Prioridades Para a Conservação De Anfíbios Da Mata Atlântica

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

Priscila Lemes de Azevedo Silva TESE DE DOUTORADO

Prioridades para a conservação de anfíbios da Mata Atlântica

Orientador: Prof. Dr. Rafael Dias Loyola

Goiânia – GO Março 2014

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

Priscila Lemes de Azevedo Silva TESE DE DOUTORADO

Prioridades para a conservação de anfíbios da Mata Atlântica

Tese apresentada à Universidade Federal

de Goiás, como parte das exigências do Programa de Pós-graduação em Ecologia e . Evolução para obtenção do título de doutor.

Orientador: Prof. Dr. Rafael Dias Loyola

Goiânia – GO Março 2014

Ficha catalográfica elaborada automaticamente com os dados fornecidos pelo(a) autor(a), sob orientação do Sibi/UFG.

Lemes de Azevedo Silva, Priscila Prioridades para a conservação de anfíbios da Mata Atlântica [manuscrito] / Priscila Lemes de Azevedo Silva. - 2014. CXIII, 113 f.

Orientador: Prof. Dr. Rafael Dias Loyola; co-orientador Dr. Atte Moilanen. Tese (Doutorado) - Universidade Federal de Goiás, Instituto de Ciências Biológicas (ICB) , Programa de Pós-Graduação em Ecologia e Evolução, Goiânia, 2014. Bibliografia. Inclui mapas, gráfico, tabelas.

1. priorização espacial. 2. mapas de extensão de ocorrência. 3. mudanças climáticas. 4. áreas prioritárias. 5. planos dinâmicos de conservação. I. Dias Loyola, Rafael, orient. II. Moilanen, Atte, co-orient. III. Título.

Dedico essa tese à minha família, em especial à minha mãe, meu amor e minha maior incentivadora…

Agradecimentos Com todo o meu coração, eu gostaria de agradecer à minha família pelo incansável incentivo aos estudos e por todo carinho dedicados a mim para que pudesse concluir essa jornada. Especialmente, gostaria de agradecer à minha mãe; foram tempos difíceis, mas estamos vencendo, juntas.

Ao meu querido John, sua chegada em minha vida foi no momento perfeito. Sou grata por todo apoio e suporte, além do amor que aumenta a cada dia. Amo você!

Pela orientação e pelo estímulo, agradeço ao Rafael Loyola. O Rafael é o tipo de orientador-inspirador; sou grata por fazer de mim uma cientista melhor. Eu realmente aprecio todo o tempo e esforço que teve para a orientação desse trabalho. Aprendi muito com você.

Agradeço também ao Adriano Melo. Foram quase dois anos até o aceite de um dos capítulos. Obrigada pela paciência.

As várias discussões entre os colegas “da sala 115” certamente possibilitaram um aprendizado e um amadurecimento científico. Para tanto, agradeço especialmente aos queridos colegas Geiziane, Marina, Nathália, Alice, Lucas

“batata”, “Douglas”, Frederico, Franciele, Leandro, Edmar, Joaquim, Maíra e

Fernando.

A todos os amigos da Pós Graduação e, principalmente, aos queridos Vanessinha,

Sid, Bruninho, Rodrigo Mello, Ricardo, Davi, Gui, Ludmila, Alessandro, Mariana,

Diogo e Luciana pela amizade, companheirismo, brincadeiras, apoio, enfim, tudo o que fizeram para tornar os anos de doutorado mais prazerosos.

Ao “Clube da Luluzinha” por me lembrar de que existe vida “além da pós”. Foram jantares e noites animadas na companhia de vocês. Em especial, agradeço às

4 minhas queridas amigas Vanessa “Van”, Dani Goeldner, Fer Cassemiro e Fer

Melo. Muito obrigada pelo apoio nos momentos difíceis, e nos felizes também.

Agradeço a co-orientação de Atte Moilanen e aos colegas do C-Big Lab, em

Helsinque na Finlândia. Feder, Peter, Joona, Enrico, Tuuli, Victoria, Johanna e

Timo, vocês tornaram a minha estadia em Helsinque mais divertida. Foi uma aventura gelada, mas de muito aprendizado. Kiitos.

Também agradeço ao Miguel Araújo pelas discussões e futuros trabalhos. Foi um curto período em Portugal, mas seus alunos David, Nacho e, em especial, minha amiga “brazuca” Fabiana, diminuíram a saudade do Brasil. Foi fixe.

A propósito, agradeço a CAPES pela bolsa-sanduíche concedida.

Aos professores do PPG em Ecologia e Evolução pelos ensinamentos ao longo desses seis anos de pós-graduação entre mestrado e doutorado.

Aos membros da banca examinadora que aceitaram o convite. Muito obrigada!

Ao CNPq, pela bolsa de doutorado concedida.

Prefácio

Esta tese é composta por cinco capítulos. O primeiro capítulo está aceito na

Revista de Biologia Neotropical. Capítulos 2-4 já estão publicados na Natureza &

Conservação, Biodiversity and Conservation e PLoS ONE, respectivamente. O capítulo 5 foi formatado nas normas da Biological Conservation. A tese conta ainda com uma compilação de suas principais novidades, apontando ainda os rumos para futuro trabalhos. Cada capítulo é um trabalho individual e alguns métodos podem se repetir ao longo da tese. As tabelas e figuras estão enumeradas e seguem a sequência de acordo com o capítulo que aparecem. A bibliografia segue imediatamente após cada capítulo. As contribuições de meus co-autores são detalhadas a seguir.

Capítulo 1: Lemes, P. & Loyola, R. Prioridades para a conservação da biodiversidade (aceito na Revista de Biologia Neotropical).

Aspectos tais como os efeitos das mudanças climáticas na biodiversidade, prioridades para a conservação da biodiversidade diante de mudanças climáticas e estratégias de conservação que incluam amplos aspectos foram discutidos nesse

6 capítulo. Também, discutimos sobre as prioridades para a conservação dos anfíbios na Mata Atlântica. PL & RL concepção teórica e escrita.

Capítulo 2: Lemes, P., Faleiro, F.V., Tessarolo, G., Loyola, R.D, 2011. Refinando dados espaciais para a conservação da biodiversidade. Natureza & Conservação

9: 240-243.

Discutimos os tipos de dados disponíveis para estudos de conservação, além das vantagens e das desvantagens do uso de mapas de extensão de ocorrência no planejamento sistemático para a conservação.

PL, FFV, GT & RDL concepção teórica e escrita.

Capítulo 3: Lemes, P., Melo, A.S., Loyola, R.D., 2014. Climate change threatens protected areas of the Atlantic Forest. Biodiversity and Conservation 23: 357-

368.

Avaliamos a eficiência das áreas protegidas em manter a riqueza de espécies no contexto de mudanças climáticas a partir de projeções consensuais dos modelos de distribuição das espécies para anfíbios da Mata Atlântica. Avaliamos se uma

área protegida poderá ganhar ou perder espécies devido às mudanças climáticas considerando a localização das mesmas no presente. Nossos resultados destacam o declínio na riqueza de espécies dentro da atual rede de áreas protegidas e, somente a altitude, é um bom preditor para perdas e ganhos do número de espécies dentro das áreas protegidas.

PL & RDL concepção teórica, ASM implementação do modelo nulo no R, PL modelagem de distribuição das espécies, análise da árvore de regressão, PL, ASM

& RDL escrita.

Capítulo 4: Lemes, P. & Loyola, R.D., 2013. Accommodating Species Climate-

Forced Dispersal and Uncertainties in Spatial Conservation Planning. PLoS ONE

8: e54323.

Usamos projeções consensuais de modelos de distribuição de espécies para identificar locais prioritários para a conservação de anfíbios da Mata Atlântica.

De maneira hierárquica, buscamos minimizar a mudança na distribuição das espécies a partir de uma medida de dispersão no contexto de mudanças climáticas. Consideramos também a atual rede de áreas protegidas estabelecidas na região, minimizando a incerteza associada aos modelos de distribuição e favorecendo locais de baixa incerteza. Nós apresentamos uma estratégia dinâmica para garantir a representação das espécies no clima atual e diante dos prováveis efeitos das mudanças climáticas.

PL & RDL concepção teórica, PL análises, PL & RDL escrita.

Capítulo 5: Integrated assessment of taxonomic, phylogenetic, and functional diversity reveals challenges and opportunities for conservation in the Brazilian

Atlantic Forest.

Integramos outras medidas de diversidade (funcional e filogenética) no planejamento para a conservação e avaliamos a coincidência entre as estratégias que visem proteger as espécies ou os processos evolutivos ou as funções 8 ecológicas desempenhadas por anfíbios. Nós encontramos alta congruência entre os planos de conservação que incluem diferentes medidas de diversidade em escala regional e por região biogeográfica. Nós destacamos oportunidades de conservação em cada região biogeográfica da Mata Atlântica, principalmente, a

Floresta de Araucária cujas soluções apresentaram maior incongruência entre as estratégias. Além disso, a Serra do Mar possui aproximadamente 62% das áreas altamente prioritárias. Nossos resultados revelam alta congruência e oportunidades para minimizar a perda de informações ecológicas e evolutivas dos anfíbios da região.

PL, RL & AM concepção teórica, PL & CFBH dados das espécies, PL e LCT modelagem de distribuição, PL & FMP análises no Zonation, TT suporte de SIG,

TFR & MGB árvore filogenética consensual e MDCC.

Resumo

Processos globais como a perda de habitat, a superexploração, a invasão de espécies exóticas e as mudanças climáticas estão conduzindo muitas espécies à extinção. Nesse contexto, o desenvolvimento de um planejamento sistemático que indique as áreas mais importantes para a conservação da biodiversidade tem sido amplamente aceito. O estabelecimento de áreas protegidas é a principal estratégia para proteção da biodiversidade e a manutenção dos processos ecossistêmicos devido à viabilidade e ao custo econômico. Todavia, a distribuição das espécies pode ser alterada pelas mudanças climáticas global e, possivelmente, a atual rede de áreas protegidas pode não ser suficiente para representar as espécies no futuro. A priorização espacial para a conservação pode antecipar os impactos das mudanças climáticas sobre a biodiversidade, além de mitigar tais impactos por meio do desenvolvimento de planos dinâmicos de conservação. No entanto, é evidente que algumas espécies têm um papel ecológico mais importante que outras devido às suas características biológicas e à história de vida, portanto, um novo desafio é adotar uma visão integrada da biodiversidade no planejamento da conservação. A Mata Atlântica é um exuberante bioma que detém 7,7% das espécies de anfíbios conhecidas do mundo e grande concentração de espécies endêmicas. Contudo, a Mata Atlântica é também um dos biomas tropicais mais ameaçados do mundo, sobretudo devido à perda e fragmentação dos hábitats naturais. Esta tese fornece propostas para os esforços de conservação, considerando os possíveis efeitos das mudanças climáticas e também os amplos aspectos da biodiversidade. Para tanto, utilizei os dados de anfíbios disponíveis na Lista Vermelha de Espécies Ameaçadas da União

Internacional para a Conservação da Natureza para toda a Mata Atlântica,

10 modelos climáticos disponíveis e informações sobre áreas protegidas oferecidas pelo World Database on Protected Areas. Os atributos das espécies e a árvore filogenética estão de acordo coma literatura específica. O capítulo 1 apresenta uma discussão sobre os impactos das mudanças climáticas e as prioridades para a conservação da biodiversidade. Ainda, é discutida a importância de incluir a diversidade funcional e filogenética nos esforços de conservação. O capítulo 2 apresenta os tipos de dados disponíveis para estudos de conservação, além das vantagens e das desvantagens do uso de mapas de extensão de ocorrência no planejamento sistemático para a conservação. O capítulo 3 apresenta a avaliação da eficiência das áreas protegidas em manter a riqueza de espécies no contexto de mudanças climáticas a partir de projeções consensuais dos modelos de distribuição das espécies para anfíbios da Mata Atlântica. Essa avaliação identifica se uma área protegida poderá ganhar ou perder espécies devido às mudanças climáticas considerando a localização da rede atual das áreas protegidas. O capítulo 4 apresenta uma abordagem alternativa para complementar a atual rede de áreas protegidas e incorpora as possíveis mudanças na distribuição das espécies. A priorização baseia-se, principalmente, na distribuição das espécies tanto no presente quanto no futuro e busca minimizar os efeitos dessa mudança na distribuição das espécies a partir de uma medida de dispersão no contexto de mudanças climáticas. Além da medida de dispersão, a solução também minimiza a incerteza associada aos modelos de distribuição, priorizando locais de baixa incerteza. O capítulo 5 apresenta a priorização espacial da conservação, incluindo diferentes aspectos da biodiversidade tal como a diversidade filogenética e funcional cujas medidas inferem tanto a história evolutiva quanto os processos ecológicos subjacentes. O principal objetivo é identificar e comparar os locais que conservam a maior informação sobre a 11 diversidade taxonômica, filogenética e funcional quanto possível. Além disso, são indicadas as prioridades para a conservação para cada região biogeográfica da

Mata Atlântica. Finalmente, em sua conclusão, são apresentadas as principais novidades da tese e discutidos os rumos para futuros trabalhos.

Palavras-chave: priorização espacial, mapas de extensão de ocorrência, mudanças climáticas, áreas protegidas, planos dinâmicos de conservação, diversidade funcional, diversidade filogenética.

Abstract

Global processes such as: habitat loss, overexploitation, invasive species and climate change are driving many species to extinction. In the face of these threats, the development of systematic planning which indicates the most important biodiversity conservation areas has become widely accepted. The establishment of protected areas is the main strategy for the protection of biodiversity and maintenance of ecosystem processes due to its feasibility and economic cost.

However, species distribution can be altered by global climate change and, possibly, the current network of protected areas may not be sufficient for species representation in future scenarios. The impact of climate change on biodiversity can be anticipated by spatial prioritization for conservation through the development of dynamic conservation plans. However, it is clear that some species have a more important ecological role than others (due to their biological and life history characteristics) which includes the unique challenge of taking an integrated view of biodiversity in conservation planning into account. The

Atlantic Forest is a lush biome which holds 7.7% of the world's known species of amphibians and high concentration of endemic species. However, the Atlantic

Forest is also one of the most threatened tropical biomes of the world, mainly due to the loss and fragmentation of natural habitats. This thesis provides proposals for conservation efforts, considering the possible effects of climate change and also the wider aspects of biodiversity. For this, I used the available amphibian data from the Red List of Threatened Species of the International Union for

Conservation of Nature, available climate models, information on protected areas by the World Database on Protected Areas, and attributes of species and the phylogenetic tree consistent with specific literature. Chapter 1 highlights a

13 discussion on the impact of climate change and priorities for biodiversity conservation and the importance of including the functional and phylogenetic diversity in conservation efforts. Chapter 2 brings a discussion about available data for conservation studies, as well as the advantages and disadvantages of using maps of extent of occurrence in systematic planning for conservation.

Chapter 3 shows the effectiveness of protected areas in maintaining species richness under climate change from consensual projections of species distribution models for amphibians inhabiting of the Atlantic Forest. In this chapter, we identified that protected areas may gain or lose species due to climate change according to the location of the current network of protected areas.

Chapter 4 presents an alternative approach to complement the existence of protected areas and incorporates possible changes in species distribution. The prioritization is based mainly on the distribution of species in both present and future scenarios. This chapter outlines a conservation plan that minimizes the effects of climate change on species dispersion. Besides these effects on species dispersion, this solution also minimizes the uncertainty associated with distribution models and prioritizing areas of low uncertainty. Chapter 5 explains the spatial prioritization of conservation, including different aspects of biodiversity, such as: phylogenetic & functional diversity measures and their influence on evolutionary history and underlying ecological processes. The main objective is to identify and compare the places that contain the most information on the taxonomic, phylogenetic and functional diversity, while also indicating the conservation priorities for each biogeographical region of the Atlantic Forest. In conclusion, we present our new and original ideas for conservation and discuss the future prospects and predictions in this area.

Keywords: Spatial prioritization, extent of occurrence maps, climate change, protected areas, dynamic conservation plans, functional diversity, phylogenetic diversity

Sumário

Capítulo 1: Lemes & Loyola. Mudanças climáticas e prioridades para a conservação da biodiversidade. Revista de Biologia Neotropical ------19 Abstract ------20 Resumo ------21 Introdução ------21 Impactos das mudanças climáticas na biodiversidade ------21 Mudanças na distribuição das espécies ------21 Prioridades para a conservação diante de mudanças climáticas ------22 O que, onde e como proteger? O caso dos anfíbios da Mata Atlântica ------23 Novos desafios para a inclusão de diferentes facetas da biodiversidade na conservação ------24 Agradecimentos ------25 Referências ------25

Capítulo 2: Lemes et al. 2011. Refinando dados espaciais para a conservação da biodiversidade. Natureza & Conservação 9: 240-243. ------30 O demônio em forma de incerteza: o déficit Walaceano ------31 Como estimar a amplitude de distribuição geográfica de uma espécie? ------32 A “Síndrome de Gabriela” ------32 Uma alternativa simples e direta para conservacionistas e acadêmicos atuantes ------33 Problema resolvido? ------33 Agradecimentos ------34 Referências ------34

Capítulo 3: Lemes et al. 2014. Climate change threatens protected areas of the Atlantic Forest. Biodiversity and Conservation 23: 357- 368. ------36 Abstract ------37

Introduction ------38 Methods ------39 Results ------41 Discussion ------42 Acknowledgements ------46 References ------46

Capítulo 4: Lemes & Loyola, 2013. Accommodating Species ClimateForced Dispersal and Uncertainties in Spatial Conservation Planning. PloS ONE 8: e54323. ------48 Abstract ------49 Introduction ------49 Methods ------50 Results ------53 Discussion ------56 Acknowledgements ------57 References ------57

Capítulo 5: Integrated assessment of taxonomic, phylogenetic, and functional diversity reveals challenges and opportunities for conservation in the Atlantic Forest ------59 Abstract ------61 Introduction ------62 Methods ------65 Results ------71 Discussion ------73 Acknowledgements ------78 References ------78 Figure Legends ------89 Supplementary Information ------90 Figure 1 ------91 Figure 2 ------92

Figure 3 ------93 Figure 4 ------94 Figure 5 ------95 Figure 6 ------96 Figure 1S ------97 Figure 2S ------98 Table 1S ------99

Conclusão: Prioridades para a conservação de anfíbios da Mata Atlântica – uma síntese ------105 Papel dos modelos de distribuição das espécies ------106 Papel da priorização para a conservação da biodiversidade ------108 Conclusões gerais ------109 Referências ------110

Nesse capítulo, discutimos aspectos tais como os efeitos das mudanças climáticas, as prioridades para a conservação, bem como as direções futuras para as estratégias de conservação.

Capítulo 1: Mudanças climáticas e prioridades para a conservação da biodiversidade

Priscila Lemes & Rafael Loyola

Artigo aceito na Revista de Biologia Neotropical

cology E Rev. Biol. Neotrop. 11(1): 47-57, 2014 / cologia E M udanças climáticas e prioridades para a conservação da biodiversidade

Priscila Lemes Universidade Federal de Goiás, Instituto de Ciências Biológicas, Programa de Pós-graduação em Ecologia e Evolução, Departamento de Ecologia, , Caixa postal 131, Goiânia, 74001- 970, Goiás, Brasil. E-mail: [email protected]

Rafael Loyola Universidade Federal de Goiás, Instituto de Ciências Biológicas, Departamento de Ecologia,Caixa postal 131, Goiânia, 74001-970, Goiás, Brasil. E-mail: [email protected]

Resumo: Processos globais como a perda de habitat, exploração excessiva de recursos naturais, in- vasão biológica e mudanças climáticas estão conduzindo muitas espécies à extinção. Nesse cenário de alto risco de extinção, qual deve ser o critério para determinar prioridades de conservação? O que, onde 47 e como proteger a biodiversidade? A resposta não é simples. Entre os efeitos esperados das mudan- ças climáticas, pode-se incluir o deslocamento das espécies para um espaço climático mais favorável, até mesmo fora de uma unidade de conservação. No entanto, eleger prioridades para a conservação da biodiversidade implica ir além das espécies, fazendo-se necessário a inclusão da história evolutiva e da manutenção dos processos nas comunidades. Aqui, apresentamos um panorama dos efeitos das mudanças climáticas sobre biodiversidade e como incluí-los em estudos de priorização espacial para a conservação. Ressaltamos a importância da conservação de anfíbios da Mata Atlântica, grupo mais ameaçado de extinção entre os vertebrados e, finalmente, apresentamos e discutimos estratégias de conservação que consideram mais que a riqueza de espécies, incluindo também informações sobre a diversidade filogenética e funcional.

Palavras-chave: priorização espacial, mudanças climáticas, diversidade filogenética, diversidade funcional, Mata Atlântica.

Abstract: Global processes such as habitat loss, overexploitation, biological invasion, and climate change are driving many species to extinction. Facing this threatening scenario, a key question is: what should be the criterion for establishing biodiversity conservation priorities? Where and how can one protect biodiversity? The answer is not so simple. Climate change could result in species moving to more favorable climatic spaces, even outside their protected areas. However, while establishing priorities for biodiversity conservation one must go beyond protecting sites with high species richness and include the evolutionary history and maintenance of community processes as well. Here, we show both an overview of the effects of climate change and their inclusion in conservation prioritization. We show a case study for the conservation of amphibians, the most endangered taxonomic group among vertebrates in the Brazilian Atlantic Forest. Finally, we discuss conservation strategies that consider phylogenetic and functional approaches.

Key words: spatial prioritization, climate change, phylogenetic diversity, functional diversity, Atlantic Forest. Introdução neamento das áreas de agricultura, de indústria ou sob influência das cidades, além da melhor alocação de recursos de conservação (Ferrier & A perda e a fragmentação de habitat, ex- Wintle, 2009). Durante esse tempo, a conver- ploração excessiva dos recursos naturais, invasão são e a degradação da terra podem continuar, e biológica e mudanças climáticas estão conduzin- a solução ideal perderá a eficiência (Meir et al., do muitas espécies à extinção. Nesse contexto, 2004; Faleiro et al. 2013). Ainda, as mudanças biólogos da conservação têm desenvolvido inú- climáticas podem reduzir a efetividade da atual meros estudos a fim de mitigar a perda dees- rede de áreas protegidas em futuras estratégias pécies (e.g. Garcia et al., 2012; Lemes & Loyola, de conservação (Hannah et al., 2007). 2013). Contudo, diante da contínua ameaça de extinção das espécies e perda dos hábitats, qual Impactos das mudanças climáticas na bio- deve ser o critério para determinar as prioridades diversidade de conservação? Há muito tempo uma questão permeia os esforços conservacionistas: o que, Não há dúvida de que estamos diante de onde e como proteger? A solução não é simples. uma crise da biodiversidade global, certamente Uma vez que não há recursos suficientes para catalisada pela influência humana (Loreau et al., proteger toda a biodiversidade do planeta (Bal- 2006). Tal crise é de grande alcance e sem prece- mford et al., 2003), fica evidente a necessidade dentes, podendo levar metade das espécies a se de eleger prioridades para a conservação. extinguirem até o final desse século. No entanto, O planejamento para a conservação é um não são apenas as taxas de extinção que estão processo de localização, configuração, implemen- aumentando, mas o âmbito geográfico das es- tação e manutenção de ações de conservação a pécies ameaçadas também está sendo ampliado fim de promover a persistência da biodiversidade (Ricketts et al., 2005). Como resultado, há uma e outros valores da natureza (Margules & Pres- enorme perda da biodiversidade com 12% de to- sey, 2000). Para tanto, o planejamento requer das as aves, 25% de todos os mamíferos e 30% o uso de técnicas quantitativas que forneçam a de todos os anfíbios, atualmente, sob algum nível informação espacial sobre as prioridades para a de ameaça de extinção (IUCN, 2013). A perda da conservação e identifiquem áreas que possuem biodiversidade é, portanto, um fenômeno global e alto valor de conservação para muitas espécies demanda esforços internacionais de conservação simultaneamente (Moilanen et al., 2009a). As- (Cardillo et al., 2006). 48 sim, diante de recursos limitados, a melhor so- As mudanças climáticas podem tornar-se a lução para a conservação deve ser compreendida maior ameaça à biodiversidade e muitos siste- como um problema geral de otimização. Tais pro- mas ecológicos já mostram seus efeitos (Garcia blemas são divididos em três grandes enuncia- et al., 2012). Nos últimos 100 anos, observam-se dos gerais: (1) o problema da cobertura mínima, alterações significativas no clima e nos extremos (2) o problema da representação máxima, e um climáticos, com impactos afetando a distribuição caso mais geral deste último, denominado (3) o geográfica de plantas e animais (Garcia et al., problema da maximização do valor de conserva- 2014). As temperaturas médias globais têm au- ção (ver Moilanen et al., 2009b). Cada um desses mentado desde a década de 70 e é muito prová- problemas é formalmente representado por uma vel que essa tendência continue no futuro (Painel função objetiva cuja equação traduz matematica- Internacional para Mudanças Climáticas - IPCC, mente o objetivo do planejamento, levando em 2007). Análises de diversos modelos climáticos consideração a complementaridade entre locais globais indicam um aumento entre 1,1 e 6,4ºC candidatos à prioritários (Moilanen et al., 2009b). até 2100 (IPCC, 2007), além de mudanças nos Diferentes fatores podem ser considerados padrões de vento, na precipitação e nas correntes no planejamento para conservação tais como as oceânicas. As projeções do IPCC (2007) também informações sobre a qualidade do hábitat, a dis- indicam um aumento da temperatura mínima tribuição das espécies, os efeitos da conectivida- diária em todos os continentes, com diminuição de, os requerimentos individuais das espécies, as dos dias com geada e ondas de frio. Tais altera- ameaças à biodiversidade e o custo econômico ções têm implicações profundas para os sistemas das ações de conservação (Margules & Sarkar naturais. As consequências relacionam-se, prin- 2007). Grande parte da literatura científica acer- cipalmente, à diminuição da aptidão da espécie, ca da priorização espacial para conservação pres- expressos em diferentes níveis e têm efeitos so- supõe que a ação necessária para a priorização é bre indivíduos, populações e comunidades (Gar- o estabelecimento de novas unidades de conser- cia et al., 2014). vação (Margules & Pressey 2000; Moilanen et al., 2009a). Teoricamente, a próxima etapa é identi- Mudanças na distribuição das espécies ficar a melhor solução dentre as disponíveis, para implementar a rede de unidades de conservação Impactos fundamentais na biodiversidade, imediatamente. Porém, na prática, raramente como as mudanças na distribuição, na abundân- esse é o caso. Outras decisões devem ser consi- cia e a variação geográfica na magnitude das res- deradas antes da implementação tal como o zo- postas às mudanças climáticas têm sido sistema- ticamente estudados (Araújo et al., 2011; Thuiller 2100 (Engler et al., 2011). Nos trópicos, os pro- et al., 2011; Garcia et al., 2012). Claramente, os váveis efeitos das mudanças climáticas também impactos humanos têm efeito na distribuição das são alarmantes. Um estudo para marsupiais do espécies. Modelos climáticos que abrangem tais Brasil predisse que, pelo menos, 67% das espé- impactos estão disponíveis (Hijmans et al., 2005) cies poderão ter a distribuição geográfica reduzi- e podem ser incluídos nos modelos de nicho eco- da em 2050 (Loyola et al., 2012). Ainda, 12% dos lógico (Peterson et al., 2011). anfíbios da Mata Atlântica poderão se extinguir Dados de sensoriamento remoto como co- regionalmente em 2080 (Lemes & Loyola, 2013). bertura do solo, índices de vegetação e mapas Temos de reconhecer, entretanto, o nível de in- de presença humana na paisagem também po- certeza dessas previsões (Diniz-Filho et al., 2009) dem ser incluídos nos modelos (Thuiller et al., e a possibilidade de que estes modelos superes- 2004). Contudo, para antecipar os efeitos das timem ou subestimem o real risco de extinção. mudanças climáticas e identificar as melhores estratégias para a conservação da biodiversi- Prioridades para a conservação diante de dade é imprescindível que se desenvolvam mo- mudanças climáticas delos que associem a distribuição das espécies aos cenários de aquecimento global e de uso do Modelos de distribuição de espécies vêm sen- solo (ver Faleiro et al., 2013 para um exemplo do usados em análises de conservação devido ao recente). Geralmente, os modelos são aplicados aumento crescente da disponibilidade de dados dentro de uma abordagem de envelope climático espaciais, da capacidade computacional dos pro- e avaliados se o nicho climático (fundamental) cessadores domésticos e da existência de pro- ocupado por uma espécie continuará, ou não, gramas específicos para a modelagem (e.g., BIO- a existir dentro da distribuição geográfica atual MOD; Thuiller et al., 2009). Assim, os modelos (Perterson et al., 2011). de distribuição de espécies podem ser úteis para Predizer o futuro, no entanto, é obviamente desenvolver planos de conservação (ver Lemes et complicado. Há uma ampla gama de modelos que al., 2011; Rangel & Loyola, 2012), especialmen- associam a ocorrência das espécies às projeções te em regiões onde a informação completa sobre no clima futuro (Peterson et al., 2011, Garcia et a distribuição das espécies não está disponível, al., 2012; Garcia et al., 2014), mas inúmeras fon- como em países megadiversos. Alguns estudos tes de incerteza dificultam a comparação entre utilizam os modelos de distribuição das espécies dois ou mais modelos de maneira definitiva. Por para o planejamento da conservação (e.g. Loisel- não haver um consenso sobre o melhor modelo le et al., 2003; Williams et al., 2005 para citar os 49 e como incluir a incerteza associada aos mesmos primeiros), mas raramente incorporam as incer- (Diniz-Filho et al., 2009), uma abordagem alter- tezas em tais exercícios (Wilson 2010). A incer- nativa e conservadora tem sido frequentemente teza diz respeito ao provável, ou possível, erro utilizada para estabelecer a direção das mudanças em relação aos dados primários que advém dos na distribuição das espécies diante de cenários de registros de ocorrência, dos dados ambientais e mudanças climáticas (Araújo & New 2007). Esta da predição do modelo de nicho ecológico (Peter- abordagem combina projeções geradas a partir son et al., 2011), e é possível explorar vários ce- de diferentes métodos de modelagem, com o in- nários de planejamento para a conservação com tuito de encontrar regiões consensuais para as diferentes níveis de incerteza (ver Meller et al., quais todos os métodos projetam presenças ou 2014, para um exemplo recente). A inclusão das ausências de espécies (Araújo & New, 2007; Gar- incertezas dos modelos de distribuição no plane- cia et al. 2012) jamento para a conservação significa basicamen- Os modelos são geralmente construídos te priorizar áreas com baixa incerteza quanto à em uma resolução espacial grosseira e podem ocorrência das espécies (Lemes & Loyola, 2013). não captar toda a variabilidade espacial de um Face à incerteza existente, a expansão de componente climático ao longo de um gradien- um sistema de unidades de conservação e mate ambiental (Lenoir et al., 2013). Contudo, os nejo, que ao mesmo tempo aumenta a área total modelos ecológicos de nicho ainda são uma im- protegida também minimize os efeitos da frag- portante ferramenta no planejamento da conser- mentação, da invasão de espécies e da poluição, vação diante de mudanças climáticas, pois avalia poderão desempenhar um papel fundamental a distribuição potencial das espécies no presente nos esforços para reduzir os impactos das mu- e no futuro (para uma revisão Peterson et al., danças climáticas na biodiversidade (Heller & Za- 2011). O cenário, entretanto, pode ser desolador. valeta, 2009). Metas de conservação atuais que Aproximadamente 57% das plantas e 34% dos têm como objetivo a conservação a longo prazo animais do mundo poderão perder mais de 50% requerem um entendimento de onde, quando e da distribuição geográfica atual em 2080 (Warren como investir nessa expansão. Para isso, méto- et al., 2013). dos quantitativos de seleção de reservas procu- Uma análise para plantas de áreas monta- ram maximizar a quantidade de biodiversidade nhosas da Europa demonstrou que entre 5% e que pode ser representada em uma rede de áreas 55% das espécies poderão perder completamen- protegidas (Margules & Pressey, 2000). Tais mé- te os hábitats com clima adequado entre 2070 e todos utilizam dados de distribuição em um de- terminado tempo e consideram que as espécies micas dentre os vertebrados da região (Haddad estarão protegidas se a rede de reservas for ade- et al., 2013). Nos últimos anos, foram descritas quadamente manejada (Moilanen et al., 2009b). pelo menos 120 novas espécies, somando mais As prioridades para a conservação são estabele- de 540 espécies para a Mata Atlântica (Haddad cidas por meio de um planejamento estático, as- et al., 2013) e, provavelmente, existem muitas sumindo que não haverá qualquer mudança dos espécies que ainda não foram descritas. A rique- recursos da biodiversidade ao longo do tempo za de espécies de anfíbios do bioma corresponde (e.g. Williams et al., 2005). a 7,7% da diversidade mundial (Frost, 2013) e, O desafio é estabelecer prioridades espaciais de todos os modos reprodutivos conhecidos, 75% para a conservação considerando os possíveis são contemplados no bioma (Haddad & Prado, efeitos das mudanças climáticas (e.g., Lemes & 2005). Tamanha diversidade pode ser atribuída Loyola, 2013; Alagador et al., 2014), uma vez à dependência da umidade, tornando as florestas que diante de cenários futuros as espécies po- o hábitat ideal para ocupação e sobrevivência de derão se deslocar para um espaço climático mais tantas espécies de anfíbios (Lion et al., 2014). favorável (Araújo, 2009). Também, a alta riqueza deve-se, em partes, pela As unidades de conservação são essenciais formação do terreno, com muitas montanhas, para desenhar estratégias de conservação em que no passado serviram como importantes bar- todo o mundo, além de manter a integridade eco- reiras para o fluxo gênico (Carnaval et al., 2009). lógica dos ecossistemas (Ladle et al., 2011). No Somado a esses fatores, existe uma elevada he- entanto, as implicações das mudanças climáticas terogeneidade de hábitats que, certamente, fa- para as unidades de conservação são múltiplas voreceu a diversificação do grupo (Haddad et al., (Hannah et al., 2007; Araújo et al., 2011). Por 2013). exemplo, os limites legais das áreas protegidas Contudo, de acordo com a lista vermelha de são fixos, mas a paisagem está em constante espécies ameaçadas de extinção (IUCN, 2013), mudança. A grande dúvida é: será que as unida- dentre as espécies conhecidas para o bioma, des de conservação de hoje são suficientes para seis espécies estão sob alguma ameaça, 93 não representar as espécies no futuro? A questão é possuem dados suficientes para ter seu nível de complexa. Primeiro porque a maioria das unida- ameaça definida, enquanto algumas, por exem- des de conservação não foi estabelecida tomando plo, Holoaden bradei (Lutz, 1958) e Ceratophrys como base os princípios ecológicos (Tsianou et ornata (Bell, 1843) já estão extintas tanto no al., 2013). Além disso, a simples presença das bioma quanto no Brasil. A vulnerabilidade apre- 50 espécies dentro de uma unidade de conservação sentada pelo grupo pode ser explicada pela de- não significa que a mesma persistirá a longo pra- pendência da integridade da floresta (Lion et al., zo (Alagador et al., 2014; Lemes et al., 2014) e, 2014), tornando-se essencial para o desenvolvi- para piorar esse quadro, muitas unidades de con- mento dos seus complexos ciclos de vida e para servação fornecem uma representação inadequa- sua sobrevivência (Haddad et al., 2013). Assim, da dos componentes da biodiversidade (Devictor o desmatamento e a fragmentação de hábitats et al., 2010). Aliás, muitas unidades de conserva- podem extinguir populações de anfíbios em am- ção são também inadequadas para a conservação bientes florestais (Lion et al., 2014), além de rede muitas espécies ameaçadas de extinção (Be- duzir a disponibilidade de abrigos e a oferta de resford et al., 2011). alimentos. Não obstante, os anfíbios não são bem representados nos estudos de conservação, O que, onde e como proteger? O caso dos quando comparados aos grupos taxonômicos me- anfíbios da Mata Atlântica nos ameaçados (Brito, 2008). Devido à urgência em delinear estratégias de A Mata Atlântica, internacionalmente conhe- conservação para os anfíbios da Mata Atlântica, cida pelo alto número de espécies e de endemis- a priorização espacial para a conservação pode mos, é um dos biomas tropicais mais ameaçados ser uma ferramenta valiosa (Trindade-Filho et al. do mundo, sobretudo pela perda e pela fragmen- 2012; Lemes & Loyola, 2013; Loyola et al., 2013; tação dos hábitats naturais (Mittermeier et al., Campos et al., 2104). Tal abordagem permite a 2004; Tabarelli et al., 2010). Historicamente, o inclusão de atributos das espécies, risco de extin- desmatamento da Mata Atlântica começou com os ção, além de remanescentes de vegetação, custo grandes ciclos econômicos do Brasil. Atualmente, da terra e áreas já protegidas. O estabelecimento 12% da vegetação natural permanece sob forma de uma rede de reservas é uma estratégia funda- de pequenos fragmentos de floresta secundária mental para a proteção de espécies (Ladle et al., e, pelo menos, 97% desses, têm menos de 250 2011), embora a eficiência da atual rede de áreas ha (Ribeiro et al., 2009). A Mata Atlântica brasi- protegidas tem sido questionada (Lemes et al., leira possui hoje ca. 2,26 milhões de hectares já 2014; Loyola et al., 2014). Por exemplo, diante protegidos (Ribeiro et al., 2009), porém, ainda das alterações globais da temperatura e da preci- não é suficiente para garantir a proteção de toda pitação, as espécies podem se deslocar para um a biodiversidade da região (Lemes et al., 2014). espaço climático mais favorável implicando que Os anfíbios da Mata Atlântica constituem o não ocorra mais em uma área já protegida (Le- grupo com a maior proporção de espécies endê- mes et al., 2014). Logo, prioridades para a con- servação devem incluir uma estratégia dinâmica sistêmicos e ao bem-estar do ser humano (e.g. de conservação que assegure a representação dispersores, polinizadores; Haines-Young & Pots- das espécies no clima atual e diante dos prová- chin, 2010) e a magnitude da perda dessa infor- veis efeitos das mudanças climáticas (Lemes & mação pode não ser facilmente predita a partir Loyola, 2013; Loyola et al., 2013). de uma ou outra espécie. Muitas medidas têm Finalmente, a conservação da biodiversidade sido desenvolvidas e permitem compreender a reside nos esforços que evitem o desmatamento estrutura da comunidade (Cadotte et al., 2010; e, consequentemente, a perda massiva de espé- Mouchet et al., 2010) indo além das clássicas cies na região (Lion et al., 2014). Nesse contex- medidas de diversidade (Magurran, 2004). Por to, uma estratégia abrangente para o estabeleci- exemplo, a diversidade funcional é uma medida mento de uma rede de áreas protegidas ao longo que reflete a diversidade de traços ecológicos, do bioma deve ser balizada por políticas públicas morfológicos e fisiológicos dentro da comunida- que incentivem e garantam a proteção e a restau- de (Petchey & Gaston, 2006). Outra medida é a ração florestal (Tabarelli et al., 2010). Como para diversidade filogenética que visa explicar o papel qualquer outro grupo, a conservação de anfíbios das interações entre as espécies e os fatores his- da Mata Atlântica depende de estratégias que en- tóricos na estrutura e composição da comunidade foquem mais do que uma espécie particular ou (Webb et al., 2002). Dessa maneira, estratégias um grupo de espécies (por exemplo, espécies eficazes de conservação devem considerar muito endêmicas ou ameaçadas ou espécies invasoras; mais que a riqueza de espécies em diferentes es- Loyola et al., 2012; Trindade-Filho et al., 2012). calas espaciais (Brooks et al., 2006). Ou seja, o Sobretudo, deve-se considerar também amplos planejamento para a conservação pode – e deve aspectos da diversidade a fim de preservar pro- – ser baseado em outras facetas da biodiversi- cessos ecossistêmicos e a história evolutiva rela- dade (Devictor et al., 2010; Zupan et al., 2014). cionadas ao grupo (Loyola et al., 2008). O conceito fundamental para a abordagem filogenética pressupõe que espécies mais- rela Novos desafios para a inclusão de diferen- cionadas (ou mais próximas) são também mais tes facetas da biodiversidade na conser- similares em suas estratégias de história de vida vação do que àquelas menos relacionadas (Winter et al., 2013). Existem evidências de que alguns cla- A ideia de urgência das ações de conservação dos filogenéticos podem ser mais suscetíveis às foi discutida por Norman Myers (1979) em seu alterações humanas do que outros (Thuiller et al., famoso conceito de “triagem”. A relação entre 2011). Distâncias evolutivas são frequentemen- 51 prioridade e grau de ameaça, no entanto, não é te correlacionadas com as potenciais diferenças linear. Por exemplo, pode se considerar inoportu- ecológicas entre espécies (Faith, 1992). Cada no o investimento em espécies sujeitas a eleva- vez mais, ecólogos têm utilizado o conhecimento dos níveis de extinção (Bottrill et al., 2008), além acerca das relações evolutivas para entender pa- das avaliações do risco de extinção contabiliza- drões ecológicos (e.g. Petchey & Gaston, 2002; rem somente a perda de cada espécie de ma- Devictor et al., 2010; Safi et al., 2011), principal- neira uniforme (Machado & Loyola, 2013). Algu- mente, devido ao aumento da disponibilidade de mas análises de priorização para a conservação árvores filogenéticas, ou dados moleculares (e.g. enfatizam áreas com grande riqueza de espécies GenBank), além das ferramentas para estimá- ou altos níveis de endemismo, nas quais diver- -las. sas espécies encontram-se sob risco iminente de Há uma série de medidas para calcular a di- extinção, ou onde já ocorreu ou está ocorrendo versidade filogenética em uma comunidade (Pa- perda substancial de habitat (Mittermeier et al., voine & Bonsall, 2011), cuja soma de todos os 2004). comprimentos dos ramos filogenéticos (PD; Fai- No entanto, esta é uma abordagem paliativa, th, 1992) ainda é a medida mais utilizada (Ca- pois corresponde à necessidade de minimizar a dotte et al., 2010). Qualquer índice pode ser apli- perda da biodiversidade em regiões onde pertur- cado tanto para diversidade filogenética quanto bações antrópicas severas já ocorreram ou ainda para diversidade funcional desde que seja basea- estão ocorrendo (Cardillo et al., 2006). Todavia, do em uma matriz de distância (Pavoine & Bon- devido às altas taxas de perda e degradação de sall, 2011). A abordagem de diversidade funcio- habitats, torna-se igualmente necessária a iden- nal considera que um atributo funcional pode ser tificação de áreas nas quais os impactos humanos qualquer característica individual mensurável que podem ser atualmente pequenos, mas o risco de afeta as relações ecológicas em qualquer locali- extinção das espécies no futuro é alto (Loyola et dade (Mouchet et al., 2010). Logo, a diversidade al., 2008; Machado & Loyola, 2013). funcional estima as diferenças entre organismos É sabido que as espécies possuem diferen- (Cianciaruso et al., 2009), independentemente tes papéis no funcionamento do ecossistema, de suas relações evolutivas. tornando-se importante considerar tais diferen- A inclusão de diferentes facetas da biodiver- ças no âmbito da conservação da natureza. Além sidade na priorização espacial para a conserva- disso, alguns grupos de espécies têm um papel ção ainda é um desafio. Por exemplo, os ecólo- ecológico mais claro quanto aos serviços ecos- gos de comunidade buscam compreender quais padrões e processos determinam a estrutura da Conservation Planning. pp. 172-184, In: A. comunidade (Petchey & Gaston, 2002; Safi et al., Moilanen, K. A. Wilson & H. P. Possingham 2011), porém, essa abordagem é raramente apli- (eds). Spatial Conservation Prioritization: cada na biologia da conservação. Recentemente, biólogos da conservação mapearam diferentes Quantitative Methods and Computational facetas da biodiversidade a fim de mitigar o ris- Tools. Oxford University Press, Oxford. co de extinção (Devictor et al., 2010), além de Araújo, M.B. & M. New. 2007. Ensemble fore- encontrar incompatibilidades espaciais entre a casting of species distributions. Trends Ecol. diversidade taxonômica, filogenética e funcional. Evol. 22: 42-47. A incongruência espacial nos padrões de diversidade foi ilustrada para aves na França (Devictor Araújo, M.B., D. Alagador, M. Cabeza, D. No- et al., 2010) e, recentemente, tal descompasso gués-Bravo & W. Thuiller. 2011. Clima- também foi demonstrado para mamíferos, aves te change threatens European conservation e anfíbios da Europa (Zupan et al., 2014). Em areas. Ecol. Lett. 14: 484-492. contrapartida, alguns estudos apontam uma alta Balmford A., K. J. Gaston, K.J., S. Blyth, A. congruência entre a diversidade taxonômica e a diversidade filogenética (Rodrigues et al., 2011; James & V. Kapos. 2003. Global variation Strecker et al., 2011) cujo padrão pode estar re- in terrestrial conservation costs, conserva- lacionado à lacuna no conhecimento das relações tion benefits, and unmet conservation nee- filogenéticas entre as espécies (Lacuna Darwinia- ds. Proc. Natl. Acad. Sci. USA. 100: 1046- na; Diniz-Filho et al., 2013). O desafio atual reside em alinhar problemas 1050. reais, como o montante de recursos disponíveis Beresford, A. E., G. M. Buchanan, P.F. Donald, para a conservação e a representação da biodi- S. H. M. Butchart, L. D. C. Fishpool & C. versidade em uma rede eficiente de unidades de Rondinini. 2011. Poor overlap between the conservação. Tendências recentes na literatura distribution of Protected Areas and globally incluem uma medida de incerteza na priorização (Meller et al., 2014) e os possíveis efeitos das threatened birds in Africa. Anim. Conserv. mudanças climáticas no processo de tomada de 14: 99-107. decisão (Lemes & Loyola, 2013). A identificação Bottrill, M. C., L. N. Joseph, J. Carwardine, M. de um conjunto abrangente de áreas naturais é C. BodeCook, E. T Game, H. Grantham, 52 apenas o primeiro passo para uma estratégia de S.Kark, S. Linke, E. McDonald-Madden, conservação da biodiversidade in situ (Devictor et al., 2010; Zupan et al., 2014) e requer um R. L. Pressey, S. Walker, K.A. Wilson & processo mais complexo desde a negociação das H. P. Possingham. 2008. Is conservation terras até a efetiva implementação das novas triage just smart decision making? Trends áreas protegidas (Knight et al., 2009). Contudo, Ecol. Evol. 23: 649-654. o estabelecimento de prioridades espaciais pode e deve ajudar tomadores de decisão embasar po- Brito, D. 2008. Amphibian conservation: Are we líticas de conservação. on the right track? Biol. Conserv. 141: 2912- 2917. Agradecimentos Brooks, T. M., R. A. Mittermeier, G. A. B. Da Fonseca, J. Gerlach, M. Hoffmann, J. F. Somos gratos a Rodrigo Daud pela gentileza do convite para submeter esse artigo à Revista de Lamoreux, C. G. Mittermeier, J. D. Pil- Biologia Neotropical. PL foi bolsista de doutorado grim & A. S. L. Rodrigues. 2006. Global do CNPq (#147345/2010-3) e bolsista de dou- Biodiversity Conservation Priorities. Science torado-sanduíche pela Capes (#12583/12-0). A 313: 58-61. pesquisa de RL vem sendo constantemente apoia- Cadotte, M. W., J. Davies, J. Regetz, S. W. da pelo CNPq (#304703/2011-7, #479959/2013- 7, #407094/2013-0), Rede CLIMA, Conservação Kembel, E. Cleland & T. H. Oakley. 2010. Internacional Brasil e Fundação Grupo Boticário Phylogenetic diversity metrics for ecological de Proteção à Natureza (PROG_0008_2013). communities: integrating species richness, abundance and evolutionary history. Ecol. Referências Lett. 13: 96-105. Campos, F. S., J. Trindade-Filho, D. Brito, G. Alagador, D., J. O. Cerdeira & M. B. Araújo. A. Llorente & M. Solé. 2014. The efficiency 2014. Shifting protected areas: scheduling of indicator groups for the conservation of spatial priorities under climate change. J. amphibians in the Brazilian Atlantic Forest. Appl. Ecol. 51: 703-713. Ecol. Evol 4: 2505-2514. Araújo, M. B. 2009. Climate Change and Spatial Cardillo, M., G. Mace, J. L. Gittleman & A. Purvis. 2006. Latent extinction risk and the pp. 172-184. In: A. Moilanen, K. A. Wilson future battlegrounds of mammal conserva- & H. P. Possingham (eds). Spatial Conser- tion. Proc. Nat. Acad. Sci. USA 103: 4157- vation Prioritization: Quantitative Methods 4161. and Computational Tools. Oxford University Carnaval, A. C., M. J. Hickerson, & C. F. B. Press, Oxford. Haddad. 2009. Stability Predicts Gene- Frost, D.R. 2013. Amphibian Species of the tic Diversity in the Brazilian Atlantic Forest World: an Online Reference. Version 6.0. Hotspot. Science 323: 785-789. American Museum of Natural History. New Cianciaruso, M. V., M. A. Batalha, K. J. Gas- York, USA. Disponível em: http://research. ton, & O. L. Petchey. 2009. Including in- amnh.org/herpetology/amphibia/index.html traspecific variability in functional diversity. Garcia, R. A., N. D. Burgess, M. Cabeza, C. Ecology 90: 81-89. Rahbek & M. B. Araújo, M. B. 2012. Ex- Devictor, V., D. Mouillot, C. Meynard, F. Ji- ploring consensus in 21st century projec- guet, W. Thuiller & N. Mouquet. 2010. tions of climatically suitable areas for African Spatial mismatch and congruence between vertebrates. Glob. Change Biol. 18: 1253- taxonomic, phylogenetic and functional di- 1269. versity: the need for integrative conserva- Garcia, M. Cabeza, C. Rahbek & M. B. Araújo. tion strategies in a changing world. Ecol. 2014. Multiple Dimensions of Climate Chan- Lett. 13: 1030-1040. ge and Their Implications for Biodiversity. Diniz-Filho, J. A. F., R. D. Loyola, P. Raia, A. Science 344:1247579. O. Mooers & L. M. Bini. 2013. Darwinian Haddad, C. F. & C. P. Prado. 2005. Reproducti- shortfalls in biodiversity conservation. Tren- ve modes in frogs and their unexpected di- ds Ecol. Evol. 28: 689-695. versity in the Atlantic Forest of Brazil. BioS- Diniz-Filho, J. A. F., L. M. Bini, T. F. L. V. B. cience 55: 207-217. Rangel, R.D. Loyola, C. Hof & M. B. Araú- Haddad, C.F.B., L. F. Toledo, C. P. A. Prado, jo. 2009. Partitioning and mapping uncer- D. Loebmann, J. L. Gasparini & I. Sazi- 53 tainties in ensembles of forecasts of species ma. 2013. Guia dos anfíbios da Mata Atlân- turnover under climate change. Ecography tica: Diversidade e Biologia. Anolisbook, São 32: 897-906. Paulo, 544p. Engler, R., C. F. Randin, W. Thuiller, S. Haines-Young, R. & M. Potschin. 2010. The Dullinger, N. E. Zimmermann, M. B. links between biodiversity, ecosystem servi- Araújo, P. B. Pearman, G. L. L. Lay, C. ces and human well-being. p:110-139. In: Piedallu, C. H. Albert, P. Choler, G. Col- R. Bardgett, M. Usher & D. Hopkins (eds) dea, X. D. Lamo, T. Dirbonk, J. C. Gé- Ecosystem Ecology: A New Synthesis 1st ed. gout, D. Gómez-García, J. A. Grytnes, E. Cambridge: Cambridge University Press. Heegaard, F. Høistad, D. Nogués-Bravo, Hannah, L., G. Midgley, S. Andelman, M. B. S. Normand, M. Puscas, M. T. Sebastià, Araújo, G. Hughes, E. Martinez-Meyer, A. Stanisci, J. P Theurillat, M. R. Trivedi, R. Pearson & P. Williams. 2007. Protec- P. Vittoz & A. Guisan. 2011. 21st Century ted area needs in a changing climate. Front. Climate Change Threatens Mountain Flora Ecol. Environ. 5: 131-138. Unequally Across Europe. Glob. Change Biol. Heller, N. E. & E. S. Zavaleta. 2009. Biodiversi- 17: 2330-2341. ty management in the face of climate chan- Faith, D. P. 1992. Conservation evaluation and ge: A review of 22 years of recommenda- phylogenetic diversity. Biol. Conserv. 61: tions. Biol. Conserv. 142: 14-32. 1-10. Hijmans, R.J., S. E. Cameron, J. L. Parra, P. Faleiro, F. V., Machado, R.B. & R. D. Loyola. G. Jones & A. Jarvis. 2005. Very high re- 2013. Defining spatial conservation priorities solution interpolated climate surfaces for in the face of land-use and climate change. global land areas. Int. J. Climatol. 25: 1965- Biol. Conserv. 158: 248-257. 1978. Ferrier, S. & B. A. Wintle. 2009. Quantitative IPCC. 2007. Climate Change 2007 – Impacts, Approaches to Spatial Conservation Prioriti- Adaptation and Vulnerability: Contribution zation: Matching the Solution to the Need. of Working Group II to the Fourth Assess- ment Report of the IPCC. Cambridge Univer- Loiselle, B. A., C. A. Howell, C. H. Graham, J. sity Press, Cambridge. M. Goerck, T. Brooks, K. G. Smith & P. H. IUCN. 2013. IUCN Red List of Threatened Spe- Williams. 2003. Avoiding Pitfalls of Using cies. Version 2013.2. < http://www.iucnre- Species-Distribution Models in Conservation dlist.org >. Consulta em 30 de novembro Planning. Conserv. Biol. 17: 1-10. 2013. Loreau, M., A. Oteng-Yeboah, M. T. K. Ar- Knight, A., R. M. Cowling, H. P. Possinghma royo, D. Babin, R. Barbault, M. Dono- & K.A. Wilson. 2009. From Theory to Prac- ghue, M. Gadgil, C. Häuser, C. Heip, A. tice: Designing and Situating Spatial Prio- Larigauderie, K. Ma, G. Mace, H. A. Moo- ritization Approaches to Better Implement ney, C. Perrings, P. Raven, J. Sarukhan, Conservation Action. pp. 249-258, In: A. P. Schei, R. J. Scholes & R. T. Watson. Moilanen, K. A. Wilson & H. P. Possingham 2006. Diversity without representation. Na- (eds). Spatial Conservation Prioritization: ture 442: 245-246. Quantitative Methods and Computational Loyola, R. D., P. Lemes, F. V. Faleiro, J. Trin- Tools. Oxford University Press, Oxford. dade-Filho, R. B. Machado. 2012. Severe Ladle, R.J., P. Jepson & L. Gillson. 2011. So- loss of suitable climatic conditions for mar- cial values and conservation biogeography. supial species in Brazil: challenges and op- In: Ladle, R.J., Whittaker, R.J (ed).Conser- portunities for conservation. PLoS ONE 7: vation Biogeography, Wiley-Blackwell. 301p e46257. Lemes, P. & R. D. Loyola. 2013. Accommo- Loyola, R. D., P. Lemes, J. C. Nabout, J. dating species climate-forced dispersal and Trindade-Filho, M. D. Sagnori, R. Do- uncertainties in spatial conservation plan- brovolski & J. A. F. Diniz-Filho. 2013. A ning. PLoS ONE 8: e54323. straightforward conceptual approach for eva- Lemes, P., F. A. M. V. Faleiro, G. Tessarolo & luating spatial conservation priorities under R. D. Loyola. 2011. Refinando Dados Espa- climate change. Biod. Conserv. 22: 483-495. 54 ciais para a Conservação da Biodiversidade. Loyola, R. D., Lemes, P., Brum, F. T., Prove- Nat. Conservacao 9: 240–243. te, D. B. & L. D. S. Duarte. 2014. Clade- Lemes, P., A. S. Melo & R. D. Loyola. 2014. -specific consequences of climate change Climate change threatens protected areas of to amphibians in Atlantic Forest protected the Atlantic Forest. Biodivers. Conserv. 23: areas. Ecography 37: 65-72. 357–368. Loyola, R.D., C. G. Becker, U. Kubota, C. F. B. Lenoir, J., B. J. Graae, P. A. Aarrestad, I. Haddad, C. R. Fonseca & T. M. Lewinso- Greve, A. S. S. Armbruster, G. Aus- hn. 2008. Hung out to dry: choice of priority trheim, C. Bergendorff, H. J. B. Birks, K. ecoregions for conserving threatened Neo- A. Brathen, J. G. Brunet, H. H. Bruun, C. tropical anurans depend on their life-history J. Dahlberg, G. Decocq, M. Diekmann, traits. PLoS ONE 3: e2120. M. Dynesius, R. Ejrnæs, J. A. Grytnes, Machado, N. & R. D. Loyola. 2013. A Com- K. Hylander, K. Klanderud, M. Luoto, prehensive Quantitative Assessment of A. Milbau, M. Moora, B. Nygaard, A. Bird Extinction Risk in Brazil. PLoS ONE 8: Odland, V. T. Ravolainen, S. Reinhar- e72283. dt, S. M. Sandvik, F. H. Schei, J. D. M. Magurran, A. E. 2004. Measuring biological di- Speed, L. U. Tveraabak, V. Vandvik, L. versity. Blackwell Publishing, Oxford, UK. 215p V. Velle, R. Virtanen, M. Zobel & J. C. Margules, C. R. & R. L. Pressey. 2000. Sys- Svenning. 2013. Local temperatures infer- tematic conservation planning. Nature 405: red from plant communities suggest strong 243–253. spatial buffering of climate warming across Margules, C.R. & S. Sarkar. 2007. Systematic Northern Europe. Glob. Change Biol. 19: Conservation Planning. Cambridge: Cam- 1470-1481. bridge University Press. 269p. Lion M. B., A. A. Garda & C. R. Fonseca. 2014. Meir, E., S. Andelman & H. P. Possingham, Split distance: a key landscape metric shaping H.P. 2004. Does conservation planning mat- amphibian population and communities in fo- ter in a dynamic and uncertain world? Ecol. rest fragments. Divers. Dist. 20: 1245-1257. Lett. 7: 615-622. Meller, L., M. Cabeza, S. Pironon, M. Bar- Rangel, T.F. & R. D. Loyola. 2012. Labeling bet–Massin, L. Maiorano, D. Georges & ecological niche models. Nat. Conservacao W. Thuiller. 2014. Ensemble distribution 10: 119-126. models in conservation prioritization: from Ribeiro, M. C., J. P. Metzger, A. C. Martensen, consensus predictions to consensus reserve F. J. Ponzoni & M. M. Hirota. 2009. The networks. Divers. Distrib. 20: 309-321. Brazilian Atlantic Forest: How much is left, Mittermeier, R.A., P. Robles-Gil, M. Hoffman, and how is the remaining forest distributed? J. Pilgrim, T. Brooks, C. G. Mittermeier, Implications for conservation. Biol. Conserv. C. G. Lamoreux & G. A. B. Da Fonseca. 142: 1141-1153. 2004. Hotspots revisited: Earth’s biologically Ricketts, T. H., Dinerstein, E., Boucher, T., richest and most endangered terrestrial eco- Brooks, T. M., Butchart, S. H.M., Hoff- regions. CEMEX, Conservation International, mann, M., Lamoreux, J.F., Morrinson, J., and Agrupación Sierra Madre, Monterrey, Parr, M., Pilgrim, J.D., Rodrigues, A.S., México Sechrest, W., Wallace, G.E., Berlin, K., Moilanen, A., H. Kujala & J. R. Leathwick. , J., Burguess, N.D., Church, D.R., Cox, 2009a. The Zonation Framework and Sof- N., Knox, D., Loucks, C., Luck, G.W., tware for Conservation Prioritization. pp. Master, L.L., Moore, R., Naidoo, R., Rid- 192-210. In: A. Moilanen, K. A. Wilson & gely, R., Schatz, G.E., Shire, G., Strand, H. P. Possingham (eds). Spatial Conserva- H., W. Wettengel & E. Wikramanayake. tion Prioritization: Quantitative Methods 2005. Pinpointing and preventing imminent and Computational Tools. Oxford University extinctions. Proc. Natl. Acad. Sci. USA 102: Press, Oxford. 18497-18501. Moilanen, A., H. P. Possingham & S. Polasky. Rodrigues, A. S. L., Grenyer, R., Baillie, J. 2009b. A Mathematical Classification of Con- E. M., Bininda-Emonds, O. R. P., Gittle- servation Prioritization Problems. In: Spatial mann, J. L., Hoffmann, M., Safi, K., Schi- Conservation Prioritization: Quantitative Me- pper, J., Stuart, S. N. & T. Brooks. 2011. 55 thods and Computational Tools. A. Moilanen, Complete, accurate, mammalian phylo- K. A. Wilson & H. P. Possingham (eds), pp. genies aid conservation planning, but not 249-258, Oxford University Press, Oxford. much. Philos. Trans. R. Soc. B 366, 2652- Mouchet, M. A., S. Villéger, N. W. H. Mason 2960. & D. Mouillot. 2010. Functional diversity Safi, K., M. V. Cianciaruso, R. D. Loyola, D. measures: an overview of their redundancy Brito, K. Armour-Marshall & J. A. F. Di- and their ability to discriminate community niz-Filho. 2011. Understanding global pat- assembly rules. Funct. Ecol. 24: 867-876. terns of mammalian functional and phyloge- Myers, N. 1979. The sinking Ark. A new look at netic diversity. Philos. T. Roy. Soc. B 366: the problem of disappearing species. Perga- 2536-2544. mon, New York. Strecker, A. L., Olden, J. D., Whittier, J. B. & Pavoine, S. & M. B. Bonsall. 2011. Measuring C. P. Paukert. 2011. Defining conservation biodiversity to explain community assembly: priorities for freshwater fishes according to a unified approach. Biol. Rev. 86: 792-812. taxonomic , functional , and phylogenetic di- Petchey, O.L. & K. J. Gaston. 2002. Functional versity. Ecol. Appl. 21, 3002-3013. Diversity (FD), species richness, and com- Tabarelli, M., A. V. Aguiar, M. C. Ribeiro, J. munity composition. Ecol. Lett. 5: 402-411. P. Metzger & C. A. Peres. 2010. Pros- Petchey, O.L. & K. J. Gaston. 2006. Functio- pects for biodiversity conservation in the nal diversity: back to basics and looking for- Atlantic Forest: Lessons from aging human- ward. Ecol. Lett. 9: 741-758. -modified landscapes. Biol. Conserv. 143: Peterson, A.T., J. Soberón, R. G. Pearson, 2328-2340. R.P. Anderson, E. Martínez-Meyer, M. Thuiller, W., L. Brotons, M. B. Araújo & S. Nakamura & M. B. Araújo. 2011. In: H.S. Lavorel. 2004. Effects of restricting envi- Horn & Levin S.A. (eds). Ecological niches ronmental range of data to project current and geographic distributions. Princeton Uni- and future species distributions. Ecography versity Press, Princeton. 316p. 27: 165-172. Thuiller, W., B. Lafourcade, R. Engler & M. B. Webb, C. O., D. D. Ackerly, M. A. McPeek & Araújo. 2009. BIOMOD – a platform for en- M. J. Donoghue. 2002. Phylogenies and semble forecasting of species distributions. Community Ecology. Annu. Rev. Ecol. Syst. Ecography 32: 369-373. 33: 475-505. Thuiller, W., S. Lavergne, C. Roquet, I. Bou- Williams, P. H., L. Hannah, S. Andelman, langeat, B. Lafourcade & M. B. Araújo. G. Midgley, M. B. Araújo, G. Hughes, L. 2011. Consequences of climate change on the Manne, E. Martinez-Meyer, R. Pearson tree of life in Europe. Nature 470: 531-534. & K. A. Wilson. 2005. Planning for Clima- Trindade-Filho, J., R. A. Carvalho, D. Brito & te Change: Identifying Minimum-Dispersal R. D. Loyola. 2012. How does the inclusion Corridors for the Cape Proteaceae. Conserv. of Data Deficient species change conserva- Biol. 19: 1063-1074. tion priorities for amphibians in the Atlantic Wilson, K. A. 2010. Dealing with Data Uncer- Forest? Biod. Conserv. 21: 2709-2718. tainty in Conservation Planning. Nat. Con- Tsianou, M. A., A. D. Mazaris, A. S. Kallima- servacao 8: 145-150. nis, P. K. Deligioridi, E. Apostolopoulou Winter, M., V. Devictor & O. Schweiger. 2013. & J. D. Pantis. 2013. Identifying the crite- Phylogenetic diversity and nature conserva- ria underlying the political decision for the tion: where are we? Trends Ecol. Evol. 28: prioritization of the Greek Natura 2000 con- 199-204. servation network. Biol. Conserv. 166: 103- Zupan, L., M. Cabeza, L. Maiorano, C. Ro- 110. quet, V. Devictor, S. Lavergne, D. Moui- Warren, R., J. VanDerWal, J. Price, J. A. Wel- llot, N. Mouquet, J. Renaud & W. Thuil- bergen, I. Atkinson, J. Ramirez-Villegas, ler. 2014. Spatial mismatch of phylogenetic T. J. Osborn, A. Jarvis, L. P. Shoo, S. E. diversity across three vertebrate groups and Williams & J. Lowe. 2013. Quantifying the protected areas in Europe. Divers. Distrib. benefit of early climate change mitigation in 20: 674-685. 56 avoiding biodiversity loss. Nat. Clim. Change Recebido em 25.VI.2014 3: 678-682. Aceito em 18.VII.2014

Nesse capítulo, discutimos sobre os tipos, as vantagens e desvantagens dos dados disponíveis para os estudos de conservação.

Capítulo 2: Refinando dados espaciais para a conservação da biodiversidade

Priscila Lemes, Frederico V. Faleiro, Geiziane Tessarolo & Rafael Loyola

Natureza & Conservação, 9: 240-243.

Fórum

Refinando Dados Espaciais para a Conservação da Biodiversidade

Refining Spatial Data for Biodiversity Conservation

Priscila Lemes1*, Frederico Augusto Martins Valtuille Faleiro1, Geiziane Tessarolo2 & Rafael Dias Loyola1

1 Laboratório de Biogeografia da Conservação, Departamento de Ecologia, Instituto de Ciências Biológicas – ICB, Universidade Federal de Goiás – UFG, Goiânia, GO, Brazil 2 Laboratório de Ecologia Teórica e Síntese, Departamento de Ecologia, Instituto de Ciências Biológicas – ICB, Universidade Federal de Goiás – UFG, Goiânia, GO, Brazil

A conservação in situ é a pedra angular da biologia da O Demônio em Forma de Incerteza: o conservação, i.e., proteger populações em locais onde Déficit Wallaceano estas ocorrem naturalmente, ainda hoje, se apresenta globalmente como a estratégia de conservação mais viável e Um dos gargalos mais apertados no momento do econômica (Loucks et al. 2008). Entretanto, desde a escala planejamento para a conservação é o fato de que, para regional até a global, a eficiência das áreas protegidas em uma infinidade de espécies, as distribuições geográficas são representar os diversos atributos da biodiversidade ainda pouco conhecidas e possuem inúmeras lacunas - problema é baixa (Rodrigues et al. 2004). Fundamentalmente, essa conhecido como déficit Wallaceano (Whittaker et al. 2005), ineficiência advém do fato de que o estabelecimento de em homenagem a Alfred R. Wallace, o pai da zoogeografia. novas reservas não tem base científica, importando mais, Diante dessa incerteza sobre onde as espécies ocorrem, na prática, a vontade política e a oportunidade imediata. usa-se, normalmente, uma inferência sobre tal distribuição. Como alternativa a esse tipo de priorização surgiu a ciência Todavia, ao inferir a distribuição de uma espécie, dois do planejamento sistemático para a conservação (Margules erros fundamentais podem ser cometidos. O primeiro é o & Pressey 2000), que visa propor o melhor conjunto de erro de omissão: omite-se a ocorrência da espécie em uma locais para a conservação e manejo da biodiversidade dentre determinada área, quando ela ocorre ali de fato. O segundo aqueles disponíveis, satisfazendo princípios-chave como é o erro de comissão (ou sobreprevisão): pressupõe-se a abrangência, adequação, representatividade e eficiência ocorrência da espécie em um local onde ela não ocorre (Wilson et al. 2009). verdadeiramente. Tais erros minam o planejamento para a O desenvolvimento e a implementação política de planos conservação e geram cenários de incerteza que atormentam de conservação que atendam a esses princípios representa os tomadores de decisão. O pior deles, chamemo-lo de um ganho de qualidade e eficiência na criação e manejo de cenário da “extinção por exagero”, é considerar (devido a reservas. Contudo, o planejamento sistemático necessita um erro de comissão) que uma espécie será protegida em de uma série de dados para seu desenvolvimento, sendo os um local caso o mesmo seja demarcado como uma reserva, dados espaciais extremamente importantes e necessários. quando ela não ocorre ali. Esse erro de julgamento poderia Particularmente em países megadiversos, dados acurados levar à extinção de populações dessa espécie em regiões sobre a distribuição geográfica das espécies são escassos e fora da reserva, onde a espécie de fato ocorre. O segundo difíceis de serem obtidos. Tradicionalmente, o planejamento pior cenário é o da “extinção por ignorância”. Nesse caso, para a conservação nesses países é feito com base em dados o número de locais disponíveis para a implementação da de baixa qualidade, normalmente disponíveis com mapas ação de conservação é substancialmente reduzido, pois a de extensão de ocorrência de espécies. ocorrência da espécie nos mesmos é desconhecida (devido a um erro de omissão), levando a uma menor eficiência *Send correspondence to: Priscila Lemes do planejamento e possível extinção das populações Departamento de Ecologia, Instituto de Ciências Biológicas, Universidade Federal de Goiás – UFG, desconhecidas. Esses problemas crescem exponencialmente CP 131, CEP 74001-970, Goiânia, GO, Brasil quando o planejamento é executado para inúmeras espécies e-mail: [email protected] - as vezes milhares - ao mesmo tempo (e.g. Loyola et al. 2009). Refinando Dados Espaciais para a Conservação 241

Como Estimar a Amplitude de de dados espaciais é parca. A desvantagem desses mapas, Distribuição Geográfica de uma além dos altos erros de comissão já mencionados, reside Espécie? na incerteza envolvida no uso da extensão de ocorrência. Tal incerteza é bastante difícil de ser avaliada, em função Atualmente, existem três alternativas básicas para estimar da subjetividade implícita ao desenho do polígono (e.g. a distribuição geográfica de uma espécie a partir de dados escala original em que foi desenhado o mapa e os critérios espaciais pré-existentes (ver Boitani et al. 2011 para uma adotados para sua delimitação). Na maioria das vezes tais discussão recente): 1) o uso de registros de ocorrência de subjetividades sequer são reportadas. Além disso, via de indivíduos de uma espécie, obtidos em campo ou a partir de regra estes mapas não são validados. coleções abrigadas em museus; 2) o uso de mapas de extensão Finalmente, mapas preditivos de distribuição, species de ocorrência, disponíveis em bases de dados acessíveis na distribution models - SDMs, (ver Franklin 2009 para uma internet; e 3) o uso de mapas preditivos de distribuição, revisão atual) vêm sendo usados como dados de entrada gerados a partir de modelos de distribuição de espécies. em análises de conservação, principalmente devido ao Todos esses dados possuem vantagens e desvantagens se aumento crescente da disponibilidade de dados espaciais considerarmos sua aplicação para a conservação. como os mencionados anteriormente, de Sistemas de De maneira geral, registros de ocorrência são bastante Informação Geográfica (SIG), da capacidade computacional enviesados. Há inúmeros problemas associados à amostragem dos processadores domésticos e da existência de programas em amplas escalas geográficas (e.g. continental), há problemas específicos para a modelagem. Apesar de sua suposta relacionados à coleta per se e relatados amiúde pela literatura sofisticação, a aplicação desses modelos em conservação pertinente (e.g. coletas em locais de fácil acesso ao coletor (assim como para outros usos) padece de problemas e próximos a estradas abundam) e há pouca ou nenhuma básicos como a ausência de consenso sobre a melhor informação sobre a ausência real de indivíduos e populações maneira de validá-los (principalmente com uma validação - não se faz um inventário para relacionar as espécies que biológica empírica e não apenas estatística) e a elevada não ocorrem em um local, senão aquelas que ocorrem. incerteza associada aos mesmos (ver Diniz-Filho et al. Devido a isso, esses dados possuem baixos erros de comissão 2009). Fontes de incerteza existem inúmeras (origem dos e altos erros de omissão, embora tais problemas sejam dados, algoritmo usado na modelagem, combinação de dependentes do esforço amostral. variáveis usadas para calibrar os modelos, métodos de validação, modelos climáticos etc.) e o busílis reside no fato Mapas de extensão de ocorrência, em contrapartida, são de que, atualmente, não é possível comparar de maneira polígonos espacializados, que tendem a incorrer em altos definitiva a qualidade de dois ou mais modelos (mas veja erros de comissão (falsas presenças) devido à interpolação Araújo & New 2007 para uma alternativa paliativa a esses geográfica de registros de ocorrência. O procedimento lógico problemas). Isso sem adicionar o fato de que os dados de elaboração destes dados envolve mapear os registros brutos para a modelagem vêm de registros de ocorrência de que presumidamente encontram-se na porção mais ou mapas de extensão de ocorrência cujas limitações já periférica da região de ocorrência de uma espécie e usar foram aqui expostas. Portanto, sua qualidade permanece uma interpolação para delinear os limites da distribuição ainda dependente da qualidade e quantidade de registros desta espécie. Existem, contudo, diferentes maneiras de ocorrência - toda discussão em contrário é simplesmente para a definição do contorno da distribuição. Assim, a debalde. Entretanto, essa constatação reforça a importância interpolação que gera o polígono dentro do qual os registros fundamental de investimentos em inventários com desenho se encontram pode ser estritamente baseada em dados de amostral cuidadosamente planejado. ocorrência (e.g. Polígono Convexo Mínimo) – ou seja, uma abordagem definida pelos próprios dados disponíveis A “Síndrome de Gabriela” (data-driven) – mas também mais informal e baseada em diferentes graus de conhecimento das condições ambientais Por que usamos dados como os mencionados acima em (e.g. limites que seguem condições ambientais inóspitas análises de conservação afinal? A resposta óbvia e ingênua para a espécie), ou ainda na opinião de especialistas (e.g. para essa pergunta é embaraçosamente simples: porque, limites que excluem áreas onde a espécie não ocorre por muitas vezes, essa é a única forma de fazer algum tipo razões históricas) – ou seja, uma abordagem definida pelo de planejamento. Especialmente em regiões com alta conhecimento prévio (knowledge-driven). É possível que biodiversidade, pobres em dados e com deficiências na tais mapas sejam ainda gerados por uma abordagem que proteção de espécies (países tropicais e emergentes, em sua reúne os dois métodos acima. A grande vantagem dos maioria), registros de ocorrência são escassos e polígonos de mapas de extensão de ocorrência reside no fato de que esses distribuição constituem a única informação espacial sobre mapas, por serem generalizações espaciais da distribuição a amplitude de distribuição das espécies. Em escala global, geográfica das espécies, estão disponíveis em escala global por exemplo, análises com dados espaciais extremamente para uma série de grupos taxonômicos (especialmente para refinados tornam-se simplesmente impossíveis de serem os vertebrados). Isso permite o delineamento de planos conduzidas. A pergunta fundamental, portanto, e de de conservação mesmo em regiões cuja disponibilidade resposta não trivial, é outra: devemos manter a popular e 242 Lemes et al. Natureza & Conservação 9(2):240-243, December 2011 inercial “síndrome de Gabriela” (imortalizada pela modinha prévio ou especializado para que a modelagem possa ser de Dorival Caymmi, cujo refrão diz: “Eu nasci assim, eu realizada a contento. cresci assim, eu sou mesmo assim, vou ser sempre assim...”) Recentemente, Rondinini et al. (2011) criaram mapas de ou, diante dos desafios e da crise atual de biodiversidade, hábitat adequado para mais de 5.000 espécies de mamíferos buscaremos soluções simples e/ou criativas para o problema? terrestres, filtrando mapas de extensão de ocorrência por O uso de modelos de distribuição de espécies é, portanto, variávies como cobertura vegetal, elevação e características extremamente promissor. Contudo, análises de conservação hidrológicas do terreno. Os novos mapas foram produzidos baseadas nos mesmos devem ser extremamente avaliadas a uma resolução espacial de 300 m, algo extremamente sob o olhar crítico da falta de consenso sobre a precisão fino para uma avaliação global. A construção desses mapas e acurácia desses modelos. Minimamente, as incertezas permitiu uma análise detalhada dos padrões atuais de associadas a eles devem ser incluídas na análise como uma riqueza de espécies e de quanto os mapas de extensão de informação a mais para o tomador de decisão (ver Wilson ocorrência foram refinados por meio dessa abordagem. 2010). Ainda assim, o uso de modelagem de distribuição é Em outro estudo prévio, tais mapas foram usados em um uma alternativa complexa, dependente de especialistas com exercício de priorização de áreas para a conservação de amplo referencial acadêmico e requer extenso treinamento vertebrados africanos (Rondinini et al. 2005) e poderiam ser igualmente utilizados para a avaliação de outros aspectos para sua aplicação de forma coerente e cuidadosa. Assim, da biodiversidade, como os avaliados por Carvalho et al. para o especialista em áreas protegidas, o conservacionista de (2010) e Safi et al. (2011). uma ONG ou um tomador de decisão que está diretamente engajado em um planejamento regional aplicável para Problema Resolvido? a conservação, o uso de tais modelos é ainda proibitivo. Infelizmente, nem tudo são flores. Mapas de hábitat adequado Uma Alternativa Simples e Direta para também estão sujeitos a críticas e têm limitações que Conservacionistas e Acadêmicos precisam ser consideradas antes de sua aplicação imediata Atuantes em um planejamento sistemático para a conservação (Rodrigues 2011). Recentemente, tem-se sugerido um refinamento dos mapas de extensão de ocorrência por variáveis ambientais que Em primeiro lugar, informações sobre a história natural determinam, grosso modo, a ocorrência das populações de das espécies, como por exemplo, uma descrição suficiente uma espécie (ver Rondinini et al. 2011 para um exemplo sobre seu hábitat preferencial não são exatamente fáceis de global). Essas variáveis costumam ser o habitat preferencial se obter, principalmente para todas as espécies incluídas no da espécie, cobertura vegetal, variação altitudinal e a planejamento. Descrições muito gerais (e.g. áreas abertas) presença e extensão de corpos d’água. A idéia básica por não são muito úteis, ao passo que descrições extremamente detalhadas (e.g. margens de riachos de cabeceira, com trás desse estratagema é a “filtragem” de um dado espacial substrato de cascalho e inclinação de terreno próxima à de grosseiro (mapa de extensão de ocorrência) para a produção 13,5°) jamais serão capturadas por dados georreferenciados de um “modelo ou mapa de hábitat adequado”, habitat de cobertura vegetal, por exemplo. suitability model - HSM, mais condizente com a realidade (Rondinini et al. 2011). Um segundo entrave reside na resposta espécie-específica das populações, que sé difícil de se prever. Até que ponto O uso desses mapas “filtrados” tem inúmeras vantagens: a especialização intraespecífica e intraindividual tornam são relativamente simples de serem confeccionados, são a associação do hábitat a uma espécie um exercício de baseados em SIG, os dados para sua criação encontram-se generalização distante da realidade ao longo de uma vasta normalmente disponíveis (mapas, imagens de satélite, fotos extensão geográfica? Espécies podem se beneficiar com aéreas) e os mapas finais podem ser refinados ainda mais, impactos antrópicos específicos? Como incluir esse tipo caso necessário. Em teoria, mapas de hábitat adequado de resposta em modelos (mapas) corrigidos por extensão reduzem os erros de comissão, pois retiram as porções dos de hábitat adequado? Como lidar com a incerteza na polígonos que não contém áreas ambientalmente adequadas própria criação do mapa de extensão de ocorrência (como para a ocorrência e sobrevivência das populações de uma mencionado anteriormente)? De fato, essas perguntas espécie. Além disso, essa “filtragem por habitat” gera permanecem sem resposta, assim como aquelas sobre a “pseudoausências” mais reais, em relação àquelas geradas verdadeira eficácia desse tipo de abordagem em relação por algoritmos de inteligência artificial, empregados na aos modelos de distribuição de espécies. Provavelmente, modelagem de distribuição de espécies. Cabe apenas a resposta para isso jaz em uma interação entre a escala ressaltar que embora boa parte dos dados para a criação do planejamento e o objetivo do mesmo. Em escalas de desses modelos encontre-se disponível, a informação sobre grande extensão geográfica, mapas de hábitat adequado o que deve ser excluído da distribuição da espécie ou sobre podem ser particularmente úteis para um planejamento o que constitui o hábitat inadequado nem sempre é fácil de mais fidedigno (ver, contudo, Rodrigues 2011). Em escalas ser obtida e permanece a dependência de conhecimento de menor extensão geográfica, modelos adequados de Refinando Dados Espaciais para a Conservação 243 distribuição de espécies podem ser uma ferramenta poderosa Franklin J, 2009. Mapping Species Distributions: Spatial para aplicação em conservação, sobretudo quando têm sua Inference and Prediction. Cambridge: Cambridge University incerteza mensurada, espacialmente mapeada e ponderada Press. 320 p. pelo planejador (Diniz-Filho et al. 2009; Wilson 2010). Margules CR & Pressey RL, 2000. Systematic conservation De qualquer maneira, independente da escala do planning. Nature, 405:243-253. PMid:10821285. http:// dx.doi.org/10.1038/35012251 planejamento, uma discussão ainda mais básica sobre a qualidade dos dados de entrada do modelo (e.g. dados Loucks C et al., 2008. Explaining the global pattern of de ocorrência bem representativos e menos enviesados, protected area coverage: relative importance of vertebrate informação sobre habitat adequado/inadequado de alta biodiversity, human activities and agricultural suitability. qualidade) permanece como central para a resolução do Journal of Biogeography, 35:1337-1348. http://dx.doi. problema aqui apresentado. Permanece em aberto também org/10.1111/j.1365-2699.2008.01899.x a questão sobre o que é mais viável hoje em dia: boas bases Loyola RD et al., 2009. Key Neotropical ecorregions for de dados com registros de ocorrência ou uma boa base de conservation of terrestrial vertebrates. Biodiversity and conhecimento espécie-específico sobre o que é e o que não Conservation, 18:2017-2031. http://dx.doi.org/10.1007/ é habitat. Como expusemos nesse artigo, infelizmente não s10531-008-9570-6 existem atalhos mágicos para refinar dados espaciais sobre Rodrigues ASL et al., 2004. Effectiveness of the global a distribuição de espécies, com vistas à sua aplicação no protected area network in representing species diversity. planejamento para a conservação. Estudos futuros dirão o Nature, 428:640-643. PMid:15071592. http://dx.doi. quanto o uso de mapas de hábitat adequado faz diferença no org/10.1038/nature02422 planejamento, acelerando a tomada de decisão ou ajudando Rodrigues ASL, 2011. Improving coarse species distribution data na manutenção das populações em ambiente natural. for conservation planning in biodiversity-rich, data-poor, regions: no easy shortcuts. Animal Conservation, 14:108-110. Agradecimentos http://dx.doi.org/10.1111/j.1469-1795.2011.00451.x PL e FAMVF são bolsistas de doutorado e de mestrado Rondinini C, Stuart S & Boitani L, 2005. Habitat suitability do CNPq, respectivamente. GT é bolsista de doutorado models and the shortfall in Conservation Planning for da CAPES. A pesquisa de RDL é financiada pelo CNPq African Vertebrates. Conservation Biology, 19:1488-1497. (processos 475886/2009-7 e 563621/2010-9), CAPES http://dx.doi.org/10.1111/j.1523-1739.2005.00204.x (processo 012/2009) e MCT/Rede CLIMA. Rondinini C et al., 2011. Global habitat suitability models of terrestrial mammals. Philosophical Transactions of the Royal Referências Society of London Series B, 366:2633-2641. P Mid:21844042. http://dx.doi.org/10.1098/rstb.2011.0113 Araújo MB & New M, 2007. Ensemble forecasting of species distributions. Trends in Ecology and Evolution, 22:42-47. Safi K et al., 2011. Understanding global patterns of http://dx.doi.org/10.1016/j.tree.2006.09.010 mammalian functional and phylogenetic diversity. Philosophical Transactions of the Royal Society of London Boitani L et al., 2011. What spatial data do we need to develop Series B, 366:2536-2544. PMid:21807734. http://dx.doi. global mammal conservation strategies? Philosophical org/10.1098/rstb.2011.0024 Transactions of the Royal Society of London Series B, 366:2623-2632. PMid:21844041. http://dx.doi.org/10.1098/ Wilson KA, 2010. Dealing with data uncertainty in conservation rstb.2011.0117 planning. Natureza & Conservação, 8:145-150. http://dx.doi. org/10.4322/natcon.00802007 Carvalho RA et al., 2011. Drafting a Blueprint for Functional and Phylogenetic Diversity Conservation in the Brazilian Wilson KA, Cabeza M & Klein CJ, 2009. Fundamental Concepts Cerrado. Natureza & Conservação, 8:171-176. http://dx.doi. of Spatial Conservation Prioritization. In: Moilanen A, org/10.4322/natcon.00802012 Wilson KA & Possingham HP (eds.). Spatial Conservation Prioritization. New York: Oxford University Press. p. 16-27. Diniz-Filho JAF et al., 2009. Partitioning and mapping uncertainties in ensembles of forecasts of species turnover Whittaker RJ et al., 2005. Conservation Biogeography: assessment under climate change. Ecography, 32:897-906. http://dx.doi. and prospect. Diversity and Distributions, 11:3-23. http:// org/10.1111/j.1600-0587.2009.06196.x dx.doi.org/10.1111/j.1366-9516.2005.00143.x

Received: September 2011 First Decision: October 2011 Accepted: October 2011

Nesse capítulo, avaliamos a eficiência da atual rede de reservas em manter a riqueza de espécies de anfíbios em cenários de mudanças climáticas.

Capítulo 3: Climate change threatens protected áreas of the

Atlantic Forest.

Priscila Lemes, Adriano S. Melo & Rafael Loyola

Biodiversity and Conservation, 23: 357-368.

501

Biodivers Conserv (2014) 23:357–368 DOI 10.1007/s10531-013-0605-2

ORIGINAL PAPER

Climate change threatens protected areas of the Atlantic Forest

Priscila Lemes • Adriano Sanches Melo • Rafael Dias Loyola

Received: 23 March 2013 / Accepted: 29 November 2013 / Published online: 6 December 2013 Ó Springer Science+Business Media Dordrecht 2013

Abstract Only 7 % of the Atlantic Forest Biodiversity Hotspot is currently protected, though it holds 18 % of all amphibian species in South America. How effective would the Atlantic Forest network of protected areas (PAs) be in a changing climate? Are there some intrinsic features of PAs that drive species loss or gain inside them? We addressed these questions by modeling the ecological niches of 430 amphibian species in the Atlantic Forest and projecting their distributions into three future climate change simulations. We then assessed changes in species richness inside PAs for different time frames and tested their signiﬁcance via null model. The number of species should decline within Atlantic Forest network of PAs under changing climate conditions. Only altitude was a good predictor of species gains or lost inside PAs. Therefore, we suggest that new PAs established in highlands would be more effective to alleviate the effects of climate change on this imperiled fauna.

Keywords Amphibians Á Dispersal Á Ecological niche models Á Global warming Á Reserve network

Electronic supplementary material The online version of this article (doi:10.1007/s10531-013-0605-2) contains supplementary material, which is available to authorized users.

P. Lemes Po´s Graduaçaõ em Ecologia e Evoluçaõ, Universidade Federal de Goia´s, Goiaˆnia, Goia´s, Brazil e-mail: [email protected]

A. S. Melo Á R. D. Loyola (&) Departamento de Ecologia, Universidade Federal de Goia´s, Caixa Postal 131, Goiaˆnia, Goia´s 74001-970, Brazil e-mail: [email protected] A. S. Melo e-mail: [email protected] 123 358 Biodivers Conserv (2014) 23:357–368

Introduction

Climate change may produce species’ range shifts altering patterns of species’ diversity and distribution, that could significantly reduce conservation efficiency of networks of protected areas (PAs) (Arau´jo et al. 2011; Monzoń et al. 2011). This is critical because PAs still figure as the cornerstone of conservation strategies and over 12 % of the Earth’s surface is currently under some form of legal protection (Jekins and Joppa 2009; Joppa and Pfaff 2011). PAs may be particularly helpful for safeguarding the global biodiversity, which continues to decline at an alarming rate (Butchart et al. 2010) in places facing continuous local loss and fragmentation of natural land cover (Ladle and Whittaker 2011). New areas may be required to accommodate species range shifts and climate-forced dispersal (Lemes and Loyola 2013, Loyola et al. 2013a, b); therefore, PAs will become the major contributor to efforts of biodiversity conservation in the future, as they are today (Arau´jo et al. 2011). If these places fail to protect species in the future, then the current biodiversity crisis would reach unprecedented levels. Climate change entails other threats to the effectiveness of a PA network (Hannah et al. 2007; Arau´jo et al. 2011). First, because PAs are fixed in the landscape, but threats are not, environmental changes in the matrix surrounding a particular PA may spread into the PA itself (Wiens et al. 2011). Second, it may foster the invasion of a new or even alien species in PAs (Loyola et al. 2012), as a result of future environmental conditions becoming more suitable for these species (Nori et al. 2011). This would otherwise jeopardize the persistence of originally established species in PAs. Third, all this is amplified by the way PAs are usually established: not within the framework of spatial conservation planning (Moilanen et al. 2009), but based on opportunistic political intentions (Tsianou et al. 2013; Ladle and Whittaker 2011) leading to a global misrepresentation of species within PA networks (Rodrigues et al. 2004). The full effectiveness of PAs depends, however, on the ability of maintaining native species in their habitats, ensuring conservation in the long- term (Arau´jo et al. 2011; Hannah et al. 2007). To evaluate how climate change might affect PA effectiveness, it is essential to develop ecological niche models (ENMs), which help to anticipate the effects of climate change at different spatial scales (Arau´jo et al. 2011; D’Amen et al. 2011). The implementation of these models, along with a careful interpretation of their results, has become a powerful way to predict the impacts of climate change for a large number of species (Lawler et al. 2009; Rangel and Loyola 2012). Geographical projections arising from ENMs can then be overlaid onto a layer of PAs to assess those that should retain suitable climatic conditions for different species (Arau´jo et al. 2004; Loyola et al. 2012). This evaluation indicates best sites for management investment or points out new areas of future conservation value (Lemes and Loyola 2013). There are, of course, considerable uncertainties and limitations for the application of ENMs in the real world (Sinclair et al. 2010). For example, a species may not be able to disperse to newly suitable habitats (Jackson and Sax 2010) and interaction among species is usually ignored (VanDerWal et al. 2009). Nevertheless, ENMs have been used successfully in conservation planning (Lawler et al. 2011) and our understanding of these models increases everyday (Peterson et al. 2011). On the other hand, one might imagine, if there are some features of PAs that could guarantee the persistence of populations in the long-term, which should be considered, such as: area, connectivity, and climatic conditions (Lindenmayer et al. 2006, Wiens et al. 2011). Consequently, the identification of correlates to explain the effectiveness of PAs becomes the first step to assess the future effectiveness of PA networks (Chape et al. 2005).

123 Biodivers Conserv (2014) 23:357–368 359

It is, therefore, necessary to understand how multiple features of PAs interact in order to quantify gains or losses of the species within these PAs in the future. According to the International Union for the Conservation of Nature (IUCN 2013), Neotropical amphibians have a higher proportion of threatened species and several records of population declines. In particular, it seems to be a group severely affected by climate change and global warming (Pounds et al. 2006). These high levels of population decline and species threats are creating demands for effective strategies for conservation to mitigate the impacts of climate change. Here, we evaluate the current and future climatic suitability of PAs in the Atlantic Forest Biodiversity Hotspot and its implications for amphibian conservation. Our approach relies on correlations between climate and amphibian species’ occurrences that estimate future shifts in a suitable climate for species to predict the effects of climate-induced shifts concerning species richness inside PAs. By understanding how multiple key intrinsic PA features interact, we were able to identify PAs that win or lose the greatest number of species and also understand what makes them more suitable or unsuitable than others.

Methods

We downloaded amphibian species data from the IUCN Red List (2013), which is the more comprehensive dataset from the extent of occurrence maps currently known. In general, these maps are typically drawn as an outline around interpolated records of occurrences and can include commission errors or false presences (Hulbert and White 2005). However, they have two important attributes: (1) they are used to deﬁne a species’ threat level (IUCN 2013) and they allow the use of presence-absence methods for modeling a species’ niche, given that every cell outside species range can be considered an absence. These maps have been used several times in ENM literature (see Diniz-Filho et al. 2009; Lawler et al. 2009; Faleiro et al. 2013), especially in places where complete and reliable information on species distribution is lacking or will not be available in a foreseeable future (Faleiro et al. 2013). We used SAM v4.0 (Spatial Analysis in Macroecology; Rangel et al. 2010)to overlay all 430 species into an equal-area grid (0.25° latitude/longitude) across all of South America, because some species have occurrence areas that extrapolate at the limits of the Atlantic Forest or could colonize the Atlantic Forest in future scenarios. We interpolated current and future climate data from the WorldClim database (Hijmans et al. 2005) downscaled to a resolution of 5-arcmin. Current climate baseline was the annual average based on the temporal series produced between years 1961 and 1990. The IPCC’s Fourth Assessment Report (AR4, IPCC 2007) developed the future climate models and variables for these models came from three coupled Atmosphere–Ocean General Circulation Models (AOGCMs: CCCMA-CGCM2, CSIRO-MK2, and HCCPR-HadCM3) for the A2a scenario for 2050. We aligned climatic data to our 0.25° 9 0.25° grid using ArcGIS 10.1 platform. For each species, we modeled its ecological niche as a function of four climatic variables: annual mean temperature, temperature seasonality (standard deviation 9 100), annual precipitation, and precipitation seasonality (coefﬁcient of variation). These variables are often used to explain geographical patterns in amphibian species richness (see Buckley and Jetz 2007).

Ecological niche models (ENMs)

We generated an ensemble of models for each of the 430 amphibian species inhabiting the Atlantic Forest. The ensemble included projections made by six widely used ENMs; 123 360 Biodivers Conserv (2014) 23:357–368 namely, genetic algorithm for rule set production (GARP), generalized additive models (GAM), generalized linear models (GLM), multivariate adaptive regression splines (MARS), maximum entropy (MaxEnt), and random forest (RF). We ran all models in the BioEnsembles platform (Diniz-Filho et al. 2009). Models were calibrated for current climate using a 75 % random sample of initial data and were evaluated against the remaining 25 % data, using area under the curve (AUC) of the receiver operation characteristic (ROC) and true skill statistic (TSS) as a measure of model fit (Liu et al. 2011). Models were evaluated based on TSS (Liu et al. 2011). Sensitivity and specificity were calculated based on the probability threshold for which their sum is maximized, not being affected by prevalence. TSS values range from -1to?1, where ?1 indicates perfect fit, minimizing overprediction and omission error rates; values close to zero indicate performance worse than randomly expected (Allouche et al. 2006). We used the TSS to weigh final estimated occurrence frequencies and models holding TSS \0.5 being discarded. Projections into future climates were repeated 10 times, each selecting a different 75 % random sample while verifying model accuracy against the remaining 25 %. Spatial patterns of change were evaluated using measures of temporal turnover for each grid cell (Peterson et al. 2002), which refers to Jaccard’s dissimilarities between current and future sets of species for which a given cell is projected to have climate suitability. Thus, the turnover was calculated by cell accounting for the number of species projected to decrease their potential range plus the number of species projected to increase their potential by the total number of species in the current scenario. Our predictions of range shifts for amphibian species in the Atlantic Forest are dependent on their limited dispersal abilities. Dispersal capability might severely limit the ability of species to track suitable climatic conditions geographically (Massot et al. 2008, Early and Sax 2011). We used three approaches to map areas with dispersal limitations on potential range shifts by restricting any potential future range expansion to 200, 100, and 50 km (Smith and Green 2005; Lawler et al. 2010), since it is unlikely that amphibians will be able to reach fully climatic conditions by shifts of their distributional ranges (Munguıá et al. 2012).

Data on protected areas

We retrieved the location of PAs (IUCN Categories I–IV) in the Atlantic Forest from the World Database on PAs (WDPA 2010). We overlaid PA polygons onto our grid considering a grid cell as ‘‘protected’’ even if only a small part of it holds a PA. We assumed, therefore, that all species occurring in that cell could potentially beneﬁt from the occurrence of a PA in that cell. We compiled data on ﬁve quantitative features for all 142 PAs currently established in Atlantic Forest (Online Resource 1) including: (1) latitudinal midpoint, (2) minimum distance from the center of a PA until the coast [km], (3) elevation [m], (4) area size [km], and (5) minimum distance between PAs accounting for the PAs’ midpoint [km].

Null model for estimating climatic suitability of PAs under climate change

We generated a null model that maintained size, form, and orientation of PAs, but removed other intrinsic effects (latitude, altitude). The null model allocated PAs randomly in the Atlantic Forest and obtained species richness both in the present and in the future, as given by our consensus ensemble model. The 142 PAs comprised 283 cells distributed in 1907 Atlantic Forest cells. Accordingly, there were many distinct possibilities to randomly 123 Biodivers Conserv (2014) 23:357–368 361 allocate PAs in the Atlantic Forest. Hence, this procedure was repeated 1,000 times and the average richness was obtained. This allowed us to compare species richness in the future and present, after isolating the effects of intrinsic properties of PAs (except size, form and orientation).

Predictors of species richness

We built a regression tree in which the response variable was the difference between modeled species richness in the future and in the present. Predictor variables were PAs’ intrinsic features described above. We obtained overgrowth trees and then pruned them to the size that signiﬁcantly reduces the variability within the subgroups. We followed the criterion suggested by De’ath and Fabricius (2000) based on cross-validation to prune overgrowth trees. This method depicts a random subset of the data to grow the tree and predict the response in a second subset. We selected the smallest tree in which its cross- validation error was within the range of the minimum cross-validation error of all tree- sizes, plus its respective standard error. As this size may vary between cross-validations, we wrote a routine and used the tree size most often obtained. After, we employed the ‘‘rpart’’ package to do the analyses 1,000 times in the R environment (Therneau and Atkinson 1997).

Results

Using the median TSS across all species as an evaluation criterion, we found that consensus projections provide the more robust results under climate change, overcoming all single ENMs. TSS values for ensemble models reached on average 0.71, indicating good model ﬁt. For all modeling methods, combined projections indicated areas of high species richness in the central and southeast part of the Atlantic Forest for now, with a decreasing number of species under future scenarios (Fig. 1a, b). Climatically suitable sites should decrease by 2050, whilst northern and southwestern portions of the biome should face many species extinctions (Fig. 1). This is because all modeling methods forecasted a general reduction in species’ ranges, which leads to a decrease in the number of sites with high species richness. Variability among ENMs was strongly affected by GLM projections that displayed higher losses and gains of climate suitability over time (Online Resource 2). Limiting species dispersal resulted in small changes in overall species richness (Fig. 2). About 33 % of species are likely to occur in the boundaries of the Atlantic Forest or even outside of the biome. Assuming unlimited dispersal, 88 % of species would experience range contraction, whereas, under a 50 km dispersal scenario, this level of contraction would reduce to 78 % (Table 1, Fig. 3). We also found high species temporal turnover (up to 78.7, 77.9, and 74.8 %, for 50, 100, and 200 km dispersal scenarios, respectively) across projections for current and future climates (Fig. 2). This means that future scenarios showed dramatic changes not only in species richness but, also in species composition. As a general pattern, few PAs are likely to harbor more species (green dots in the graph and green PAs in the map) than expected by our null model. Those PAs in that situation are usually located in the in the cooler southern region of the Atlantic Forest (Fig. 4). In addition to the general trend of species loss, our null model of random allocation of PAs (blue dots) indicated great variation on how many species would be lost (and also gained) in the future. However, when the null model is projected for current and future climates 123 362 Biodivers Conserv (2014) 23:357–368

Fig. 1 Maps of modeled amphibian species richness based on geographic range overlap of 430 species predicted to occur in the Atlantic Forest Biodiversity Hotspot for present (a) and 2050 (b) using real locations of PAs, species richness was generally lower (red dots in Fig. 4) than those obtained for random locations of PAs (blue dots in Fig. 4). PAs along the coast or in adjacent mountainous regions would loose species at higher rates than expected by our null model (PAs indicated in red). Further, the western region of the biome, which includes semi-deciduous forests, would experience a severe reduction in species richness owing to climate change. Protected area mean altitude was the only feature capable of explaining species gains or losses in future scenarios. This means that we were unable to actually build a tree, because there is no split after all. Anyway, PAs that should gain species are often located on steep slopes (Fig. 4; Online Resource 1). Lowland PAs, in particular those bellow 161 m above sea level, will lose more species than those located in highlands.

Discussion

Protected areas in the Atlantic Forest would become less effective to protect amphibians in the future. It is important, therefore, to anticipate how climatic changes will lead to a decreasing species representation across the entire network of PAs. Further, the inclusion of new areas in the currently established network of PAs should observe quantitative thresholds associated with altitude, in particular, for this could improve conservation investments by focusing on sites that are more likely to protect species in the long-term.

123 Biodivers Conserv (2014) 23:357–368 363

Fig. 2 Spatial patterns of species richness predicted by different dispersal scenarios for the current time and 2050. Right maps show spatial patterns of species turnover under climate change as forecasted by a consensus model of species’ ecological niche and projection of their geographic ranges

Table 1 Number of Atlantic Forest amphibian species that should experience climate-induced range contractions or expansion out of 430 amphibian species for four different dispersal scenarios Range response Dispersal distance

50 km 100 km 200 km Unlimited

Contraction 317 320 321 377 Expansion 63 73 82 44 Extinction 20 20 17 9 Unchanged 8 6 3 – Total 408 419 423 430 Extinction stands for a range contraction of 100 %

Amphibian species are more susceptible to climate change because of their dependence on microhabitats, hydrological regimes, diseases, and limited ability to disperse (Pounds et al. 2006; Early and Sax 2011), and these factors can act in synergy. Our ﬁndings indicate that the most likely impact of climatic changes should occur in the central portion of the biome. A high turnover rate in these regions is likely to happen, as forecasted by our models, because these regions should face the largest climatic changes in the future (Lemes and Loyola 2013). For example, the northeast should hold high annual mean temperatures in future scenarios (Online Resource 3), and biome boundaries should have dramatic 123 364 Biodivers Conserv (2014) 23:357–368

Fig. 3 The percentage of 430 amphibian species projected to experience four different levels of geographic range loss under climate change changes in precipitation patterns. Moreover, turnover patterns can be explained by species distribution, landscape connectivity, and species’ dispersal capability. Areas with high turnover rates are more likely to have a high density of peripheral species ranges. Lowland regions, however, should not exhibit great changes in species richness due to the short latitudinal temperature gradient of the tropics (Colwell et al. 2008). Consequently, range shifts will be likely more pronounced toward higher elevations than higher latitudes (Klorvuttimontara et al. 2011). Indeed, our projections for the future predict many species ranges moving upward in elevation, where a large number of endemic species in the Atlantic Forest occur (Carnaval et al. 2009). Even under limited dispersal scenarios, our ENMs predicted potential distributions for many species outside the Atlantic Forest. This could have happened for different reasons. First, although the extent of occurrence maps used in our models have been recently updated and revised by experts (IUCN 2013), there may be overestimated of species occurrences in the data. Second, amphibian species mainly associated with open areas or with high environmental plasticity, such as the red-snouted treefrog (Scinax ruber) might expand their range on future scenarios (Dahanukar 2012). Third, species occurring in the border of the biome can ﬁnd suitable climatic areas in the Atlantic Forest in the future; this is the case of the orange-legged leaf frog (Phyllomedusa hypocondrialis). Even when assuming unlimited dispersal, our ﬁndings indicate that many species should experience range contractions more than range expansion (Table 1). Limiting dispersal ability, however, tend to generate range expansion. For example, when there was a limited dispersal the potential future ranges in Atlantic Forest, the northeastern pepper frog (Leptodactylus vastus) and milky treefrog (Trachycephalus venulosus) might reach until 108 and 104 %, respectively, than it current potential range.

123 Biodivers Conserv (2014) 23:357–368 365

(A1) Dispersal up to 200 Km (A2) Dispersal up to 200 Km Latitude −30 −25 −20 −15 −10 −5 0 50 100 150 Species Richness (Future) 0 50 100 150 −55 −50 −45 −40 −35

(B1) Dispersal up to 100 Km −5 (B2) Dispersal up to 100 Km −10 Latitude −30 −25 −20 −15 0 50 100 150 Species Richness (Future) 0 50 100 150 −55 −50 −45 −40 −35

(C1) Dispersal up to 50 Km (C2) Dispersal up to 50 Km Latitude −25 −20 −15 −10 −5 −30 0 50 100 150 Species Richness (Future) 0 50 100 150 −55 −50 −45 −40 −35 Species Richness (Present) Longitude

Fig. 4 Negative impacts are expected for most PAs and the loss of species is more consistent than expected in random selected PA. Protected areas projected to gain (green cells in maps and dots in graphs) or lose (red cells in maps and dots in graphs) species in future changing climate. Orange cells in maps and dots in graphs indicate no signiﬁcant change according to the null model. Blue dots stand for null model expectation

In the tropics, mountainous regions are major areas of endemism and are expected to serve as important refuges for species under climate change (Williams et al. 2003; Arau´jo et al. 2011; Klorvuttimontara et al. 2011). For example, the Serra do Mar State Park is an important center of endemism with an elevation up to 2,316 m, which likely became a large climatic refuge for Neotropical amphibians from the late Pleistocene (Carnaval et al. 2009). Elevation, latitude, distance to coast, vegetation cover, and solar radiation could explain regional climate patterns (Shoo et al. 2010). However, speciﬁc climates will become scarcer or disappear entirely under climate change (Williams et al. 2007). Climate change should induce species’ range shifts moving species out of PAs (Monzo´n et al. 2011), whereas local extinction could alter community composition within protected sites (Wiens et al. 2011), suggesting that climate effects should be considered when setting conservation priorities. 123 366 Biodivers Conserv (2014) 23:357–368

Amphibian species in Atlantic Forest PAs are more vulnerable to habitat destruction (IUCN 2013), mainly because this area is the economic core of Brazil, with highly fragmented areas (Ribeiro et al. 2009) due to high human population density, mining, and logging. Our analysis does not provide quantitative estimates of species extinction risk, but show evidence of a possible regional extinction within unlimited and limited dispersal models. In addition, our findings indicate that many PAs will become less effective in future scenarios, decreasing species protection. Moreover, changes in community phylogenetic structure are also expected in the Atlantic Forest, as recently showed (Loyola et al. 2013a). Before concluding, we must highlight some caveats of our study. Our models are based on two important assumptions: the current climate species distribution is a good indicator of a suitable climate (Arau´jo et al. 2011) and the absence of biological interactions that mediate species range (Soberoń 2007). Model predictions are fraught with uncertainties to obtain range changes (Arau´jo and New 2007). However, ensemble forecasting is an alternative and conservative approach for reducing uncertainties among models, which combines the central tendency of multiple ENMs (Arau´jo and New 2007). It was recently suggested that this approach might produce more reliable models for conservation purposes (Loyola et al. 2012; Rangel and Loyola 2012). Furthermore, the number of protected species here tend to be overestimated, mainly because PAs have different degrees of effectiveness (Chape et al. 2005) and the species are considered protected if any part of their range overlaps into a PA area (Rodrigues et al. 2004). The same network might not safeguard conservation of another species with limited dispersal abilities (see Online Resource 1), which the simple presence within a PA is insufficient to ensure the long-term persistence of many species (Rodrigues et al. 2004). In addition, some PAs are unsuitable for the conservation of many threatened species (Beresford et al. 2011). Most PAs in the Atlantic Forest were created without any ecological criterion and species representation in this network is highly variable. Our findings indicate that the current protection network is unsuitable to cover the species representation of amphibian species that inhabit the Atlantic Forest (Loyola et al. 2013b), which may be a misrepresentation in future scenarios (Hannah et al. 2007, see also Lemes and Loyola 2013, and Loyola et al. 2013a, b for recent advances on this issue). Our analyses do not incorporate connectivity of a protected network. Furthermore, if a PA is to be isolated or less connected, they can become locally unsuitable, but their global efficiency in the full network is not affected (Mazaris et al. 2013). In any case, to recover all species in a future changing climate, the establishment of additional PAs is imperative increasing the effectiveness and ecological reliability of the regional network PAs.

Acknowledgments The authors acknowledge T.F. Rangel for providing access to the BioEnsembles platform used in the analyses. PL received a PhD scholarship from CNPq. RDL and ASM received research productivity fellowships granted by CNPq (Grants #304703/2011-7 and #307479/2011-0, respectively). RDL’s work is funded by the Brazilian Research Network on Global Climate Change (Rede CLIMA), Con- servation International Brazil, and by Fundaçaõ Grupo O Botica´rio de Proteçaõa` Natureza (Prog #08_2013). We would also like to thank two anonymous reviewers for critical comments and editing of this paper.

References

Allouche O, Tsoar A, Kadmon R (2006) Assessing the accuracy of species distribution models: prevalence, kappa and the true skill statistic (TSS). J Appl Ecol 43:1223–1232 Arau´jo MB, New M (2007) Ensemble forecasting of species distributions. Trends Ecol Evol 22:42–47 Arau´jo MB, Cabeza M, Thuiller W, Hannah H, Williams PH (2004) Would climate change drive species out of reserves? An assessment of existing reserve selection methods. Glob Chang Biol 10:1618–1626 123 Biodivers Conserv (2014) 23:357–368 367

Arau´jo MB, Alagador D, Cabeza M, Nogue´s-Bravo D, Thuiller W (2011) Climate change threatens European conservation areas. Ecol Lett 14:484–492 Beresford AE, Buchanan GM, Donald PF, Butchart SHM, Fishpool LDC et al (2011) Poor overlap between the distribution of protected areas and globally threatened birds in Africa. Anim Conserv 14:99–107 Buckley AH, Jetz W (2007) Environmental and historical constraints on global patterns of amphibian richness. Proc R Soc Lond B 7:1167–1173 Butchart SHM, Walpole M, Collen B, van Strien A, Schanlemann JPW et al (2010) Global biodiversity: indicators of recent declines. Science 328:1164–1168 Carnaval AC, Hickerson MJ, Haddad CFB, Rodrigues MT, Moritz C (2009) Stability predicts genetic diversity in the Brazilian Atlantic Forest Hotspot. Science 323:785–789 Chape S, Harrison J, Spalding M, Lysenko I (2005) Measuring the extent and effectiveness of protected areas as an indicating for meeting global biodiversity targets. Philos Trans R Soc B 360:443–455 Colwell RK, Brehm G, Cardelu´s CL et al (2008) Global warming, elevational range shifts, and lowland biotic attrition in the wet tropics. Science 322:258–261 D’Amen M, Bombi P, Pearman PB, Schmatz DR, Zimmermann NE, Bologna MA (2011) Will climate change reduce the efficacy of protected areas for Amphibians conservation in Italy? Biol Conserv 144:989–997 Dahanukar N (2012) Badis ruber. In: IUCN 2013. IUCN Red List of threatened species. Version 2013.1. www.iucnredlist.org. Accessed on 22 Aug 2013 De’ath G, Fabricius KE (2000) Classification and regression trees: a powerful yet simple technique for ecological data analysis. Ecology 81:3178–3192 Diniz-Filho JAF, Bini LM, Rangel TFLVB, Loyola RD, Hof C, Nogue´s-Bravo D, Arau´jo MB (2009) Partitioning and mapping uncertainties in ensembles of forecasts of species turnover under climate change. Ecography 32:897–906 Early R, Sax DF (2011) Analysis of climate paths reveals potential limitations on species range shifts. Ecol Lett 14:1125–1133 Faleiro FV, Machado RB, Loyola RD (2013) Defining spatial conservation priorities in the face of land-use and climate change. Biol Conserv 158:248–257 Hannah L, Midgley G, Andelman S, Arau´jo M, Hughes G, Martinez-Meyer E, Pearson R, Williams P (2007) Protected area needs in a changing climate. Front Ecol Environ 5:131–138 Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A (2005) Very high resolution interpolated climate surfaces for global land areas. Int J Climatol 25:1965–1978 Hulbert AH, White EP (2005) Disparity between range map and survey-based analyses of species richness: patterns, processes and implications. Ecol Lett 8:319–327 IPCC (2007) Climate change 2007: impacts, adaptation and vulnerability: Contribution of Working Group II to the Fourth Assessment Report of the IPCC. Cambridge University Press, Cambridge IUCN (2013) The IUCN Red List of Threatened Species. Version 2012.2. http://www.iucnredlist.org. Accessed on 22 August 2013 Jackson ST, Sax DF (2010) Balancing biodiversity in a changing environment: extinction debt, immigration credit and species turnover. Trends Ecol Evol 25:153–160 Jekins CN, Joppa L (2009) Expansion of the global terrestrial protected area system. Biol Conserv 142:2166–2174 Joppa L, Pfaff A (2011) Global protected area impacts. Proc Biol Sci 278:1633–1638 Klorvuttimontara S, McClean CJ, Hill JK (2011) Evaluating the effectiveness of protected areas for conserving tropical forest butterflies of Thailand. Biol Conserv 144:2534–2540 Ladle RJ, Whittaker RJ (eds) (2011) Conservation biogeography. Wiley-Blackwell, Oxford Lawler JJ, Shafer SL, White D, Kareiva P, Maurer EP, Blaustein AR, Bartlein PJ (2009) Projected climate- induced faunal change in the Western Hemisphere. Ecology 90:588–597 Lawler JJ, Shafer SL, Bancroft BA, Blaustein AR (2010) Projected climate impacts for the amphibians of the western hemisphere. Conserv Biol 24:38–50 Lawler JJ, Wiersma YF, Huettmann F (2011) Using species distribution models for conservation planning and ecological forecasting. In: Drew A, Wiersma YF, Huettmann F (eds) Predictive modeling in landscape ecology, 1st edn. Springer, New York, pp 271–290 Lemes P, Loyola RD (2013) Accommodating species climate-forced dispersal and uncertainties in spatial conservation planning. PLoS ONE 8:e54323 Lindenmayer DB, Franklin JF, Fischer J (2006) General management principles and a checklist of strategies to guide forest biodiversity conservation. Biol Conserv 131:433–445 Liu C, White M, Newel G (2011) Measuring and comparing the accuracy of species distribution models with presence–absence data. Ecography 34:232–243

123 368 Biodivers Conserv (2014) 23:357–368

Loyola RD, Lemes P, Faleiro FV, Trindade-Filho J, Machado RB (2012) Severe loss of suitable climatic conditions for marsupial species in Brazil: challenges and opportunities for conservation. PLoS ONE 7:e46257 Loyola RD, Lemes P, Brum FT, Provete DB, Duarte LDS (2013a) Clade-specific consequences of climate change to amphibians in Atlantic Forest protected areas. Ecography. doi:10.1111/j.1600-0587.2013.00396.x Loyola RD, Lemes P, Nabout JC, Trindade-Filho J, Sagnori MD, Dobrovolski R, Diniz-Filho JAF (2013b) A straightforward conceptual approach for evaluating spatial conservation priorities under climate change. Biodiver Conserv 22:483–495 Massot M, Clobert J, Ferrie`re R (2008) Climate warming, dispersal inhibition and extinction risk. Glob Change Biol 14:461–469 Mazaris AD, Papanikolaou AD, Barbet-Massin M, Kallimanis AS, Jiguet F et al (2013) Evaluating the connectivity of a protected areas’ network under the prism of global change: the efficiency of the european natura 2000 network for four birds of prey. PLoS ONE 8:e50640 Moilanen A, Wilson KA, Possingham HP (2009) Spatial conservation prioritization: quantitative methods and computational tools, 1st edn. Oxford University Press, Oxford Monzoń J, Moyer-Horner L, Palamar MB (2011) Climate change and species range dynamics in protected area. Bioscience 61:752–761 Munguıá M, Rahbek C, Rangel TF et al (2012) Equilibrium of global amphibian species distributions with climate. PLoS One 7:e34420 Nori J, Urbina-Cardona JN, Loyola RD, Lescano JN, Leynaud GC (2011) Climate change and American bullfrog invasion: what could we expect in South America? PLoS ONE 6:e25718 Peterson AT, Ortega-Huerta MA, Bartley J, Sanchez-Cordero V, Soberoń J et al (2002) Future projections for Mexican faunas under global climate change scenarios. Nature 416:626–629 Peterson AT, Soberoń J, Pearson RG, Anderson RP, Martıńez-Meyer E et al (2011) In: Horn HS, Levin SA (eds) Ecological niches and geographic distributions. Princeton University Press, Princeton Pounds JA, Bustamente MR, Coloma LA, Consuegra JA, Fodgen MPL et al (2006) Widespread amphibian extinctions from epidemic disease driven by global warming. Nature 439:161–167 Rangel TF, Loyola RD (2012) Labeling ecological niche models. Nat Conserv 10:119–126 Rangel TF, Diniz-Filho JAF, Bini LM (2010) SAM: a comprehensive application for spatial analysis in Macroecology. Ecography 33:46–50 Ribeiro MC, Metzer JP, Martensen AC, Ponzoni FJ, Hirota MM (2009) The Brazilian Atlantic Forest: how much is left, and how is the remaining forest distributed? implications for conservation. Biol Conserv 142:1141–1153 Rodrigues ASL, Andelman SL, Bakarr MI, Boitani L, Brooks TM et al (2004) Effectiveness of the global protected area network in representing species diversity. Nature 428:640–643 Shoo LP, Storlie C, Vanderwal J, Little J, Williams SE (2010) Target protection and restoration to conserve tropical biodiversity in a warming world. Glob Change Biol 17:186–193 Sinclair SJ, White MD, Newell GR (2010) How useful are species distribution models for managing biodiversity under future climates. Ecol Soc 15:8 Smith MA, Green DM (2005) Dispersal and the metapopulation paradigm in amphibian ecology and conservation: are all amphibian populations metapopulations? Ecography 28:110–128 Soberoń J (2007) Grinnellian and Eltonian niches and geographic distributions of species. Ecol Lett 10:1115–1123 Therneau TM, Atkinson B (1997) An introduction to recursive partitioning using the rpart routines. Techinical Report 61, Mayo Clinic, Section of Biostatistics. Rochester, MN Tsianou MA, Azaris AD, Kallimanis AS, Deligioridi PK, Apostolopoulou E et al (2013) Identifying the criteria underlying the political decision for the prioritization of the Greek Natura 2000 conservation network. Biol Conserv 166:103–110 VanDerWal J, Shoo LP, Johnson CN, Williams SE (2009) Abundance and the environmental niche: environmental suitability estimated from niche models predicts the upper limit of local abundance. Am Nat 174(2):282–291 Wiens JA, Seavy NE, Jongsomjit D (2011) Protected areas in climate space: what will the future bring? Biol Conserv 144:2119–2125 Williams SE, Bolitho EE, Fox S (2003) Climate change in Australian tropical rainforests: an impending environmental catastrophe. Proc R Soc Lond B 270:1887–1892 Williams JW, Jackson ST, Kutzbach JE (2007) Projected distribution of novel and disappearing climates by 2100 AD. Proc Natl Acad Sci USA 104:5738–5742 World Database on Protected Areas Consortium (2010) World database on protected areas annual release 2010 (United Nations Environment Programme, Nairobi, Kenya, and World Conservation Monitoring Centre, Cambridge, UK)

123

Nesse capítulo, identificamos locais prioritários para a conservação de anfíbios da Mata Atlântica diante de mudanças climáticas.

Capítulo 4: Accomodating Species

Climate-Forced Dispersal and

Uncertainties in Spatial

Conservation Planning

Priscila Lemes & Rafael Loyola

PLoS ONE, 8: e54323.

501

Accommodating Species Climate-Forced Dispersal and Uncertainties in Spatial Conservation Planning

Priscila Lemes1, Rafael Dias Loyola2* 1 Programa da Po´s-graduaçaõ em Ecologia e Evoluçaõ, Universidade Federal de Goia´s, Goiaˆnia, Goia´s, Brazil, 2 Departamento de Ecologia, Universidade Federal de Goia´s, Goiaˆnia, Goia´s, Brazil

Abstract Spatial conservation prioritization should seek to anticipate climate change impacts on biodiversity and to mitigate these impacts through the development of dynamic conservation plans. Here, we defined spatial priorities for the conservation of amphibians inhabiting the Atlantic Forest Biodiversity Hotspot that overcome the likely impacts of climate change on the distribution of this imperiled fauna. First, we built ecological niche models (ENMs) for 431 amphibian species both for current time and for the mid-point of a 30-year period spanning 2071–2099 (i.e. 2080). For modeling species’ niches, we combined six modeling methods and three different climate models. We also quantified and mapped model uncertainties. Our consensus models forecasted range shifts that culminate with high species richness in central and eastern Atlantic Forest, both for current time and for 2080. Most species had a significant range contraction (up to 72%) and 12% of species were projected to be regionally extinct. Most species would need to disperse because suitable climatic sites will change. Therefore, we identified a network of priority sites for conservation that minimizes the distance a given species would need to disperse because of changes in future habitat suitability (i.e. climate-forced dispersal) as well as uncertainties associated to ENMs. This network also maximized complementary species representation across currently established protected areas. Priority sites already include possible dispersal corridors linking current and future suitable habitats for amphibians. Although we used the a top-ranked Biodiversity Hotspot and amphibians as a case study for illustrating our approach, our study may help developing more effective conservation strategies under climate change, especially when applied at different spatial scales, geographic regions, and taxonomic groups.

Citation: Lemes P, Loyola RD (2013) Accommodating Species Climate-Forced Dispersal and Uncertainties in Spatial Conservation Planning. PLoS ONE 8(1): e54323. doi:10.1371/journal.pone.0054323 Editor: Alessandro Flammini, Indiana University, United States of America Received January 29, 2012; Accepted December 11, 2012; Published January 22, 2013 Copyright: ß 2013 Lemes, Loyola. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: P.L. received a CNPq doctoral scholarship (project # 147345/2010-3). R.D.L. research is funded by CAPES-FCT Program (Brazil-Portugal) and by the Brazilian Research Network on Global Climate Change (Rede CLIMA). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have declared that no competing interests exist. * E-mail: [email protected]

Introduction However, model uncertainty arise from many sources, such as the methods and the climate projections used to generate ENMs [9], A wide range of evidences indicate climate change as one the not to mention the limitations to model extrapolation in space and th greatest threats to biodiversity in the 21 century [1]. Climate time (i.e. model transferability), and how to evaluate model change impacts, which may have already resulted in several recent performance [10]. species extinction [2], are species-specific and produce shifts in The ongoing debate on model performance, statistical fit, and species phenology, ecological interactions, and geographical transferability indicates that it is still difficult to determine the best distributions [3–4]. Global climate change poses new challenges method for modeling species’ ecological niche, because the to biodiversity conservation especially because it induces species outcome of these methods is strongly dependent on data range shifts yielding additional complexity and uncertainty to availability and geographic scale for which they have been definition and implementation of spatially oriented actions for projected [11]. To cope with these issues, a combination of conservation investment [5]. Here we address this challenge by different projections built upon different climatic conditions and developing spatial conservation plans that consider the likely modeling methods – the ensemble forecasting approach – has species’ range shifts under baseline and future climate scenarios. been suggested as more conservative than single model analysis Climate change effects on biodiversity depend on how species’ [9]. Ensembles of forecasts should be used when it is impossible to distribution will respond to such changes. These responses are determine which type of model should produce most accurate usually inferred trough ecological niche models (henceforth predictions. Predictions from multiple models or from multiple ENMs) [6]. Currently, there are several methods for modeling input data sets are usually averaged and weighted by model species occurrences as a function of environmental variables, accuracy. Thus, by combining different model projections, final which is the standard approach used by ENMs (see Franklin [7] consensus may benefit from accurate models, although depending and Peterson et al. [8], for recent reviews). Techniques for on how model predictions are combined, poor model predictions generating ENMs range from very simple bioclimatic envelope may cancel accurate models [12]. When applying ensemble of models up to complex machine learning-based methods [7]. forecasts the final solution is an unique consensus, weighted by

PLOS ONE | www.plosone.org 1 January 2013 | Volume 8 | Issue 1 | e54323 Spatial Conservation Planning under Climate Change overall statistical fit (e.g. TSS statistics, AUC values) of combined data to model species ecological niche is still incipient in the ENMs models, from which is also possible to quantify and map model literature (but see Diniz-Filho et al. [13], Lawler et al. [25], for uncertainties [13]. good examples). However, in regions of poor knowledge on species Ecological niche models can be useful to develop conservation distribution, and under high threat to biodiversity, such approach plans, especially in regions where complete information on species may provide an initial identification of general priorities, which distribution is not available or will not be in the future, as expected can be revised after data improvement [26–28]. in megadiverse countries [14]. Other studies have used ENMs for We compiled current climatic data from the WorldClim conservation planning, however ENM uncertainties are rarely database (worldclim.org/current), and future climatic scenarios incorporated (but see Carroll et al. [15], Wilson [16], for recent from the International Center for Tropical Agriculture (CIAT, examples). Therefore, it is still necessary to develop science-based http://ccafs-climate.org), developed by IPCC’s Fourth Assessment portfolios of spatial priorities in which species’ range shifts driven Report (AR4). We modeled species’ ecological niche as a function by climatic changes are incorporated. of four climatic variables: annual mean temperature, temperature To improve spatial conservation planning one can map and seasonality (standard deviation *100), annual precipitation, and quantify species’ range shifts driven by climate change, measuring precipitation seasonality (coefficient of variation). These variables how much (and in which direction) a species is expected to move, are often used to explain patterns of amphibian species richness and include this specific response in priority-setting analyses. For and distribution [29]. We used the following Atmosphere-Ocean example, a species that is highly sensitive to changes in climate Global Circulation Models (hereafter AOGCMs) projected to the would either need (1) a larger conservation area compared to a less mid-point of a 30-year period spanning 2071–2099 (i.e. 2080): sensitive species or (2) the conservation of an area that is currently CCCMA-CGCM2 – developed by the Canadian Centre for out of its geographic distribution. Species climate-forced dispersal Climate Modeling Analysis, CSIRO-MK2 – developed by the takes place when some species need to disperse to sites that will Australia’s Commonwealth Scientific and Industrial Research become climatically suitable in the future because those in which Organization, and HCCPR-HadCM3 – built by the Hadley they currently occur are becoming unsuitable. Centre for Climate Predictions and Research’s General Circula- Here we used consensual projections of ENMs to generate a tion Model. We choose these AOGCMs because they are widely nested ranking of priority sites for species conservation that used in the literature, having also different equilibrium climate considers species climate-forced dispersal by minimizing the sensitivity values ranging from 3.1uC to 4.4uC (see also Diniz-Filho distance a species would need to move to find a climatically et al. [13], Nori et al. [30]). Equilibrium climate sensibility is the suitable site. We measured uncertainty associated to ENMs and annual mean surface air temperature change experienced by used it to minimize model uncertainties, favoring the inclusion of climate system after it has attained a new equilibrium in response low-uncertainty sites in conservation plans. We also considered the to a doubling of CO2 concentration and are within the range of all current network of protected areas established in the region in our AOGCMs available from International Panel on Climate Change analyses. Hence our plans complement the level of protection (IPCC) [31]. already achieved in the region. We projected species’ distribution in a 0.160.1 latitude- longitude grid (ca. 11 km size in the equator, totaling 11,461 Methods equal-area grid cells). We used six modeling methods to built ENMs: Generalized Linear Models (GLMs [32]), Generalized Geographic extent of the study Additive Models (GAMs [33]), Multivariate Adaptive Regression We focused our study in the Atlantic Forest. This natural Splines (MARS [34]), Genetic Algorithm for Rule Set Production domain is a Biodiversity Hotspot given its high level of plant (GARP [35]), Random Forest (RF [36]), and Maximum Entropy endemism and a massive loss of it natural vegetation cover [17]. (MaxEnt [37]). These methods are commonly used to generate 2 Originally extending over 1.5 million km along eastern Brazilian ENMs, and details on each one of them can be found in Franklin coast, now only ca.11% of its natural cover remains [18], and only [7]. For each species, data were randomly divided into calibration 7.2% of its remaining habitats are strictly protected in Brazil (I–IV and validation sets, comprising 75 and 25% of the species’ range, IUCN protected areas categories; [19]). Here we used the respectively. This procedure was repeated 50 times, maintaining historical Atlantic Forest domain extension [20] to acquire the observed prevalence of species in each partition (i.e. for information on species original climatic conditions. presence-only methods, 75% of the cells within the species’ range, randomly defined; for presence-absence methods, we did the Ecological niche models analyses using a random sample of 75% of cells both inside and We gathered information on geographic distribution (extent of outside species’ range). occurrence maps downloaded from iucnredlist.org/technical- We established a threshold of pseudo-absences for each model documents/spatial-data) of 431 amphibian species inhabiting the to allow building the receiving operating curve (ROC) and Atlantic Forest. We used amphibians as our case study because transforming quantitative predictions of models into a binary they are the most threatened vertebrate group on Earth [21], vector of 0/1, indicating forecasted presences or absences in each being particularly sensitive to climate change [2]. They also need grid cell [38]. We established the cut-off point by using the urgent conservation actions in the Neotropics [22–23]. delimitation of bioclimatic envelope of 95%. We used True Skill Systematic conservation planning demands spatially extensive Statistics (TSS) as our measure of model statistical fit. Sensitivity information on species distributions [24]. Although usually used, and specificity were calculated based on the probability threshold point location data are sparse and often biased in their sampling for which their sum is maximized, not being affected by toward areas that are easily accessible, thus increasing omission prevalence. TSS values range from 21to+1, where +1 indicates errors. For these reasons, here we used digital range maps to perfect fit, minimizing overprediction and omission error rates; generate a presence-absence matrix of amphibian occurrence in values close to zero indicate performance worse than randomly the Atlantic Forest. This matrix, along with climate variables (see expected [38]. We combined all model outputs generating bellow), was then used as our input data for building species’ ensemble-based frequencies of species distributions both for ecological niche models (Fig. 1). The use of range maps as input current and future climates. We considered species as occurring

PLOS ONE | www.plosone.org 2 January 2013 | Volume 8 | Issue 1 | e54323 Spatial Conservation Planning under Climate Change

Figure 1. Flowchart of stages on analysis and data inputs used in this study. Flowchart of stages on analysis and data inputs used in spatial conservation planning approach designed to the Atlantic Forest Biodiversity Hotspot, Brazil, in the face of climate change. doi:10.1371/journal.pone.0054323.g001 in a given cell if at least 50% of models predict its occurrence there replication [40] to quantify variation associated to each source, (i.e. a majority consensus rule, see Diniz-Filho et al. [13]). Finally, using species richness as response variable and modeling methods we also calculated species turnover for each combination of and AOGCMs as factors [13]. We then obtained the sum of modeling method, and AOGCM, which was based on the number squares, which can be attributed to each of these sources. As we of potential species gained (G) or lost (L) within each cell, and did the analyses for each grid cell covering the whole Atlantic given by (G+ L)/(S+G), where S is the species richness of the cell in Forest, it was possible to map each variance component and the present [12,39]. identify sites of low and high uncertainty [13]. We used the We averaged the projections of species distributions across each estimated proportion of the sum of squares attributable to the two grid cell generating a species richness consensus map, as well as sources in respect to total sum of squares (i.e. model uncertainty) as coefficients of variation that allow mapping where uncertainty in a constraint in spatial prioritization analyses (see below). model projections is larger. To map uncertainties associated with SDMs, we did a two-way Analysis of Variance (ANOVA) without

PLOS ONE | www.plosone.org 3 January 2013 | Volume 8 | Issue 1 | e54323 Spatial Conservation Planning under Climate Change

Defining spatial conservation priorities for current time specific value of this weigh is arbitrary, although the transforma- and for 2080 tion of IUCN Red List categories to an ordinal scale has been For solving the ‘‘utility maximization’’ problem we identified already used in conservation [45], as well as in spatial planning priority areas using the Zonation framework and software [41–42]. analyses (e.g. Gittleman [44], Loyola et al. [45]). Therefore, Zonation algorithm identifies sites primarily important for sensitivity analyses to test the importance of weighting would be retaining high-quality and connected habitats for several features required prior to utilization in a real-world conservation context. (e.g. species). It establishes a hierarchical ranking of conservation Further, we included restricted protected areas (IUCN I-IV priorities for all cells throughout the geographic space, minimizing categories) already established in the Atlantic Forest (data from the loss of conservation value [41]. A cell is defined as ‘‘more UNEP-WCMC [21]) in all spatial planning analyses (totaling 820 important’’ when its relative contribution to total conservation out of 11,461 grid cells). This means that our plans consider the value is the highest along the entire planning region. This level of current network of protected areas and indicate sites in which importance is the conservation value of that cell. conservation investment should take place to complement the Mathematically, the function for calculating the marginal loss actual system. (i.e. relative contribution of each cell to total conservation value) In addition, we used information on the percentage of natural transforms the representation of feature j in site i into a general vegetation cover for each grid cell in all analyses as another conservation value. There are different ways to calculate marginal attribute to be represented. We assigned weight five to this feature. loss of a given site i (d ). The basic way is shown in equation 1: We applied a weight for retaining natural vegetation higher than i those for threatened species for two main reasons. First, given the lack of detailed information of habitat preferences for all species qij : wj di~ max ð1Þ we studied, and the coarse spatial resolution of our vegetation j Qj(S) : ci layer (ca. 11 km), it was impossible to clip consensus species’ distributions to natural vegetation remnants (which would where q is the representation level of feature j at site i (e.g. ij diminish commission errors). Nevertheless, we controlled for proportion of species’ distribution in site i, in our case), w is the j model overprediction securing that only sites with currently weight (or priority) of feature j, and c is the cost of site i (ENM i available habitat to species survival and reproduction would be uncertainty associated to site i, in our case). The term Q (S) M S j i included in the spatial plan. Second, indicating that sites with high Sq , is the proportion of original distribution of species j at the ij percentage of natural vegetation cover need to be retained means remaining set of sites S. that every result indicates priority sites only in areas in which there Here we used a variant of Zonation (the additive benefit are large remnants of natural vegetation to ensure real effective- function) that promotes representation of all species, favoring sites ness of spatial plans, if they would be applied. with high species richness while considering species’ proportional As explained above, we also used model uncertainties arising distribution in a given cell [41]. Then, marginal loss of each site i both from modeling methods and AOGCMs as a constraint in our was defined as a function of the sum for species-specific values that analyses. This component is useful to assess the exclusion/ occur in the grid cell [i.e. summing the result of equation (1) for all inclusion effect of cell in priority site selection [46]. In our case, species occurring at the cell, for all cells]. Here we calculate the estimated proportion of sum of squares attributable to the two marginal loss using additive benefit function [43], as follows: sources in respect to total sum of squares, for each cell, was included as a constraint (or a cost, technically), c (see equation 1 X X i 1 1 and 2). Including uncertainties in ENMs as a constraint means that d ~ :w DV ~ w V Q ðÞS {V Q ðÞS{fig ð2Þ i j j j j j j j sites for which there is low concordance among model projections ci j ci j produced by different modeling methods or climate models should In equation (2), marginal loss of cell i is simply the difference in not be prioritized because there is risk of misallocating scarce conservation value of cell i found between the value (Vj)in conservation resources in places where the certainty about the remaining set as a priority (S) and the value (Vj) when the site i is occurrence of species is low. removed from solution (S - {i}). Repeated iteration of equation (2) and removal of sites that generate the smallest loss of conservation Accommodating climate-forced dispersal in spatial value (i.e. smallest marginal loss) produce a rank based on conservation plans complementarity, over the geographic space [42]. This rank is Developing a dynamic spatial plan requires protecting impor- used to map priority sites for conservation. The last removed site is tant areas for conservation both in current and future climates. the one with the highest value of marginal loss, that is, the one that Here, assigned a high-conservation value only to sites that are contributes the most to feature conservation (see Moilanen et al. suitable for each species both in current and future climates [41], [42], for more details). Note that this is a heuristic algorithm that because species distribution tends to be limited by contiguous does not necessarily achieve a solution that is optimal, but often suitable habitats [47]. For this, we obtained centroids of each near optimal. Yet, our problem is non-linear and very complex species’ distribution projected for present and future. Then, we (with 431 species and .11,000 sites). In such cases, the degree of assumed that Euclidean distance between current and future near-optimality associated to solutions is much less relevance given centroids of species’ projected distribution could act as a measure that the plan would not be implemented at once. of the dispersal ability of a species (in time). We established weights to species according to their conserva- In this case, the Euclidean distance between these centroids tion status defined by the IUCN Red List [37]: non-threatened corresponds to a negative exponential function that describes species = 1, vulnerable and data deficient species = 2, endangered species dispersal from present to future, forced by climate change species = 3, critically endangered species = 4. All other species had [48], i.e. the distance that a species would need to move to find weight = 1. Thus, weighting a critically endangered species as 4, suitable climatic conditions in the future. We then used species- means that maximizing representation of this species in priority specific Euclidean distance between current and future distribution sites is four times more important than doing so for a common centroids to derive a bi-dimensional model estimated on the basis species (because weights are multiplicative, see equation 2). The of the width of species-specific smoothing kernels (i.e. the distance

PLOS ONE | www.plosone.org 4 January 2013 | Volume 8 | Issue 1 | e54323 Spatial Conservation Planning under Climate Change

Figure 2. Different amphibian richness patterns for current and future climates in the Atlantic Forest, Brazil. Patterns of amphibian species richness projected by ecological niche models generated from different modeling methods (GLM, GAM, MARS, GARP, RF, and MaxEnt), and climate models (AOGCMs) both for current time and the year 2080 in the Atlantic Forest Biodiversity Hotspot, Brazil. doi:10.1371/journal.pone.0054323.g002 a species would disperse under climatic changes, see Moilanen & methods were responsible for 72% of variation in projections, Kujala [41]). Hence, the connectivity value among neighbor sites generating very distinct patterns of species richness. In general, all is directly proportional to occupancy level of species at the focal methods indicate high species richness in the eastern part of site. We used the distribution smoothing method available in Atlantic Forest, with low species richness in southern and western Zonation to connect areas in agreement to surrounding area portions of the biome. Projections of GLM provide a clear suitability. The method considers species-specific requirements in exception to this pattern, with richest areas concentrated in climate and dispersal capacity (based on smoothing kernels). The northeast, both for current time and for 2080 (Fig. 2). All models result is a set of priority sites that are more clumped in space [41]. forecasted a general reduction in species’ ranges, which leads to a There are few reasons for which aggregating sites at this spatial decrease in the number of sites with high species richness. resolution would benefit species (especially considering extinction Variation among models projected into the future (i.e. within risk and metapopulation dynamics [41]), but fundamentally, this is AOGCMs) was low, corresponding to only 0.5% of difference the only way to include species climate-forced dispersal in analyses among maps (Fig. 2). For most species, TSS values were relatively done here. high (TSS 6 SD = 0.6361.33) indicating good model fit. Hereafter, we will focus our attention in the consensus map, Results derived from the combination of all above-mentioned projections, weighted by their model fit (models with higher TSS value have ENMs differed according to modeling methods and climate more weight). Our consensus model forecasted range shifts that models used to project species’ distributions (Fig. 2). Modeling

PLOS ONE | www.plosone.org 5 January 2013 | Volume 8 | Issue 1 | e54323 Spatial Conservation Planning under Climate Change

Figure 3. Consensus species richness patterns for current and future climates in the Atlantic Forest, Brazil. Consensus maps of amphibian species richness patterns for current time (A), and the year 2080 (B). Uncertainty level associated to ensemble of ecological niche models (C), and spatial patterns of species turnover (D) in the Atlantic Forest Biodiversity Hotspot, Brazil. doi:10.1371/journal.pone.0054323.g003 culminate with high species richness in central and eastern portion and eastern Atlantic forest is expected to have high turnover rates of the biome, both for current time and for 2080 (Fig. 3 A–B, (Fig. 3D). respectively). Most species had a significant range contraction (up For practical purposes, here we show the top 17% of cells that to 72%; mean 6 SD = 3862.38%) and 12% of species were contribute the most to our conservation goal. The 17% of land projected to be regionally extinct. The western part of the biome is area target was recently proposed for conservation of terrestrial expected to have fewer species in the future. ecosystems by the Convention on Biological Diversity [49]. It is a The interaction between modeling methods and AOGCMs general target aimed at providing a fixed level of protection shows a different contribution to the geographically structured worldwide to be achieved by 2020. As it is a time-bound target variation around consensus solution. ENM uncertainty is higher in recommended to encourage countries to increase their level of southwest, but also along coastline (Fig. 3C). This means that, protected area coverage in the coming decade, coverage targets while there is no expected gain in species richness in the west, our may be higher in the future. Hence, our proposed priority area models do not forecast the same effects of climate change in this network seeks only to meet the minimum target recommended for particular region. Although future scenarios did not show dramatic 2020 [49]. changes in species richness, mean projected turnover was relatively Priority sites for investment in amphibian conservation that high throughout the biome, ranging from 0.04 to 0.89. Southern complements current established network of protected areas

PLOS ONE | www.plosone.org 6 January 2013 | Volume 8 | Issue 1 | e54323 Spatial Conservation Planning under Climate Change

Figure 4. Priority sites for amphibian species conservation in the Atlantic Forest, Brazil. Top 17% of cell that should be prioritized if conservation of amphibian species inhabiting the Atlantic Forest Hotspot, Brazil, is planned for the present (A), future (B), and if we consider species climate-forced dispersal (from present to future, C). A full combination of these solutions is shown in Fig. 4D. doi:10.1371/journal.pone.0054323.g004 differed when built upon present or future species distribution we would reach such target. However, because of climate-driven (Fig. 4A–B). We reached a compromised solution when we species’ range shifts, we would still need to protect an additional included species climate-forced dispersal in optimization proce- 1% of the Atlantic Forest to safeguard all amphibian species. For dure (Fig. 4C, compare similar areas in Fig. 4A and B). This later this reason, the solution presented in Fig. 4C is the best option solution indicates a set of sites that are climatic suitable both in among the ones we presented. present and future, and that are connected by the likely dispersal In addition to priority maps, curves plotting the performance of distance species would be able to comprise during climatic solutions (Fig. 5) provide valuable insights on the relative changes. Hence, it already includes possible dispersal corridors protection attained under different climatic contexts. The figure linking current and future suitable climate in priority sites for shows the fraction of species distribution remaining against the amphibian conservation. Moreover, it also includes a minimiza- fraction of remaining sites in the Atlantic forest, as the algorithm tion of errors associated with ENMs at remaining areas of native gradually eliminates cells with the smallest marginal loss. The vegetation (Fig. 4C). A full combination of these solutions is shown arrow in Fig. 5 indicates the tipping point where 83% of the biome in Fig. 4D. is lost (therefore retaining the best 17% of its land surface for Our results indicate that today we still need to protect at least an protection). additional 9.8% of the biome to meet the 17% target (blue sites in Fig. 4A). If all priority sites proposed here only for the current time were to be converted in protected areas during the next 70 years

PLOS ONE | www.plosone.org 7 January 2013 | Volume 8 | Issue 1 | e54323 Spatial Conservation Planning under Climate Change

any changes in shape and size of the biome itself [54]. Likewise, we did not included neighboring species outside of biome that can immigrate to Atlantic Forest due to climate change. Finally, coarse scale data can over-exaggerate the impact of climate change on species distribution [55]. Our results show that most species will likely experience a significant range contraction in the future, and many others could face extinction. The contraction of a species’ range is a reasonable concern since it increases the probability of species extinction [56]. The clearest impact should occur along the coastal line and the limits of the biome where we expect more changes in distribution. In fact, changes in species richness are not expected to be high in tropical regions given relatively short environmental gradients. However, given that species inhabiting the tropics may have narrower niches than their temperate counterparts, they are potentially susceptible to even small alterations in climatic conditions [57–58]. Although projections for the future cannot be truly validated given the dynamic processes of natural systems [6], ENMs still figure as the best strategy to obtain data that will be used into conservation planning [59]. Predicting the future is obviously not Figure 5. Performance curves of different spatial conservation trivial, as it requires model extrapolation, so the effectiveness of solutions under climate change. Performance of conservation plans conservation strategies will depend on suitable habitats for species for amphibian species inhabiting the Atlantic Forest Hotspot, Brazil. both now and in the future [60]. Here, we addressed the problem Line colored in magenta represents the prioritization considering of incorporating species range shifts (climate-driven dispersal) in species climate-forced dispersal and model uncertainties. Blue and spatial prioritization conservation. Our approach included the yellow lines stand, respectively, for the prioritizations based on current and future species distribution models. Colors as in Fig. 4. minimization of distance between centroid of the range maps in doi:10.1371/journal.pone.0054323.g005 present and future. Spatial conservation plans will be obviously more effective if the effects of climatic changes can be anticipated Discussion [59]. Today, the number and location of networks of protected area are still mostly based on current species distributions [61–62], Our analyses use a conservation biogeography approach [50] to ignoring range shifts that should happen with climate changes. evaluate changes in species’ ranges as a function of climate change, Recently, some authors planned for species persistence over and to optimize priorities sites for conservation under such threat. time, considering a dynamic environment and planning for reserve Instead of focusing on a particular species (e.g. an invasive species, connectivity in fragmented landscape [4,14]. Game et al. [63], for Nori et al. [30]), we evaluated the efficiency of conservation plans example, described strategies for climate change adaptation as part as a function of environmental variables for a large number of of national conservation assessment in Papua New Guinea. In species. We believe our results convey important recommenda- particular they demonstrate that inclusion of climate refugia and tions for environmental management and policy. Although we cross-environment connectivity would make possible to reduce the used a top-ranked Biodiversity Hotspot and amphibians as a case amount of environmental change expected to take place inside study for illustrating our approach, we believe this could help protected areas [63]. We support their recommendations. For the developing effective conservation actions under a dynamic Atlantic Forest, most remaining habitats figure as small and assessment (like those expected in face of climate change). isolated forest remnants [17], which highlights the importance of Therefore, our approach may be applied to different spatial connectivity between different habitats to accommodate species scales, geographic regions, and taxonomic groups. climate-forced dispersal. All this should be considered in future Some authors showed greater robustness of consensual models conservation assessments for the region. when compared to a particular model [51–52]. The ensemble The best 17% of the Atlantic Forest covers different proportions forecasting approach minimizes the difficulty of establishing the of species ranges when planning is made for different time periods, best criterion to evaluate performance of ecological niche models or to accommodate climate-forced dispersal. Within the best 17% [10]. Moreover, the central tendency of selected forecasts has of sites (see Fig. 5), the apparently inferior performance of the greater precision of species distribution since this consensus model spatial solution based on species climate-forced dispersal (line covers a full range of uncertainties [51]. Notice that 72% of colored in magenta) depends on the way marginal loss was variation around of consensus model is due to different techniques calculated: it is constrained to select cells that may include used to model species’ ecological niche and project their combinations of site with both high and low conservation value, distribution. Moreover, it gives an objective measurement of while trying to represent most of species distribution in conserva- uncertainty in the process, which can be mapped or, as done here, tion planning. Thus, analysis based only on future geographic used to generate a weighting scheme in spatial conservation distribution is apparently better in terms of species representation prioritization (see below). – because distributions are smaller. However, investing in Application of species’ ENMs assumes that species exhibit an conservation plans based only on future distribution models is unlimited dispersal ability and absence of biological interactions problematic, since there is no guarantee that species will in fact [53]. Thus, our models result from the interaction between shift their geographic range to the predicted location. Thus, we mechanisms operating at a broad spatial scale, given that species highlight the importance of the solution shown in Fig. 4C, which distributions are driven only by environmental or climatic shows the climate-forced dispersal scenario. As expected, the conditions. Further, our modeling approach does not consider performance of the climate-forced dispersal solution is intermedi-

PLOS ONE | www.plosone.org 8 January 2013 | Volume 8 | Issue 1 | e54323 Spatial Conservation Planning under Climate Change ate between current and future solution, as it represents a [70] and opens a strategy to the establishment of dynamic compromise between these two. programs and conservation planning analyses that may help to Several sources of uncertainty arise in the process of conserva- better allocate scarce resources for biodiversity conservation [71]. tion planning [64,65]. Available species distribution data are We hope that our approach provides insights on the establishment incomplete or with high-commission errors (false presence) due to of conservation priorities within sites of high biological importance the interpolation of occurrence records [66]. Secondly, ecological in the face climate change. niche modeling techniques introduce uncertainties, because model projections vary [13]. Of course, it is desirable to achieve a Acknowledgments compromise between low ENM uncertainty and the conservation value of a given site [64]. In addition, other aspects of uncertainty We thank JAF Diniz-Filho, LM Bini, RP Bastos, and four anonymous can also be considered when proposing the establishment of new reviewers for helpful comments on this manuscript. We owe a debt of protected areas, such as extinction risk related to patch area [67– thanks to Thiago Rangel for discussions about the ensemble forecasting approach applied to biodiversity conservation, and for revising the final 68] and availability of land for immediate acquisition [69]. Here version of this paper. we developed a more general conceptual model for establishing a dynamic spatial conservation prioritization analysis (see Fig. 1) that Author Contributions help planners to identify locations that are important both for the current time and for future scenarios of climate change. This is one Conceived and designed the experiments: PL RDL. Analyzed the data: PL. of the top-priority questions in spatial conservation prioritization Wrote the paper: PL RDL.

References 1. Brook BM, Sodhi NS, Bradshaw CJA (2008) Synergies among extinction drivers 22. Loyola RD, Becker CG, Kubota U, Haddad CFB, Lewinsohn TM (2008) Hung under global change. Trends Ecol Evol 23: 453–460. out to dry, choice of priority ecoregions for conserving threatened Neotropical 2. Pounds JA, Bustamente MR, Coloma LA, Consuegra JA, Fogden MPL, et al. anurans depends on lifehistory traits. PLoS ONE 3 (5): e2120. (2006) Widespread amphibian extinctions from epidemic disease driven by 23. Becker CG, Loyola RD, Haddad CFB, Zamudio KR (2010) Integrating species global warming. Nature 439:161–67. lifehistory traits and patterns of deforestation in amphibian conservation 3. Root TL, Price JT, Hall KR, Schneider SH, Rosenzweig C, et al. (2003) planning. Divers Distrib 16 (1): 10–19. Fingerprints of global warming on wild animals and plants. Nature 421: 57–60. 24. Elith J, Graham CH, Anderson RP, Dudı´k M, Ferrier S, et al. (2006) Novel 4. Arau´jo MB, Cabeza M, Thuiller W, Hannah L, Williams PH (2004) Would methods improve prediction of species’ distributions from occurrence data. climate change drive species out of reserves? An assessment of existing reserve Ecograpphy 29: 129–151. selection methods. Glob Change Biol 10: 1618–1626. 25. Lawler JJ, Shafer SL, White D, Kareiva P, Maurer EP, et al. (2009) Projected 5. Margules CR, Pressey RL (2000) Systematic conservation planning. Nature 405: climate-induced faunal change in the Western Hemisphere. Ecology 90 (3): 588– 243–53. 597. 6. Pearson RG, Dawson TP (2003) Predicting the impacts of climate change on the 26. Rondinini C, Di Marco M, Chiozza F, Santulli G, Baisero D, et al. (2011) distribution of species: are bioclimate envelopes useful? Glob Ecol Biogeogr 12: Global habitat suitability models of terrestrial mammals. Philos Trans R Soc B 361–371. 366 (1578): 2633–2641. 7. Franklin J (2009) Mapping species distributions: spatial inference and 27. Lemes PL, Faleiro FAMV, Tessarolo G, Loyola RD (2011) Refinando dados predictions. 1st edition. Cambridge University Press, Cambridge. espaciais para a conservaçaõ da biodiversidade. Nat Conservaçaõ 9(2): 240–243. 8. Peterson AT, Soberoń J, Pearson RG, Anderson RP, Martinez-Meyer ER, et al. 28. Whittaker RJ, Arau´jo MB, Jepson P, Ladle RJ, Watson JEM, et al. (2005) (2011) Ecological Niches and Geographical Distributions. Monographs in Conservation Biogeography: assessment and prospect. Divers Distrib 11: 3–23. Population Biology 49, Princeton University Press. 29. Buckley AH, Jetz W (2007) Species richness, hotspots, and the scale dependence 9. Arau´jo MB, New M (2007) Ensemble forecasting of species distributions. Trends of range maps in ecology and conservation. P Natl Acad Sci USA 104: 13384– Ecol Evol 22: 42–47. 13389. 10. Lobo JM, Jimeńez-Valverde A, Real R (2008) AUC: a misleading measure of the 30. Nori J, Urbina-Cardona JN, Loyola RD, Lescano JN, Leynaud GC, et al. (2011) performance of predictive distribution models. Glob Ecol Biogeogr, 17: 145– Climate Change and American Bullfrog Invasion: What Could We Expect in 151. South America? PLoS ONE 6: e25718. 11. Elith J, Leathwick JR (2009) Species Distribution Models: Ecological 31. IPCC (2007) Climate Change 2007 – Impacts, Adaptation and Vulnerability: Explanation and Prediction across Space and Time. Annu Rev Ecol Syst 40: Contribution of Working Group II to the Fourth Assessment Report of the 677–697. IPCC. Cambridge University Press, Cambridge. 12. Lawler JJ, Wiersma YF, Huettmann F (2011) Using species distribution models 32. McCullagh P, Nelder JA (1989) Generalized linear models. 2nd ed. Chapman for conservation planning and ecological forecasting. In: Drew AC, Wiersma and Hall/CRC. YF, Huettmann F. Predictive Species and Habitat Modeling in Landscape 33. Yee TW, Mitchell ND (1991) Generalised additive models in plant ecology. Ecology. Springer New York, New York, USA. pp 271–290. J Veg Sci 2: 587–602. 13. Diniz-Filho JAF, Bini LM, Rangel TFLVB, Loyola RD, Hof C, et al. (2009) ´ Partitioning and mapping uncertainties in ensembles of forecasts of species 34. Munõz J, Fellicısimo AM (2004) Comparison of statistical methods commonly turnover under climate change. Ecography 32: 897–906. used in predictive modeling. J Veg Sci 15: 285–292. 14. Kamino LHY, Stehmann JR, Amaral S, De Marco P Jr, Rangel TF, et al. (2011) 35. Stockwell DRB, Noble IR (1992) Induction of sets of rules from animal Challenges and perspectives for species distribution modelling in the neotropics. distribution data: a robust and informative method of data analysis. Math Biol Lett doi:10.1098/rsbl.2011.0942. Comput Simulat 33: 385–390. 15. Carroll C, Dunk JR, Moilanen A (2010) Optimizing resiliency of reserve 36. Breiman L (2001) Random forest. Mach Learn 45: 5–32. networks to climate change: multispecies conservation planning in the Pacific 37. Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modeling of Northwest, USA. Glob Change Biol 16: 891–904. species geographic distributions. Ecol Model 190: 231–259. 16. Wilson KA (2010) Dealing with Data Uncertainty in Conservation Planning. 38. Allouche O, Tsoar A, Kadmon R (2006) Assessing the accuracy of species Nat Conservacao 8: 145–150. doi: 10.4322/natcon.00802007. distribution models: prevalence, kappa and the true skill statistic (TSS). J Appl 17. Mittermeier RA, Robles-Gil P, Hoffman M, Pilgrim J, Brooks T, et al. (2004) Ecol 43: 1223–1232. Hotspots revisited: Earth’s biologically richest and most endangered terrestrial 39. Thuiller W (2004) Patterns and uncertainties of species’ range shifts under ecoregions. Chicago, IL: University of Chicago Press. 392p climate change. Glob Change Biol 10: 2020–2027. 18. Ribeiro MC, Metzger JP, Martensen AC, Ponzoni FJ, Hirota MM (2009) The 40. Sokal RR, Rohlf FJ (1995) Biometry, 3rd edn. W.H. Freeman, San Francisco, Brazilian Atlantic forest: how much is left, and how is the remaining forest California. distributed? Implications for conservation. Biol Conserv 142:1141–1153. 41. Moilanen A, Kujala H (2008) Zonation: software for spatial conservation 19. UNEP-WCMC (2010) Data Standards for the World Database on Protected prioritization. User Manual v2.0. Metapopulation Research Group, University Areas, UNEP-WCMC. of Helsinki, Finland. 20. SOS Mata Atlaˆntica, Instituto Nacional de Pesquisas Espaciais (2008) Atlas dos 42. Moilanen A, Franco AMA, Early R, Fox R, Wintle B, et al. (2005) Prioritising remanescentes florestais da Mata Atlaˆntica, perıódo de 2000 a 2005. Available: multiple-use landscapes for conservation: methods for large multi-species http://www.sosmatatlantica.org.br. planning problems. Proc R Soc B 272: 1885–1891. 21. Stuart SN, Chanson JS, Cox NA, Young BE, Rodrigues ASL, et al. (2004) Status 43. Arponen A, Heikkinen RK, Thomas CD, Moilanen A (2005) The value of and tends of Amphibian Declines and Extinction Worldwide. Science 306: biodiversity on reserve selection: representation species, species weighting, and 1783–1785 benefit functions. Conserv Biol 21: 527–533.

PLOS ONE | www.plosone.org 9 January 2013 | Volume 8 | Issue 1 | e54323 Spatial Conservation Planning under Climate Change

44. Gittleman JL (2001) Carnivore conservation. Cambridge University Press, 59. Hannah L, Midgley G, Andelman S, Arau´jo M, Hughes G, et al. (2007) Cambridge. Protected area needs in a changing climate. Front Ecol Environ 5: 131–138. 45. Loyola RD, Oliveira G, Diniz-Filho JAF, Lewinsohn TM (2008) Conservation of 60. Heller NE, Zavaleta ES (2009) Biodiversity management in the face of climate Neotropical carnivores under different prioritization scenarios: mapping species change: a review of 22 years of recommendations. Biol Conserv 142:14–32. traits to minimize conservation conflicts. Divers Distrib 14: 949–960. 61. Pressey RL, Cabeza M, Watts ME, Cowling RM, Wilson KA (2007) 46. Cabeza M, Moilanen A (2006) Replacement cost: a practical measure of site Conservation planning in a changing world. Trends Ecol Evol 22: 583–592. value for cost-effective reserve planning. Biol Conserv 132: 336–342. 62. Gaston KJ, Charman K, Jackson SF, Armsworth PR, Bonn A, et al. (2006) The 47. Arau´jo MB, Thuiller W, Pearson RG (2006) Climate warming and the decline of ecological effectiveness of protected areas: the United Kingdom. Biol Conserv amphibians and reptiles in Europe. J Biogeogr 33: 1712–1728. 132: 76–87. 48. Rayfield B, Moilanen A, Fortin MJ (2009) Incorporating consumer-resource 63. Game ET, Lipsett-Moore G, Saxon E, Peterson N, Sheppard S (2011) spatial interactions in reserve design. Ecol Model 220: 725–733. Incorporating climate change adaptation into national conservation assessment. 49. Mittermeier R, Baiaõ PC, Barrera L, Buppert T, McCullough J, et al. (2010) O Glob Change Biol 17: 3150–3160. Protagonismo do Brasil no Histo´rico Acordo Global de Proteçaõ a` Biodiversi- 64. Moilanen A, Runge MC, Elith J, Tyred A, Carmele Y, et al. (2006) Planning for dade. Nat Conservacao 8: 197–200. doi: 10.4322/natcon.00802017 robust reserve networks using uncertainty analysis. Ecological Model 199: 115– 50. Ladle RJ, Whittaker RJ, eds. (2011) Conservation Biogeography.Wiley-Black- 124. well. 301p 65. McCarthy MA, Thompson CJ, Moore AL, Possingham H (2011) Designing 51. Arau´jo MB, Whittaker RJ, Ladle R, Erhard M (2005) Reducing uncertainty in nature reserves in the face of uncertainty. Ecol Lett 14 (5): 470–475. projections of extinction risk from climate change. Glob Ecol Biogeogr 14: 529– 66. Rondinini C, Stuart S, Boitani L (2005) Habitat suitability models and the 538. 52. Marmion M, Parviainen M, Luoto M, Heikkinen RK, Thuiller W (2009) shortfall in conservation planning for African vertebrates. Conserv Biol 19: Evaluation of consensus methods in predictive species distribution modeling. 1488–1497. Divers Distrib 15: 59–69. 67. Hole DG, Huntley B, Arinaitwe J, Butchart SHM, Collingham Y, et al. (2011) 53. Soberoń J (2007) Grinnellian and Eltonian niches and geographic distributions Toward a Management Framework for Networks of Protected Area in the Face of species. Ecol Lett 10: 1115–1123. of Climate Change. Conserv Biol 25: 305–325 54. Salazar LF, Nobre CA, Oyama MD (2007) Climate change consequences on the 68. Nicholson E, Possingham HP (2007) Making conservation decisions under distribution in tropical South America. Geophys Res Lett 34: 1–6. uncertainty for the persistence of multiple species. Ecol Appl 17: 251–165. 55. Willis KJ, Bhagwat SA (2009) Biodiversity and Climate Change. Science 69. Meir E, Andelman S, Possinghma HP (2004) Does conservation planning matter 326:806–807. in a dynamic and uncertainty world? Ecol Lett 7: 615–622. 56. Jetz W, Wilcove DS, Dobson AP (2007) Projected impacts of climate and land- 70. Arau´jo MB (2009) Climate change and spatial conservation planning. In: use change on the global diversity of birds. PLoS Biol 5:e157. Moilanen A, Wilson KA, Possingham HP, editors. Spatial conservation 57. Bush MA, Hooghiemstra H (2005) Tropical biotic responses to climate change. prioritization, quantitative methods & computational tools. Oxford University In: Lovejoy TE, Hannah H, editors. Climate change and biodiversity. New Press, Oxford. pp 172–184. Haven and London, Yale University Press. pp 125–156. 71. Moilanen A, Kujala H, Leathwick J (2009) The Zonation Framework and 58. Carnaval AC, Hickerson MJ, Haddad CFB, Rodrigues MT, Moritz C (2009) Software For Conservation Prioritization. In: Moilanen A, Wilson KA, Stability predicts genetic diversity in the Brazilian Atlantic Forest hotspot. Possingham HP. Spatial Conservation Prioritization: Quantitative Methods Science 323:785–789. and Computational Tools, Oxford University Press, Oxford, UK. pp 196–210.

PLOS ONE | www.plosone.org 10 January 2013 | Volume 8 | Issue 1 | e54323

Nesse capítulo, avaliamos a congruência entre as estratégias de conservação considerando a diversidade taxonômica, funcional e filogenética.

Capítulo 5: Integrated assessment of taxonomic, phylogenetic, and functional diversity reveals challenges and opportunities for conservation in the Brazilian

Atlantic Forest

501

1 Integrated assessment of taxonomic, phylogenetic, and functional diversity

2 reveals opportunities for conservation in the Brazilian Atlantic Forest

3 Priscila Lemes1, Federico Montesino Pouzols2, Tuuli Toivonen2, Mariana Gomes Batista1, Levi 4 Carina Terrible3, Thiago Fernando Rangel4, Célio Fernando Baptista Haddad5, Atte Moilanen2 & 5 Rafael Loyola4,*

7 Authors’ Affiliations:

8 1 Pós Graduação em Ecologia e Evolução, Universidade Federal de Goiás, Goiânia, Goiás, 9 Brazil.

10 2 Department of Biosciences, Finnish Centre of Excellence in Metapopulation Biology, PO Box

11 65, FI-00014 University of Helsinki, Finland.

12 3 Departamento de Ecologia, Universidade Federal de Goiás, Campus Jataí, BR 364, Km 192,

13 Campus II, Jataí, Goiás, 75801-615, Brazil.

14 4 Departamento de Ecologia, Universidade Federal de Goiás, Caixa Postal 131, Goiânia, Goiás,

15 74001-970, Brazil.

16 5 Departamento de Zoologia, Instituto de Biociências, UNESP, Rio Claro, São Paulo, 35264-100,

17 Brazil.

19 Corresponding author’s address:

20 Rafael Loyola. Departamento de Ecologia, Universidade Federal de Goiás, Caixa Postal 131, 21 Goiânia, Goiás, 74001-970, Brazil. E-mail: [email protected]

60 23 24 25 Abstract

26 Spatial conservation prioritization should broadly embrace different aspects of biodiversity to 27 protect nature in a well-balanced manner. Phylogenetic and functional diversity are powerful 28 measures of evolutionary history and their underlying ecological processes. However, it still is 29 unclear how to include these diversity measures into spatial planning and how to evaluate 30 conservation outcome with respect to them. Here we implemented high resolution conservation 31 prioritization for the coverage of species distributions and taxonomic, phylogenetic and 32 functional diversity of amphibian communities in the Brazilian Atlantic Forest. We found 33 relatively high congruence between spatial conservation plans that were based on different 34 diversity measures. We also identified high-priority conservation areas for each of the 35 ecoregion's of the Brazilian Atlantic Forest. Conservation priority outputs by biogeographic 36 region also show high congruence between prioritization scenarios, with exceptions to be found 37 for example in the Alto Paraná Atlantic Forest. All previously existing protected areas were 38 identified as high-priority areas. Overall, our results reveal high spatial congruence of 39 conservation opportunities that minimize loss of both evolutionary potential and species 40 distributions simultaneously. Given the level of urgency and the need for action in Brazil, the 41 present conservation prioritization can inform decision makers about high priority areas for 42 anuran conservation.

43 Keywords: amphibians, complementarity, conservation planning, conservation policy, 44 congruence, decision analysis.

47 48 Introduction

49 The ongoing global biodiversity crisis requires a comprehensive conservation effort that

50 challenges conservation scientists to develop ecologically meaningful methods that can be

61 51 applied to the real world (Brooks et al., 2009; Butchart et al., 2010). Losing species may result in

52 the loss of evolutionary information and underlying ecological processes (Purvis, 2008).

53 Consequently, ecologists have come to recognize that conservation strategies should embrace all

54 the many aspects of diversity in biodiversity conservation (Devictor et al., 2010; Strecker et al.,

55 2011; Winter et al., 2013; Mazel et al., 2014). Today, limited resources and incomplete data

56 make it challenging to develop proposals broadly cover biodiversity in a balanced manner

57 (Naidoo et al., 2008; Rodrigues et al., 2011). Prioritization has frequently relied on taxonomic

58 diversity as a proxy for overall biodiversity, without accounting for the influence on the relative

59 functional attributes of different organisms or their ecological similarities correlated to periods of

60 evolutionary divergence (but see Devictor et al., 2010; Trindade-Filho et al., 2012; Tucker et al.,

61 2012; Sobral et al., 2014; Zupan et al., 2014). One way to do so is to account for phylogenetic

62 diversity in prioritization, in addition to the most common component of biodiversity, species

63 distributions.

64 Phylogenetic approach fundament is the assumption that closely related taxa will tend to be more

65 similar in their physiologies and life history strategies than more distantly related taxa (Harvey

66 and Pagel, 1991). Therefore, maximizing the preservation of phylogenetic diversity will also tend

67 to maximize ecologically important traits and the potential relationship between functional

68 diversity and ecosystem functioning will also be considered as a byproduct (Cadotte et al., 2011).

69 A functional trait can be any measurable individual feature of a species that potentially affects

70 the performance of the species at any locality. Based on functional traits, functional diversity can

71 strongly influence ecosystem functioning (Tilman et al., 1997; Mouchet et al 2010) and it can

72 provide additional complementary information about species occurring in the location (Cadotte

73 et al., 2013). For instance, if there is agricultural intensification, the functional and taxonomic

74 diversity of animal communities may be reduced (Flynn et al., 2009). Therefore, areas with high 62 75 functional diversity are especially relevant for biodiversity conservation (Trindade-Filho et al.,

76 2012).

77 Phylogenetic and functional diversity may differ substantially between species assemblages with

78 the same taxonomic richness (Petchey and Gaston, 2006). Beyond taxonomic diversity, species

79 composition strongly influences ecosystem processes (Tilman, 1997). Some studies have

80 proposed that diversity measures are acceptable surrogates for taxonomic diversity (Schipper et

81 al., 2008; Safi et al., 2011), but there are conflicting patterns (Rodrigues et al., 2011). Because

82 there are still no unified measures that include all aspects of biodiversity (Pavoine and Bonsall,

83 2011; Tucker and Cadotte, 2013; but see Mazel et al., 2014), the value of each component should

84 be properly recognized. Although taxonomic diversity has been well-employed in conservation

85 plans (Rodrigues et al., 2011), conservation biologists have found spatial mismatches between

86 taxonomic, phylogenetic, and functional diversity (Devictor et al., 2010; Strecker et al., 2011).

87 Thus, including evolutionary relationships between species (phylogenies) and ecosystem

88 functioning into spatial conservation prioritization remains a substantial challenge.

89 Systematic conservation planning is an operational model for cost-efficient and effective design

90 and implementation of biodiversity conservation (Margules and Pressey, 2000; Knight et al.

91 2006). Within systematic conservation planning, spatial conservation prioritization is an analysis

92 that identifies conservation priority areas based on a host of considerations related to the

93 occurrence of various components of biodiversity (e.g., species, habitats, ecosystem services),

94 connectivity, threats and costs and other socioeconomic and political criteria (Di Minin et al.,

95 2013; Faleiro et al., 2013; Kukkala and Moilanen 2013; Nori et al., 2013; Thomas et al., 2013).

96 Because of the need to balance numerous partially conflicting considerations, it can be said that

97 systematic conservation planning is based on the concept of complementarity (Kukkala and

63 98 Moilanen 2013). As an important consideration, conservation scientists are concerned about

99 ensuring adequate conservation effort under ongoing global change (Carvalho et al. 2011; Kujala

100 et al., 2013; Lemes and Loyola, 2013). Thus, different methods have been developed to help

101 identify optimal sets of sites for protection (Strecker et al., 2011; Trindade-Filho et al., 2012;

102 Sobral et al., 2014).

103 The focal species group of this study, Neotropical amphibians, have an alarming proportion of

104 threatened species and several instances of severe population declines (Stuart et al., 2004; IUCN,

105 2014). In fact, amphibians are globally the group of vertebrates with the highest proportion of

106 threatened species and the strongest IUCN Red List Index decline (Hoffman et al., 2010). Habitat

107 loss has accounted for more declines than any other factor (Lion et al., 2014), although the

108 spread of pathogenic fungal disease is also a significant factor (Fisher et al., 2009). Some studies

109 have illustrated that both amphibian extinction risk and population declines are taxonomically

110 non-random (Cooper et al., 2008; Becker & Loyola 2008) at different geographic scales (Bielby

111 et al., 2010; Batista et al., 2013). Therefore, there is demand for information relevant for

112 amphibian conservation. Recent amphibian conservation efforts typically target protection,

113 conservation conflicts (Nori et al., 2013), and climate change (Lemes and Loyola, 2013; Loyola

114 et al., 2013), but have lacked conservation plans that account for phylogenetic and/or functional

115 diversity.

116 Here we compare common spatial prioritization based on species distribution with analyses that

117 incorporate functional traits and evolutionary information. We verify the hypothesis that spatial

118 prioritization based on different diversity measures (taxonomic, functional and phylogenetic)

119 deliver different solutions. We integrate these solutions into a more inclusive and meaningful

120 strategy for spatial conservation prioritization. Relevant for on-the-ground conservation effort in

64 121 Brazil, we identify high-priority areas for anuran conservation in each biogeographic region of

122 the Brazilian Atlantic Forest.

123

124 Methods

125 Study area

126 We focused our analyses on the Brazilian Atlantic Forest, which includes several biogeographic

127 regions (Olson et al., 2001) that host distinct assemblages of amphibians (Haddad et al. 2013).

128 Internationally recognized as a Biodiversity Hotspot (Mittermeier et al. 2005), the Brazilian

129 Atlantic Forest is a lush biome that holds 18% of all South American amphibian species (IUCN,

130 2014). The original vegetation cover of the forest was approximately 1,300,000 km2 (Ribeiro et

131 al., 2009), but only 7.2% of remaining habitats are strictly protected (I-IV IUCN categories;

132 IUCN and UNEP, 2013). In addition, 11.7% of the forest remaining is in the form of small

133 fragments (97% are smaller than 250 ha) of secondary growth forest (Ribeiro et al., 2009). As it

134 is, the current network of protected areas has failed to conserve a significant sample of

135 amphibian biodiversity (Lemes et al., 2014), and further conservation effort is needed.

136 Ecological Niche Modeling (ENM)

137 There are approximately 600 amphibians species in the Brazilian Atlantic Forest (Haddad et al.,

138 2013), but the final dataset used here was restricted both by availability of occurrence records

139 and ecological traits data (see subsection Evolutionary and Ecological Distinctiveness). We were

140 able to use distribution information for 328 species, which were obtained as conservative

141 estimates of the extent of occurrence from the International Union for Conservation of Nature

142 (IUCN, 2014). While commonly used, point locality data can be inadequate or biased to easily

65 143 accessible areas, thus increasing omission errors. For these reasons, we used digital range maps

144 to generate a presence-absence matrix of amphibian anuran occurrence in the Brazilian Atlantic

145 Forest. We used SAM v4.0 (Spatial Analysis in Macroecology; Rangel et al., 2010) to overlay all

146 of the species into an equal-area grid (0. latitude/longitude, 27 km) that covers the full extent of

147 the biome. Given that species-climate niche models have a limited ability to cope with few

148 occurrence records, we considered analyses of only those species that had more than ten

149 withingrid-cells records.

150 The climatic variable used here were those of Hijmans et al. (2005), publicly available for

151 download at . The four variables used in this analysis were: mean annual

152 temperature (°C), temperature seasonality (standard deviation * 100), mean annual summed

153 precipitation (mm), and precipitation seasonality (coefficient of variation); which have often

154 been used to explain patterns of amphibian species richness and distribution (Buckley and Jetz,

155 2007).

156 A key assumption in our modeling is that ensemble forecasts generate more accurate or at least

157 more conservative projections of species distribution than do single models (Araújo and New,

158 2007; Marmion et al., 2009). Therefore, five modeling methods were used to build the

159 Ecological Niche Models (ENMs), including: General Additive Models (GAM), General Linear

160 Models (GLM), Multivariate Adaptive Regression Splines (MARS), Random Forest (RF) and

161 Artificial Neutral Network (ANN) (for a review see Peterson et al., 2011). For all models, we

162 randomly partitioned the presence and absence data of each species into calibration and

163 validation data (75:25) and repeated this process for 10× cross-validation, maintaining the

164 observed prevalence of each species. We combined all model outputs to generate an

165 ensemblebased estimate of species distributions. Species were considered to occur in a given cell

66 166 if at least 50% of the ENMs predict species occurrence there (i.e. a majority consensus rule)

167 (Diniz-Filho et al., 2009). Meanwhile, our final distribution estimate for a species was a

168 consensus prediction weighted by the overall statistical fit of combined models (TSS – True Skill

169 Statistics; Allouche et al., 2006). The TSS range was from -1 to +1, where values equal to plus

170 one (=+1) indicate perfect prediction and values equal or less than zero (≤0) imply that

171 predictions are no better than random (Allouche et al., 2006).

172

173 Prioritization rules

174 For solving the “maximum utility” conservation problem, we carried out a spatial conservation

175 prioritization analysis using the Zonation framework and software (Moilanen et al., 2012).

176 Zonation can be used with many (up to tens of thousands) spatial distributions of biodiversity

177 features described by GIS raster maps (Pouzols et al. 2014). Zonation can identify areas that are

178 important for retaining habitat quality and connectivity for multiple features (species, habitats,

179 ecosystem servicesetc.) simultaneously, by performing a complementary-based iterative

180 prioritization process that generates a ranking map where the rank of a sites (grid cells) reflects

181 its conservation priority (Moilanen et al., 2012). Here, the cells with highest rank would have

182 highest conservation priority for amphibians in the Brazilian Atlantic Forest; below, we explain

183 how phylogenetic and functional diversity were accounted for in the process. We used an

184 analysis variant called the Additive Benefit Function, which bases selection on a cell’s weighted

185 normalized occurrence for all species, accounting for how much would be lost for the species in

186 areas not allocated for conservation (Moilanen et al., 2007). In total, we used a data set of 656

187 conservation features, including species-specific ENMs and their corresponding uncertainty

188 maps in Zonation analysis (see Fig. 1S). 67 189 The main input of our spatial analysis was the ENM consensus models downscaled to high

190 resolution (0.05 degrees). Some studies have demonstrated that the greatest source of

191 uncertainty in forecasts of species range shifts arises from using alternative methods for ENM

192 estimation (Diniz-Filho et al., 2009; 2010; Loyola et al., 2012). To account for the spatially

193 explicit pattern of uncertainty in the ENMs, we used the distribution-discounting technique

194 (Moilanen et al., 2006), which penalizes site and species -specific measures of occurrence by an

195 associated error measure. In effect, we subtracted one standard deviation off the mean, thereby

196 emphasizing locations where models consistently predicted high occurrence levels (Carroll et al.

197 2010; Kujala et al. 2013).

198 In addition, for each riverine species, we filtered ENM outputs and habitat preferences in high

199 resolution accounting for rivers data in Brazilian Atlantic Forest from Brazilian National Water

200 Agency (available for download at < http://hidroweb.ana.gov.br>, see the flowchart in Fig. 1S).

201 For taxonomic diversity, we included quantitative criterion with the conservation status of all the

202 species from IUCN Red List of Threatened Species (IUCN, 2014) as species-specific priorities into

203 our spatial prioritization: 0 (least concern), 1 (near threatened), 2 (vulnerable), 3 (endangered), and

204 4 (critically endangered and data deficient) following the precautionary principle (Mace et al.,

205 2008). These quantitative representations of conservation status was used to weight species’

206 importance in spatial prioritizations.

207

68 208 Evolutionary and ecological distinctiveness as a proxy for the phylogenetic and functional

209 diversity

210 The measure of phylogenetic species isolation describes the phylogenetic relationship of a

211 species in relation to other species of its clade regardless of whether they co-occur (Isaac et al.,

212 2007; 2012). Therefore, the evolutionary distinctiveness (ED) was calculated from the sum of the

213 branch lengths along the phylogenetic tree to a given species (Isaac et al., 2007). (Note that the

214 sum of all ED-species is equivalent to the basic-phylogenetic diversity (PD) approach (Faith,

215 1992).) Analogous to ED, we calculated the ecological distinctiveness for measuring how unique

216 the functional traits of a species in a given assemblage are. We extracted the phylogenetic

217 relationships for 328 anuran species from an undated phylogeny that only provides the

218 relationship among species (topology). Our second source of topological information for 122

219 anuran species was the molecular phylogeny of Pyron and Wiens (2011). This phylogeny is

220 based on molecular data containing 2,871 species represented by up to 12,871bp of sequence

221 data from 12 genes, of which three of them are mitochondrial and nine are nuclear. We pruned

222 the Pyron and Wiens phylogeny (2011), to produce a ‘main taxa tree’ to which according to

223 Frost (2013), would be relatively uncontroversial. For the other 206 species, we used the Most

224 Derived Consensus Clade (MDCC; Rangel et al., 2013) that provided possible topologies of the

225 phylogenetic tree (Martins et al., 2013). Using a Monte Carlo procedure, we generated 1,000

226 possible topologies for amphibian phylogenetic tree that include 122 species already present in

227 available phylogeny plus phylogenetic uncertainty taxa (PUT). This approach maximizes the use

228 of incomplete phylogenetic information while considering PUT in statistical inferences (Fig. 1S).

229 Both ED and MDCC were calculated using the software Phylogenetic Analyses in Macroecology

230 (PAM) (Rangel and Diniz-Filho, 2013).

69 231 For measuring ecological distinctiveness, we collected information on 29 anuran ecological

232 traits: body size (average snout-vent length, in mm); reproductive mode (eggs and exotrophic

233 tadpoles in still water; eggs and exotrophic tadpoles in running water; eggs and exotrophic

234 tadpoles in water in tree holes or arterial plants; exotrophic tadpoles in ponds; eggs in floating

235 foam nests and exotrophic tadpoles in ponds; eggs imbedded in dorsum of an aquatic female

236 hatching into exotrophic tadpoles; eggs on the ground, on rocks, or in burrows; eggs hatching

237 into endotrophic tadpoles that complete their development in the nest; direct development of

238 terrestrial eggs; arboreal eggs; eggs in foam nests; eggs carried by adults; unknown; habit

239 (arboreal; terrestrial; cryptozoic; rupicolous; reophilic; aquatic); activity period (diurnal;

240 nocturnal); calling site (bamboo; canopy; tree branches; stream, swamp or lake; rocks;

241 bromeliad; forest floor), and habitat (open areas; forest). This dataset includes biological traits

242 describing the major ecological anuran reproductive strategies and resource use of habitat

243 (Haddad et al., 2008). Then, we built a functional dendrogram from a modification of Gower’s

244 distance (Pavoine et al., 2009) that can accommodate different types of variables and the

245 unweighted pair-group method using arithmetic averages (UPGMA) by using the function

246 dist.ktab() in the ade4 package (Dray and Dufour, 2007) and evol.distinct() in picante package

247 all implemented in R (R Development Core Team, 2010).

248 Our approach is to weight the priority of taxonomic, phylogenetic and functional diversity

249 differentially, influencing the relative balance that emerges between features in the solution (e.g.

250 Moilanen and Arponen, 2011). Raising the weight of a feature will lead to an increased

251 representation level for it across the priority ranking. Zonation produces a hierarchical ranking of

252 throughout the landscape with increasingly important biodiversity remaining in top ranked

253 locations (Moilanen et al., 2012). We compared the spatial concordance between the top 10%

254 and at successive 10% intervals for all pairwise combinations of diversity measures. In order to 70 255 determine whether congruence was greater than expected by chance, we ran a randomization

256 procedure with n = 999 permutations in R (R Development Core Team, 2010).

257

258 Results

259 Distribution models developed under current climatic conditions have accurately predicted

260 species distribution throughout the Brazilian Atlantic Forest. ENMs differed according to the

261 modeling method used to project the potential species’ distribution (see Fig. 1). This mainly

262 occurred for projections of GLM with the highest species richness in the northeast (see Fig. 1).

263 Nevertheless, all ecological niche models indicated high species richness in the eastern part of

264 the Brazilian Atlantic Forest, with decreased species richness in the southern and western region

265 of the biome. In general, the fit statistic of the models was high (TSS±SD=0.94±0.49; see Table

266 1S), and the higher variation among ENMs was in the northeast and at the biome’s boundaries

267 (see Fig. 1).

268 Phylogenetic and functional diversity measures the summarized the degree of relatedness among

269 species. Obviously, our quantitative results are dependent on the ecological traits and on the

270 phylogeny considered in the analysis. However, both evolutionary and ecological distinctiveness

271 is not evenly distributed across species of conservation urgency (see Fig. 2). Amphibian ED

272 scores for evolutionary diversity ranged from 0.008 (Physalaemus signifer group) to 0.178

273 (Myersiella microps) while ED scores for ecological diversity ranged from 0.012 (Scinax alter)

274 to 0.364 (Limnomedusa macroglossa).

275 Priority scenarios found highest congruence between functional and phylogenetic diversity

276 scenarios (95.75%), and the lowest congruence between phylogenetic diversity and taxonomic

277 diversity (91.5%) when the top 10% highest ranked sites were compared (see Fig. 3). In a 71 278 simultaneous comparison of all three scenarios, the congruence was 98.06% for the highest 10%

279 ranking of each scenario. All combinations of diversity scenarios were significantly more

280 congruent than random expectations (n = 999 permutations; P < 0.001).

281 Additionally, we examined the patterns of spatial mismatches for taxonomic, functional, and

282 phylogenetic diversity scenarios (see Fig. 4). The highest conservation priorities are for all

283 scenarios concentrated along the east coast, mainly from the southern portion of the state of

284 Pernambuco to the state of Rio de Janeiro, including areas such as Pernambuco and Bahia

285 Forest, and Serra do Mar Forests. These are well-known biogeographic regions, which hold a

286 high number of endemic species (see Fig. 5). In addition, we have examined the top of 17%

287 priorities, following a recent target for global conservation (Mittermeier et al., 2010). For each

288 biogeographic region, there was a high congruence between prioritization scenarios, similar to

289 the Pernambuco Forests (see Fig. 6). However, both the lowest congruence of these and the

290 highest mismatches were found in the Alto Parana Atlantic Forest. All protected areas have been

291 ranked as high-priority areas, among which 62% are in the Serra do Mar Forests.

292

293 Discussion

294 We found that taxonomic diversity might be a good diversity surrogate contrary to our original

295 hypothesis that different biodiversity measures could possibly deliver mismatches in spatial

296 prioritization. Such results are critical because they reveal high-priority congruence and

297 opportunities to minimize the loss of evolutionary and ecological information for amphibian

298 conservation.

299

72 300 Methodological aspects

301 Our spatial conservation plans were based on consensual models of species distributions that

302 were corrected by penalizing those locations displaying the highest uncertainty of the predicted

303 occurrence of species (Moilanen et al., 2006). Consensual models already combine single

304 models’ predictions thereby minimizing the uncertainty of their outputs (Araújo and New, 2007)

305 and increasing the accuracy of ENMs (Marmion et al., 2009). Key assumptions of these models

306 commonly include unlimited dispersal ability of species and the absence of biological

307 interactions (Soberón, 2007), while the broad outlines of species range have been limited by

308 climate niche. Consensual models have been frequently applied in Conservation Biology,

309 biodiversity assessment and spatial prioritization for conservation (Loyola et al., 2012; 2014;

310 Lemes et al., 2014) and can provide a valuable and meaningful tool especially in poorly surveyed

311 regions that are under increasing human pressure (see Faleiro et al., 2013; for example). However,

312 a model that overpredicts presences could lead to spending limited resources protecting species

313 where they do not occur or where unsuitable conditions exist for the species. Many studies have

314 reported that the greatest source of uncertainty in ENMs arises from alternative modeling methods

315 (Diniz-Filho et al., 2009; 2010). Therefore, the quantification of underlying uncertainties in ENMs

316 is relevant information for stakeholders and decision makers

317 (De Ornellas et al., 2011) and should therefore be accounted for in spatial conservation plans

318 (Regan et al., 2009; Faleiro et al., 2013; Lemes and Loyola, 2013). Our analyses use the

319 distribution discounting technique to account for uncertainty in spatial distributions (Moilanen et

320 al., 2006). This increases the robustness of our assessment in the face of uncertainty inherent in

321 the models.

73 322 Ecologists and evolutionary biologists have recently been discussing the importance of

323 evolutionary information in nature conservation (Mouquet et al., 2012; Winter et al., 2013).

324 Here, we have incorporated different diversity measures into complementarity-based spatial

325 prioritization to determine the best areas to expand conservation effort. Not surprisingly, our

326 spatial prioritizations reveal that spatial congruence of priority areas depends on the degree of

327 overlap between the different diversity measures underlying analysis (see Fig. 4). Our reults

328 suggest that variation exists in ecosystem functioning beyond what can be explained by

329 taxonomic diversity. Nevertheless, some studies have appointed high congruence between

330 taxonomic and phylogenetic diversity because little is known about these phylogenetic

331 relationships among species (Rodrigues et al., 2011; Diniz-Filho et al. 2013). Due to the

332 imminence of further species extinctions globally (Butchart et al., 2010), molecular biologists

333 have been motivated to develop more accurate phylogenies based on molecular data (Pyron and

334 Wiens, 2011; Diniz-Fiho et al. 2013). Although a complete molecular phylogeny of amphibians

335 is unavailable (Pyron and Wiens, 2011), new statistical approaches have been considered.

336 Accordingly, some studies have provided insights about how to include a consensual

337 phylogenetic tree accounting for uncertainties of possible trees (Batista et al., 2012; Isaac et al.,

338 2012). Our analyses based on the MDCC’s approach are consistent with the principles of a

339 phylogenetic ‘supertree’ construction and can handle substantial amounts of uncertainty

340 regarding the phylogenetic positions of species (Bininda-Emonds et al., 2002).

341

342 Opportunities

343 The Brazilian Atlantic Forest is an area of high biodiversity value that is extremely threatened by

344 human development. The area has been appointed a global conservation priority (Myers et al., 74 345 2000). Consequently, ecologists have advocated for conservation planning to protect different

346 taxonomic groups within this area (Trindade-Filho et al., 2012; Loyola et al., 2013), to facilitate

347 species range-shifts under climate change (Lemes and Loyola, 2013; Loyola et al., 2013), and to

348 assess the current protected areas (Lemes et al., 2014; Loyola et al., 2014). Understanding

349 priority areas for amphibians is one piece in the puzzle of understanding conservation priorities

350 in this area. Amphibian species are relatively threatened due to their dependence on both specific

351 microhabitats and hydrological regimes, which makes them sensitive to habitat destruction

352 (Haddad et al., 2008; Lion et al., 2014). There is insufficient information to allow the evaluation

353 of the conservation status of many amphibians inhabiting the Brazilian Atlantic Forest (“Data

354 Deficient” species, IUCN, 2014); nevertheless, these species each have vast and complex

355 evolutionary histories (Isaac et al., 2012; Morais et al., 2013).

356 One of the most surprising outcomes of the present analysis was the high level of concordance

357 between spatial prioritization based on taxonomix (species), phylogenetic and functional (trait)

358 data. The top of 10% priorities-ranked was congruent up 91.5% across taxonomic, phylogenetic,

359 and functional diversity (see Fig. 3). According to the Convention on Biological Diversity, 17%

360 of terrestrial ecosystems should be protected (Mittermeier et al., 2010), but this target has not

361 been reached yet in the Brazilian Atlantic Forest. Our results were shocking when the ecoregion

362 subdivision is taken into account: at top-ranked 17% priority the Serra do Mar Forests should

363 have 26.2% of land under protection while the Pernambuco Forests would have less than 1%

364 (see Fig. 5). Also, the lack of protected areas (I-IV IUCN categories; IUCN and UNEP, 2013) in

365 the Alto Parana Atlantic Forests and Araucaria Forests should be highlighted to achieve better

366 balanced conservation coverage across the entire biome (see Fig. 6). The Brazilian Atlantic

367 Forest is a region that hosts some of the most endemic amphibian fauna in the world, and the

75 368 Serra do Mar Forest is an area with particularly high amphibian richness and endemism

369 (Villalobos et al. 2013). Such patterns also coincide with historical habitat stability (Carnaval et

370 al., 2009) and underscore how unique patterns of diversification should be accounted for in

371 spatial prioritization.

372 Challenges

373 There are significant challenges for amphibian conservation, especially in the Brazilian Atlantic

374 Forest. Unfortunately, humans have historically altered the Brazilian Atlantic Forest, which has

375 resulted in a severe reduction of its original extent (Ribeiro et al., 2009). Major current threats to

376 biodiversity include invasive species, unsustainable resource usage and roadway development

377 (“economic growth”), and high population density (Tabarelli et al., 2010). The ongoing

378 degradation of pristine habitats (Ribeiro et al., 2009) defines the biome as one of the most

379 threatened and fragmented tropical forests in the world (Myers et al., 2000). Consequently, many

380 species of different taxa are currently threatened by global extinction as populations are

381 collapsing locally and regionally (Butchart et al., 2010; IUCN 2013).

382 Complementarity-based spatial prioritization has been providing increasingly useful information

383 for amphibian conservation (Lemes and Loyola 2013; Lemes et al., 2014, Loyola et al., 2013;

384 2014). Our analyses are useful for conservation planners thinking about the establishment of

385 protected areas considering both evolutionary and ecological perspectives. The identification of a

386 comprehensive set of biodiversity features describing the multifaceted nature of biodiversity, is

387 only the first step in the systematic conservation approach (Margules and Pressey, 2000; Pressey

388 and Bottrill, 2009). If functional and phylogenetic diversity components are good surrogates, an

389 integrative approach to connect biogeography, evolution, and functional ecology should be

390 feasible in conservation plans. Further, our analyses might help decision makers address priority 76 391 areas with high congruence for amphibian conservation. Our results are encouraging in the sense

392 that taxonomic diversity appears a good surrogate for both phylogenetic and functional diversity

393 in spatial conservation planning. This is good news, as measuring phylogenetic and functional

394 diversity is often time-consuming and highly dependent on expert training and knowledge.

395 Future studies will be required to confirm whether improvements in the amphibian phylogenetic

396 tree will maintain this close relationship between diversity measures into spatial conservation

397 plans.

398

399 400 Acknowledgements

401 PL received fellowship from CNPq (grant #147345/2010-3 and #150480/2014-8) and a CAPES

402 Sandwich-PhD fellowship (grant #12583/12-0). FMP, TT, and AM thank the ERC-StG grant

403 260393 (GEDA) for support. RL research has been constantly funded by CNPq (grants

404 #304703/2011-7, 479959/2013-7, 407094/2013-0), Conservation International Brazil, the O

405 Boticário Group Foundation for the Protection of Nature (PROG_0008_2013). We are grateful to

406 M.V. Cianciaruso, C.E.V. Grelle, and R.P. Bastos for useful comments and discussion. We also

407 grateful to John Karpinski for editing.

408

409 References

410 Allouche O, Tsoar A, Kadmon R (2006) Assessing the accuracy of species distribution models:

411 prevalence, kappa and the true skill statistic (TSS). J. Appl. Ecol. 43, 1223–1232.

77 412 Araújo MB, New M (2007) Ensemble forecasting of species distributions. Trends Ecol. Evol. 22,

413 42–47.

414 Batista MCG, Gouveia SF, Silvano DL, Rangel TF (2013) Spatially explicit analyses highlight

415 idiosyncrasies: species extinctions and the loss of evolutionary history. Divers. Distrib. 19,

416 1543–1552. doi:10.1111/ddi.12126

417 Becker CG, Loyola RD, (2008) Extinction risk assessments at the population and species level:

418 implications for amphibian conservation. Biodiver. Conserv. 17,2297–2304.

419 Ben-Haim Y (2006) Info-Gap Decision Theory - Decisions under Severe Uncertainty. Oxford,

420 UK: Academic Press. 421 Bielby J, Cardillo M, Cooper N, Purvis A (2010) Modelling extinction risk in multispecies data

422 sets: phylogenetically independent contrasts versus decision trees. Biodivers. Conserv. 19,

423 113–127. doi: 10.1007/s10531-009-9709-0

424 Bininda-Emonds ORP, Gittleman JL, Steel MA (2002) The (Super) Tree of Life: Procedures,

425 Problems, and Prospects. Annu. Rev. Ecol. S. 33, 265–289.

426 Brooks TM, Wright SJ, Sheil D (2009) Evaluating the success of conservation actions in

427 safeguarding tropical forest biodiversity, Conserv. Biol. 23, 1448–1457.

428 Buckley LB, Jetz W (2007) Environmental and historical constraints on global patterns of

429 amphibian richness. P. R. Soc. B. 274, 1167–1173.

430 Butchart SHM, Walpole M, Collen B, et al. (2010) Global biodiversity: indicators of recent

431 declines. Science 328, 1164–1168. doi:10.1126/science.1187512

432 Cadotte MW, Carscadden K, Mirotchnick N (2011) Beyond species: functional diversity and the

433 maintenance of ecological processes and services. J. Appl. Ecol. 48, 1079–1087. doi:

434 10.1111/j1365-2664.2011.02048.x

435 Cadotte MW, Albert CH, Walker SC (2013) The ecology of differences: assessing community

436 assembly with trait and evolutionary distances. Ecol. Lett. 16, 1–11. doi:10.1111/ele.12161 78 437 Cardinale BJ, Duffy JE, Gonzalez A, et al. (2012) Biodiversity loss and its impact on humanity.

438 Nature 59–67. doi: 10.1038/nature.11148

439 Carnaval AC, Hickerson MJ, Haddad CFB (2009) Stability Predicts Genetic Diversity in the

440 Brazilian Atlantic Forest Hotspot. Science 323, 785–789.

441 Carvalho SB, Brito JC, Crespo EG, Watts ME, Possingham HP (2011) Conservation planning

442 under climate change: Toward accounting for uncertainty in predicted species distributions to

443 increase confidence in conservation investments in space and time. Biol. Conserv. 144, 2020–

444 2030.

445 Cooper N, Bielby J, Thomas GH, Purvis A (2008) Macroecology and extinction risk correlates of

446 frogs. Global Ecol. Biogeogr. 17, 211–221.

447 De Ornellas P, Milner-Gulland EJ, Nicholson E (2011) The impact of data realities on

448 conservation planning. Biol. Conserv. 144, 1980–1988. doi: 10.1016/j.biocon.2011.04.018

449 Devictor V, Mouillot D, Meynard C, Jiguet F, Thuiller W, Mouquet N (2010) Spatial mismatch

450 and congruence between taxonomic, phylogenetic and functional diversity: the need for

451 integrative conservation strategies in a changing world. Ecol. Lett. 13, 1030–1040. doi:

452 10.1111/j.1461-0248.2010.01493.x

453 Díaz S, Lavorel S, de Bello F, Quétier F, Grigulis K, Robson TM (2007) Incorporating plant

454 functional diversity effects in ecosystem service assessments. P Natl. Acad. Sci. USA 104,

455 20684–20689.

456 Diniz-Filho JAF, Bini LM, Rangel TF, Loyola RD, Hof C, Nogués-Bravo D, Araújo MB (2009)

457 Partitioning and mapping uncertainties in ensembles of forecasts of species turnover under

458 climate change. Ecography 32, 897–906

459 Diniz-Filho JAF, Nabout JC, Bini LM, Loyola RD, Rangel TF, Nogués-Bravo D, Araújo MB

460 (2010) Ensemble forecasting shifts in climatically suitable areas for Tropidacris cristata

79 461 (Orthoptera: Acridoidea: Romaleidae). Insect Conserv. Divers. 3, 213–221. doi:

462 10.1111/j.1752-4598.2010.00090.x.

463 Diniz-Filho JAF, Loyola RD, Raia P, Bini LM (2013) Darwinian shortfalls in biodiversity

464 conservation. Trends Ecol. Evol. 28, 689–695. doi:10.1016/j.tree.2013.09.003

465 Di Minin E, Macmillan DC, Goodman PS, Escott B, Slotow R, Moilanen A (2013)

466 Conservation businesses and conservation planning in a biological diversity hotspot. Conserv.

467 Biol., 27, 808–820. doi:10.1111/cobi.12048

468 Dray S, Dufour AB (2007) The ade4 package: implementing the duality diagram for ecologists.

469 J. Stat. Softw. 22, 1–20.

470 Faleiro FV, Machado RB, Loyola RD (2013) Defining spatial conservation priorities in the face

471 of land-use and climate change. Biol. Conserv. 158, 248–257. doi:

472 10.1016/j.biocon.2012.09.020.

473 Fisher MC, Garner TWJ, Walker SF (2009) Global emergence of Batrachochytrium

474 dendrobatidis and amphibian chytridiomycosis in space, time, and host. Annu. Rev.

475 Microbiol. 63, 291–310.

476 Flynn DFB, Gogol-Prokurat M, Nogeire T et al. (2009) Loss of functional diversity under land

477 use intensification across multiple taxa. Ecol. Lett. 12, 22–33.

478 Frost DR (2013) Amphibian Species of the World: an Online Reference. Version .0. (accessed

479 10 April 2013). Electronic Database accessible at

480 http://research.amnh.org/herpetolgy/amphibia/index.html. American Museum of Natural

481 History, New York, USA.

482 Haddad CFB, Toledo LF, Prado CPA (2008) Atlantic Forest amphibians. Editora Neotropica,

483 São Paulo, Brazil.

484 Harvey PH, Pagel M (1991) The comparative method in evolutionary biology. Oxford, UK:

80 485 Oxford University Press.

486 Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A (2005) Very high resolution interpolated

487 climate surfaces for global land areas. Int. J. Climatol. 25, 1965–1978.

488 Hoffmann M, Hilton-Taylor C, Angulo A et al. (2010) The Impact of Conservation on the Status

489 of the World’s Vertebrates. Science 330, 1503-1509. doi: 10.1126/science.1194442

490 Hooper DU, Chapin FSIII, Ewel JJ et al. (2005) Effects of biodiversity on ecosystem

491 functioning: a consensus of current knowledge. Ecol. Monogr. 75, 3–35.

492 Isaac NJB, Turvey ST, Collen B, Waterman C, Baillie JEM (2007) Mammals on the EDGE:

493 Conservation Priorities Based on Threat and Phylogeny. PLoS ONE 2, e296. doi:

494 10.1371/journal.pone.0000296

495 Isaac NJB, Redding DW, Meredith HM, Safi K (2012) Phylogenetically-informed priorities for

496 amphibian conservation. PLoS ONE 7, e43912. doi: 10.1371/journal.pone.0043912

497 IUCN and UNEP (2013) The World Database on Protected Areas (WDPA). UNEP-WCMC.

498 Cambridge, UK. www.protectedplanet.net

499 IUCN (2014) The IUCN Red List of Threatened Species. Version 2.2. Available at:

500 http://www.redlist. org. (accessed 14 June 2014).

501 Kembel SW, Cowan PD, Helmus MR et al. (2010) Picante: R tools for integrating phylogenies

502 and ecology. Bioinformatics 26, 1463–1464.

503 Kukkala AS, Moilanen A (2013) Core concepts of spatial prioritisation in systematic

504 conservation planning. Biol. Rev. 88, 443–464. doi: 10.1111/bio.12008

505 Lemes P, Loyola RD (2013) Accommodating species climate-forced dispersal and uncertainties

506 in spatial conservation planning. PLoS ONE 8, e54323. doi:10.1371/jornal.pone.0054323

507 Lemes P, Melo AS, Loyola RD (2014) Climate change threatens protected areas of the Atlantic

508 Forest. Biodivers. Conserv. 23, 357–368. doi:10.1007/s10531-013-0605-2

81 509 Lion MB, Garda AA, da Fonseca CR (2014) Split distance: a key landscape metric shaping

510 amphibian populations and communities in forest fragments. Divers. Distrib. doi:

511 10.1111/ddi.12228

512 Loyola RD, Lemes P, Faleiro FV, Trindade-Filho J, Machado RB (2012) Severe loss of suitable

513 climatic conditions for marsupial species in Brazil: challenges and opportunities for

514 conservation. PLoS ONE 7, e46257. doi:10.1371/journal.pone.004625

515 Loyola RD, Lemes P, Nabout JC, Trindade-Filho J, Sagnori MD, Dobrovlski R, Diniz-Filho JAF

516 (2013) A straightforward conceptual approach for evaluating spatial conservation priorities

517 under climate change. Biodivers. Conserv. 22, 483–495. doi: 10.1007/s10531-012-0424-x

518 Loyola RD, Lemes P, Brum FT, Provete DB, Duarte LDS (2014) Clade-specific consequences of

519 climate change to amphibians in Atlantic Forest protected areas. Ecography 37:65–72. doi:

520 10.1111/j.1600-0587.2013.00396.x

521 Mace GM, Collar NJ, Gaston KJ et al. (2008) Quantification of extinction risk: IUCN’s system

522 for classifying threatened species. Conserv. Biol. 22, 1424–1442.

523 Margules CR, Pressey RL (2000) Systematic conservation planning. Nature 405, 243–253.

524 Marmion M, Parviainen M, Luoto M, Heikkinen RK, Thuiller W (2009) Evaluation of consensus

525 methods in predictive species distribution modelling. Divers. Distrib. 15, 59–69.

526 Martins WS, Carmo WC, Longo HJ, Rosa TC, Rangel TF (2013) SUNPLIN: Simulation with

527 Uncertainty for Phylogenetic Investigations. BMC Bioinformatics 14, 324.

528 doi:10.1186/14712105-14-324

529 Mazel F, Guilhaumon F, Mouquet N et al. (2014) Multifaceted diversity-area relationship reveal

530 global hotspots of mammalian species, trait and lineage diversity. Global Ecol. Geograph. 23,

531 836−847. doi: 10.1111/geb.12158 532 Mittermeier R, Baião PC, Barrera L et al. (2010) O Protagonismo do Brasil no Histórico Acordo

533 Global de Proteção à Biodiversidade. Nat. Conservação 8, 197–200. doi: 82 534 10.4322/natcon.00802017

535 Moilanen A (2007) Landscape Zonation, benefit functions and target-based planning: Unifying

536 reserve selection strategies. Biol. Conserv. 134, 571–579.

537 Moilanen A, Arponen A (2011) Administrative regions in conservation: Balancing local

538 priorities with regional to global preferences in spatial planning. Biol. Conserv. 144,

539 1719−1725. doi:10.1016/j.biocon.2011.03.007

540 Moilanen A, Meller L, Leppanen J, Pouzols FM, Arponen A, Kujala H (2012) Zonation Spatial

541 Conservation Planning Framework and Software V3.1, User Manual.

542 Moilanen A, Wintle BA, Elith J, Burgman M (2006) Uncertainty analysis for regional-scale

543 reserve selection. Conserv. Biol. 20, 1688–1697.

544 Morais AR, Siqueira MN, Lemes P, Maciel NM, de Marco P, Brito D (2013) Unraveling the

545 conservation status of Data Deficient species. Biol. Conserv. 166, 98−102. doi:

546 10.1016/j.biocon.2013.06.010

547 Mouchet MA, Villéger S, Mason NWH, Mouillot D (2010) Functional diversity measures : an

548 overview of their redundancy and their ability to discriminate community assembly rules.

549 Funct. Ecol. 24, 867–876.

550 Mouillot D, Bellwood DR, Baraloto C, et al. (2013) Rare species support vulnerable functions in

551 high-diversity ecosystems. PLoS Biol. 11, e1001569.

552 Mouquet N, Devictor V, Meynard CN, et al. (2012) Ecophylogenetics: advances and

553 perspectives. Biol. Rev. 87, 769–785.

554 Myers N, Mittermeier RA, Mittermeier CG, da Fonseca GA, Kent J (2000) Biodiversity hotspots

555 for conservation priorities. Nature 403, 853–858.

556 Naidoo R, Balmford A, Costanza R, et al. (2008) Global mapping of ecosystem services and

557 conservation priorities. P. Natl. Acad. Sci. USA 105, 9495–9500.

83 558 Nori J, Lescano JN, Illoldi-Rangel P, Frutos N, Cabrera MR, Leynaud GC (2013) The conflict

559 between agricultural expansion and priority conservation areas : Making the right decisions

560 before it is too late. Biol. Conserv. 159, 507–513. doi:10.1016/j.biocon.2012.11.020

561 Olson DM, Dinerstein E, Wikramanayake ED, et al. (2001) Terrestrial ecoregions of the world: a

562 new map of life on Earth. Bioscience 51, 933–938.

563 Ortega YK, Pearson DE (2005) Weak vs. Strong Invaders of Natural Plant Communities :

564 Assessing Invasibility and Impact. Ecol. Appl. 15, 651–661.

565 Pavoine S, Bonsall MB (2011) Measuring biodiversity to explain community assembly: a unified

566 approach. Biol. Rev. 86, 792–812. doi: 10.1111/j.1469-185X.2010.00171.x 1

567 Pavoine S, Vallet J, Dufour AB, Gachet S, Daniel H (2009) On the challenge of treating various

568 types of variables: application for improving the measurement of functional diversity. Oikos

569 118, 391–402.

570 Petchey OL, Gaston KJ (2006) Functional diversity: back to basics and looking forward. Ecol.

571 Lett. 9, 741–758.

572 Peterson AT, Soberón J, Pearson RG, Anderson RP, Martínez-Meyer E, Nakamura M, Araújo

573 MB (2011) In: Horn HS, Levin SA (eds) Ecological niches and geographic distributions.

574 Princeton University Press, Princeton. 575 Pouzols, FM, Toivonen, T, Di Minin, E, Kukkala, A, Kullberg, P, Kuusterä, J, Lehtomäki, J,

576 Tenkanen, H., Verburg, PH, Moilanen, A (2014) Global protected area expansion is

577 compromised by projected land-use and parochialism. Nature, in press.

578 Pressey RL, Bottrill MC (2008) Opportunism, threats and the evolution of systematic

579 conservation planning. Conserv. Biol. 22, 1340–1345.

580 Purvis A (2008) Phylogenetic Approaches to the Study of Extinction. Annu. Rev. Ecol. Evol. S.

581 39, 301–319.

84 582 Pyron RA, Wiens JJ (2011) A large-scale phylogeny of Amphibian including over 2,800 species,

583 and a revised classification of extant frogs, salamanders, and caecilians. Mol. Phylogenet.

584 Evol. 61, 543–583. doi: 10.1016/j.ympev.2011.06.012

585 R Development Core Team (2010) R: a language and environment for statistical computing.

586 Vienna, Austria: R Foundation

587 Rangel TF, Diniz-Filho JAF (2013) PAM: Phylogenetic Analysis in Macroecology. Version 0.9.

588 User’s Guide and application, available upon request from the authors.

589 Rangel TFLVB, Diniz-Filho JAF, Bini LM (2010) SAM: a comprehensive application for spatial

590 analysis in macroecology. Ecography 33, 46–50.

591 Regan H, Burgman M, Carey J (2009) Conservation prioritization and uncertainty in planning

592 inputs. In: Moilanen A, Wilson KA, Possingham HP (Eds.), Spatial Conservation

593 Prioritization: Quantitative Methods and Computational Tools. Oxford University Press,

594 Oxford, pp. 145–157.

595 Ribeiro MC, Metzger JP, Martensen AC, Ponzoni FJ, Hirota MM (2009) The Brazilian Atlantic

596 Forest: How much is left, and how is the remaining forest distributed? Implications for

597 conservation. Biol. Conserv. 142, 1141–1153.

598 Rodrigues ASL, Grenyer R, Baillie JEM, et al. (2011) Complete, accurate, mammalian

599 phylogenies aid conservation planning, but not much. Philos. T.R. Soc. B. 366, 2652–2660.

600 Safi K, Cianciaruso MV, Loyola RD, Brito D, Armour-Marshall K, Diniz-Filho JAF, (2011)

601 Understanding global patterns of mammalian functional and phylogenetic diversity. Philos. T.

602 R. Soc. B. 366, 2536–2544. doi: 10.1098/rstb.2011.0024

603 Schipper J, Chanson JS, Chiozza F et al., (2008) The Status of the World's Land and Marine

604 Mammals: Diversity, Threat, and Knowledge. Science 322, 225–230.

605 Soberón J (2007) Grinnellian and Eltonian niches and geographic distributions of species. Ecol.

85 606 Lett. 10, 1115–1123. Sobral FL, Jardim L, Lemes P, Machado N, Loyola R, Cianciaruso MV

607 (2014) Spatial conservation priorities for top predators reveal mismatches among taxonomic,

608 phylogenetic, and functional diversity. Nat. Conserv. (in press)

609 Strecker AL, Olden JD, Whittier JB, Paukert CP (2011) Defining conservation priorities for

610 freshwater fishes according to taxonomic, functional, and phylogenetic diversity. Ecol. Appl.

611 21, 3002–3013.

612 Stuart SN, Chanson JS, Cox NA, Young BE, Rodrigues ASL, Fischman DL, Waller RW (2004)

613 Status and tends of Amphibian Declines and Extinction Worldwide. Science 306, 1783–1785.

614 Tabarelli M, Aguiar AV, Ribeiro MC, Metzger JP, Peres CA (2010) Prospects for biodiversity

615 conservation in the Atlantic Forest : Lessons from aging human-modified landscapes. Biol.

616 Conserv. 143, 2328–2340.

617 Thomas CD, Anderson BJ, Moilanen A et al. (2013) Reconciling biodiversity and carbon

618 conservation. Ecol. Lett. 16, 39–47. doi: 10.1111/ele.12054

619 Tilman D, Knops J, Wedin D, Reich P, Ritchie M, Siemann E (1997) The Influence of

620 Functional Diversity and Composition on Ecosystem Processes. Science 277, 1300–1302. 621 Trindade-Filho J, Sobral FL, Cianciaruso MV, Loyola RD (2012) Using indicator groups to

622 represent bird phylogenetic and functional diversity. Biol. Conserv., 146, 155–162.

623 doi:10.1016/j.biocon.2011.12.004

624 Tucker CM, Cadotte MW (2013) Unifying measures of biodiversity: understanding when

625 richness and phylogenetic diversity should be congruent. Divers. Distrib. 19, 845–854. doi:

626 10.1111/ddi.12087

627 Villalobos F, Dobrovolski R, Provete DB, Gouveia SF (2013) Is Rich and Rare the Common

628 Share? Describing Biodiversity Patterns to Inform Conservation Practices for South American

629 Anurans. PLoS ONE 8, e56073. doi:10.1371/journal.pone.0056073

86 630 Winter M, Devictor V, Schweiger O (2013) Phylogenetic diversity and nature conservation:

631 where are we? Trends Ecol. Evol. 28, 199–204. doi: 10.1016/j.tree.2012.10.015

632 633 Figure Legends

634 Figure 1. Patterns of species richness projected by ecological niche models (ENMs) from

635 different modeling methods for amphibians inhabiting in the Brazilian Atlantic Forest.

636 Figure 2. Conservation status along of evolutionary and ecological distinctiveness for

637 amphibians inhabiting in the Brazilian Atlantic Forest.

638 Figure 3. Spatial congruence (%) between the top of 10% and at successive 10% intervals for

639 outcomes of different spatial conservation prioritization analyses based on taxonomic,

640 functional, and phylogenetic measures.

641 Figure 4. Spatial mismatches in priority areas for all pairwise combinations of diversity

642 measures for amphibians inhabiting in the Brazilian Atlantic Forest.

643 Figure 5. Top 17% of priorities for amphibians in each biogeographic region in the Brazilian

644 Atlantic Forest.

645 Figure 6. Congruence and mismatches between the top of 17% of all scenarios of spatial

646 prioritization for conservation.

647 Supplementary Information

648 Figure 1S. Flowchart of steps followed to process and analyze data inputs, taking into account

649 different biodiversity measures: 1. taxonomic diversity; 2.phylogenetic diversity; and 3.

650 functional diversity. 87 651 Figure 2S. Brazilian Atlantic Forest amphibian anurans phylogeny with P.U.T. shown in red,

652 placed polytomically in their respective MDCC.

653 Table 1S. The True Skill Statistic (TSS) was employed to evaluate the accuracy of ensemble

654 forecasting models for each amphibian species inhabiting in the Brazilian Atlantic Forest. The

655 TSS range was from -1 to +1, where values equal to one (+1) is a perfect prediction and values

656 equal or less than zero (≤0) is a prediction no better than random.

µ Ecological Niche Models (ENMs)

ENMs' Ensemble ENMs' Uncertainty

111

● Least Concern ● Near Threatened ● Vulnerable ● Endangered ● Data Deficient

Ecological Distinctiveness

112

Taxonomic, Functional Taxonomic, Phylogenetic Functional, Phylogenetic All scenarios

0 20 40 60 80 100

113

114

Congruence Mismatches Protected Areas

Ecological Niche Models

5 Ecological Niche Models (ENMs ) Environmental variables GAM,GLM,MARS,RF, and ANN Range maps 10 repetitions 328 amphibian species

Consensual models Habitat filter majority consensus rule

Scenarios

Incomplete 3 3 Ecological traits phylogenetic tree 2 body size, reproductive mode, habit, activity period, habitat, calling site

1 Gower’s distance IUCN Status modification

M.D.C.C. + P.U.T.

Distinctiveness picante

Spatial Prioritization

Table 1S. The TSS was employed to evaluate the accuracy of ensemble forecasting models for each amphibian species inhabit in the Brazilian Atlantic Forest. The TSS range was from -1 to +1, where values equal one (+1) is a perfect prediction and values equal or less than zero (≤0) is a prediction no better than random Species TSS Aplastodiscus callipygius 0.882 Adelophryne pachydactyla 0.8509 Aparasphenodon brunoi 0.9702 Aplastodiscus albofrenatus 0.961 Aplastodiscus cavicolus 0.9558 Aplastodiscus cochranae 0.9237 Aplastodiscus ehrhardti 0.872 Aplastodiscus arildae 0.9795 Aplastodiscus flumineus 0.9128 Aplastodiscus ibirapitanga 0.852 Aplastodiscus sibilatus 0.9522 Aplastodiscus musicus 0.8436 Bokermannohyla caramaschii 0.8976 Aplastodiscus albosignatus 0.915 Bokermannohyla astartea 0.9687 Bokermannohyla carvalhoi 0.97 Bokermannohyla claresignata 0.8808 Bokermannohyla clepsydra 0.9362 Bokermannohyla gouveai 0.8035 Bokermannohyla martinsi 0.9772 Bokermannohyla luctuosa 0.9522 Bokermannohyla circumdata 0.9845 Brachycephalus didactylus 0.9394 Brachycephalus hermogenesi 0.9314 Bokermannohyla hylax 0.9234 Brachycephalus vertebralis 0.9789 Chiasmocleis atlantica 0.9367 Chiasmocleis capixaba 0.9727 Chiasmocleis carvalhoi 0.9018 Chiasmocleis schubarti 0.9663 Crossodactylus caramaschii 0.9665 Crossodactylus grandis 0.8378 Chiasmocleis leucosticta 0.9238 Crossodactylus dispar 0.9566 Crossodactylus schmidti 0.9334 Crossodactylus trachystomus 0.9489 Cycloramphus acangatan 0.9538 Cycloramphus asper 0.9674 Crossodactylus aeneus 0.9512 Cycloramphus brasiliensis 0.9342 Cycloramphus boraceiensis 0.9391 Crossodactylus gaudichaudii 0.9744 Cycloramphus diringshofeni 0.9547 Cycloramphus dubius 0.9747 Cycloramphus duseni 0.936 Cycloramphus eleutherodactylus 0.9782 Cycloramphus granulosus 0.9233 Cycloramphus izecksohni 0.9269 Cycloramphus migueli 0.9736 Cycloramphus ohausi 0.9489 Cycloramphus rhyakonastes 0.954 Cycloramphus semipalmatus 0.9571 Cycloramphus lutzorum 0.9699 Brachycephalus ephippium 0.9686 Allobates olfersioides 0.9671 Cycloramphus valae 0.8401 Dasypops schirchi 0.9657 Dendrophryniscus berthalutzae 0.8708 Dendrophryniscus carvalhoi 0.9185 Cycloramphus fulginosus 0.954 Dendrophryniscus brevipollicatus 0.9162 Dendrophyniscus proboscideus 0.7673 Dendrophryniscus leucomystax 0.9066 Dendropsophus dutrai 0.8945 Aplastodiscus perviridis 0.9627 Dendropsophus giesleri 0.9176 Dendropsophus meridianus 0.9299 Dendropsophus haddadi 0.9657 Dendropsophus nahdereri 0.9011 Dendropsophus oliveirai 0.9803 Dendropsophus pseudomeridianus 0.9181 Dendropsophus ruschii 0.8545 Dendropsophus nanus 0.924 Dendropsophus soaresi 0.8978 Dendropsophus anceps 0.9385 Elachistocleis erythrogaster 0.9674 Euparkerella cochranae 0.9403 Euparkerella brasiliensis 0.9509 Dendropsophus werneri 0.9042 Fritiziana fissilis 0.9401 Fritiziana ohausi 0.9393 Fritiziana goeldi 0.9641 Frostius erythrophthalmus 0.9766 Gastrotheca albolineata 0.9234 Gastrotheca ernestoi 0.9196 Dendropsophus berthalutzae 0.9507 Gastrotheca fissipes 0.9915 Gastrotheca fulvorufa 0.9291 Frostius pernambucensis 0.9466 Dendropsophus bipunctatus 0.9698 Holoaden luederwaldti 0.9663 Dendropsophus seniculus 0.9683 Hylodes babax 0.9911 Hylodes charadranaetes 0.9275 Hylodes heyeri 0.9106 Hylodes magalhaesi 0.977 Hylodes meridionalis 0.952 Hylodes mertensi 0.9666 Dendropsophus branneri 0.99 Hylodes asper 0.9321 Dendropsophus sanborni 0.9926 Hylodes ornatus 0.966 Hylodes lateristrigatus 0.9161 Hylodes perplicatus 0.9425 Hylodes phyllodes 0.9236 Hylodes regius 0.8075 Hylodes sazimai 0.7659 Hylomantis granulosa 0.982 Hylomantis aspera 0.9109 Gastrotheca microdiscus 0.9646 Dendropsophus decipiens 0.9661 Hypsiboas atlanticus 0.9435 Hypsiboas caingua 0.9135 Hypsiboas albopunctatus 0.9161 Hylodes nasus 0.8905 Ceratophrys aurita 0.9743 Hypsiboas joaquini 0.9349 Hypsiboas guentheri 0.9654 Hypsiboas marginatus 0.9127 Hypsiboas leptolineatus 0.9911 Hypsiboas polytaenius 0.9691 Hypsiboas prasinus 0.9563 Dendropsophus elegans 0.9825 Hypsiboas pombali 0.988 Elachistocleis bicolor 0.9944 Hypsiboas secedens 0.7345 Hypsiboas semiguttatus 0.8266 Ischnocnema bilineata 0.9743 Hypsiboas bischoffi 0.9481 Ischnocnema bolbodactyla 0.922 Ischnocnema erythromera 0.9572 Hypsiboas pardalis 0.9744 Ischnocnema gualteri 0.9271 Ischnocnema hoehnei 0.9201 Ischnocnema henselii 0.962 Ischnocnema izecksohni 0.9635 Ischnocnema juipoca 0.9688 Ischnocnema lactea 0.9453 Ischnocnema manezinho 0.979 Hypsiboas pulchellus 0.9941 Ischnocnema octavioii 0.9485 Dendropsophus microps 0.9337 Ischnocnema pusilla 0.9208 Ischnocnema paulodutrai 0.9819 Haddadus binotatus 0.963 Ischnocnema ramagii 0.9717 Ischnocnema spanios 0.8836 Ischnocnema venancioi 0.9783 Ischnocnema verrucosa 0.8658 Ischnocnema parva 0.9846 Ischnocnema vinhai 0.9842 Leptodactylus ajurauna 0.9104 Ischnocnema nasuta 0.9836 Leptodactylus furnarius 0.9733 Leptodactylus flavopictus 0.9672 Leptodactylus plaumanni 0.9906 Ischnocnema guentheri 0.9623 Leptodactylus mystaceus 0.9769 Leptodactylus thomei 0.9816 Leptodactylus bokermanni 0.9825 Leptodactylus viridis 0.9709 Hypsiboas albomarginatus 0.954 Leptodactylus vastus 0.9438 Leptodactylus notoaktites 0.9578 Megaelosia goeldii 0.9454 Lithobates palmipes 0.9436 Limnomedusa macroglossa 0.9753 Hypsiboas semilineatus 0.9836 Leptodactylus troglodytes 0.9582 Leptodactylus marmoratus 0.9099 Megaelosia massarti 0.9793 Melanophryniscus admirabilis 0.9503 Melanophryniscus cambaraensis1 0.9654 Melanophryniscus dorsalis 0.8741 Leptodactylus natalensis 0.9824 Melanophryniscus simplex 0.9098 Melanophryniscus tumifrons 0.9839 Leptodactylus spixii 0.9869 Odontophrynus carvalhoi 0.9764 Myersiella microps 0.9267 Paratelmatobius poecilogaster 0.9349 Phasmahyla exilis 0.9765 Phasmahyla cochranae 0.9864 Phasmahyla spectabilis 0.9415 Phasmahyla guttata 0.9725 Itapotyhila langsdorffii 0.9692 Leptodactylus gracilis 0.9914 Phrynomedusa appendiculata 0.9791 Macrogenioglottus alipioi 0.9724 Phrynomedusa vanzolinii 0.986 Phyllodytes acuminatus 0.9306 Phrynomedusa marginata 0.8246 Phyllodytes edelmoi 0.936 Phyllodytes gyrinaetes 0.7787 Hypsiboas crepitans 0.9638 Phyllodytes kautskyi 0.9782 Phyllodytes maculosus 0.9628 Phyllodytes melanomystax 0.9715 Phyllodytes tuberculosus 0.9873 Phyllodytes wuchereri 0.9873 Phyllomedusa bahiana 0.957 Phyllodytes luteolus 0.9591 Phyllomedusa iheringii 0.9989 Phyllomedusa nordestina 0.9455 Phyllomedusa distincta 0.9365 Phyllomedusa rohdei 0.9346 Hypsiboas faber 0.9104 Phyllomedusa tetraploidea 0.9614 Physalaemus aguirrei 0.9847 Physalaemus caete 0.988 Physalaemus crombiei 0.9592 Physalaemus erikae 0.9665 Physalaemus irroratus 0.9952 Physalaemus jordanensis 0.961 Physalaemus kroyeri 0.9577 Physalaemus lisei 0.9597 Physalaemus maculiventris 0.9536 Physalaemus maximus 0.8896 Physalaemus moreirae 0.9735 Physalaemus nanus1 0.9562 Physalaemus obtectus1 0.8322 Physalaemus spiniger 0.9399 Phyllomedusa burmeisteri 0.9872 Proceratophrys avelinoi 0.982 Proceratophrys bigibbosa 0.9248 Pipa carvalhoi 0.9923 Proceratophrys brauni 0.9715 Proceratophrys cristiceps 0.9271 Proceratophrys melanopogon 0.9768 Proceratophrys moehringi 0.9708 Proceratophrys palustris 0.9162 Physalaemus olfersi 0.941 Physalaemus signifer 0.9475 Proceratophrys laticeps 0.9688 Proceratophrys appendiculata 0.9067 Pseudis bolbodactyla 0.9719 Pseudis cardosoi 0.9774 Pseudis fusca 0.9553 Proceratophrys schirchi 0.9484 Rhinella abei 0.9429 Proceratophrys subguttata 0.945 Rhinella dorbignyi 0.9273 Pseudis minuta 0.9618 Rhinella henseli 0.967 Rhinella hoogmoedi 0.9602 Rhinella pygmaea 0.9332 Scinax agilis 0.9652 Scinax albicans 0.9701 Scinax angrensis 0.8202 Rhinella crucifer 0.9812 Scinax ariadne 0.927 Scinax atratus 0.9731 Rhinella ornata 0.9622 Scinax auratus 0.9541 Scinax brieni 0.9517 Scinax caldarum 0.9878 Scinax cardosoi 0.8563 Scinax carnevallii 0.9892 Scinax crospedospilus1 0.9757 Rhinella fernandezae 0.9183 Scinax catharinae 0.9496 Scinax duartei 0.9842 Scinax heyeri 0.981 Scinax hiemalis 0.9794 Scinax fuscomarginatus 0.9283 Pseudopaludicola falcipes 0.9814 Scinax humilis 0.9018 Scinax littoreus 0.9598 Scinax longilineus 0.9946 Scinax littoralis 0.9414 Scinax luizotavioi 0.9915 Scinax machadoi 0.9682 Scinax melloi 0.941 Scinax pachycrus 0.9787 Scinax obtriangulatus 0.9812 Scinax argyreornatus 0.9454 Scinax hayii 0.9408 Proceratophrys boiei 0.9642 Scinax ranki 0.9648 Scinax flavoguttatus 0.9585 Scinax perpursillus 0.9118 Scinax cuspidatus 0.9227 Scinax eurydice 0.9759 Scinax trapicheiroi 0.9482 Scinax uruguayus 0.9825 Scinax similis 0.911 Scythrophrys sawayae 0.9357 Scinax v-signatus 0.9345 Scinax rizibilis 0.9456 Scinax granulatus 0.9934 Scinax perereca 0.984 Leptodactylus mystacinus 0.937 Sphaenorhynchus caramaschii 0.9063 Sphaenorhynchus orophilus 0.98 Odontophrynus americanus 0.909 Sphaenorhynchus pauloalvini 0.9585 Sphaenorhynchus surdus 0.9436 Sphaenorhynchus palustris 0.9753 Stereocyclops parkeri 0.9327 Thoropa lutzi 0.9493 Thoropa petropolitana 0.9781 Thoropa saxatilis 0.918 Sphaenorhynchus planicola 0.9385 Trachycephalus dibernardoi 0.9533 Stereocyclops incrassatus 0.9675 Sphaenorhynchus prasinus 0.9729 Rhinella icterica 0.9368 Trachycephalus imitatrix 0.9753 Rhinella jimi 0.9356 Xenohyla eugenioi 0.9842 Zachaenus parvulus 0.9736 Xenohyla truncata 0.9158 Scinax alter 0.9296 Thoropa miliaris 0.9712 Trachycephalus nigromaculatus 0.975 Vitreorana eurygnata 0.9363 Scinax fuscovarius 0.9724 Dendropsophus minutus 0.7248 Trachycephalus mesophaeus 0.9696 Vitreorana uranoscopa 0.9545 Scinax squalirostris 0.945 Rhinella schneideri 0.8328 Physalaemus cuvieri 0.9925 Leptodactylus latrans 0.7768 Leptodactylus fuscus 0.9848 Trachycephalus venulosus 0.948 Scinax x-signatus 0.994 Rhinella granulosa 0.9726

Conclusão: Prioridades para a conservação de anfíbios da Mata

Atlântica – uma síntese

501

Prioridades para a conservação de anfíbios da Mata Atlântica – uma síntese

Esta tese discutiu conceitos e abordagens alternativas que devem ser consideradas em estratégias de conservação para anfíbios da Mata Atlântica diante dos efeitos das mudanças climáticas, incluindo informações sobre a diversidade filogenética e funcional para o grupo. Nós apresentamos uma discussão sobre a eficiência da atual rede de reservas e também propusemos uma solução que inclui a distribuição das espécies, no clima atual e em cenários futuros, aliado à baixa incerteza dos modelos. Ainda, comparamos as soluções que retém a maior informação sobre a diversidade taxonômica, filogenética e funcional das comunidades de anfíbios na Mata Atlântica brasileira.

Papel dos modelos de distribuição das espécies

Os modelos de distribuição têm sido amplamente utilizados para o delineamento de áreas protegidas, a restauração ecológica, o manejo de espécies invasoras, além de antecipar os potenciais efeitos das mudanças ambientais globais sobre padrões biogeográficos (Peterson et al., 2011). O uso de modelos de distribuição das espécies é uma ferramenta fundamental, fornecendo uma estimativa da ocupação das espécies em áreas em que o conhecimento está ausente ou escasso

(Elith & Leathwick 2009).

Geralmente, os modelos são aplicados numa abordagem de envelope climático (Peterson et al. 2011), todavia, o efeito das interações bióticas também pode serconsiderado (Araújo & Rozenfeld 2014). É importante ressaltar que compreender como as espécies respondem ao seu ambiente permite prever como as espécies vão responder a ameaças e mudanças ambientais (Guisan & Thuiller

2005). No entanto, essa ferramenta deve ser usada com cautela uma vez que dados detalhados e completos de distribuição das espécies não estão disponíveis

106

(Lemes et al., 2011; Rangel & Loyola 2012). Como resultado, são muitas fontes de incerteza que incluem a origem dos dados, a escolha das varáveis ambientais, o algoritmo usado na modelagem, métodos de validação, modelos climáticos, entre outras (Araújo & New 2007) e a qualidade dos modelos sempre será dependente da qualidade – e quantidade – dos dados.

Ainda, como não é possível comparar um ou mais modelos de maneira definitiva, o consenso entre os modelos tem sido amplamente utilizado e essa abordagem foi aplicada ao longo dessa tese (ver seção Métodos de cada capítulo).

Devido à urgência em delinear estratégias de conservação para inúmeras espécies, o uso da modelagem de distribuição das espécies no planejamento sistemático tornou-se uma prática comum (Elith & Leatwick, 2009). Além dos modelos de distribuição das espécies, outros dados de entrada podem ser incluídos na priorização tais como remanescentes de vegetação, custo da terra e

áreas já protegidas.

O estabelecimento de uma rede de reservas é uma estratégia fundamental para a proteção de espécies (Ladle et al., 2011), embora a eficiência da atual rede de áreas protegidas tem sido questionada (Hannah et al., 2007; Araújo et al.,

2011; Lemes et al., 2014). Por exemplo, diante das alterações globais da temperatura e da precipitação, as espécies podem se deslocar para um espaço climático mais favorável implicando que não ocorra mais em uma área já protegida (Lemes et al., 2014). Nesse contexto, o ideal são planos dinâmicos de conservação, ou seja, estabelecimento de áreas protegidas que considere a distribuição das espécies no clima atual e diante de cenários futuros. Muitos estudos têm incluído o modelo consensual na priorização (Faleiro et al. 2013;

Lemes & Loyola, 2013) e, recentemente, biólogos da conservação exploraram

107

como o consenso de algoritmos de modelagem de distribuição afeta na solução do planejamento sistemático (Meller et al., 2013).

Papel da priorização para a conservação da biodiversidade

São mais de 30 anos desde o primeiro estudo de priorização espacial para a conservação (Kirkpatrick 1983) e um algoritmo simples era capaz de resolver o problema da cobertura mínima (minimum set coverage problem) com base em uma única população. O desenvolvimento inicial dessa técnica tem sido revisado várias vezes, sob a denominação de planejamento sistemático para a conservação

(Margules & Pressey, 2000; Margules & Sarkar, 2007). No início dos anos 90, a priorização espacial para a conservação foi caracterizada pelo desenvolvimento de conceitos fundamentais que constituíram a base dessa ciência (Pressey et al.,

1994). Com um pequeno conjunto de dados e relativa simplicidade, o algoritmo fundamentava-se na complementaridade em suas soluções (Moilanen et al.,

2009).

As análises tornaram-se mais complexas, envolvendo um conjunto maior de espécies e, por sua vez, exigindo algoritmos mais sofisticados. Assim, ao identificar o problema corretamente implica na escolha adequada da estratégia de conservação (Margules & Sarkar 2007). Nesse sentido, para resolver o problema de conservação da maximização da utilidade (utility maximization problem) nós utilizamos o Zonation (ver seção Métodos do capítulo 5 e Lemes &

Loyola, 2013). O algoritmo identifica locais importantes para a retenção de habitats de alta qualidade, com alta conectividade para vários recursos (por exemplo, espécies), estabelecendo uma hierarquização de prioridades de

108

conservação para todas as células, minimizando a perda do valor de conservação

(Moilanen & Kujala 2008).

Aqui, usamos a função benefício aditivo que promove a representação de todas as espécies, favorece locais com alta riqueza de espécies e considera a distribuição proporcional das espécies em uma determinada célula da grade

(Moilanen & Kujala 2008). Ainda, incluímos na priorização espacial a distância que uma espécie deveria deslocar para as mesmas condições climáticas no futuro como uma medida de dispersão entre o clima atual e futuro (Lemes & Loyola,

2013) a fim de estabelecer uma priorização dinâmica. Também, incluímos medidas de diversidade filogenética e funcional uma vez que as prioridades de conservação devem incluir uma visão integrada da biodiversidade para compreender as facetas complementares à diversidade (DeVictor et al., 2010;

Capítulo 5).

Portanto, o conhecimento básico dos aspectos qualitativos e quantitativos do problema de conservação, e as possíveis respostas a ele, são fundamentais para a decisão. O desafio agora reside em harmonizar os problemas reais – como os recursos disponíveis para a conservação – e a representação da biodiversidade.

Nesse sentido, as mais recentes tendências incluem reduzir a incerteza (Meller et al., 2013) e antecipar os efeitos das mudanças climáticas (Lemes & Loyola, 2013).

Conclusões

Não há dúvida quanto à importância das áreas protegidas para proteção da biodiversidade (Ladle et al., 2011), e aproximadamente 12% da superfície da Terra está sob alguma proteção (Loucks et al., 2008). A Mata Atlântica, no entanto, tem

109

7,7% do bioma sob alguma proteção, o que não é suficiente para proteger toda a biodiversidade (Lemes et al., 2014). Historicamente, o estabelecimento das áreas protegidas tem sido embasado apenas na vontade política, no custo econômico ou na beleza cênica (Ladle et al., 2011) o que não contribui idealmente para a representação da biodiversidade (embora obviamente cumpram um papel importante no contexto nacional). Poucas áreas protegidas da atual rede de reservas são capazes de abrigar mais espécies do que o esperado ao acaso (Lemes et al., 2014). Uma abordagem sistemática para o planejamento da conservação fornece a base necessária para atingir tal objetivo (Margules & Sarkar, 2007).

Assim, apresentamos um cenário sob o ponto de vista da Biogeografia

Conservação que incluiu as mudanças na distribuição das espécies diante de mudanças climáticas e forneceu locais com alta representação e baixa incerteza

(Lemes & Loyola, 2013). Nossos cenários são parte de um esforço amplo para fortalecer a base científica das decisões de conservação, sabendo que é passível a extrapolação para outras regiões e outros grupos. Nossas análises podem ser úteis para os planejadores de conservação uma vez que incluem o potencial efeito das mudanças climáticas e aspectos da diversidade tais como a história evolutiva e o funcionamento do ecossistema. A identificação de um conjunto abrangente de

áreas naturais é apenas o primeiro passo para uma estratégia de conservação da biodiversidade in situ (Capítulo 5) e requer um processo mais complexo desde a negociação das terras até a efetiva implementação das novas áreas protegidas

(Knight et al., 2009).

Contudo, espera-se que a abordagem descrita ao longo dessa tese possa ajudar tomadores de decisão embasar políticas de conservação. Muito embora, as ferramentas apresentadas aqui são de apoio à decisão, e não de tomada de

110

decisão. Os pressupostos matemáticos e ecológicos devem ser entendidos e os métodos aplicados criticamente. Portanto, as estratégias de conservação devem abalizar a proteção da biodiversidade, mas considerando os interesses de todas as partes interessadas.

Referências

Araújo, M.B. & New, M. 2007. Ensemble forecasting of species distributions. Trends Ecol. Evol. 22, 42-47

Araújo, M.B., Alagador, D., Cabeza, M., Nogués-Bravo, D. & Thuiller, W. 2011. Climate change threatens European conservation areas. Ecol. Lett. 14, 484-492.

Araújo, M.B. & Rozenfeld, A. 2014, The geographic scaling of biotic interactions, Ecography 37, 1-10.

Devictor, V. Mouillot, D., Meynard, C., Jiguet, F., Thuiller, W. & Mouquet, N. 2010. Spatial mismatch and congruence between taxonomic, phylogenetic and functional diversity: the need for integrative conservation strategies in a changing world. Ecol. Lett. 13, 1030-1040.

Diniz-Filho, J.A.F., Bini, L.M., Rangel, T.F.L.V.B., Loyola, R.D., Hof, C. & Araújo, M.B. 2009. Partitioning and mapping uncertainties in ensembles of forecasts of species turnover under climate change. Ecography 32, 897-906.

Elith, J. & Leathwick, J. 2009. The Contribution of Species Distribution Modelling to Conservation Prioritization. In: Spatial Conservation Prioritization: Quantitative Methods and Computational Tools. Moilanen, A., Wilson, K.A. & Possingham, H.P.(eds), pp. 70-92, Oxford University Press, Oxford.

Faleiro, F.V., Machado, R.B., Loyola, R.D., 2013. Defining spatial conservation priorities in the face of land-use and climate change. Biological Conservation 158, 248-257.

Guisan, A. & Thuiller, W. 2005. Predictng Species Distributon: Offering More than Simple Habitat Models. Ecol. Lett. 8, 993-1009.

111

Hannah, L., Midgley, G., Andelman, S., Araujo, M.B., Hughes, G., Martinez- Meyer, E., Pearson, R. & Williams, P. 2007. Protected area needs in a changing climate. Front. Ecol. Environ. 5, 131-138.

Kirkpatrick, J.B. 1983. An Iterative Method for Establishing Priorities for the Selection of Nature Reserves – An Example from Tasmania. Biol. Conserv. 25, 127-134.

Knight, A., Cowling, R.M., Possinghma, H.P., Wilson, K.A. 2009. From Theory to Practice: Designing and Situating Spatial Prioritization Approaches to Better Implement Conservation Action. In: Spatial Conservation Prioritization: Quantitative Methods and Computational Tools. Moilanen, A., Wilson, K.A. & Possingham, H.P. (eds), pp. 249-258, Oxford University Press, Oxford.

Ladle, R.J., Jepson, P. & Gillson, L. 2011. Social values and conservation biogeography. In: Conservation Biogeography, Ladle, R.J., Whittaker, R.J (eds).Wiley-Blackwell. 301p

Lemes, P. & Loyola, R.D., 2013. Accommodating species climate-forced dispersal and uncertainties in spatial conservation planning. PloS ONE 8, e54323.

Lemes, P., Faleiro, F. A. M. V., Tessarolo, G. & Loyola, R.D. 2011. Refinando Dados Espaciais para a Conservação da Biodiversidade. Nat. Con. 9, 240-243.

Lemes, P., Melo, A.S., Loyola, R.D., 2014. Climate change threatens protected areas of the Atlantic Forest. Biod. Conserv. 23, 357-368.

Loucks, C., Ricketts, T.H., Naidoo, R., Lamoreux, J. & Hoekstra, J. 2008. Explaining the global pattern of protected area coverage, relative importance of vertebrate biodiversity, human activities and agricultural suitability. J Biogeogr 35, 1337-1348.

Margules, C.R. & Pressey, R.L., 2000. Systematic conservation planning. Nature 405, 243-253.

Margules, C.R. & Sarkar, S. 2007. Systematic Conservation Planning. Cambridge: Cambridge Unversity Press. 269p.

Meller, L., Cabeza, M., Pironon, S. Barbet-Massin, M., Maiorano, L., Georges, D., Thuiller, W. 2013. Ensemble distribution models in conservation prioritization:

112

from consensus predictions to consensus reserve networks. Div. Distrib. 20, 309- 321.

Moilanen, A., Possingham, H.P. & Polasky, S. 2009. A Mathematical Classification of Conservation Prioritization Problems. In: Spatial Conservation Prioritization: Quantitative Methods and Computational Tools. Moilanen, A., Wilson, K.A. & Possingham, H.P. (eds), pp. 249-258, Oxford University Press, Oxford.

Moilanen, A., Meller, L., Leppanen, J., Pouzols, F.M., Arponen, A., Kujala, H., 2012. Zonation Spatial Conservation Planning Framework and Software V3.1, User Manual.

Peterson, A.T., Soberón, J., Pearson, R.G., Anderson, R.P., Martínez-Meyer, E., Nakamura, M., Araújo, M.B., 2011. In: Horn H.S., Levin S.A. (eds) Ecological niches and geographic distributions. Princeton University Press, Princeton.

Pressey, R.L.; Bedward, M. & Keith, D.A. 1994. New Procedures for Reserve Selection in New South Wals: Maximising the Chances of Achieving a Representative Network. In: Systematics and Conservation Evaluation. Forey, P.L. Humphries, C.J. & Vane- Wright R.I. (eds). Oxford: Clarendon Press, pp. 351-373.

Rangel, T.F., & Loyola, R.D., 2012. Labeling ecological niche models. Nat. Conserv., 10, 119-126.

113