pleistocene stack of 57 globally distributed benthic δ18O records. Paleoceanography, v. 20, n. 1, 2005.

LYONS, S. Kathleen; SMITH, Felisa A. Macroecological patterns of body size in mammals across time and space. In: Body size: linking pattern and process across space, time, and taxonomy. University of Chicago Press, Chicago, p. 116-146, 2013.

MAIORANO, Luigi et al. Building the niche through time: using 13,000 years of data to predict the effects of climate change on three tree species in Europe. Global Ecology and Biogeography, v. 22, n. 3, p. 302-317, 2013.

70

MULITZA, Stefan et al. Temperature: δ18O relationships of planktonic foraminifera collected from surface waters. Palaeogeography, Palaeoclimatology, Palaeoecology, v. 202, n. 1-2, p. 143-152, 2003.

OKSANEN, Jari. et al. vegan: Community Ecology Package. R package version 2.5-1. 2018. Available at: https://CRAN.R-project.org/package=vegan.

PAYNE, Jonathan L.; FINNEGAN, Seth. The effect of geographic range on extinction risk during background and mass extinction. Proceedings of the National Academy of Sciences, v. 104, n. 25, p. 10506-10511, 2007.

PBDB. The Paleobiology database, 2017. Available from: https://paleobiodb.org/.

PEARSON, Richard G. et al. Predicting species distributions from small numbers of occurrence records: a test case using cryptic geckos in Madagascar. Journal of biogeography, v. 34, n. 1, p. 102-117, 2007.

PECL, Gretta T. et al. Biodiversity redistribution under climate change: Impacts on ecosystems and human well-being. Science, v. 355, n. 6332, p. eaai9214, 2017.

PIGOT, Alex L.; TOBIAS, Joseph A. Species interactions constrain geographic range expansion over evolutionary time. Ecology letters, v. 16, n. 3, p. 330-338, 2013.

POWERS, Ryan P.; JETZ, Walter. Global habitat loss and extinction risk of terrestrial vertebrates under future land-use-change scenarios. Nature Climate Change, p. 1, 2019.

PURVIS, Andy et al. Predicting extinction risk in declining species. Proceedings of the Royal Society of London B: Biological Sciences, v. 267, n. 1456, p. 1947- 1952, 2000.

R CORE TEAM. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Viena, Austria, 2016. Available from: https:// www.R-project.org/. Downloaded in 7th July 2017.

REIMER, Paula J.; et al. IntCal13 and Marine13 radiocarbon age calibration curves 0– 50,000 years cal BP. Radiocarbon, v. 55, n. 4, p. 1869-1887, 2013.

REVELL, Liam J.; HARMON, Luke J.; COLLAR, David C. Phylogenetic signal, evolutionary process, and rate. Systematic Biology, v. 57, n. 4, p. 591-601, 2008.

RODRÍGUEZ-REY, Marta et al. A comprehensive database of quality-rated fossil ages for Sahul’s Quaternary vertebrates. Scientific data, v. 3, p. 160053, 2016.

RUMPF, Sabine B. et al. Elevational rear edges shifted at least as much as leading edges over the last century. Global Ecology and Biogeography, v. 28, n. 4, p. 533-543, 2019.

71

SERRA‐DIAZ, Josep M. et al. Bioclimatic velocity: the pace of species exposure to climate change. Diversity and Distributions, v. 20, n. 2, p. 169-180, 2014.

STEPHENS, Philip A.; SUTHERLAND, William J.; FRECKLETON, Robert P. What is the Allee effect?. Oikos, p. 185-190, 1999.

STUIVER, M.; REIMER, P. J.; REIMER, R. W. CALIB 7.1, 2018. Available from: http://calib.org/. Accessed in: 12th March 2018.

THE STAGE THREE PROJECT. Institute for Quaternary Research at Cambridge University, 2017. Available from: https://www.esc.cam.ac.uk/research/research- groups/research-projects/stage-three-project/stage-three-project-database- downloads. Accessed in: 5th October 2017.

VARELA, Sara; LOBO, Jorge M.; HORTAL, Joaquín. Using species distribution models in paleobiogeography: a matter of data, predictors and concepts. Palaeogeography, Palaeoclimatology, Palaeoecology, v. 310, n. 3-4, p. 451-463, 2011.

VARELA, Sara; LIMA-RIBEIRO, Matheus S.; TERRIBILE, Levi Carina. A short guide to the climatic variables of the last glacial maximum for biogeographers. PloS one, v. 10, n. 6, p. e0129037, 2015.

WALLACE, A. R. The Geographical Distribution of Animals. Cambridge Univ. Press, Cambridge, 1876.

ZACAÏ, Axelle et al. Phylogenetic conservatism of species range size is the combined outcome of phylogeny and environmental stability. Journal of Biogeography, v. 44, n. 11, p. 2451-2462, 2017.

72

Supplementary Information – Chapter 02

Table S1. Atmosphere-ocean general circulation models (AOGCM’s) with references of their respective modeling group(s) and Institute ID’s.

Model Name Institute ID Modeling Group(s)

CCSM RSMAS University of Miami - RSMAS Japan Agency for Marine- Earth Science and Technology, Atmosphere and Ocean Research MIROC MIROC Institute (The University of Tokyo), and National Institute for Environmental Studies

73

Table S2. Algorithms, modeling methods and the type of data used to build the ENM’s.

Method Statistical feature Data type R package

Euclidean distance Environmental Presence-only stats1 distance Bioclim Climatic envelope Presence-only dismo2 Domain Environmental Presence-only dismo2 (Gower distance) distance Maximum entropy Machine learning Presence/background dismo2 (Maxent) Support vector Machine learning Presence/background kernlab3 machine (SVM)

1R Core Team (2017); 2Hijmans et al. (2017); 3Karatzoglou et al. (2004)

74

Table S3. Bioclimatic variables selected to represent the climatic characteristics of Wallacean biogeographical realms. bio1 – Annual mean temperature; bio2 – Mean diurnal range (Mean of monthly (maximal temperature – minimal temperature)); bio3 – Isothermality ((bio2 / bio7) * 100); bio4 – Temperature seasonality (standard deviation * 100); bio5 – Maximum temperature of warmest month; bio7 – Temperature annual range (bio5 - bio6); bio13 – Precipitation of wettest month; bio14 – Precipitation of driest month; bio15 – Precipitation seasonality (coefficient of variation); bio16 – Precipitation of wettest quarter; bio17 – Precipitation of driest quarter.

Biogeographical realm Bioclimatic variables

African bio1, bio2, bio16, bio17 Australian bio1, bio2, bio16, bio17 Indo-malayan bio1, bio4, bio16, bio17 Neartic bio5, bio7, bio13, bio15 Neotropic bio1, bio2, bio3, bio16, bio17 Paleartic bio2, bio4, bio5, bio13, bio14

Literature Cited

HIJMANS, Robert J.; PHILLIPS, Steven; LEATWICK, John; ELITH, Jane. dismo: Species distribution modeling. R package version 1.1-4. https://CRAN.R- project.org/package=dismo, 2017.

KARATZOGLOU, Alexandros et al. kernlab-an S4 package for kernel methods in R. Journal of statistical software, v. 11, n. 9, p. 1-20, 2004.

R CORE TEAM. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Viena, Austria, 2016. Available from: https:// www.R-project.org/. Downloaded in 7th July 2017.

75

CAPÍTULO 3

76

A Macroecological Perspective on Late Quaternary Megafauna Extinction

Vinícius Silva Reis1,2, Sara Varela3, Levi Carina Terribile1, Matheus Souza Lima- Ribeiro1

1 – Laboratório de Macroecologia, Universidade Federal de Jataí, Jataí, Brasil; 2 – Instituto de Ciências Biológicas V, Laboratório de Ecologia Teórica e Síntese, Universidade Federal de Goiás, Goiânia, Brasil; 3 - Museum für Naturkunde - Leibniz Institute for Research on Evolution and Biodiversity, Berlin, Germany.

Abstract

The extinction of late-Quaternary Megafauna has been calling attention for which causes might entailed such wave of loss of biodiversity. In order to survive from climate change, species should adapt to new environmental conditions or track the displacement of suitable habitats. Since smaller landmasses have lost more megafaunal species than larger ones, we ask here whether the landmasses influenced the extinction of late-Quaternary Megafauna. We used ecological niche modeling to generate information about range shift and displacement for 154 species of extinct mammal Megafauna. We analyzed the relation between the proportion of extinct Megafauna with the size of the landmasses directly and also via habitat tracking and range shift with a more usual ordinary least squares (OLS) approach and also with a quantile regression (QR) approach. We found that the late-Quaternary extinctions represent a multi- factorial and complex event that is better captured by the usage of the QR approach, since the majority of relations analyzed can be categorized as being constraint envelopes sensu Brown and Maurer (1987, 1989). The landmasses constrained species to follow suitable habitats for survival more strongly than to shift their geographical ranges. Future macroecological studies on late-Quaternary Megafauna extinctions should benefit from the characterization of emerging constraint envelopes using quantile regression approaches.

Introduction

During the late Quaternary, the terrestrial biota witnessed the extirpation of large-bodied mammals, spread throughout the globe and occupying since primary consumer trophic levels untill top predators (Koch; Barnosky, 2006). Inside the Megafauna extinction debate, two main – sometimes antagonistic – lines of investigation arise: anthropogenic and climate hypotheses. The anthropogenic hypothesis considers the direct predatory impact by Homo sapiens, a species endowed with sufficient cognition for social organization, competition and production of hunting technologies such as and arrows [see Martin (1973, 1984); Johnson et al. (2013); Sandom et al. (2014); Surovell et al. (2016)]. On the other side of the debate, the

77

climate hypothesis predicts that climate change over the last ice age was the major stressor rising the extinction risk of Megafauna during the late Quaternary [see review on Grayson (1984) and Nogués-Bravo et al. (2008); Lima-Ribeiro et al. (2013); Villavicencio et al. (2016); Broughton; Weitzel (2018) for evidences on coupled climate-human influences on Megafauna extinction].

When considering climate as a great trigger of extinctions (Thomas et al., 2004; Mclean; Wilson, 2011; Pearson et al., 2014; Urban, 2015), species are expected to move to places where new suitable conditions appear (the so-called habitat tracking, Eldredge; Gould, 1972; Lyons, 2003), in order to survive or to stay and adapt enough to tolerate a range of new environmental conditions (evolutionary rescue, see Bell, 2017 for a review of the theme and Diniz-Filho et al., 2019 for recent advances). Except for these two possibilities, species would get extinct. This spatial view on climate change places the movement of the geographical range of a species in a central role of analysis: If a species is not able to displace the geographical range enough, the species may face a continuous abatement of the range size and the increase of extinction probability (Lima- Ribeiro et al., 2014; Darroch; Saupe, 2018). Once terrestrial species are unable to colonize marine environments, the space available for the geographical range to spread might be a limiting factor. Thus, landmasses might constraint species to completely follow suitable climate conditions for survival. Although it has been proposed that landmasses might limit geographical ranges to displace or change in size, to our knowldege, the influence of the landmasses on the extinction of late-Quaternary mammal fauna has never been tested before.

Our work departs from the observation that narrower landmasses (as Australia and South America) experienced greater relative loss on mammal Megafauna species during the Pleistocene-Holocene transition (see Koch; Barnosky, 2006; Stuart, 2015). Here we aim to test whether the late-Quaternary great wave of mammal diversity loss was boosted by limitation imposed by the landmasses size. We hypothesize that the landmasses precluded Megafauna species to track suitable habitats and limited their geographical range at narrow and unsuitable habitats over the last ice age. For this, we carried out ecological niche models and derived distribution maps for 154 late- Quaternary extinct Megafauna species.

78

Material and Methods

Paleodistribution modeling

We gathered georeferenced and dated fossil records for species of extinct Megafauna from all continents in online datasets and published articles, namely: The Paleobiology Database (https://paleobiodb.org; PBDB, 2017), a global dataset, with the package “paleobioDB” (version 0.5.0, Varela et al. 2015), the FosSahul Database (Rodríguez-Rey et al., 2016), the New and Old Worlds Database (http://www.helsinki.fi/science/now/; Fortelius, 2015), the faunal database from The Stage Three Project (The Stage Three Project, 2017), the FaunMap (Graham; Lundelius, 2014), Card (2011) Databases, and the occurrences in Lima-Ribeiro et al. (2013b) for South America, and Faith (2014) for Africa. We inspected all occurrences and eliminated the non-specific taxonomic records and non-dated occurrences, as well as those occurrence points located within the same 0.5º cell (~55 Km) of spatial resolution (lat/long) and with the same age at 1 ky of temporal resolution. We calibrated all radiocarbon dated records for calendrical aging with the CALIB Radiocarbon Calibration software (version 7.1, Stuiver et al., 2018). We used IntCal13 calibration curve (Reimer et al., 2013) for north hemisphere records and the SHCal13 calibration curve (Hogg et al., 2013) for south hemisphere records.

We obtained 19 bioclimatic layers from the ecoClimate database (www.ecoClimate.org; Lima-Ribeiro et al. 2018) for the Last Glacial Maximum (LGM, ~21 ky BP) and mid-Holocene (~6 ky BP). We followed Varela et al. (2015b) and selected the three AOGCMs that maximize uncertainty among climate simulations at LGM, namely: CCSM, GISS and MIROC (Table S2). To minimize multicollinearity among the bioclimatic variables used for niche modeling, we performed a factorial analysis for each biogeographical realm. The number of factors was determined from scree plot, and the bioclimatic variables with the greatest loadings within each factor were selected (Table S3). To ensure comparable predictors for niche modeling. we standardized the selected bioclimatic variables for them ranging between 0 and 1 by using the function “decostand” from “vegan” package (version 2.5-1, Dixon, 2003; Oksanen et al., 2018).

79

In order to acquire climatic information for occurrences that escaped the temporal range of standard LGM and mid-Holocene time bins, we performed a temporal interpolation using the δ18 Oxygen curve from Lisiecki and Raymo (2005) at 1 ky resolution as covariate. The δ18 Oxygen curve is the ratio curve between the quantity of the 18O and 16O stable isotopes, obtained from biogenic compounds and is commonly used as an indicator of climatic oscillations through time (Mulitza et al., 2003). Because interpolation accuracy decreases longer back ago, we eliminated all fossil records aged older than 700 ky BP.

We combined climate conditions from fossil occurrences in a multitemporal calibration approach (Nogués-Bravo, 2009) to predict the potential distributions of the 154 late Quaternary extinct Megafauna species onto LGM and mid-Holocene. We opted to use multiple methods encompassing distinct statistical approaches to take ENM uncertainty into account: Bioclim, Domain (Gower distance), Maximum entropy (Maxent) and Support Vector Machine (SVM, see Table S4). We did not account for modeling methods based on both presence and absence information once real absences are conceptually unattainable for extinct species. We randomly drawn 1000 points from the background areas (biogeographical realm where each species occurs) to fit the presence-background methods (Maxent and SVM). We considered the simplest parameterization for machine learning presence-background methods, as linear kernel function with probabilistic output for SVM and only the linear feature to combine climatic predictors in Maxent, with logistic output, according recommendations for modeling ecological niches with fossil records [Varela et al. (2011)]. Additionally, we used the Euclidean environmental distance method alone to build ENMs for species with less than five occurrence points.

Then we generated consensus maps for LGM and mid-Holocene from distinct methods and AOGCMs following the ensemble forecasting by Araújo and New (2007). We standardized all ENM predictions for the suitability values vary from 0 to 1, and averaged them to generate standardized consensus maps. Because fossil records are temporally dispersed, ENMs were not evaluated and ensembles obtained from arithmetic means. To convert continuous suitability predictions into binary distribution maps, we adopted a generic approach searching for the threshold value that enforces species to distribute on 50% of the background area at the LGM. Given we want

80

accessing relative changes in the geographical distribution of species over time and not absolute values (occupied distributions), this strategy becomes reasonable once it guarantees the same chance (50%) for species both to reduce or increase geographical range through time. Moreover, this strategy avoids spurious extreme range size and shift by stablishing extreme thresholds.

Geographic range dynamics

To evaluate if landmasses acted as dispersal barriers over the last ice age determining the late Quaternary Megafaunal extinction risk, we computed the magnitude that species range size changed from LGM to mid-Holocene (range shift) and the distance across which species should displace to track suitable climatic conditions through time (habitat tracking). Both range shift and habitat tracking were obtained for every species under two scenarios: 1) restricted range dynamics into continents (hereafter, “realized scenario”), and 2) unlimited range dynamics throughout the Earth’s surface, not considering continental barriers (hereafter, “potential scenario”). Taking into account we are dealing with terrestrial non-volant species of Megafauna, the potential scenario releases species from continents and mimics the geographical dynamics of species range if there were no physical barrier imposed by the geographical limits of landmasses.

We measured the species range size as the number of occupied grid-cells from binary maps and then estimated the range shift for both realized and potential scenarios as the difference between range sizes at mid-Holocene and LGM. Positive range shifts indicate that species expanded its range, whereas negative values indicate the magnitude of range contraction. To estimate habitat tracking, we applied the bioclimatic velocity algorithm by Serra-Díaz et al. (2014) to compute the distance by which the species should track across space to follow suitable habitats over the last ice age. The bioclimatic velocity measures the species exposure to climate change. It benefits from inclusion of species-specific climatic tolerance to deal with habitat tracking via ecological niche models rather than generalized climate change velocities (Carroll et al., 2015). From binary maps, we calculated the distance of every occupied grid-cell in the LGM to the nearest occupied grid-cell in the mid-Holocene. Then, we calculated the habitat tracking (HT) by adding cell-by-cell distances weighted by its respective climatic suitability throughout the background, as follows: 81

HT = ∑ di Si

th where di and Si are the spatial distance and climatic suitability for the i grid-cell. Thus, species were more exposed to climate change across suitable habitats (grid-cells) that displaced for longer distances across space and time.

Statistical analysis

To access the magnitude in which landmasses limited habitat tracking (our working assumption), we computed the difference between habitat tracking distances from realized and potential scenarios. Next, we used a path analysis scheme to summarize the magnitude that commonly tested range dynamics (realized range shift and habitat tracking) and our potential habitat tracking limitation explain Megafauna extinctions across continents.

At the extremes of path analysis, we used the size of landmasses, computed as the number of 0.5o lat/long grid-cells, as general predictor variable, and the percentage of extinct genera in each landmass (a proxy for extinction rate) as general response variable. The extinction rate was gathered from Koch and Barnosky (2006), except for Indo-Malayan realm, which was gathered from Stuart (2015). Across the path analysis, we analyzed the general effect of land size on extinction rate via both the common realized range dynamics and the potential habitat tracking limitation. Finally, we compared all the effects predicted from usual regressions, the linear ordinary least squares (OLS) approach, and from the quantile regression (QR) approach, an appropriated method to fit complex non-linear relationships (Cade; Noon, 2003) such as range dynamics versus extinction risk (see details in Lima-Ribeiro et al., 2014). From quantile regressions, we estimated the relationships at the center (quantile 50%, similar to OLS) and extremes of scatterplot (both lower - 5%, 10%, 25% - and upper quantiles - 75%, 90%, 95% - whichever was significantly stronger). The median relationships (quantile 50%) are comparable with OLS, whereas the lower/upper quantiles depict non-linear relationships such as triangular-shaped envelopes (see Lima-Ribeiro et al. 2014). The standardized slopes (beta) from OLS and QR represent the effect size along the path analysis scheme. We performed both the species distribution modelling and further statistical analysis in R program (R Core Team, 2017).

82

Results

The OLS and QR methods revealed general concordance regarding the direction (negative or positive) of the effects from all predictors of extinction rates across the path analysis. However, such relations differ as the magnitude of such effects (Figure 1). Except for direct land size effect on the extinction rate, the QR approach revealed generalized non-linear, triangular-shaped relationships, by showing stronger effects across either upper or lower boundaries of the scatterplot when compared to median- 50% quantile or OLS approach (Figure 1). In average, QR showed almost two times higher effects than OLS (bQR = 0.55; bOLS = 0.32; paired t-test: t = 2.04, df = 6, p = 0.08). The QR analysis also revealed the strongest effect of land size on the extinction rate indirectly via habitat tracking limitation, supporting our working assumption; smaller landmasses mostly limited Megafauna to potentially track their habitats, increasing their extinction rates. Both the direct relationship between land size and extinction rate and the indirect ones via the usual predictors linked to realized range dynamics (realized habitat tracking and range shift) consistently showed weaker effects than potential continental constraints (habitat tracking limitation; Figure 1).

Moreover, the OLS and QR results showed contradictory predictions regarding usual realized range dynamics. Besides QR predicts an effect of realized habitat tracking an order of magnitude stronger than OLS, the lower and upper quantiles showed opposite predictions with positive and negative relationships, respectively. Yet, while the OLS indicates that greater realized range shifts relate with higher extinction rates, the QR approach indicates neutral relationships irrespective of the quantile considered (median, upper or lower boundary, Figure 1).

83

A

B

C

84

Figure 1. Path analysis scheme summarizing multiple hypotheses testing and their respective effect sizes as predicted by quantile regression (QR, panels A and B) and usual OLS regression (panel C). The arrows width is proportional to their respective effect size represented by slopes (b) from regression models. Bold values represent significant slopes at 5%. Usual hypotheses testing from common predictors of climate change effect on Megafauna extinction risk are highlighted inside grey boxes.

Discussion

Our findings clearly support our prediction and show the potential constraints of terrestrial species habitat tracking imposed by continental limits as the closest climate predictor of megafaunal extinction rates across biogeographical realms. Additionally, our comparisons between QR and OLS regression approaches and through different climate predictors reflect the contrasting nature of the late Quaternary extinction debate, as extensively observed in literature (eg. Grayson; Meltzer, 2003; Fiedel; Haynes, 2004; Nogués-Bravo et al., 2008, 2010; Prescott et al., 2012; Wroe et al., 2013; Lima-Ribeiro et al., 2012, 2013; Sandom et al., 2014; Bartlett et al., 2016; Saltré et al., 2016; Araújo et al., 2017; Di Febbraro et al., 2017; Řičánková et al., 2017). The generalized stronger QR effects suggest that the late Quaternary extinctions had a non-linear dynamic as a consequence of the climate change through the last ice age. Besides supporting our prediction, the quantile regressions largely reported macroecological constraint envelopes, the founding landmark of the macroecology research (Brown; Maurer,1987, 1989). Macroecological constraint envelopes are not captured by usual, simple OLS regressions (Lima-Ribeiro et al. 2014). Overall, our findings suggest that usual methods and predictors handled to test climate change effects on Megafauna extinction should be updated to advance with the debate of late Quaternary extinctions.

Our results clearly show that changing climates did not essentially affect Megafauna by imposing long distances for habitat tracking, nor contracting its geographical ranges by itself, as often claimed in literature (see Schloss et al., 2012; Williams; Blois, 2018, for habitat tracking effects; and Nogués-Bravo et al. 2008; Lima- Ribeiro et al. 2013, for range shift effects). The weak influence of realized habitat tracking on continental extinction rates would be expected given that large-bodied mammals often present good dispersal capacity (Bowmann et al., 2002; Angert et al., 2011), such that extinct Megafauna would be able to follow suitable habitats for long distances into the continents. 85

Instead, the non-linear climate impacts, associated with the geographical barriers imposed by the continents, potentiated the extinction risk of Pleistocene Megafauna worldwide. Smaller landmasses mostly limited terrestrial species to track their climatically suitable habitats displacing over oceans and stocked higher amounts of extinctions along the last ice age. Climate change did not induce the collapse of megafaunal geographical ranges by disappearing suitable climates from Earth’s surface (non-analogous climates, see Nogués-Bravo et al. 2010), but by mediating geographical constraints that landmasses imposed to their habitat tracking. Once large-bodied mammals occupy broader geographical ranges, it seems natural that Megafauna would become more impacted with displacement of suitable climate conditions toward the oceans by restricting their occupied ranges to proportionally narrower areas through the time.

Species should sustain a geographical range large enough to acquire energy for maintaining population above the minimum viable size (Brown; Maurer,1987, 1989). Thus, large-bodied animals must attain exclusively large geographical distributions (Brown; Maurer, 1987, 1989). With the geographical displacement of suitable areas during the last ice age climate change, the Megafauna species present in smaller landmasses faced more restriction in following climatically suitable areas. This highest strength of restriction led to proportionally higher extinction rates in regions like Australia.

Literature cited

ANGERT, A. L. et al. Do species’ traits predict recent shifts at expanding range edges?. Ecology Letters, v. 14, n. 7, pp. 677-689, 2011.

ARAÚJO, Miguel B.; NEW, Mark. Ensemble forecasting of species distributions. Trends in ecology & evolution, v. 22, n. 1, p. 42-47, 2007.

ARAÚJO, Bernardo B. A. et al. Bigger kill than chill: The uneven roles of humans and climate on late Quaternary megafaunal extinctions. Quaternary International, v. 431, p. 216-222, 2017.

BARTLETT, Lewis J. et al. Robustness despite uncertainty: regional climate data reveal the dominant role of humans in explaining global extinctions of Late Quaternary megafauna. Ecography, v. 39, n. 2, p. 152-161, 2016.

86

BELL, Graham. Evolutionary rescue. Annual Review of Ecology, Evolution, and Systematics, v. 48, p. 605-627, 2017.

BOWMAN, J., JAEGER, J. A., FAHRIG, L. Dispersal distance of mammals is proportional to home range size. Ecology, v. 83, n. 7, pp. 2049-2055, 2002.

BROUGHTON, Jack M.; WEITZEL, Elic M. Population reconstructions for humans and megafauna suggest mixed causes for North American Pleistocene extinctions. Nature communications, v. 9, n. 1, p. 5441, 2018.

BROWN, James H.; MAURER, Brian A. Evolution of species assemblages: effects of energetic constraints and species dynamics on the diversification of the North American avifauna. The American Naturalist, v. 130, n. 1, p. 1-17, 1987.

BROWN, James H.; MAURER, Brian A. Macroecology: the division of food and space among species on continents. Science, v. 243, n. 4895, p. 1145-1150, 1989.

CADE, Brian S.; NOON, Barry R. A gentle introduction to quantile regression for ecologists. Frontiers in Ecology and the Environment, v. 1, n. 8, p. 412-420, 2003.

CARD – The Canadian Archaeological Radiocarbon Database. Available from: http://www.canadianarchaeology.ca/. Accessed in: 15th January 2011.

CARROLL, Carlos et al. Biotic and climatic velocity identify contrasting areas of vulnerability to climate change. PLoS One, v. 10, n. 10, p. e0140486, 2015.

DARROCH, Simon AF; SAUPE, Erin E. Reconstructing geographic range-size dynamics from fossil data. Paleobiology, v. 44, n. 1, p. 25-39, 2018.

DI FEBBRARO, Mirko et al. Does the jack of all trades fare best? Survival and niche width in Late Pleistocene megafauna. Journal of biogeography, v. 44, n. 12, p. 2828-2838, 2017.

DINIZ‐FILHO, José Alexandre F. et al. A macroecological approach to evolutionary rescue and adaptation to climate change. Ecography, 2019.

DIXON, Philip. VEGAN, a package of R functions for community ecology. Journal of Vegetation Science, v. 14, n. 6, p. 927-930, 2003.

ELDREDGE, Niles; GOULD, Stephen Jay. Punctuated equilibria: an alternative to phyletic gradualism. In: Schopf, T. J. M. (eds) Models in Paleobiology. Freeman, Cooper, Sna Francisco, p. 82-115, 1972.

FAITH, J. Tyler. Late Pleistocene and Holocene mammal extinctions on continental Africa. Earth-Science Reviews, v. 128, p. 105-121, 2014.

FIEDEL, Stuart; HAYNES, Gary. A premature burial: comments on Grayson and Meltzer's “Requiem for overkill”. Journal of Archaeological Science, v. 31, n. 1, p. 121-131, 2004. 87

FORTELIUS, M. New and Old Worlds database of fossil mammals (NOW), 2015. Available from: http://www.helsinki.fi/science/now/. Accessed in: 4th October 2017.

GRAHAM, R. W.; LUNDELIUS-Jr, E. L. FAUNMAP II: New data for North America with a temporal extension for the Blancan, Irvingtonian and early Rancholabrean. FAUNMAP II Database, version 1.0, 2010. Available from: http://www.ucmp.berkeley.edu/faunmap/. Acessed in: 15th August 2014.

GRAYSON, Donald K. Nineteenth century explanations of Pleistocene extinctions: A review and analysis. In: MARTIN, Paul S.; KLEIN, Richard G. (Ed.). Quaternary extinctions: a prehistoric revolution. Press,1984.

GRAYSON, Donald K.; MELTZER, David J. A requiem for North American overkill. Journal of Archaeological Science, v. 30, n. 5, p. 585-593, 2003.

HOGG, Alan G. et al. SHCal13 Southern Hemisphere calibration, 0–50,000 years cal BP. Radiocarbon, v. 55, n. 4, p. 1889-1903, 2013.

JOHNSON, Chris N. et al. Rapid megafaunal extinction following human arrival throughout the New World. Quaternary International, v. 308, p. 273-277, 2013.

KOCH, Paul L.; BARNOSKY, Anthony D. Late Quaternary extinctions: state of the debate. Annual Review of Ecology, Evolution, and Systematics, v. 37, 2006.

LIMA-RIBEIRO, Matheus Souza et al. Potential suitable areas of giant ground sloths dropped before its extinction in South America: the evidences from bioclimatic envelope modeling. Natureza & Conservação, v. 10, 145–151, 2012.

LIMA-RIBEIRO, Matheus Souza et al. Climate and humans set the place and time of Proboscidean extinction in late Quaternary of South America. Palaeogeography, Palaeoclimatology, Palaeoecology, v. 392, p. 546- 556, 2013.

LIMA-RIBEIRO, Matheus S. et al. Constraint envelope analyses of macroecological patterns reveal climatic effects on Pleistocene mammal extinctions. Quaternary research, v. 82, n. 1, p. 260-269, 2014.

LIMA-RIBEIRO, Matheus S. et al. The ecoClimate Database, Available from: http://ecoclimate.org. Accessed in: 20th March 2018.

LISIECKI, Lorraine E.; RAYMO, Maureen E. A Pliocene‐Pleistocene stack of 57 globally distributed benthic δ18O records. Paleoceanography, v. 20, n. 1, 2005.

LYONS, S. Kathleen. A quantitative assessment of the range shifts of Pleistocene mammals. Journal of Mammalogy, v. 84, n. 2, p. 385-402, 2003.

88

MARTIN, Paul S. The Discovery of America: The first Americans may have swept the Western Hemisphere and decimated its fauna within 1000 years. Science, v. 179, n. 4077, p. 969-974, 1973.

MARTIN, Paul S. Prehistoric overkill: the global model. In: MARTIN, P. S.; KLEIN, R. G. Quaternary extinctions: a prehistoric revolution, p. 354-403, 1984.

MACLEAN, Ilya MD; WILSON, Robert J. Recent ecological responses to climate change support predictions of high extinction risk. Proceedings of the National Academy of Sciences, v. 108, n. 30, p. 12337-12342, 2011.

MULITZA, Stefan et al. Temperature: δ18O relationships of planktonic foraminifera collected from surface waters. Palaeogeography, Palaeoclimatology, Palaeoecology, v. 202, n. 1-2, p. 143-152, 2003.

NOGUÉS-BRAVO, David et al. Climate change, humans, and the extinction of the woolly . PLoS biology, v. 6, n. 4, p. e79, 2008.

NOGUÉS‐BRAVO, David. Predicting the past distribution of species climatic niches. Global Ecology and Biogeography, v. 18, n. 5, p. 521-531, 2009.

NOGUÉS‐BRAVO, David et al. Climate predictors of late Quaternary extinctions. Evolution: International Journal of Organic Evolution, v. 64, n. 8, p. 2442-2449, 2010.

OKSANEN, Jari. et al. vegan: Community Ecology Package. R package version 2.5-1. https://CRAN.R-project.org/package=vegan, 2018.

PBDB. The Paleobiology database, 2017. Available from: https://paleobiodb.org/.

PEARSON, Richard G. et al. Life history and spatial traits predict extinction risk due to climate change. Nature Climate Change, v. 4, n. 3, p. 217, 2014.

PRESCOTT, Graham W. et al. Quantitative global analysis of the role of climate and people in explaining late Quaternary megafaunal extinctions. Proceedings of the National Academy of Sciences, v. 109, n. 12, p. 4527-4531, 2012.

R CORE TEAM. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Viena, Austria, 2016. Available from: https:// www.R-project.org/. Downloaded in 7th July 2017.

REIMER, Paula J.; et al. IntCal13 and Marine13 radiocarbon age calibration curves 0– 50,000 years cal BP. Radiocarbon, v. 55, n. 4, p. 1869-1887, 2013.

ŘIČÁNKOVÁ, Věra Pavelková et al. Environmental correlates of the Late Quaternary regional extinctions of large and small Palaearctic mammals. Ecography, v. 41, n. 3, p. 516-527, 2018.

RODRÍGUEZ-REY, Marta et al. A comprehensive database of quality-rated fossil ages for Sahul’s Quaternary vertebrates. Scientific data, v. 3, p. 160053, 2016. 89

SALTRÉ, Frédérik et al. Climate change not to blame for late Quaternary megafauna extinctions in Australia. Nature Communications, v. 7, p. 10511, 2016.

SANDOM, Christopher et al. Global late Quaternary megafauna extinctions linked to humans, not climate change. Proceedings of the Royal Society B: Biological Sciences, v. 281, n. 1787, p. 20133254, 2014.

SCHLOSS, Carrie A.; NUÑEZ, Tristan A.; LAWLER, Joshua J. Dispersal will limit ability of mammals to track climate change in the Western Hemisphere. Proceedings of the National Academy of Sciences, v. 109, n. 22, p. 8606-8611, 2012.

SERRA‐DIAZ, Josep M. et al. Bioclimatic velocity: the pace of species exposure to climate change. Diversity and Distributions, v. 20, n. 2, p. 169-180, 2014.

STUART, Anthony John. Late Quaternary megafaunal extinctions on the continents: a short review. Geological Journal, v. 50, n. 3, p. 338-363, 2015.

STUIVER, M.; REIMER, P. J.; REIMER, R. W. CALIB 7.1, 2018. Available from: http://calib.org/. Accessed in: 12th March 2018.

SUROVELL, Todd A. et al. Test of Martin’s overkill hypothesis using radiocarbon dates on extinct megafauna. Proceedings of the National Academy of Sciences, v. 113, n. 4, p. 886-891, 2016.

THE STAGE THREE PROJECT. Institute for Quaternary Research at Cambridge University, 2017. Available from: https://www.esc.cam.ac.uk/research/research- groups/research-projects/stage-three-project/stage-three-project-database- downloads. Accessed in: 5th October 2017.

THOMAS, Chris D. et al. Extinction risk from climate change. Nature, v. 427, n. 6970, p. 145, 2004.

URBAN, Mark C. Accelerating extinction risk from climate change. Science, v. 348, n. 6234, p. 571-573, 2015.

VARELA, Sara; LOBO, Jorge M.; HORTAL, Joaquín. Using species distribution models in paleobiogeography: a matter of data, predictors and concepts. Palaeogeography, Palaeoclimatology, Palaeoecology, v. 310, n. 3-4, p. 451-463, 2011.

VARELA, Sara et al. paleobioDB: An R package for downloading, visualizing and processing data from the Paleobiology Database. Ecography, v. 38, n. 4, p. 419- 425, 2015.

VILLAVICENCIO, Natalia A. et al. Combination of humans, climate, and vegetation change triggered Late Quaternary megafauna extinction in the Última Esperanza region, southern , Chile. Ecography, v. 39, n. 2, p. 125-140, 2016.

90

WILLIAMS, J. Eric; BLOIS, Jessica L. Range shifts in response to past and future climate change: Can climate velocities and species’ dispersal capabilities explain variation in mammalian range shifts?. Journal of biogeography, v. 45, n. 9, p. 2175-2189, 2018.

WROE, Stephen et al. Climate change frames debate over the extinction of megafauna in Sahul (Pleistocene Australia-). Proceedings of the National Academy of Sciences, v. 110, n. 22, p. 8777-8781, 2013.

91

Supplementary Information – Chapter 03

Table S1. Extinct species considered in this study.

Extinct species

Plesiorycteropus madagascariensis Macrauchenia patachonica Xenorhinotherium bahiense Neolicaphrium recens Mixotoxodon larensis Toxodon platensis Cynotherium sardous Dusicyon avus Protocyon troglodytes Homotherium serum Homotherium latidens Miracinonyx trumani Smilodon fatalis Brachyprotoma obtusata Arctodus simus Arctotherium tarijense Arctotherium wingei Capromeryx minor Stockoceros conklingi Bootherium bombifrons Megalotragus priscus Megalovis guangxiensis Pelorovis antiquus Soergelia minor Spirocerus kiakhtensis Camelops hesternus Hemiauchenia macrocephala Hemiauchenia paradoxa Palaeolama major Palaeolama mirifica Agalmaceros blicki Antifer ultra Megaceroides algericus Megaloceros giganteus Morenelaphus brachyceros Morenelaphus lujanensis Navahoceros fricki

92

Paraceros fragilis cazioti Sangamona fugitiva Kolpochoerus majus Metridiochoerus compactus Eutatus seguini Propraopus sulcatus Propraopus punctatus Protemnodon hopei Protemnodon nombe Protemnodon tumbuna Sthenurus atlas Sthenurus stirlingi Sthenurus tindalei Troposodon minor Borungaboodie hatcheri Caloprymnus campestris Propleopus oscillans Phascolonus gigas Nesiotites similis Aztlanolagus agilis Megalibgwilia ramsayi Hippidion principale Hippidion devillei Coelodonta antiquitatis Stephanorhinus hemitoechus Stephanorhinus kirchbergensis Diabolotherium nordenskioldi Megalocnus rodens Neocnus comes Parocnus browni Caipora bambuiorum Protopithecus brasiliensis Pachylemur insignis Mammuthus exilis Cuvieronius hyodon Stegomastodon waringi Stegodon trigonocephalus Stegodon orientalis Stegodon florensis Castoroides ohioensis Neochoerus aesopi

93

Pliomys lenki Microgale macpheei Canis dirus Cryptoprocta spelea Leopardus amnicola Tremarctos floridanus Antidorcas australis Antidorcas bondi Bos primigenius Bubalus palaeokerabau Damaliscus niro Damaliscus hypsodon Euceratherium collinum Gazella atlantica Gazella tingitana Hippotragus leucophaeus Alces scotti Hexaprotodon sivalensis Hippopotamus madagascariensis Phanourios minutes Catagonus stenocephalus Dasypus bellus Macropus ferragus Macropus greyi Onychogalea lunata Thylogale christenseni Vombatus hacketti Ochotona whartoni Zaglossus hacketti Perameles eremiana Equus hydruntinus Equus semiplicatus Tapirus augustus Tapirus copei Elephas antiquus Elephas cypriotes Elephas iolensis Elephas mnaidriensis Hystrix refossa Conilurus albipes Pseudomys gouldii Hypogeomys australis

94

Doedicurus clavicaudatus Glyptodon clavipes Glyptodon reticulatus Glyptotherium floridanum Glyptotherium cylindricum Hoplophorus euphractus Panochthus tuberculatus Holmesina occidentalis Holmesina paulacoutoi Pampatherium humboldti Pampatherium typum Diprotodon optatum Maokopia ronaldi Prolagus sardus Chaeropus ecaudatus Eremotherium laurillardi Megatherium tarijense Catonyx cuvieri Glossotherium robustum Lestodon armatus Mylodon darwinii Scelidodon chiliensis Scelidotherium leptocephalum Valgipes deformis Nothrotherium maquinense Archaeolemur majori Archaeolemur edwardsi Hadropithecus stenognathus Megaladapis edwardsi Megaladapis madagascariensis Palaeopropithecus ingens

95

Table S2. Atmosphere-ocean general circulation models (Aogcm’s) with references of their respective modeling group(s) and institute ID’s.

Model Name Institute ID Modeling Group(s)

CCSM RSMAS University of Miami - RSMAS GISS NASA GISS NASA Goddard Institute for Space Studies Japan Agency for Marine- Earth Science and Technology, Atmosphere and Ocean Research MIROC MIROC Institute (The University of Tokyo), and National Institute for Environmental Studies

96

Table S3. Bioclimatic variables selected to represent the climatic characteristics of Wallacean biogeographical realms. bio1 – Annual mean temperature; bio2 – Mean diurnal range (Mean of monthly (maximal temperature – minimal temperature)); bio3 – Isothermality ((bio2 / bio7) * 100); bio4 – Temperature seasonality (standard deviation * 100); bio5 – Maximum temperature of warmest month; bio7 – Temperature annual range (bio5 - bio6); bio13 – Precipitation of wettest month; bio14 – Precipitation of driest month; bio15 – Precipitation seasonality (coefficient of variation); bio16 – Precipitation of wettest quarter; bio17 – Precipitation of driest quarter.

Biogeographical Realms Bioclimatic variables

African bio1, bio2, bio16, bio17 Australian bio1, bio2, bio16, bio17 Indo-malayan bio1, bio4, bio16, bio17 Neartic bio5, bio7, bio13, bio15 Neotropic bio1, bio2, bio3, bio16, bio17 Paleartic bio2, bio4, bio5, bio13, bio14

97

Table S5. Algorithms, modeling methods and the type of data used to build the ENM’s.

Method Statistical feature Data type R package

Euclidean distance Environmental Presence-only stats1 distance Bioclim Climatic envelope Presence-only dismo2 Domain Environmental Presence-only dismo2 (Gower distance) distance Maximum entropy Machine learning Presence/background dismo2 (Maxent) Support vector Machine learning Presence/background kernlab3 machine (SVM)

1R Core Team (2017); 2Hijmans et al. (2017); 3Karatzoglou et al. (2004)

Literature Cited

HIJMANS, Robert J.; PHILLIPS, Steven; LEATWICK, John; ELITH, Jane. dismo: Species distribution modeling. R package version 1.1-4. https://CRAN.R- project.org/package=dismo, 2017.

KARATZOGLOU, Alexandros et al. kernlab-an S4 package for kernel methods in R. Journal of statistical software, v. 11, n. 9, p. 1-20, 2004.

R CORE TEAM. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Viena, Austria, 2016. Available from: https:// www.R-project.org/. Downloaded in 7th July 2017.

98

CONSIDERAÇÕES FINAIS

Ao longo deste trabalho pude concluir que:

- O tamanho do corpo é o melhor preditor do risco de extinção em mamíferos quando comparado a idade filogenética e mudança no tamanho da distribuição geográfica;

- A inclusão de espécies extintas é importante nos trabalhos que buscam causas para o risco de extinção das espécies, pois os fatores que influenciam a extinção se tornam ainda mais evidenciados;

- As massas de terra limitam tanto a mudança do tamanho quanto o deslocamento da distribuição geográfica das espécies de mamíferos;

- Apesar disso, apenas a restrição sobre a dispersão das espécies em seguirem ambientes adequados à sobrevivência esta relacionada com o risco de extinção;

- Espécies ameaçadas e extintas percorreram distâncias menores em busca de ambientes adequados do que as não-ameaçadas em relação à distância que poderiam percorrer sem a influência da barreira continental;

- A restrição das massas de terra sobre a dispersão dos mamíferos deve ser vista como um mecanismo de seleção a nível de espécie no qual espécies capazes de acompanhar ambientes favoráveis à sobrevivência possuem maior probabilidade de sobreviver à mudança do clima;

- Pelo caráter multifatorial e não-linear, a extinção da Megafauna de mamíferos no Quaternário tardio é melhor analisada utilizando a técnica da regressão quantílica em detrimento a abordagens convencionais de quadrados mínimos;

- Os continentes restringiram a distância possível que as espécies de Megafauna puderam percorrer para acompanhar ambientes mais favoráveis à sobrevivência;

- Essa restrição influenciou os padrões de extinção de Megafauna observados mais que a restrição imposta pelos continentes à mudança do tamanho da distribuição geográfica;

Em suma, evidenciei nesta tese a utilização de descritores do risco de extinção que se apresentam em amplas escalas espaciais e temporais. Mostrei a importância do tamanho corporal como melhor preditor do risco de extinção em mamíferos; mostrei 99

que o tamanho dos continentes afeta o risco de extinção neste grupo, assim espécies em massas de terra menores estão mais sujeitas a se extinguirem pela mudança do clima; pude mostrar ainda que além de moldar o padrão de risco de extinção observado atualmente nos mamíferos, o tamanho dos continentes ainda influenciou grandes extinções pretéritas como o desaparecimento da Megafauna. Por fim, evidenciei o papel central da capacidade das espécies de seguirem ambientes adequados à sobrevivência (i.e. habitat tracking) na probabilidade de sobrevivência à mudança do clima em nível global. Sugiro que estudos futuros possam focar no papel do habitat tracking em relação à sobrevivência das espécies de mamíferos também em diferentes cenários de mudança climática futura, bem como na influência dos continentes e de barreiras intra- continentais no risco de extinção dos mamíferos também no futuro.

100