UNIVERSIDADE ESTADUAL DE CAMPINAS INSTITUTO DE BIOLOGIA
LUDMILA MARIA RATTIS TEIXEIRA
ON THE HABITAT AND LANDSCAPES: THE EFFECT OF SCALE AND DATA ON SPECIES RESPONSE
SOBRE O HÁBITAT E A PAISAGEM: EFEITOS DA ESCALA E DOS DADOS SOBRE A RESPOSTA DAS ESPÉCIES
CAMPINAS
2016
LUDMILA MARIA RATTIS TEIXEIRA
ON THE HABITAT AND LANDSCAPES: THE EFFECT OF SCALE AND DATA ON SPECIES RESPONSE
SOBRE O HÁBITAT E A PAISAGEM: EFEITOS DA ESCALA E DOS DADOS SOBRE A RESPOSTA DAS ESPÉCIES
Thesis presented to the Institute of Biology of the University of Campinas in partial fulfillment of the requirements for the degree of Doctor in Ecology
Tese apresentada ao Instituto de Biologia da Universidade Estadual de Campinas como parte dos requisitos exigidos para a obtenção do Título de Doutora em Ecologia.
ESTE ARQUIVO DIGITAL CORRESPONDE À VERSÃO FINAL DA TESE DEFENDIDA PELA LUDMILA MARIA RATTIS TEIXEIRA E ORIENTADA PELO PROF. DR. RAFAEL DIAS LOYOLA.
Orientador: RAFAEL DIAS LOYOLA
CAMPINAS
2016
Campinas, 16 de junho de 2016
COMISSÃO EXAMINADORA
Prof. Dr. Rafael Dias Loyola
Prof. Dr. Carlos Eduardo Viveiro Grelle
Prof. Dr Paulo De Marco Junior
Profa. Dra Eleonore Zulnara Freire Setz
Prof. Dr. Leonardo Ré Jorge
Os membros da Comissão Examinadora acima assinaram a Ata de defesa, que se encontra no processo de vida acadêmica do aluno.
À Danielle, Nubia, Vanessa e Genevieve e a tudo que elas me ensinaram...
Bate forte até sangrar a mão os tambores de Minas soarão seus tambores nunca se calarão... (Márcio Borges, mineiro)
AGRADECIMENTOS
Agradeço... ao Ensino Público Brasileiro, do qual desfrutei durante todo o ensino fundamental, parte do ensino médio e toda a Pós-Graduação;
Ao Grupo Tal Pai Tal Filho, à Coordenação de Aperfeiçoamento Profissional do Ensino Superior e à Fundação de Amparo à Pesquisa do Estado de São Paulo pelo apoio financeiro;
À Universidade Estadual de Campinas (UNICAMP), Universidade Federal de Goiás (UFG), Universidade Federal da Bahia (UFBA) e Carleton University, todos seus professores e funcionários;
Ao Rafael Loyola por confiar em mim. O Rafa sabe que a Academia é formada por seres humanos, com problemas e idiossincrasias, e respeita cada um de seus colegas. Ele quem me encorajou a fazer os cursos de Frugivoria e Dispersão de Sementes e Análises Espaciais em Macroecologia, que foram tão importantes na minha formação como ecóloga. À Lenore Fahrig que me ensinou a remar, cortar lenha, acampar, acender fogueira, fazer bolo de chocolate na casca de laranja, assar marshmallows, testar hipóteses e escrever artigos científicos. No dia que entramos no barco para ela me ensinar a remar no Grant Lake, ela disse: eu direciono e você dá a força, nosso objetivo é alcançar aquela pedra, na outra margem... Obedeci e fizemos dois capítulos juntas durante o ano que estive na Carleton University.
Aos membros da pré-banca: Sidney F. Gouveia, Marcus Cianciaruso e Divino Silvério pelas discussões e comentários. Aos membros da Banca Examinadora por aceitarem o convite e contribuírem para o crescimento deste trabalho. Ao Pedro Jordano por sua competência, elegância e simplicidade a ponto de “fazer ciência e soltar pipa” o que me faz lembrar outra pessoa a quem tenho muito a agradecer: meu avô científico Thomas Lewinsohn, o qual sempre me confundiu... Nunca soube dividir em quais momentos ele estava soltando pipa ou fazendo ciência, tamanha sua satisfação em lecionar Ecologia de Comunidades e discutir ciência nas horas que nunca são vagas quando se está ao seu lado.
Ao Mauro Galetti pelas boas ideias durante o curso de Frugivoria e Dispersão de Sementes e às demais discussões que se seguiram. Ao Sérgio Furtado dos Reis pela compreensão e cordialidade.
Ao Paulo de Marco que, no início de 2012, quando cheguei a Goiânia, ao mesmo tempo que o belo Otávio veio ao mundo, discutiu comigo meu trabalho, disse-me que, para ele, aluno de doutorado deveria ser tratado como um colega e, quando resolvi mudar-me para Goiânia me deu um forte abraço acolhedor. Valeu Paulo!
À Sandra Martins, Ivan Correto, Poliana Mendes, Benjamin e Genevieve Perkins por terem cuidado de mim após meu acidente no Canadá. À Genevieve de novo, de novo e de novo. A Gen me ensinou a falar inglês, estudar estatística, tomar café e comer Vegemite, o que significa que devo tudo a ela.
Ao Mafalda (Luis Francisco Mello Coelho) que sem dúvida é o responsável por minha entrada na Academia.
Ao Leonardo Ré Jorge e Janaina Cortinóz por discutirem ciência e vida e me ajudarem nos trabalhos das disciplinas na Unicamp.
Aos funcionários e professores de Carleton U e aos companheiros do Geomatics Landscape Ecology Lab pelas discussões e ajudas, em especial à Margareth, Fernanda Zimmerman, Sandra Martins, Sara Collins, Igor Pfeifer e Koreen Millard. Aos companheiros de UFG, do Laboratório de Biogeografia da Conservação (CB-Lab), Laboratório de Comunidades, LETS e The MetaLand, em especial à Fernanda Brum, Raísa Vieira, Lívia Laureto, Fernanda Fava, Leandro Maracahipes, Fernando Sobral, Marcos Carlucci, Ricardo Dobrovolski, Sidney Gouveia, Fabricio Villalobos, Diogo Samia, Luciano Sgarbi, Bruno Vilela, Miguel Ollala-Tárraga, Kelly Souza, Marcelo Weber, Welma, Daniel Paiva e Cristina Araújo. Cada uma dessas pessoas sabe o quanto me ajudou nesses quatro anos e agradeço a cada um por fazer parte da tese. À Capoeira Angola e a tudo que ela me deu pelas mãos dos Mestres Leninho e Guaxini do Mar.
Ao pessoal do IPAM, em especial ao Paulo Brando pelas discussões, comentários e grande incentivo na reta final. Aos amigos com quem tive o prazer de dividir o grande teto dos alunos do programa de Ecologia da UFG, o edifício Barbosa Lima, com pessoas incríveis. Dentre eles, Tia Ísis, Eduardo Pacífico, Caroline Nóbrega, Sidney Gouveia, Fabricio Villalobos, Ricardo Dobrovolski, Diogo Provete, Bruno Vilela, Fábio Carvalho, Vanessa Staggemeier, Alexander Florez, Leandro Maracahipes, Thiago Bernardi, Poliana Mendes, Daniel Paiva e Karina Silva.
Tenho que agradecer especialmente à Be e à Beubinha. Compartilhamos a essência, somos diferentes em todo o resto. Obrigada por dividirem comigo o café, vinho, o dramin e a co-autoria da filosofia “apamerrda”. Adoro discutir ciência com vocês! É sempre muito frutífero!
Aos meus brilhantes amigos Gabriela Venturini e Vinicius Cardoso, pelo exemplo. À Danielle Miranda por discutir comigo o que realmente importa desde os tempos da graduação em Minas Gerais. À minha família, por ser tudo na minha vida. Eu sobrevivi e devo cada superação às orações dos meus avós Luzia e Oswaldo, à memória da vovó Geralda, à culinária do Rogerinho e à ajuda dos cunhados.
Ao apoio incondicional do Paulinho Teixeira e da Fatinha Rattis, de seus filhos lindos e às filhas de seus filhos... Obrigada aos Teixeira por trabalharem tanto. Obrigada aos Rattis por acreditarem no impossível.
ABSTRACT
The environment of a given species is always changing and studying the causes and consequences of these changes is a challenging task in Ecology. This is because the very definition of suitable habitat is species-specific and even local-specific. Besides that, some factors can affect the results and conclusions coming from the studies carried out in landscapes. The habitat surrogate, the measurement of its effect on biodiversity, and the spatial extent of the investigation could change according to data availability and the purpose of research. Here we tested some hypothesis about the effect of habitat amount, patch size and isolation on species richness; spatial extent and species traits affecting the relationship between landscape structure and biodiversity and proposed a method to evaluate carrying capacity and connectivity on species’ range scale: (1) is habitat amount in the surrounding landscape as strong a predictor of species richness, as the combined predictive strength of patch isolation and patch size? (Chapter 1); (2) how does the scale of effect change with species traits and the measured population response? (Chapter 2); we also proposed a method to evaluate connectivity and carrying capacity at species range scale (Chapter 3). In chapter 1, our results supported the habitat amount hypothesis. Patch size and isolation did not have effects on species richness independent of habitat amount; they can be replaced by habitat amount alone. In chapter 2, we found that the spatial extent at which landscape structure mostly affect plant community depends on the movement range of the species, reproduction rate and the particular response analyzed. Species that disperse farther and produce more seeds showed a larger scale of effect, and species abundance responded to the surrounding landscape at a smaller scale than species occurrence. In Chapter 3, we highlighted that habitat amount and carrying capacity were more dependent on species’ features, whereas fragmentation and functional connectivity were more scale dependent, except when the species has high dispersal ability. The proposed approach can lead to lower risks of incurring in commission errors arising from landscape-scale underestimation of species’ occurrences and provide a new approach to assess species-habitat relationship. We also highlighted the importance of considering range-scaled landscape, landscape carrying capacity, and patch functional connectivity in a synthetic and integrated framework for conservation studies.
Keywords: carrying capacity; deductive habitat suitability modeling; functional connectivity; habitat amount; herbaceous plant species; landscape ecology; scale of effect.
RESUMO
O ambiente das espécies está em constante mudança. Estudar as causas e consequências dessa mudança é uma das tarefas mais desafiadoras em Ecologia. Isto porque a definição de hábitat adequado é espécie-específica, podendo ser diferente até mesmo para uma mesma espécie que ocorre em diferentes locais. Além disso, alguns fatores podem afetar os resultados e conclusões adquiridas por meio de estudos sobre a mudança de hábitat na escala da paisagem. O conceito de hábitat, a avaliação do seu efeito sobre a biodiversidade e a extensão espacial do estudo podem mudar segundo a disponibilidade de dados e os objetivos almejados. Nesta tese, testamos a predição de algumas hipóteses sobre o efeito da quantidade de hábitat, tamanho e isolamento das manchas na riqueza de espécies, sobre a extensão espacial e características das espécies que afetam a relação entre a estrutura da paisagem e a biodiversidade e dessa relação na escala da distribuição da espécie. Respondemos à duas perguntas e propomos um método que versam sobre as mudanças de hábitat e a resposta das espécies à essas mudanças, em diferentes escalas: (1) A quantidade de hábitat na paisagem local explica tanto da variação na riqueza de espécies, quanto o efeito combinado de área e isolamento das manchas de hábitat? (Cap. 1); (2) como a extensão espacial em que a paisagem afeta as espécies muda devido às características das espécies e a resposta biológica medida? (Cap. 2); também propusemos um método para avaliar a conectividade e capacidade de suporte das manchas na escala da área de distribuição das espécies (Cap. 3). No capítulo 1, os resultados corroboram a hipótese da quantidade de hábitat. Tamanho da mancha e isolamento não tiveram efeitos sobre a riqueza de espécies independentemente da quantidade de habitat; podendo ser substituídos com uma única variável, que é a própria quantidade habitat. No capítulo 2, descobrimos que a extensão espacial em que a estrutura da paisagem afeta as comunidades vegetais depende da capacidade de movimentação das espécies, suas taxas de reprodução e da resposta biológica utilizada no modelo. Espécies que dispersam e se reproduzem mais tendem a ser afetadas pela estrutura da paisagem em uma extensão espacial maior. Além disso, dados de abundância das espécies respondem à paisagem circundante em uma escala menor do que os dados de ocorrência. No Capítulo 3, mostramos que as medidas de quantidade de habitat e da capacidade de suporte são mais dependentes das características intrínsecas das espécies, enquanto que as análises de fragmentação e conectividade funcional são mais influenciadas pela escala, exceto quando a espécie tem alta capacidade de dispersão. A abordagem que propomos no capítulo 3 pode levar a menores riscos de incorrer em erros de comissão, que poderiam resultar de uma subestimativa da escala em que a paisagem afeta a potencial ocorrência das espécies. Além disso, propomos uma nova abordagem para avaliar a relação espécie-habitat. Destaca-se também a importância de se considerar toda a área de distribuição da espécie em análises de capacidade de suporte e conectividade funcional dos fragmentos, em uma estrutura simples e integrada para estudos de conservação. Palavras-chave: capacidade suporte; conectividade funcional; ecologia de paisagens; escala de efeito; modelagem dedutiva de adequabilidade de habitat; quantidade de hábitat; plantas herbáceas.
SUMÁRIO ABSTRACT ...... 9 RESUMO...... 10 Introdução geral ...... 13 REFERÊNCIAS ...... 14 No effect of patch size and isolation after removing habitat amount component ...... 16 ABSTRACT ...... 17 INTRODUCTION ...... 18 METHODS ...... 19 RESULTS ...... 26 DISCUSSION ...... 28 ACKNOWLEDGEMENTS...... 32 REFERENCES ...... 33 Factors determining the spatial extent in which species respond to landscape structure ...... 35 ABSTRACT ...... 36 INTRODUCTION ...... 38 METHODS ...... 42 RESULTS ...... 48 DISCUSSION ...... 51 ACKNOWLEDGEMENTS...... 55 REFERENCES ...... 55 A method to evaluate habitat connectivity and carrying capacity at the scale of species' range ...... 59 ABSTRACT ...... 60 INTRODUCTION ...... 61 METHODS ...... 63 RESULTS ...... 69 DISCUSSION ...... 73 ACKNOWLEDGEMENTS...... 76 REFERENCES ...... 77 Conclusão geral ...... 83 REFERÊNCIAS ...... 87
13
Introdução geral
Uma espécie é influenciada diretamente pela distribuição e pela qualidade das áreas adequadas para a sobrevivência e reprodução dos indivíduos, áreas estas conhecidas como hábitat. O hábitat deixa de sê-lo, integral ou parcialmente, à medida que influências internas e externas o modificam. Diante dessa situação, os indivíduos podem adaptar-se, dispersar-se ou morrer; a espécie pode migrar, colonizar novas áreas ou simplesmente extinguir. À medida que organismos reagem às mudanças no hábitat, os destinos de sua espécie e de outras que com ele interagem serão modificados. Podemos avaliar, quantificar e tomar decisões a respeito das mudanças de hábitat. Uma boa avaliação da relação entre as distribuições das espécies e seus hábitats dependerá de três fatores: a escolha da escala em que a relação de interesse é avaliada (Thornton & Fletcher, 2014), os fatores que serão considerados de ambas as partes – componentes de hábitat e de biodiversidade – e como a relação entre esses fatores serão avaliados (Dunning et al., 1992). Se a avaliação é legítima, o tamanho, a direção e a precisão podem ser quantificados e sintetizados à luz do conhecimento já obtido. Fatores intrínsecos e extrínsecos às espécies determinam como elas respondem ao ambiente e às mudanças que nele ocorrem. Dentre as características intrínsecas estão a posição na cadeia trófica, a densidade populacional, o tamanho da área de distribuição da espécie, o tempo e a área de vida, o tamanho do corpo e a amplitude de exploração de recursos (e.g. espécies generalistas e especialistas) (Bender et al., 1998; Purvis et al., 2000; Thuiller et al., 2004). Dentre os fatores extrínsecos, estão as ameaças antropogênicas que explicam boa parte da variação do total do risco de extinção das espécies (Purvis et al., 2000). Dentre as ameaças antropogênicas, Diamond (1989) definiu as quatro principais fontes de risco como a sobre- exploração de recursos naturais, a introdução de espécies exóticas, as extinções em cadeia e a perda de hábitat. Brook et al. (2008) adiciona à essa lista as mudanças climáticas que, em sinergismo com o “quarteto demoníaco” proposto por Diamond (1989), podem acelerar as já existentes ameaças à biodiversidade. A supressão de áreas naturais pode gerar, além da perda, a fragmentação de hábitats antes contínuos. Desde a publicação da teoria de Biogeografia de Ilhas (MacArthur & Wilson 1967), e sua posterior aplicação em áreas fragmentadas, um número crescente de pesquisas tenta elucidar os principais efeitos da perda de hábitat e da fragmentação em comunidades biológicas (Fahrig, 2003; Prugh et al., 2008; Pardini et al., 2009). Contudo, a aplicação do equilíbrio 14
dinâmico de ilhas à áreas continentais, em que fragmentos de hábitat diferem em tamanho, grau de isolamento e matriz circundante, é questionada (Bender et al., 1998; Mendenhall et al., 2014; Quinn et al., 2016). Inclusive propõem-se que um arcabouço teórico seja construído para áreas continentais com base, porém paralelamente, à teoria do equilíbrio dinâmico de ilhas (Mendenhall et al., 2014). Para que esse arcabouço seja construído é necessária maior compreensão dos processos que afetam a diversidade de espécies em fragmentos continentais e de como o delineamento experimental afeta nossa compreensão de tais processos. As varáveis empregadas nos modelos, a escala do estudo, o objetivo da investigação e as características intrínsecas das espécies estudadas afetam tais respostas (Thornton & Fletcher, 2014). Precisamos, portanto, quantificar tais diferenças para compreender a importância relativa dos dados e das escalas empregadas nos modelos dependendo do objetivo do estudo. Nesta tese respondemos à duas perguntas e propomos um método que versam sobre as mudanças de hábitat e a resposta das espécies a essas mudanças, em diferentes escalas. No capítulo 1, a pergunta central é sobre as variáveis preditoras que usamos em modelos de riqueza de espécies: quantidade de hábitat é tão boa quanto o efeito combinado de tamanho e isolamento dos fragmentos para predizer riqueza e abundância das espécies em paisagens locais? No capítulo 2, estudamos como a escala de efeito muda de acordo com as características de mobilidade, demografia e resposta biológica das espécies estudadas. No terceiro e último capítulo propomos um método de estudo do efeito da estrutura do hábitat sobre a ocorrência potencial das espécies e comparamos os resultados obtidos em escala local e na escala da distribuição da espécie para responder a seguinte questão: Quanto da área de distribuição da espécie é potencialmente adequada e funcionalmente conectada e como os resultados obtidos mudam com a escala estudada?
REFERÊNCIAS
Bender D.J., Contreras T.A., & Fahrig L. (1998) Habitat loss and population decline: a meta- analysis of the patch size effect. Ecology, 79, 517–533. Brook B.W., Sodhi N.S., & Bradshaw C.J.A. (2008) Synergies among extinction drivers under global change. Trends in Ecology & Evolution, 23, 453–60. Diamond J.M. (1989) Overview of recent extinctions. Conservation for the Twenty-First Century (ed. by D. Western and M.C. Pearl), pp. 37–41. Oxford University Press, Dunning J.B., Danielson B.J., Pulliam H.R., & Oct N. (1992) Ecological Processes That Affect 15
Populations in Complex Landscapes. Oikos, 65, 169–175. Fahrig L. (2003) Effects of habitat fragmentation on biodiversity. Annual Review of Ecology, Evolution, and Systematics, 34, 487–515. MacArthur, R H Wilson E.O. (1967) The Theory of Island Biogeography. Princeton University Press., Mendenhall C.D., Karp D.S., Meyer C.F.J., Hadly E.A., & Daily G.C. (2014) Predicting biodiversity change and averting collapse in agricultural landscapes. Nature, 509, 213– 217. Pardini R., Faria D., Accacio G.M., Laps R.R., Mariano-Neto E., Paciencia M.L.B., Dixo M., & Baumgarten J. (2009) The challenge of maintaining Atlantic forest biodiversity: A multi-taxa conservation assessment of specialist and generalist species in an agro-forestry mosaic in southern Bahia. Biological Conservation, 142, 1178–1190. Prugh L.R., Hodges K.E., Sinclair A.R.E., & Brashares J.S. (2008) Effect of habitat area and isolation on fragmented animal populations. Proceedings of the National Academy of Sciences of the United States of America, 105, 20770–5. Purvis A., Gittleman J.L., Cowlishaw G., & Mace G.M. (2000) Predicting extinction risk in declining species. Proceedings of the Royal Society B: Biological Sciences, 267, 1947– 1952. Quinn J.F., Harrison S.P., Quinn J.F., & Harrison S.P. (2016) International Association for Ecology Effects of Habitat Fragmentation and Isolation on Species Richness : Evidence from Biogeographic Patterns Published by : Springer in cooperation with International Association for Ecology Stable URL : http://www.jstor. 75, 132–140. Thornton D.H. & Fletcher R.J. (2014) Body size and spatial scales in avian response to landscapes: A meta-analysis. Ecography, 37, 454–463. Thuiller W., Araújo M.B., Pearson R.G., Whittaker R.J., Brotons L., & Lavorel S. (2004) Biodiversity conservation: Uncertainty in predictions of extinction risk. Nature, 430, 145– 148.
16
No effect of patch size and isolation on species richness after removing habitat
amount component
Ludmila Rattis1,2*, Rafael Loyola2 & Lenore Fahrig3
Authors’ affiliations: 1*Programa de Pós-Graduação em Ecologia, Universidade Estadual de Campinas, CP6109, 13083-970 Campinas, São Paulo, Brazil. Email: [email protected] 2Laboratório de Biogeografia da Conservação, Departamento de Ecologia, Universidade Federal de Goiás, CP 131, 74001-970 Goiânia, Goiás, Brazil. 3Geomatics and Landscape Ecology Research Laboratory (GLEL), Department of Biology, Carleton University, Ottawa, ON, K1S 5B6, Canada
17
ABSTRACT
The habitat amount hypothesis predicts that the effects of both patch size and isolation on species richness at a site can be replaced by the effect of a single variable, habitat amount in the local landscape surrounding that site. We tested the habitat amount hypothesis using herbaceous plant species data from Ontario, Canada. We gathered species occurrence data and data on forest cover from 189 sites belonged to 29 landscapes. If the habitat amount hypothesis is correct, the effect of forest amount in the local landscape surrounding a forest plant plot encapsulates both the effects of local forest patch size and the effects of local forest patch isolation. We tested six predictions: A and B) forest amount in the local landscape should have a significant positive effect on plant species richness at a sample site when the local patch area and local nearest neighbor patch distance are controlled. Conversely, C) local patch area and/or
D) nearest neighbor distance should have no effect on plant species richness at a site when the effect of forest amount in the local landscape is controlled. Plant species richness at a site should increase with E) increasing forest amount in the local landscape even if the local patch size decreases or F) nearest neighbor distance increases. Our results support predictions A to E.
What some people call spatial configuration effects – patch area and isolation effects – is essentially the habitat amount effect, driven mainly by a single underlying process, the sample area effect. Patch area and patch isolation can be replaced by habitat amount to predict species richness, because they are essentially habitat amount effect.
18
INTRODUCTION
The 'habitat patch' is a common spatial unit for measuring species richness.
Richness in a habitat patch increases with patch size (Almeida-Gomes et al., 2016) and decreases with patch isolation (Petit et al., 2004). That species richness in a patch increases with patch size is not surprising because a larger patch will contain more individuals (all species combined), and it has long been known that the number of species in a given area is a positive function of the number of individuals there (Preston, 1962; Connor & McCoy, 1979, 2001; reviewed in Chapters 2 and 3 in Hubbell, 2001). It is also true, though less obvious, that species richness should decrease with patch isolation. Patch isolation is a function of the amount of habitat in the surrounding landscape (Tischendorf & Fahrig, 2000; Bender et al., 2003; Saura et al., 2014). If a patch is less isolated, the landscape surrounding it contains more habitat than the landscape around a more isolated patch.
The landscape surrounding the less isolated patch will therefore contain more individuals (as there is more habitat) and therefore more species due to the individuals-species relationship. The species-individual relationship is the species richness increase as a function of the number of individuals. Larger areas probably receive more immigrants, that in turn can represent a wider range of species, increasing species richness (for a synthesis: Connor &
McCoy, 2001).
One of the proposed mechanisms behind the ubiquitous species-area relationship is the passive sample hypothesis, proposed by Simberlof (1976). This hypothesis predicts that…
Here we argue that patch size effect and the patch isolation effect are both at least partly driven by the individuals-richness relationship. Both are functions of habitat amount.
More habitat means more individuals, which means more species.
These ideas form the basic arguments leading to the habitat amount hypothesis
(Fahrig, 2013). The habitat amount hypothesis proposes that the effects of both patch size and 19
isolation on species richness at a site are due only or primarily to the relationships between (i) habitat amount and number of individuals and (ii) number of individuals and species richness.
The habitat amount hypothesis predicts that the effects of both patch size and isolation on species richness at a site can be replaced by the effect of a single variable, habitat amount in the local landscape surrounding that site.
We tested the habitat amount hypothesis using data on shade-tolerant herbaceous plant species from Ontario, Canada. The species richness data are from 189 forested sites, and habitat amount is the amount of forest surrounding each site. If the habitat amount hypothesis is correct, the effect of forest amount in the local landscape surrounding a forest plant plot encapsulates both the effects of local forest patch size and the effects of local forest patch isolation. In this case we predict that forest amount in the local landscape should have a significant positive effect on plant species richness at a plot when either the local patch area or the nearest neighbor distance (patch isolation) is controlled (predictions A and B, Table 1).
Conversely, local patch area and/or nearest neighbor distance should have no effect on plant species richness at a plot when the effect of forest amount in the local landscape is controlled
(predictions C and D, Table 1). Finally, plant species richness at a plot should increase with increasing forest amount in the local landscape even if the local patch size decreases or nearest neighbor distance increases with increasing forest amount in the local landscape (predictions E and F, Table 1). Note, hereafter, we shorted ‘forest amount in the local landscape’ to simply
‘forest amount’.
METHODS
Overview
We estimated the effects of forest amount, local forest patch area and forest patch nearest neighbor distance on species richness of shade-tolerant herbaceous plants in 189 forest 20
plots (two 1m2 quadrats per plot) in Eastern Ontario, Canada. More information about the initial selection of plots is provided below.
To estimate the independent effects of forest amount, patch area and patch isolation we created different sub-sets of the 189 plots, that maximized variation in the predictor of interest while minimizing variation in the other predictors. For prediction A, we selected plots with varying forest amount and constant patch area. For prediction B, we selected plots with varying forest amount and constant patch isolation. For prediction C, we selected plots with varying patch area and constant forest amount. For prediction D, we selected plots with varying patch isolation and constant forest amount. For prediction E, we selected plots such that patch area decreased with increasing forest amount. For prediction F we selected plots such that patch isolation increased with increasing forest amount (Table 1).
Table 1. The predictions tested in this study. Each prediction was tested using a specific sub- set of the data and all of them derive from the main prediction, that habitat amount can replace patch area and isolation in predicting species richness and abundance. In the third column, the circles represent the local landscapes, the green blocks represent forest cover, and the red squares in the center of each landscape are the plots in which species richness of shade-tolerant herbaceous plants was sampled. The red arrows indicate the nearest neighbor distances from the focal patch. Gray polygons indicate that they were kept constant. Schematic of data subset used to test the Prediction prediction
Habitat amount should have a A significant effect on species richness when the local patch area is controlled.
Habitat amount should have a significant effect on species richness B when the local patch nearest neighbor
distance is controlled. 21
Local patch area should have no effect C on species richness when the habitat amount is controlled
Local patch nearest neighbor distance should have no effect on species D richness when the habitat amount is
controlled. Species richness should increase with E increasing habitat amount even if the local patch area decreases.
Species richness should increase with increasing habitat amount even if the F local patch nearest neighbor distance
increases.
Study plot and sample design
Plots were sampled in a forest within approximately 60 km of Ottawa, Ontario,
Canada between 1999 and 2001 (Figure 1). The dominant tree species include Acer saccharum,
Tilia americana, Fraxinus sp. and Pinus sp. The forest is primarily second growth and the original forest was cut between 160 and 90 years ago (Brommit et al., 2004a).
The original data set includes 290 vegetation plots (two 1-m2 quadrats per plot), clustered with 10 plots in each of 29 1km2 areas. Of these we selected the 189 forested plots
(Figure 1). Details of plot selection and sampling are in Brommit et al., (2004). 22
Figure 1: The study area in Southeast Ontario, Canada (B) and the sample design (A). Shade- tolerant herbaceous plants were sampled in 189 forest plots (red dots in panel C) distributed in 29 1-km2 areas; details of plot selection and sampling are in Brommit et al (2004). To find the spatial scale of effect of forest amount on plant species richness we modeled the species richness in the plots with the forest amount in circular areas of different radii: 50 m, 100 m, 200 m, 500 m, 1000 m and 2000 m around the plots (panel A). The scale of effect was inferred to be the radius for which model Akaike information criterion (AIC) was minimized (Miguet et al., 2015): 100 m.
Herbaceous plant data – Response variable
In each plot, the herbaceous plants were sampled in two 1m2 quadrats in each of three years, 1999-2001. Field guides (Hosie, 1990; Legasy et al., 1995) were used to identify all herbaceous species present in both 1-m2 quadrats of each plot. Species richness per plot was the total number of species present across the three years in both quadrats.
Patch size, patch isolation and forest amount – Independent variables 23
We calculated the area and nearest neighbor distance of the local forest patch containing each sample plot using Fragstats v.4.0.4 (McGarigal et al., 2012), based on the maps provided by T. Contreras (2002). Contreras (2002) estimated the forest cover using both satellite imagery (unclassified Russian MK-4 satellite images (7.5m x 7.5m ground pixel size; courtesy of Parks Canada) and topographic maps (at 1: 50,000 scale provided by the National
Topographic Data Base, Geomatics Canada).
To identify the appropriate spatial extent for measuring forest amount around each sample plot, (see section “Caution 2: appropriate spatial scale” in Fahrig 2013), we fit the relationship between forest amount and species richness within each of six nested circular areas around the 189 plots: 50 m, 100 m, 200 m, 500 m, 1000 m and 2000 m. The local landscape scale was inferred to be the radius for which model Akaike information criterion (AIC) was lowest (Miguet et al., 2015). In the tests of the six predictions (Table 1), forest amount effect was estimated at that extent.
To test the six predictions we sub-selected plots to create the following specific gradients in predictor variables (see also Table 1 and Figure 2): 1) constant local patch area and varying forest amount (for prediction A); 2) constant nearest neighbor distance and varying forest amount (for prediction B); 3) varying local patch area and constant forest amount (for prediction C); 4) varying nearest neighbor distance and constant forest amount (for prediction
D); 5) decreasing local patch area with increasing forest amount (for prediction E) and 6) increasing nearest neighbor distance with increasing forest amount (for prediction F).
Plots selected to test prediction A – the effect of forest amount on species richness controlling for local patch area – ranged from 2% to 98% of forest amount and from 1 to 12 hectares in local patch area (data subset A1; n=19) and from 20 to 148 hectares (data subset
A2; n=69). Plots selected to test prediction B – the effect of forest amount on species richness controlled by nearest neighbor distance – also ranged from 2% to 98% and nearest neighbor 24
distance varied from 403 to 1096 meters (data subset B1; n=92) and from 1096 to 2980 meters
(data subset B2; n=23).
Plots selected to test prediction C – the effect of local patch area on species richness controlled by forest amount – ranged between 30% - 54% of forest amount (data subset C1; n=17) and 55% - 81% (data subset C2; n=55) and local patch area varying from 1 to 22026 hectares. Plots selected to test prediction D – the effect of nearest neighbor distance on species richness controlled by forest amount – ranged between 37% - 60% of forest amount (data subset
D1; n=23) and 60% - 90% (data subset D2; n=66) and nearest neighbor distance varying from
36 to 10080 meters.
The 25 plots selected to test prediction E – the effect of forest amount even when local patch area decreases – ranged from 51% to 100% of habitat amount and of decreased local patch area from 97 hectares to 3 hectares. Finally, the 39 plots selected to test prediction F – the effect of forest amount even when nearest neighbor patch distance increases – varied from
36% to 99% of forest amount and increased nearest neighbor distance from 39 to 4953 meters. 25
Figure 2: Subsets of the sampled plots used to test each of the six predictions of the habitat amount hypothesis in Table 1. The predictor variable gradients selected are shown 26
diagrammatically in column 3 (see also Table 1). In the fourth column each point represents a sampled plot and the rectangles enclose the points selected to test each prediction. For each of the first four predictions, we conducted two tests, based on the two data subsets represented by the rectangles with solid and dashed borders.
Analysis
We modeled plant species richness using Markov Chain Monte Carlo Generalized
Linear Mixed effect Models (MCMC-GLMM) with a gaussian distribution and log link function. We entered the predictor variable(s) (forest amount, patch size, and/or nearest neighbour distance) as fixed effects. As the 189 sample plots were nested within 29 landscapes, we also entered landscape ID as a random effect. We visually inspected quantile plots, and residuals vs fitted values plots to check for normality of predicted values and any remaining lack of fit. We plotted correlograms based on Moran’s I of the raw herbaceous plant data and modelled residuals to check for spatial autocorrelation, as recommended by Zuur et al., (2009).
All analyses were conducted in R version 3.0.1 (R Core Team, 2015) using the following packages: reshape2 (Wickham, 2007), MuMIn (Barton, 2015), glmmADBD (Skaug et al.,
2012), MCMCglmm (Hadfield, 2010) and pgirmess (Giraudoux, 2015).
RESULTS
From the 3432 recorded shade-tolerant herbaceous plant specimens, at forested plots representing 16 species in 189 plots, distributed among the 29 landscapes, we limited our analyses to the 16 species that met the following two criteria. First, we limited the species to shade-tolerant species as these are the species for which forest cover was an appropriate surrogate of habitat. Second, we omitted any species present at fewer than 10% of the plots to avoid zero-inflated models.
Habitat amount always had a significant effect on species richness when the local patch area was controlled, unlike the local patch area, which in turn did not affect species 27
richness when the effect of habitat amount is controlled (Table 2). Also, species richness increases with increasing habitat amount even if local patch area decreases.
Similarly, habitat amount had a significant effect on species richness when nearest neighbor distance was controlled, and patch isolation had no effect on species richness when habitat amount effect was controlled. However, contrary to the hypothesis postulates, species richness decreases with increasing habitat amount and increasing nearest neighbor patch distance.
Table 2: Habitat amount test results of each formulated prediction. Each prediction (from A to F) was tested with species richness data. To test predictions A – D we created two subsets (for details of subset criteria, see Fig. 2 and landscape structure – independent variables section). We present the bayes estimate as mean of posterior distribution (posterior mean), as well as the lower and upper confidence intervals of the posterior mean. pMCMC values under 0.05 indicate the model significance.
Posterior CI 95% - CI95% - prediction subset N pMCMC mean lower upper
A 1 19 0.41 0.20 0.58 < 0.0001
A 2 69 0.42 0.32 0.53 < 0.0001
B 1 92 0.09 -0.02 0.20 0.12
B 2 23 0.43 0.25 0.63 < 0.0001
C 1 17 -0.16 -1.00 0.54 0.69
C 2 55 0.17 -0.26 0.61 0.45
D 1 23 -0.18 -1.30 0.79 0.76 28
D 2 66 0.01 -0.17 0.20 0.96
E 1 25 1.00 0.75 1.25 < 0.0001
F 1 39 -2.52 -2.92 -2.11 < 0.0001
Figure 3: Results of habitat amount hypothesis tests for shade tolerant herbaceous plants. The headers and refer to predictions tested (Table 1). The colors refer to the data subsets. Stars represent significant effects (*** p < 0.001; ** p < 0.01; * p < 0.05), points represent marginally significant effects (p < 0.07). Each bar represents one prediction test. There are pairs of bars in predictions A-D because it was possible to select more than one dataset to test the predictions. The effect sizes for forest amount are always higher than the effect sizes for patch area and nearest neighbor distance.
DISCUSSION
Patch area and isolation had no effect on species richness independent of habitat amount; this result suggests habitat amount can replace these predictors as predicted by the habitat amount hypothesis. We argue that even when a correlation between patch size and species richness is found, with additional analysis, in some cases, it is possible to see additional landscape processes and patterns behind the assumed patch area effect. 29
Predictions related to patch area effect
Habitat amount had a significant effect on species richness when the local patch area was controlled and local patch area had no effect on species richness when the habitat amount was controlled. To confirm that patch area effect is essentially habitat amount effect, species richness increased with increasing habitat amount even if the local patch area decreased.
By selecting plots with varying habitat and constant fragmentation and vice-versa, we were able to isolate the effects of habitat amount from the effects of habitat spatial configuration. With
(2015) also found no difference between clumped and fragmented landscapes when analyzing the relationship between total species richness and total habitat area. He found a difference between species-area relationship measured on the patch-scale and landscape-scale, which means that the landscape is more than the sum of its patch areas. Also, abundance of bumblebees slightly declined with increasing distance from natural habitat when the habitat amount also reduced slightly.
Predictions related to patch isolation effect
Habitat amount had a significant effect on species richness when the local patch nearest neighbor distance was controlled in one of the two analyzed cases. But local patch nearest neighbor distance had no effect on species richness when the habitat amount was controlled. Surprisingly we found a negative effect of habitat amount with the increasing isolation and we do not know why. Especially for forest-interior species, the number of connecting elements around the focal patch plays an important role in determining species occurrence (van Dorp & Opdam, 1987), probably because these areas provide additional feeding spots and corridors for dispersal. Also, the community was probably composed by transient and successional species due to a disturbance caused by an ice storm one year before the sampling starts. Isolated patches could be under more disturbance than connected ones, 30
increasing the number of pioneer species, which increased the number of species in more isolated patches, even with less habitat amount. These unexpected results reinforce our lack of knowledge about the effects of landscape structure on biodiversity. In a recent meta-analysis
Mendenhall et al. (2014) found a variety of bat responses to agricultural landscapes, ranging from positive strong effects of habitat amount on species richness to an increase of individuals in open areas. The authors suggest that the direct application of island biogeographic theory to agricultural landscapes is distorting our understanding and conservation strategies in human affected landscapes. They advocate the need of a countryside biogeography framework.
The scale of effect concept
The habitat amount hypothesis was tested here on the scale wherein the habitat amount has an effect on species richness for the particular group of species studied, the scale of effect (Holland et al., 2004; Jackson & Fahrig, 2014). Hanski (2015) called it a narrow perspective because it does not consider the entire area of species occurrence. But the scale of effect is the scale in which the species-habitat relationship is biologically and statistically most important. Most of the studies evaluating species responses to habitat do not consider different spatial scales (Delciellos et al., 2016), which is necessary, given that even species belonging to the same taxon or guild can respond to the habitat at different scales (Holland et al., 2004). The effect of landscape structure on biodiversity is spatially dependent, since the results of this effect change according to the extent. As proposed by Holland et al., (2004) and tested by many other authors (Holland et al., 2005; see Miguet et al., 2015 for a review), determine the scale of effect is important to include several spatial scales around the sampling locations, a multi-scale study. The scale of effect is that in which the species response is stronger, independent how small or large the radius around the plot is. If the species in a plot is mostly affected by the spatial structure in a given extent, it means that this is the scale in which the effect should be analyzed. 31
Some processes and patterns behind the so-called effect of area and isolation
We, not exhaustively, described below the landscape processes and patters that could be behind the so-called effect of area and isolation.
The number of cover types is important. Patch density and corridors (van Dorp &
Opdam, 1987) play an important role by enabling individuals to move and occupy vacant patches, allowing dispersal flow and metapopulation viability. A large number of cover types in the landscape can positively affect mammals’ species richness by enabling organisms to move among the remnants. Diversity of cover types and shape index explained much more of the variance in carnivores and non-carnivores respectively than patch size, which was a weak predictor of species richness in South Africa (Ramesh et al., 2016). Johnson et al. (2016) also conclude that the landscape context was more important to maintain estuarine and open coastal species than the patch size alone. By the way, these authors found a negative relationship between patch size and species richness for two of the three studied groups.
Another thing to consider is the habitat association and patch quality. Habitat association (i.e. the preferential habitat to each species) (Bender et al., 1998) and patch quality
(Alstad & Damschen, 2016) seems to explain most of the variation of patch size effect on animals and plants, respectively, in patchy landscapes. To tease apart each variable’s effect on species responses, each predictor must be controlled when testing the effect of another one.
This leads us to habitat suitability for species using any given landscape, one expects that total amount of habitat will be of greater importance than its spatial configuration because the effect of patch area and isolation on species will mostly depend on the suitability of the surrounding habitats (Andrén, 1994). The surrounding area seems to be more important in explaining species richness than patch area per se; and this pattern also holds for estuarine and open coast environments (Johnson et al., 2016). 32
Further, there is also the geometric effect. The patch area effect tends to be weak or null when controlled by geometric effect (Bender 1998). The geometric effect occurs when the species density is related to total patch size rather than the inhabited area, underestimating values for edge species (because core area is not habitat for these species) and interior species
(because they do not inhabit edges). A more biology-related aspect is the effect of habitat heterogeneity. The number of species from equal sized plots is greater the more heterogeneous are the habitats and greater the habitat amount (Haila, 1983). Some effects of area could be, in fact, the effect of increasing habitat heterogeneity. That can be seem where a changing of a habitat predictor by another one in the model (e.g. mean trunk circumference by shape of wood) there was no significant drop in the variance (van Dorp & Opdam, 1987). Haila (1983), for example, found a high correlation between island area (90% of the variance) and species richness. But two other variables positively correlated with island area: number of habitat types and habitat diversity had almost equally high correlations.
Although largely used as predictors, here we found that patch size and isolation do not seem to explain species' occurrence in a given site by their components per se. We highlight that the so-called spatial configuration effects – patch area and isolation effects – is essentially habitat amount effect.
ACKNOWLEDGEMENTS
Ludmila Rattis is funded by Paulo Teixeira & Co, CAPES (Process 04100/2014- 00) and Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP scholarship 2012/02207-9). Rafael Loyola’s research has been constantly funded by Conselho Nacional de Pesquisa (grants #304703/2011-7, 479959/2013-7, 407094/2013-0), Conservation International Brazil, and the O Boticário Group Foundation for the Protection of Nature (PROG_0008_2013) and CNCFlora.
33
REFERENCES
Almeida-Gomes M., Vieira M.V., Rocha C.F.D., Metzger J.P., & De Coster G. (2016) Patch size matters for amphibians in tropical fragmented landscapes. Biological Conservation, 195, 89–96. Barton K. (2015) MuMIn: Multi-Model Inference R package version 1.10. 5. . Bender D.J., Tischendorf L., & Fahrig L. (2003) Using patch isolation metrics to predict animal movement in binary landscapes. Landscape Ecology, 18, 17–39. Brommit A.G., Charbonneau N., Contreras T.A., & Fahrig L. (2004a) Crown Loss and Subsequent Branch Sprouting of Forest Trees in Response to a Major Ice Storm. Journal of the Torrey Botanical Society, 131, 169. Brommit A.G., Charbonneau N., Contreras T.A., & Fahrig L. (2004b) Crown Loss and Subsequent Branch Sprouting of Forest Trees in Response to a Major Ice Storm. Journal of the Torrey Botanical Society, 131, 169. Connor E.F. & McCoy E.D. (1979) The Statistics and Biology of the Species-Area Relationship. The American Naturalist, 113, 791–833. Connor E.F. & McCoy E.D. (2001) Species-area relationships. Encyclopedia of biodiversity pp. 397–411. Contreras T. (2002) Changes in the Abundance and Distribution of Woody Plants Related to Dispersal Mechanisms Along a Forest Cover Gradient. Carleton University, Ottawa, Ontario. Darwin A.T., Ladd D., Galdins R., Contreras T.A., & Fahrig L. (2004) Response of forest understory vegetation to a major ice storm. Journal of the Torrey Botanical Society, 131, 45–52. Fahrig L. (2013) Rethinking patch size and isolation effects: the habitat amount hypothesis. Journal of Biogeography, 40, 1649–1663. Giraudoux P. (2015) pgirmess: Data Analysis in Ecology. R Package version 1.9. . Hadfield J. (2010) MCMC Methods for Multi-response Generalized Linear Mixed Models: The MCMCglmm R Package. Journal of Statistic Software, 33, 1–22. Hosie R.C. (1990) Native Trees of Canada. Fitzhenry and Whiteside, Markham, Ontario. Hubbell S.P. (2001) The unified neutral theory of biodiversity and biogeography. Princeton University Press, Legasy K., LaBelle-Beadman S., & Chambers B. (1995) Forest Plants of Northeastern Ontario. Lone Pine Publishing, Edmonton. 34
McGarigal K., Cushman S.A., & Ene E. (2012) FRAGSTATS v4: Spatial Pattern Analysis Program for Categorical and Continuous Maps. . Mendenhall C.D., Karp D.S., Meyer C.F.J., Hadly E.A., & Daily G.C. (2014) Predicting biodiversity change and averting collapse in agricultural landscapes. Nature, 509, 213– 217. Miguet P., Jackson H.B., Jackson N.D., Martin A.E., & Fahrig L. (2015) What determines the spatial extent of landscape effects on species? Landscape Ecology, 1–18. Olthof I., King D.J., & Lautenschlager R. a. (2004) Mapping deciduous forest ice storm damage using Landsat and environmental data. Remote Sensing of Environment, 89, 484–496. Pasher J. & King D.J. (2006) Landscape fragmentation and ice storm damage in eastern ontario forests. Landscape Ecology, 21, 477–483. Petit S., Griffiths L., S. Smart S., M. Smith G., C. Stuart R., & M. Wright S. (2004) Effects of area and isolation of woodland patches on herbaceous plant species richness across Great Britain. Landscape Ecology, 19, 463–472. Pisaric M.F.J., King D.J., MacIntosh A.J.M., & Bemrose R. (2008) Impact of the 1998 ice storm on the health and growth of sugar maple (Acer saccharum Marsh.) dominated forests in Gatineau Park, Quebec. Journal of the Torrey Botanical Society, 135, 530–539. Preston F. (1962) The canonical distribution of commonness and rarity: Part I. Ecology, 43, 185–215. R Core Team (2015) R: A Language and Environment for Statistical Computing. . Saura S., Bodin Ö., & Fortin M.-J. (2014) Stepping stones are crucial for species’ long-distance dispersal and range expansion through habitat networks. Journal of Applied Ecology, 51, 171–182. Skaug H., Fournier D., Nielsen A., Magnusson A., & Bolker B. (2012) Generalized Linear Mixed Models using AD Model Builder. R package version 0.7. 2.12. R CRAN, . Tischendorf L. & Fahrig L. (2000) On the usage and measurement of landscape connectivity. Oikos, 90, 7–19. Wickham H. (2007) Reshaping Data with the reshape Package. Journal of Statistical Software, 21, 1–20. Zuur A.F., Ieno E.N., Walker N., Saveliev A.A., & Smith G.M. (2009) Mixed effects models and extensions in ecology with R. Springer, New York.
35
Factors determining the spatial extent in which species respond to
landscape structure
Ludmila Rattis1,2*, Rafael Loyola2, Lenore Fahrig3
Authors’ affiliations:
1*Programa de Pós-Graduação em Ecologia, Universidade Estadual de Campinas,
CP6109, 13083-970 Campinas, São Paulo, Brazil. Email: [email protected]
2Laboratório de Biogeografia da Conservação, Departamento de Ecologia,
Universidade Federal de Goiás, CP 131, 74001-970 Goiânia, Goiás, Brazil.
3Geomatics and Landscape Ecology Research Laboratory (GLEL), Department of
Biology, Carleton University, Ottawa, ON, K1S 5B6, Canada
36
ABSTRACT
The relationship between species and their habitat must be measured at a scale that is appropriate for the species and phenomena being studied. Here we tested if mobility, demography and biological response are predictive of the spatial extent in which herbaceous plant species are affected by forest habitat amount, the called scale of effect. We tested five predictions derived from three hypothesis explaining which factors affect the scale of effect:
(1) If the scale of effect depends mostly on movement range, then the scale of effect should be weak or absent if the response integrates across a group of species with widely differing movement ranges; (2) the scale of effect should be smaller if the species reproductive rate is higher, and (3) the scale of effect should be smaller if the response variable is measured as population abundance as opposed to species occurrence. When species with varying seed masses and dispersal modes were modeled together there was no clear scale of effect of forest amount and species richness, but when species are analyzed separately in groups with similar seed masses and dispersal mode we found a scale of effect for zoochoric light-seeded species and anemochoric heavy-seeded species. Seed mass explained the scale of effect: the heavier the seeds, the smaller the scale of effect, with no effect of dispersal mode. Species with higher reproduction rate showed a larger scale of effect. Species abundance responded to the surrounding landscape at a smaller scale than species occurrence. The scale of effect of landscape structure on a plant community depends on the mobility, demography and the biological response of herbaceous plant species. Restoration plans including those species must consider the narrow spatial pattern of seed rain and recruitment and the positive relationship with reproductive rate. Also, landscape management plan applied at an appropriate extent for capturing the response in species abundance will not ensure species occurrence.
37
Key-words: Dispersal mode; habitat amount; habitat loss; local landscape; plant demography; scale of effect.
38
INTRODUCTION
Scaling is a critical aspect for Landscape Ecology (Wiens 1989). Determining the scale of effect is not only important, but crucial (Jackson & Fahrig, 2012a, 2015a; Martin &
Fahrig, 2012; Miguet et al., 2015), because once it is determined, then the effects of the structural components of the landscape could be correctly estimated.
To determine the extent to which the habitat amount affects the species, it is necessary to measure the landscape structure and its effects at the scale of effect. The scale of effect approach consists in measuring habitat amount or any predictor in local landscapes of different radii from sample sites where the biological response is measured. Then the relationship between the biological response and landscape predictor for different landscape sizes is estimated. The scale of effect is where the relationship between the biological response and the landscape predictor is strongest (Brennen et al. 2002; Holland et al. 2004; Fig. 1).
The occurrence and density of a species in a given area is affected by habitat amount surrounding the area (Wiens et al., 1993; Brückmann et al., 2010) However, habitat amount affects the species to a limited spatial extent, from which its effects are diminished until there is little or no significant effect (Jackson & Fahrig, 2012b). Specific traits like dispersal ability
(Ozinga et al., 2005), reproduction rate (Murray et al., 2002; Boedeltje et al., 2003), and interaction with other species are some of these intrinsic factors that also affect the species occurrence and density. Different dispersal abilities, reproduction rates and tolerance to harsh environments can affect individuals’ movement and colonization (Levin, 1992). Even the nature of the biological response (i.e. genetic diversity, species occurrence, species abundance) can affect the results, because they are subject to distinct temporal and spatial processes
(Jackson & Fahrig, 2014).
Miguet et al. (2015) summarized a series of hypotheses and related predictions of factors that may influence the scale of effect. We provide the first tests of three of these 39
hypotheses concerning the mobility, demography and employed response variable. The first hypothesis suggests that, if the scale of effect depends mostly on movement range (Jackson &
Fahrig, 2012a), then it should be weak or absent if the response integrates across a group of species with widely differing movement ranges.
The second hypothesis proposes that the scale of effect should be smaller for species with higher reproductive rates (Jackson & Fahrig, 2012a). Populations of a species with a high reproduction rate are more likely to be near or above carrying capacity than populations with lower reproductive rates. If a population is near or at its carrying capacity, immigration has a much smaller effect on abundance than for a population well below carrying capacity. In a
'near-full' site, the only other sites whose emigrants are likely to influence the abundance are the closest sites, because these immigrants will arrive first. Therefore, distant sites will have little or no influence on abundance, reducing the scale of effect of the landscape.
The third hypothesis we test suggests that the scale of effect should be smaller if the response variable is measured as population abundance than if it is measured as species occurrence, because these two population outcomes are regulated at different temporal scales
(Jackson & Fahrig, 2014). Species abundance can fluctuate generation by generation, whereas occurrence is a pattern determined by local extinction and local colonization, processes that play out over many generations. The more generations a mechanism takes to produce a local response, the more important are the geographically distant individuals in affecting that response.
Here we used shade-tolerant herbaceous species data from Ontario, Canada to investigate how does the scale of effect change with species traits and the measured population response. We tested five predictions derived from those three presented hypothesis. The first three predictions are derived from the first hypothesis, while predictions 4 and 5 are derived from the second and third hypotheses, respectively. 1) There should be a distinct scale of effect 40
of forest amount (percent cover of the landscape surrounding the sample site) on plant total abundance or richness if total abundance and richness are measured across a group of species with similar seed mass and dispersal mode. In contrast, there should be only weak, if any, evidence for a scale of effect for the relationship between forest amount and total abundance or species richness, when these measures includes species with very different seed masses and dispersal modes. 2) Seed mass should negatively influence the scale of effect, because species with heavier seeds usually have limited dispersal distance compared to species with lighter seeds (McEuen & Curran, 2004b). 3) The scale of effect should be smaller for animal-dispersed plant species compared to wind-dispersed plant species because animal-dispersed plants show spatially limited seed dispersal in comparison to the wide-spread seed dispersal of wind- dispersed species (McEuen & Curran, 2004b). 4) There should be a negative cross-species correlation between the scale of effect of forest amount on occurrence and species seed production. 5) Finally, for a given forest herbaceous plant species, the scale of effect of forest amount on abundance should be smaller than the scale of effect of forest amount on occurrence.
41
Figure 1. How the scale of effect is determined. First one calculates the habitat amount or other landscapes predictor in local landscapes of different radii from the sample sites where the ecological response, here species richness, is measured (panel A). Here the radii are 1km, 2km,
3km, 4km, 5km. Then one estimates the relationship between the ecological response and habitat amount for each of the different landscape sizes. The scale of effect is where the relationship between the ecological response variable and the landscape predictor is stronge r
(panel B).
42
METHODS
Overall approach
The data for this study were shade-tolerant herbaceous plant species occurrences and abundances measured at 189 forested plots in Eastern Ontario, Canada, and forest amount in the landscapes surrounding each plot (details in Plant and landscape data below). In addition, we obtained species trait information from the literature, as specified below. For prediction 1 we investigated the scale of effect of the total forest amount on abundance of species with large variation in dispersal ability and then do the same analysis with groups of species with similar dispersion capabilities. We used seed mass and dispersal mode as surrogates of movement range. For prediction 2 we investigated if seed mass affects the scale of effect. For prediction 3 we investigated if dispersal mode and the interaction between dispersal mode and seed mass affect the scale of effect. For prediction 4 we investigated if seed production affects the scale of effect. For prediction 5 we investigated if the response variable – species occurrence versus species abundance – affect the scale of effect.
Plant and landscape data
The plots were sampled between 1999 and 2001 in Ontario, Canada, in a region within approximately 60 km of Ottawa (Fig. 2). The dominant tree species include Acer saccharum, Tilia americana, Fraxinus sp. and Pinus sp. The forest is primarily second growth and the original forest was cut between 160 and 90 years ago (Brommit et al., 2004). The original data set includes 290 vegetation plots (two 1m2 quadrats per plot), with 10 plots in each of 29 1km2 landscapes. Of these we selected the 189 forested plots (Fig. 2). Details of plot selection and sampling are in Brommit et al. (2004).
The herbaceous plants were sampled in two 1m2 per plot. Field guides (Hosie, 1990;
Legasy et al., 1995) were used to identify all herbaceous species present in both 1m quadrats of 43
each plot. Using 10% increments, the percentage of the quadrat covered by each herbaceous species was estimated. We averaged the number of specimens per plot across the three sampling years, 1999-2001. Species abundance was the averaged specimens number present across the three years. Species occurrence was based on abundance table, with presence indicating more than one specimen and absence when there is no individual representing the species. The total species abundance is the sum of the average number of specimens per plot across the three sampling years, 1999-2001.
We used seed mass (g) and dispersal mode (wind/animal) as surrogates of movement range and seed production (in weight – grams – of seed produced per m2 per year: g x m-2 x year-1) as surrogate of reproduction rate. We searched the published and grey literature to obtain these data for all the shade tolerant herbaceous species present in more than 10% of the sampled sites (Waldron, 1965; Bierzychudek, 1982; Feret et al., 1982; Leckie et al., 2000;
Whigham, 2004).
44
Figure 2. The study area in Eastern Ontario, Canada and the sample design. Shade-tolerant herbaceous plants were sampled in 189 forested sample sites (black dots) distributed in 29 landscapes; details of site selection and sampling are in Brommit et al. (2004). To find the scale of effect we modeled the species richness and species abundance for different sub-sets of the species, in each site on the forest amount in buffers of different radius: 50m, 100m, 200m,
500m, 1000m and 2000m (buffers on the top right).
Estimating the scale of effect of forest amount
To estimate the scale of effect, we fit the relationship between forest amount and the appropriate response variable depending on the prediction (see below), within each of six nested circular areas around the 189 plots: 50m, 100m, 200m, 500m, 1000m and 2000m. The 45
scale of effect was the landscape radius for which the model Akaike information criterion (AIC) was lowest (Jackson & Fahrig, 2012b).
Prediction 1: Distinct scale of effect for species group with similar dispersal
We used seed mass and dispersal mode as surrogates of movement range. We tested the relationship between forest amount (%) and total abundance, including 30 species with very different seed masses (min=0.6g; max=3,143g; mean=157g) and dispersal modes (animal=15 species; wind=14 species; explosive=1 species). So, we got the effect size of each model, which is based on the correlation coefficient and their 95% confidence interval and compared those effect sizes across all the investigated scales.
We investigated if there was a more distinct scale of effect if total abundance was measured across a group of species with similar seed mass and dispersal mode. To do that, we grouped the species into four groups. The species belonging to the first group (8 species) included the animal seed-dispersed with seed masses ranging from 1.2g to 8.8g. The second group (5 species) included anemochorous species with seed masses ranging from 0.6g to 7.6g.
The species included in the third group (4 species) comprised all the animal seed-dispersed with seed masses ranging from 16.1g to 55.6g. The fourth group (8 species) included anemochorous species with seed masses ranging from 15.4g to 55.3g).
Prediction 2: The effect of seed mass on the scale of effect
We tested if species with heavier seeds usually have limited dispersion distance compared to lighter seeds. We used the scale of effect of forest amount on 25 species occurrence as response variable to test the effect of seed mass. We performed generalized linear mixed models with binomial distribution and log link between forest amount and species occurrence. 46
So, the scale of effect was inferred as described at Estimating the scale of effect of forest amount section.
Prediction 3: The effect of dispersal mode on the scale of effect
We investigated if the scale of effect is smaller for animal-dispersed plant species compared to wind-dispersed plant species and the effect of seed mass and dispersal mode on the scale of effect. We used the scale of effect of forest amount on 25 species occurrence as response variable to test the effect of dispersal mode. We performed generalized linear mixed models with binomial distribution and log link between forest amount and species occurrence.
So, the scale of effect was inferred as described at Estimating the scale of effect of forest amount section. We found out the effect of dispersal mode on scale of effect by a two sample Wilcoxon test. The effect of both dispersal ability surrogates was investigated through an analysis of covariance.
Prediction 4: The effect of reproduction rate on the scale of effect
We tested if there is a negative cross-species correlation between the scale of effect of forest amount and species seed production. We used 14 species to test the effect of seed production on the scale of effect. We performed generalized linear mixed models with binomial distribution and log link between forest amount and species occurrence. So, the scale of effect was inferred as described at Estimating the scale of effect of forest amount section. The effect of seed production on the scale of effect was investigated by generalized linear model with
Gaussian distribution.
Prediction 5: Relationship between the effect of effect and population outcome 47
To investigate if the correlation between abundance and forest amount is strongest at smaller scales than the correlation between species occurrence and forest amount, we compared both scales of effect. We obtained the scale of effect of forest amount on species occurrence and forest amount on species abundance for each of the 23 species present in 30%
- 90% of the sites. Finally, as we knew the spatial extents in which the forest amount affects the population response – abundance and occurrence – of the same species, we tested the prediction about the biological response using a t-test in which the scales of effect were compared, using
ΔAIC of the best model explaining the variation of each population outcome as a weight in the analysis.
Data Analysis
We used mixed models instead of a simple linear model because the landscape describes the data sample as a subset of the entire data set. Since there are 29 landscapes within which the 189 sample units are clustered, we entered forest amount as fixed term and landscape
ID as a random term for predictions of species total abundance and richness.
To find what probability distribution best fitted our data we modelled the observations and the quantiles with five different distributions – normal, lognormal, negative binomial, poisson and gamma. Total abundance best fit the lognormal distribution and data transformation was not needed. We estimated the parameters using maximum likelihood because the random effects – landscapes – are nested and our data are balanced (similar sample size in each factor group). We estimated the effect of forest amount on species total abundance by linear mixed model with lognormal distribution. The estimates of the fixed effect (forest amount) is showed as regression estimates and their confidence intervals estimated with bootstrap. The local landscape scale was inferred to be the radius for which model Akaike information criterion (AIC) was lowest. 48
All analyses were done in R version 3.0.1 (R Core Team, 2015) using the packages
MuMIn (Barton, 2015), glmmADBD (Skaug et al., 2012) and pgirmess (Giraudoux, 2015), car and MASS. We inspected normality, lack of fit, independence and dispersion of the residuals.
We visually inspected quantile plots, and residuals vs fitted values plots to check for normality of predicted values and any remaining lack of fit. We plotted correlograms based on Moran’s I of the raw herbaceous plant data and modelled residuals to check for spatial autocorrelation, as recommended by Zuur et al., (2009).
RESULTS
Prediction 1: Distinct scale of effect for species group with similar dispersal
There is no best scale of effect when we analyzed the effect of forest amount on species total abundance, as the models at 200, 100, 500 and 2000 are all within 2 AICc (Fig. 3).
Figure 3. Effect size of the relationship between of forest amount and total abundance of a group of 30 plant species with widely varying movement ranges ranges, with forest amount 49
calculated at six different scales. There is no clear best scale as all the models are within 2 AIC of the best model (at 200m).
When the assemblage was decomposed into groups of similar seed masses and dispersal modes (surrogates of dispersal distance), we found the scale of effect for species with lighter seeds dispersed by animals (100m – 200m within 4.74 AICc) and heavier seeds dispersed by the wind (100m – 200m within 3.37 AICc) (Fig. 4).
Figure 4. Effect size of the relationship between forest amount and total abundance of four groups of species separated by seed mass and dispersal mode. At top left panel are the results of eight species dispersed by animals and seed mass ranging from 1.2g to 8.8g (mean=3.8g).
At top right panel are the results of five species with same seed mass range, but dispersed by the wind. At bottom left panel are the results of four species dispersed by animals and seed mass ranging from 15.4g to 55.6g (31.2g). At the bottom right panel are the results of eight species with the same seed mass range but dispersed by the wind. Forest amount was calculated at six different scales: 50m, 100m, 200m, 500m, 1000m and 2000m. 50
Prediction 2: The effect of seed mass on the scale of effect
Seed mass affect negatively the scale of effect (r = -0.44; CI95%= -0.84 – -0.04; p =
0.0159).
Prediction 3: The effect of dispersal mode on the scale of effect
Dispersal mode did not affect the scale of effect (two sample Wilcoxon test W =
90.5, p-value = 0.458). There is a significant negative effect of seed mass on the scale of effect, even after controlling for seed dispersal mode (F(1,22) = 4.83; p-value = 0.03).
Prediction 4: The effect of reproduction rate on the scale of effect
The scale of effect is also affected by seed production. The relationship between the scale of effect of forest amount on abundance of 14 forest herbaceous plants and species seed production is significant (Pearson’s r = 0.89 CI95%= 0.14 – 1.63; p-value=0.0191).
Prediction 5: Relationship between the effect of effect and population outcome
The weighted t-test did show that the scale of effect for population abundance were smaller than for species occurrence (t= 2.28; df= 35; p-value= 0.028; Fig. 5). This indicates that population abundance respond more locally to the forest amount than species occurrence. 51
Figure 5. The scale of effect (radius measured in meters from the site) of the relationship between forest amount and forest plant species occurrence/species abundance. The 23 species included were those present in 30 – 90% of the 189 sites. The green circles represent all the scales evaluated: 50m, 100m, 200m, 500m, 1000m and 2000m. Red circles represent the scale of effect for occurrence (dotted lines) and abundance (solid lines). A t-test of scales of effect weighted by the ΔAICc of the abundance model at the scale of effect compared to the occurrence model supports the prediction that the scale of effect is smaller if the response variable is population abundance than if it is occurrence (t-value= 2.28; weighted df= 35; p- value= 0.028).
DISCUSSION
The spatial extent of the relationship between species and their habitat is affected by mobility, demography and biological response. Multiple species studies must include multiple spatial scales, since the trends being analyzed for each species would be weakened for many species (Holland 2005).
Prediction 1: Distinct scale of effect for species group with similar dispersal 52
We found no scale of effect of forest amount on species abundance of an assemblage with wide range movement ability. But, once we accounted for seed masses and dispersal mode, we found the scale of effect between 100 – 200 meters for zoochoric light- seeded species as well as for anemochoric heavy-seeded species. This similar scale of effect for two groups of species can be explained by their similar colonization patterns and mechanisms.
Zoochoric small-seeded species and heavy wind-dispersed species are similar in their patterns of arrival and survival: 1) their arrival are more likely in maturing and recent gaps, respectively, than beneath forest canopy; 2) their survival probabilities are highest in recent gaps because of the reduced incidence of disease (for a full explanation about this issue see Schupp, 1989).
Prediction 2: The effect of seed mass on the scale of effect
According to our results, seed mass is predictive of the scale of effect. As seed mass increases, the scale of effect decreases. We included just herbaceous shade-tolerant species in order to correctly define forest as habitat for the studied species (see Fahrig 2013) and it has some implications. Shade-tolerant species usually have larger seeds that can germinate beneath the forest canopy (Schupp et al., 1989) with narrowly dispersed seeds. Seeds falling at forest spots in different successional stages have different probabilities of arrival, survival and recruitment, which depends on their mass and dispersal mode.
Prediction 3: The effect of dispersal mode on the scale of effect
As opposed to McEuen and Curran (2004), we found no support for the prediction about the dispersal mode. According to these authors, animal dispersed species exhibited lower dispersion compared to wind dispersed species. We grouped birds, bats and ants dispersed species into a single group due to analysis constrains (i.e. small number of species belonging to each group), but their different dispersal behaviors may have hidden effects that dispersers have 53
on the scale of effect. For example, large vertebrates are unlikely to deposit large seeds in gaps, while birds and bats are more active in and around maturing gaps (Schupp et al., 1989).
In a simulation study, Jackson and Fahrig (2012) developed and supported a quantitative prediction on the relationship between dispersal distance and the scale of effect of habitat amount on abundance. According to these authors ‘landscape should be measured at a radius that is 4-9 times the median dispersal distance or 30-50% the maximum dispersal distance of the species of interest’ (Jackson and Fahrig 2012, p. 938). But an empirical study investigating what spatial scales the songbird abundances are most affected by landscapes did not support the prediction of the positive effect of dispersal distance on the scale of effect
(Tittler, 2008).
Prediction 4: The effect of reproduction rate on the scale of effect
We predicted that the scale of effect is smaller for species with a higher reproductive rate, which was supported by a simulation (Jackson & Fahrig, 2012a) and an empirical study
(Kallio, 2014), but not supported by a meta-analysis performed by Jackson and Fahrig (2015).
Here, we found that the scale at which landscape structure affects the occurrence of species of herbaceous shade-tolerant species is also influenced by the reproductive rate of the species, but in the opposite way to what we expected. We expected that species with higher reproductive rates were closer to reaching the carrying capacity, which would make immigration had a much smaller effect on abundance than for a species far below carrying capacity. In a 'near-full' site, the only other sites whose emigrants would influence the abundance would be the closest sites, because these immigrants would arrive first. But we found here that the higher the reproductive rate, the greater the scale of effect in which forest amount affects species occurrence. Our prediction was similar to what Miguet et al. (2015) expected and both are all about immigration.
But maybe what is happening is more related to emigration processes. Once the carrying 54
capacity is met, individuals will disperse to immediate surrounding, and since the neighbor patches are getting full, the individuals disperse farther, which makes high reproductive rate translating into an increase of the scale of effect.
Prediction 5: Relationship between the effect of effect and population outcome
We found that the spatial extent in which forest amount affects species abundance is narrower than the extent in which forest amount affects species occurrence. A landscape management plan applied at an appropriate extent for abundance will not ensure occurrence, and this finding has an important implication. If local populations go extinct by chance, it is likely that the area managed around them will be insufficient to ensure re-colonization, at least if the management area is determined using the abundance-based scale of effect.
Future studies could investigate the effect of single dispersal agents (for example birds, bats, invertebrates and large mammals) on the relationship between dispersal distance and the scale of effect of habitat amount on different biological responses. Another issue needing more attention is about the landscape variable tested and its relationship with the species trait measured, since habitat fragmentation seems to have a smaller scale of effect than habitat amount (Miguet et al., 2015). The effect of landscape structure on plant community depends on mobility-related traits, reproduction rate and the biological response. We must include dispersal distance and reproduction parameters to find a clear scale of effect for plant communities, and bear in mind that species abundance responds to a smaller scale than species occurrence. Our results did support the three hypothesis we formulated, but the study of spatial scale in which landscapes influence the species response still lacks more simulation and empirical tests (Miguet et al., 2015).
55
ACKNOWLEDGEMENTS
Ludmila Rattis is funded by Paulo Teixeira & Co, CAPES (Process 04100/2014-
00) and Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP scholarship
2012/02207-9). Rafael Loyola’s research has been constantly funded by Conselho Nacional de
Pesquisa (grants #304703/2011-7, 479959/2013-7, 407094/2013-0), Conservation International
Brazil, and the O Boticário Group Foundation for the Protection of Nature (PROG_0008_2013) and CNCFlora. Lenore Fahrig supported by Natural Sciences and Engineering Research
Council of Canada (NSERC) grant.
REFERENCES
Barton K. (2015) MuMIn: Multi-Model Inference R package version 1.10. 5. .
Bierzychudek P. (1982) Life histories and demography of shade-tolerant temperate forest herbs:
a review. The New Phytologist, 90, 757–776.
Boedeltje G., Bakker J.P., Bekker R.M., Van Groenendael J.M., & Soesbergen M. (2003) Plant
dispersal in a lowland stream in relation to occurence and three specific life-hystory traits
of the species in the species pool. Journal of Ecology, 91, 855–866.
Brennen J., Bender D., Contreras T., & Fahrig L. (2002) Focal patch landscape studies for
wildlife management: Optimizing sampling effort across scales. Integrating Landscape
Ecology into Natural Resource Management pp. 68–91. Cambridge University Press
Cambridge,
Brommit A.G., Charbonneau N., Contreras T.A., & Fahrig L. (2004) Crown Loss and
Subsequent Branch Sprouting of Forest Trees in Response to a Major Ice Storm. Journal
of the Torrey Botanical Society, 131, 169.
Brückmann S. V., Krauss J., & Steffan-Dewenter I. (2010) Butterfly and plant specialists suffer
from reduced connectivity in fragmented landscapes. Journal of Applied Ecology, 47, 56
799–809.
Fahrig L. (2013) Rethinking patch size and isolation effects: the habitat amount hypothesis.
Journal of Biogeography, 40, 1649–1663.
Feret P.P., Kreh R.E., Merkle S.A., & Oderwald R.G. (1982) Flower Abundance, Premature
Acorn Abscission, and Acorn Production in Quercus alba L. Botanical Gazette, 143, 216–
218.
Giraudoux P. (2015) pgirmess: Data Analysis in Ecology. R Package version 1.9. .
Holland J.D., Bert D.G., & Fahrig L. (2004) Determining the Spatial Scale of Species’
Response to Habitat. BioScience, 54, 227.
Hosie R.C. (1990) Native Trees of Canada. Fitzhenry and Whiteside, Markham, Ontario.
Jackson H.B. & Fahrig L. (2012a) What size is a biologically relevant landscape? Landscape
Ecology, 27, 929–941.
Jackson H.B. & Fahrig L. (2012b) What size is a biologically relevant landscape? Landscape
Ecology, 27, 929–941.
Jackson H.B. & Fahrig L. (2015a) Are ecologists conducting research at the optimal scale?
Global Ecology and Biogeography, 24, 52–63.
Jackson H.B. & Fahrig L. (2015b) Are ecologists conducting research at the optimal scale?
Global Ecology and Biogeography, 24, 52–63.
Jackson N.D. & Fahrig L. (2014) Landscape context affects genetic diversity at a much larger
spatial extent than population abundance. Ecology, 95, 871–881.
Kallio S.L. (2014) Relationship between species traits and landscape extent in ground beetles.
Carleton University, Ottawa, Ontario, Canada,
Leckie S., Vellend M., Bell G., Waterway M.J., & Lechowicz M.J. (2000) The seed bank in an
old-growth, temperate deciduous forest. Canadian Journal of Botany, 78, 181–192.
Legasy K., LaBelle-Beadman S., & Chambers B. (1995) Forest Plants of Northeastern Ontario. 57
Lone Pine Publishing, Edmonton.
Martin A.E. & Fahrig L. (2012) Measuring and selecting scales of effect for landscape
predictors in species-habitat models. Ecological Applications, 22, 2277–2292.
McEuen a B. & Curran L.M. (2004a) Seed dispersal and recruitment limitation across spatial
scales in temperate forest fragments. Ecology, 85, 507–518.
McEuen A.B. & Curran L.M. (2004b) Seed dispersal and recruitment limitation across spatial
scales in temperate forest fragments. Ecology, 85, 507–518.
Miguet P., Jackson H.B., Jackson N.D., Martin A.E., & Fahrig L. (2015) What determines the
spatial extent of landscape effects on species? Landscape Ecology, 1–18.
Murray B.R., Thrall P.H., Gill a. M., & Nicotra A.B. (2002) How plant life-history and
ecological traits relate to species rarity and commonness at varying spatial scales. Austral
Ecology, 27, 291–310.
Ozinga W.A., Schaminée J.H.J., Bekker R.M., Bonn S., Poschlod P., Tackenberg O., Bakker
J., & Groenendael J.M. van (2005) Predictability of plant species composition from
environmental conditions is constrained by dispersal limitation. Oikos, 108, 555–561.
Schupp E.W., Howe H.F., Augspurger C.K., & Levey D.J. (1989) Arrival and survival in
tropical treefall gaps. Ecology, 70, 562–564.
Skaug H., Fournier D., Nielsen A., Magnusson A., & Bolker B. (2012) Generalized Linear
Mixed Models using AD Model Builder. R package version 0.7. 2.12. R CRAN, .
Tittler R. (2008) Source sink dynamics, dispersal, and landscape effects on North America
songbirds. Carleton University, Ottawa, Ontario, Canada,
Waldron R.M. (1965) Cone Production and Seedfall in a Mature White Spruce Stand. 41, .
Whigham D.F. (2004) Ecology of Woodland Herbs in Temperate Deciduous Forests*. Annual
Review of Ecology, Evolution, and Systematics, 35, 583–621.
Wiens J.A., Stenseth N.C., Vanhorne B., & Ims R.A. (1993) Ecological mechanisms and 58
landscape ecology. Oikos, 66, 369–380.
Zuur A.F., Ieno E.N., Walker N., Saveliev A.A., & Smith G.M. (2009) Mixed effects models
and extensions in ecology with R. Springer, New York.
59
A method to evaluate habitat connectivity and carrying capacity at the scale
of species' range
Ludmila Rattis1,*; Ricardo Dobrovolski2; Maurício Talebi3; Rafael Loyola4
Authors’ affiliations: 1,4*Programa de Pós-Graduação em Ecologia, Universidade Estadual de Campinas, CP6109, 13083-970 Campinas, São Paulo, Brazil. Email: [email protected] 2Departamento de Zoologia, Universidade Federal da Bahia, Rua Barão de Geremoabo, 147, Campus Ondina, 40170-180 Salvador, Bahia, Brazil. 3Departamento de Ciências Biológicas, Universidade Federal de São Paulo, Rua Professor Artur Riedel, 275, Campus Diadema, 09972-270 São Paulo, Brazil 4Laboratório de Biogeografia da Conservação, Departamento de Ecologia, Universidade Federal de Goiás, CP 131, 74001-970 Goiânia, Goiás, Brazil.
60
ABSTRACT
We propose a method to evaluate habitat connectivity and carrying capacity of the whole species range based on land use maps, habitat preferences, home range size and movement ability. Within geographic range of species, we applied a deductive habitat suitability model and ecologically scaled landscape indices to assess the carrying capacity and functional connectivity and other two metrics that correlate habitat amount and fragmentation patterns within the landscape. We analyzed five species with different habitat requirements and movement patterns to exemplify our approach and quantified scale dependence of landscape patterns doing the same analyses in different landscape extents. Based on the deductive habitat suitability analysis, the total habitat amount can be 83% smaller than their extension of occurrence (EOO). Carrying capacity was more dependent on species’ features, whereas fragmentation and functional connectivity were more scale dependent, except when the species has high dispersal ability. Our approach can lead to lower risks of incurring in commission errors arising from landscape-scale underestimation of species’ occurrences and provide a new approach to assess species-habitat relationship. We highlight the importance of considering range-scaled landscape, landscape carrying capacity, and patch functional connectivity in a synthetic and integrated framework for conservation studies.
Key-words: Carrying capacity; Deductive Habitat Suitability Models; Functional
Connectivity; Landscape Structure; Movement; Scale Dependency.
61
INTRODUCTION
The consequences of land use changes on biodiversity are pervasive at different levels of ecological organization, and geographic scales. Existing studies range from works on metapopulation established in fragmented landscapes (Ovaskainen & Hanski, 2003; Burkey,
1997), to the investigation of landscape features that modulate the species-area relationship
(Benchimol & Peres, 2013; Hanski et al., 2013; Rybicki & Hanski, 2013), and the perceived effects of habitat loss on biodiversity at regional and even broader geographic scales (Brooks et al., 2002; Gaston et al., 2003).
It is known that species’ response to habitat change depends primarily on their ability to deal with environmental dynamics (Tilman et al., 1994; Dellinger et al., 2013). At the local scale – the usual scale in which conservation actions are implemented – habitat suitability models have been suggested as a tool to refine information on species distribution and help guiding conservation assessment and decision more precisely (da Fonseca et al., 2000; Ottaviani et al., 2004; Rondinini et al., 2011). Habitat suitability models are important tools to evaluate species distribution based on their potential habitat remnants. They rely on Hutchinson’s concept of ecological niche (Hirzel & Arlettaz, 2003) and are designed to predict species’ occurrence based on environmental data and species habitat preferences. Among several possible applications, these models have been used to assess species recovery (Cianfrani et al.,
2013a), to estimate extinction debt when coupled with species-area relationships (Olivier et al.,
2013), to map the potential distribution of invasive species (Crall et al., 2013), and to evaluate global patterns of species richness (Rondinini et al., 2011).
Habitat suitability models lack information about the interaction between the species features and spatial structure of habitat patches. Studying so, landscape ecologists have developed a set of methods to assess the effects of landscape structure on species responses
(McGarigal & Cushman, 2005). Methods employed to study the effect of patch size and 62
isolation (for instance, the landscape metrics) on species outcomes usually (1) lack information on species’ ecological features and their ecological meaning, (2) do not consider species- specific responses to landscape change, (3) are not able to transpose the results obtained from one landscape to another, or (4) require a large amount of data. In this regard, Vos et al. (2001) proposed two indices linking species to habitat; both called Ecologically Scaled Landscape
Indices: ESLIc and ESLIk. ESLIc associates patch area, pairwise distance between patches, and species’ movement ability resulting in an index that reflects landscape functional connectivity.
ESLIk associates patch area and individual area requirement, resulting in an index that describes the landscape carrying capacity for a given metapopulation.
We employed a well-known, yet seldom used concept that classifies species according to their response to environment, specifically to habitat fragmentation and habitat amount sensitivity: the ecological profile. Grimm et al. (1996) developed this concept to incorporate different colonization responses to environment when empirical data is lacking. The strength of this approach is the employment of this concept in answering general questions about how a given group of organisms with some intrinsic features can respond to differences in the composition and configuration of natural environments. For instance, species with similar dispersal ability and similar total areas required to reproduce, even belonging to different taxa, could have the responses to landscape structure evaluated together.
Here we propose a method to evaluate connectivity and carrying capacity at species range scale, coupling deductive habitat suitability model and landscape connectivity and carrying capacity through ESLIs for five species that represent different ecological profiles. Our approach has four improvements over others because 1) it includes a realistic and intuitive definition of habitat; 2) it is a more factual way to know how much habitat is there based not only on remaining area, but also on species’ habitat preferences; 3) it quantifies how much habitat would be potentially occupied by a metapopulation, given patches’ carrying capacity 63
that, would be connected by individual moving ability, 4) it allows one to perform a validation test for the obtained models. Using this approach, we were able to compare (i) habitat structure in local landscape and in entire species range and (ii) define how much area is suitable and functionally connected for the species. To exemplify our approach, we applied it to five species distributed in different places around the world. Moreover, we quantified the scale dependence of landscape studies running the same analyses in different landscape extents.
METHODS
Case studies
We chose five species belonging to four different ecological profiles (sensu Vos et al. 2001) distributed around the world (Fig. 1). The first one was the Spanish Imperial Eagle
Aquila adalberti (Falconiformes). This is a species with highly sensible to habitat fragmentation given its large individual area requirements and short dispersal ability. Other factors can increase its extinction risk, like sedentary, low reproductive rates (an average of 0.75 offspring per reproductive unit), a relative delay in the first reproduction (reproductive age at 4 or 5 years old), and high mortality caused by electrocution and poisoning, the latest resulting in a decreasing fecundity of adults (Ferrer et al., 2003, 2013a). In fact, according to the IUCN Red
List of Threatened Species (IUCN, 2015), A. adalberti is vulnerable to extinction (VU), though its population is increasing due to some conservation initiatives (Ferrer et al., 2013b).
The second species was the South American Muriqui Brachyteles arachnoides
(Primates, Atelidae). Some features like small individual area requirement and intermediary dispersal ability would put this species at a low extinction risk level. However, according to the
IUCN Red List, B. arachnoides is assigned to the endangered category (EN). The major threat to B. arachnoides is the residential and commercial development, agriculture and aquaculture especially annual and perennial non-timber crops (Mendes et al., 2008). 64
The third species, the Malagasy Blue-eyed black lemur Eulemur flavifrons
(Primates, Lemuridae), also has small individual area requirement and intermediary dispersal ability. The major threat to E. flavifrons is forest conversion by slash-and-burn agriculture and hunting (Andrainarivo et al., 2011). Eulemur flavifrons living in disturbed habitat is usually under stress, as shown by parasitological analysis of population habiting in areas with different levels of degradation (Schwitzer et al., 2010a). This species is currently considered as critically endangered (CR) by the IUCN Red List.
The fourth species was the North American Gila Monster Heloderma suspectum
(Squamata, Helodermatidae). This species has survival strategies that combines timing and duration of activity, resource storage, economical use of this resources and high tolerance to physiological disturbances (Davis & Denardo, 2010). These features could make this species more robust to habitat disturbances. According to the IUCN Red List, this species is considered near threatened (NT).
Our fifth species was the Tasmanian Devil Sarcophilus harrisii (Marsupialia,
Dasyuridae). This species presents large dispersal distance and large individual area requirement regardless of its habitat generalism. This species is victim of a disease that has rapidly annihilated local populations, the devil facial tumor disease parasitic cancer. According to the IUCN Red List, this species belongs to the endangered category (EN).
65
Figure 1. Five case studies belonging to four ecological profiles sensu Vos et al (2001). Species were sorted by habitat amount sensitivity and habitat fragmentation sensitivity based on the amount of area required for a reproductive unit (individual area requirement) and average ability to move per day (daily path length). In the upper left Heloderma suspectum, upper right Aquila adalberti, in the left middle Eulemur flavifrons, in the right middle Brachyteles arachnoides, and in the lower right Sarcophilus harrisii. Species with short daily path length are more sensible to fragmentation, as well species with large area required for reproductive units are more sensible to habitat amount.
Deductive habitat suitability models
We built deductive habitat suitability models based on two spatial variables: land use and elevation. We extracted land use information from the Glob Cover 2.1 , a global land use map with 0.0028° x 0.0028° lat x long resolution (ca. 300m at the Equator) (IONIA 2009).
We obtained data on elevation from the elevation map Shuttle Radar Topography Mission
(SRTM) originally with 90 m resolution (US Geological Survey, 2006) and resampled to
0.0028°. We also considered three intrinsic features of species: species’ range (maps obtained from IUCN 2012), habitat use and elevation preferences, the latter two obtained from literature
(Appendix S1). Based on species’ preferences we clipped all types of land use and elevation range in which species do not occur from the original geographic range of species, a process known as habitat filtering (Fig. 2; see also Faleiro et al., 2013).
To validate our distribution mapping (i.e. level of omission and commission errors) we ran a randomization test using point-locality data acquired for each species (i.e. local species records). We obtained species’ records from the Global Biodiversity Information Facility
(GBIF, available at http://gbif.net) for three species (A. adalberti, H. suspectum and S. harrisii), and from two unpublished databases: one from Maurício Talebi, for B. arachnoides, and the other from Réseau de la Biodiversité de Madagascar (REBIOMA), for E. flavifrons.
We overlaid species’ records to a raster map of the geographic distribution of species in which grid cells were classified as suitable or unsuitable according to our model. We 66
considered only one record for each species matching a grid cell, even when more than one record for the same species matched the grid cell. When records were inside a suitable grid cell, we consider a positive match. We defined the proportion of positive matches in relation to all available species’ records as our observed value. We randomized species’ records within the species’ range 999 times. After this procedure, we calculated the proportion of randomized records that have matched cells with suitable habitat. We defined P-value for our analysis as the number of times in which the proportion of positive matches was equal or higher than the observed one.
Remaining habitat, functional connectivity and carrying capacity of landscapes
We randomly selected six points within the geographic range from which we draw smaller landscapes in order to evaluate the difference between the proposed scale in this study
- a range-scaled landscape - and scales usually employed in landscape studies: three landscapes of 5km radius and three landscapes of 10km radius - which together with the proposed scale totaled seven landscapes at three different scales. For each landscape, we assessed total remaining habitat, habitat fragmentation, functional connectivity, and carrying capacity based on three variables and two species features. We evaluated total amount of habitat and the percentage of habitat cover that corresponds to the largest patch - relative largest patch size – rLPS (Rueda et al., 2013a). We calculated landscape metrics using Fragstats (McGarigal et al.,
2012). We also considered two species features: home-range size of one reproductive unit (in km2), and movement ability (daily length path – dlp, in meters). We used the reproductive unit area to define the patches that can be occupied, thus they must have area equal or larger than it.
We used dlp as our dispersal measure because it has a clear ecological meaning in the processes that connects local and regional populations and allows for comparison between species because it considers time. 67
We applied two ESLIs (Vos et al., 2001) to assess the functional connectivity and carrying capacity of the landscape. ESLIk uses data on individual area requirement (IAR). We calculated IAR for A. adalberti based on the home-range size of a breeding couple (Ferrer, 1993;
Ferrer et al., 2004; González et al., 2006). For B. arachnoides (Milton, 1984; Strier, 1987) and
H. suspectum (Jones, 1983; Beck, 1990) IAR was based on female home-range size because for these species, females establish territories while males move among flocks. For S. harrisii we calculated IAR based on female home range because male and female do not take care of offspring together and, therefore, females live alone (Guiler, 1970; Pemberton, 1990). Finally, we calculated IAR for E. flavifrons based on flock home range size (Schwitzer et al., 2010b;
Volampeno et al., 2011). For more details on our dataset, see Appendix S1 (Table S2).
The ESLI that measures landscape carrying capacity, ksi, was calculated as follow:
where ai is the patch area of each habitat patch i and IARsi is the individual area requirement of the species s in a given patch i. From (1), we calculated the proportion of patches within the landscape that supports more than one reproductive unit, i.e. the potentially occupied patches. To enable comparison among all five species we used average patch carrying capacity as follow:
where n is the total number of patches. To calculate landscape functional connectivity we used two metrics, dlps and Euclidian Nearest Neighbor distance (dij)
68
when csij is below 1, the patch is not functionally connected. In this case, the individual is not able to reach the nearest patch due to its dispersal limitation. The second metric can be described as
Equation (4) considers the focal patch area, Ai, the Euclidian distance between all patches within the landscape in relation to focal patch, Cij. The dispersal kernel (α) is species- specific and it was set based on the dlp, that has a value that yields close to 0 contributions at distances beyond the maximum observed dlp, under a negative exponential distribution. Species with the lowest dlp has the highest α (see details in Appendix S1, Table S1). To enable comparison among all species we used average patch connectivity as follow:
where n is the total number of patches in the landscape.
69
RESULTS
Habitat suitability models
Total habitat amount based on habitat suitability analysis varied among different scales. The average habitat amount decreased 39% (min=12%, max=83%) when we applied deductive habitat suitability models to the range-scaled landscape (Fig. 2, Table 1). The average habitat amount at 10-km landscapes decreased 46% (min=16%, max=72%), whereas at 5-km landscapes it decreased 65% on average (min=25%, max=96%). To four species, our models were successful in predicting species’ occurrence, as shown by model validation: A. adalberti
(193 species’ records, p<0.001); B. arachnoides (106 records, p<0.001); E. flavifrons (11 records, p<0.001), and H. suspectum (185 records, p=0.021). However, for S. harrisii, a species inhabiting open areas and with high dlp, our model was not robust (326 records, p=0.99).
Figure 2. Deductive habitat suitability models for five species around the world. For each species we showed the geographic range and the suitable habitat after we applied the deductive 70
habitat suitability model. From the upper left, in clockwise, Heloderma suspectum, Aquila adalberti, Eulemur flavrifrons, Sarcophilus harrisii and Brachyteles arachnoides.
Habitat carrying capacity
Habitat patches had small carrying capacity for species with high sensitivity to habitat amount, always below 34% at any given scale (Fig.3 and Fig. 4). For species with intermediate or low sensitivity to total habitat amount, landscape carrying capacity were always high, even where habitat amount was below 40% (Fig.3 and Fig. 4). Carrying capacity almost did not change across the scales.
Habitat connectivity
Functional connectivity was high for all species at the range-scaled landscape, and at 10-km landscapes for S. harrisii, the species with highest dlp; and for H. suspectum, the species found in landscapes with highest habitat amount and lowest fragmentation (Fig. 4).
Functional connectivity was low in 5-km landscapes for H. suspectum and 10-km landscapes for both B. arachnoides and E. flavifrons, even with large habitat amount and low fragmentation
(Fig. 4). This is because those species have low dispersal abilities to the current fragmentation degree in their area of occurrence. There was also an effect of scale dependence of functional connectivity in landscapes of A. adalberti in which all studied scales had high fragmentation, but for 5-km and 10-km landscapes, functional connectivity was low. Connectivity is scale dependent as showed in Figure 4: as the spatial extent decreases, the functional connectivity exponentially dropped.
71
Figure 3. Species combination results of ecologically scaled landscape indices (log- transformed) for average patch carrying capacity and average patch connectivity using equations (1) and (3). Each point on the graph represents a species combination. The first number is the percentage of habitat amount and the second number is the fragmentation degree described by relative Largest Patch Size (rLPS). AA: Aquila adalberti; BA: Brachyteles arachnoides; EF: Eulemur flavifrons; HS: Heloderma suspectum and SH: Sarcophilus harrisii. The silhouette positions are the same as the Fig. 1 and are showed just to enable a direct comparision.
72
73
Figure 4: The results of the fragmentation landscape metric (rLPS) and two ecologically scaled landscape indices (ESLIc and ESLIk) for the five case studies. Fragmentation and functional connectivity changed according to the scale. Fragmentation is higher at 10-km landscapes and lower at 5-km and range-scaled landscapes. Functional connectivity dropped down as the spatial extent decreased. Carrying capacity is not sensible to spatial extent, being constant across scales.
Table 1. Total area of extension of occurrence (EOO), number of patches (NP) and proportion of area remaining from deductive habitat suitability model and two Ecologically Scaled Landscape Indices for three scales: range scaled (RANGE), 5-km and 10-km landscapes.
EOO Total habitat suitability area % patches functionally Species % patches k≥1 (km2) (km2) connected and k≥1
10-km 5-km 10-km 5-km 10-km 5-km RANGE RANGE RANGE (mean) (mean) (sum) (sum) (sum) (sum)
38.7%; 46.8%; 4.3%; 33.9%; 25.7%; 0.0%; 0.0%; 0%; 0%; A. adalberti 142,030 NP=7399 NP=29 NP=8 NP=31 NP=1 NP=0 NP=0 NP=0 NP=0
B. 11.9%; 61.2%; 33.6%; 11.9%; 61.2%; 33.6%; 11.5%; 49.1%; 32.3%; 292,824 arachnoides NP=4877 NP=16 NP=2 NP=4877 NP=16 NP=2 NP=2315 NP=9 NP=1
28.8%; 57.9%; 36.8%; 28.8%; 57.9%; 35.9%; 27.6%; 56.5%; 34.5%; E. flavifrons 7,923 NP=309 NP=34 NP=6 NP=309 NP=34 NP=4 NP=139 NP=18 NP=2
83.4%; 93.6%; 81.4%; 83.4%; 93.6%; 81.4%; 83.4%; 33.7%; 0%; H. suspectum 368,912 NP=3634 NP=4 NP=3 NP=3634 NP=4 NP=3 NP=3634 NP=1 NP=0
29.6%; 34.8%; 32.1%; 21.5%; 24.9%; 7.6%; 21.5%; 24.9%; 7.6%; S. harrisii 81,512 NP=8022 NP=54 NP=14 NP=178 NP=1 NP=1 NP=178 NP=1 NP=3
DISCUSSION
Habitat spatial structure affects ecological processes like species extinction, and the extinction involve the whole population network. That is why population persistence should be evaluated at the geographic range scale, the area where the network populations of a particular species inhabit. The understanding of the effects of habitat changes on the species-habitat relationship should involve the application of methods that consider 1) aspects of the life history 74
of the species and 2) the patterns of size and isolation of habitat patches and 3) that are able to synthesize all this information, as the applied ESLI's. That because we might evaluate the fine scale patterns - like patch size and the area required for reproduction of a reproductive unit, which determines landscape carrying capacity at different scales - and also how dispersal dynamics and patch isolation affect functional connectivity - to understand the broad scale phenomena (Gouveia et al., 2016).
Here we showed that three aspects must be considered in evaluating species-habitat relationship: habitat amount, landscape configuration, and scale dependence. We also found that landscape habitat amount and carrying capacity are more dependent on species’ features, whereas fragmentation and functional connectivity are more scale dependent, except when species have great dispersal ability.
First of all, we need a realistic definition and applicable data about the habitat concept (Didham et al., 2012; Fahrig, 2013; Rueda et al., 2013b) and landscape scale. The scale in which conservation planning is usually done must consider the focal patch and its neighborhood as well other patches that act as sources of individuals. Since metapopulation is a complex network of habitat patches, landscape supplementation, must be studied in a more broadly scale, under a holistic point of view. A holistic analysis of the environment and loc al conservation initiatives need to rely on approaches that are able to deal with high amounts of data and that can be easily applied. Habitat suitability modeling is an approach increasingly used for different purposes, including in connectivity assessment (Cianfrani et al., 2013b).
Here, we were able to diagnose the status of different species’ geographic distribution, getting closer to the actual area of occupancy (AOO) of species and landscape supplementation. Habitat suitability models based on a free global land use map allowed us to assess the suitable areas within specie’s range, being a starting point to analyze species-habitat relationship. It was a more factual way to know how much and where are the habitat based not 75
only on the area, but also on species’ habitat preferences. This study helps in the recent discussion about the downscaling of biodiversity distribution data, based not only on GIS information, but also on intrinsic features of the species, which have more ecological meaning, confirming what was raised by Jenkins et al. (2013).
Another important issue is patch size and isolation. We propose that landscape metrics employed to study size and isolation of habitat patches should consider not only landscape features, but also intrinsic biological features of species, given that some metrics fail to account for scale-dependent variation within species due to the lack of ecological significance. The metrics proposed by Vos et al. (2001) is an accordant initial step in that way since they are easy to interpret and have an understandable ecological meaning (Swihart &
Verboom, 2004). In this study, we showed that the habitat amount and fragmentation effects on landscape carrying capacity and functional connectivity can vary greatly, and even species occurring in landscapes with less than 20% of habitat amount and fragmentation degree below
50% can experience high functional connectivity. These results lead us to think about the importance of the differences in sensitivity of each ecological profile, those respond in a similar way to landscape patterns, irrespective to the ecological processes (fragmentation and habitat amount).
We also considered the scale dependency of landscape studies. Didham et al. (2012) argued that there are differences in responses to fragmentation due to differences between sets of species traits, indeed generating similar species responses in groups of related species. Here we show that species with different habitat preferences, dispersal abilities and sensitivity to habitat amount and fragmentation have differences in responses until a given point where these species could be grouped in ecological profiles. In the same way, Rueda et al. (2013) opened a venue for investigation about ‘what extent are species’ responses to fragmentation’. Here we showed that as the extent increase, the carrying capacity remains species dependent and the 76
functional connectivity becomes scale dependent, and at range scaled landscapes it is high, even in high fragmentation degree.
We must highlight some caveats of our approach. First, species with dispersal ability under the spatial resolution of our analyses cannot have their functional connectivity assessed by the ratio between dispersal distance and distance between patches, but would be by the ESLIc. Still, species that have small home ranges, like some amphibians, are not sensible to
ESLIk or any landscape metric because their home range does not go beyond the grid cell scale.
Generalist species like S. harrisii are also not sensible to the analysis, since the habitat preferences are too broad. Another limitation is to deal with more than one species in a given region. If the geographic range can be used as the landscape scale, how we can deal with more than one non-coincident geographic range? We suggest that the first step to deal with this kind of problem is to take into account (1) the proportion of geographic range of each species will be evaluated in the given region, (2) how much habitat is enough to maintain viable populations in patches with minimum carrying capacity and functionally connected, and (3) where is the area most vulnerable to habitat loss (Fahrig, 2003).
Our approach can lead to lower risks of incurring in commission errors arising from landscape-scale underestimation of species’ occurrences and provide a new approach to assess species-habitat relationship. We highlight the importance of considering range-scaled landscape, landscape carrying capacity, and patch functional connectivity in a synthetic and integrated framework for conservation studies.
ACKNOWLEDGEMENTS
We are grateful to Paulo Brando and Lenore Fahrig for very helpful input. We thank
Dr. Andriamandimbisoa Razafimpahanana and the REBIOMA Team for providing Eulemur flavifrons point locality data. Ludmila Rattis is funded by Paulo Teixeira & Co and Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP scholarship 2012/02207-9). Ricardo 77
Dobrovolski work is funded by CAPES, CNPq (Process 461665/2014-0) and PRODOC/UFBA
(Process 5849/2013). Rafael Loyola’s research has been constantly funded by Conselho
Nacional de Pesquisa (grants #304703/2011-7, 479959/2013-7, 407094/2013-0), Conservation
International Brazil, and the O Boticário Group Foundation for the Protection of Nature
(PROG_0008_2013).
REFERENCES
Andrainarivo C., Andriaholinirina V.N., Feistner A., Felix T., Ganzhorn J., Garbutt N., Golden
C., Konstant B., Louis Jr. E., Meyers D., Mittermeier R.A., Perieras A., Princee F.,
Rabarivola J.C., Rakotosamimanana B., Rasamimanana H., Ratsimbazafy J.,
Raveloarinoro G., Razafimanantsoa A., Rumpler Y., Schwitzer C., Sussman R.,
Thalmann U., Wilmé L., & Wright P. (2011) Eulemur flavifrons. IUCN Red List of
Threatened Species. IUCN 2013. Version 2013.1., .
Beck D.D. (1990) Ecology and Behavior of the Gila Monster in Southwestern Utah. Journal of
Herpetology, 24, 54–68.
Benchimol M. & Peres C.A. (2013) Anthropogenic modulators of species-area relationships in
Neotropical primates: a continental-scale analysis of fragmented forest landscapes.
Diversity and Distributions, 19, 1339–1352.
Brooks T.M., Mittermeier R.A., Mittermeier C.G., Fonseca G.A.B., Rylands A.B., Konstant
W.R., Flick P., Pilgrim J., Oldfield S., Magin G., & Hilton-Taylor C. (2002) Habitat loss
and extinction in the hotspots of biodiversity. Conservation Biology, 16, 909–923.
Burkey T.V. (1997) Metapopulation extinction in fragmented landscapes: using bacteria and
protozoa communities as model ecosystems. The American Naturalist, 150, 568–91.
Cianfrani C., Maiorano L., Loy A., Kranz A., Lehmann A., Maggini R., & Guisan A. (2013a)
There and back again? Combining habitat suitability modelling and connectivity analyses 78
to assess a potential return of the otter to Switzerland. Animal Conservation, 16, 584–
594.
Cianfrani C., Maiorano L., Loy A., Kranz A., Lehmann A., Maggini R., & Guisan A. (2013b)
There and back again? Combining habitat suitability modelling and connectivity analyses
to assess a potential return of the otter to Switzerland. Animal Conservation, 16, 584–
594.
Crall A.W., Jarnevich C.S., Panke B., Young N., Renz M., & Morisette J. (2013) Using habitat
suitability models to target invasive plant species surveys. Ecological Applications, 23,
60–72.
Davis J.R. & Denardo D.F. (2010) Seasonal Patterns of Body Condition, Hydration State, and
Activity of Gila Monsters (Heloderma suspectum) at a Sonoran Desert. Journal of
Herpetology, 44, 83–93.
Dellinger J.A., Proctor C., Steury T.D., Kelly M.J., & Vaughan M.R. (2013) Habitat selection
of a large carnivore, the red wolf, in a human-altered landscape. Biological Conservation,
157, 324–330.
Didham R.K., Kapos V., & Ewers R.M. (2012) Rethinking the conceptual foundations of
habitat fragmentation research. Oikos, 121, 161–170.
Fahrig L. (2003) Effects of habitat fragmentation on biodiversity. Annual Review of Ecology,
Evolution, and Systematics, 34, 487–515.
Fahrig L. (2013) Rethinking patch size and isolation effects: the habitat amount hypothesis.
Journal of Biogeography, 40, 1649–1663.
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.
Ferrer M. (1993) Ontogeny of dispersal distances in young Spanish imperial eagles. Behavioral
Ecology and Sociobiology, 32, 259–263. 79
Ferrer M., Newton I., & Muriel R. (2013a) Rescue of a small declining population of Spanish
imperial eagles. Biological Conservation, 159, 32–36.
Ferrer M., Newton I., & Muriel R. (2013b) Rescue of a small declining population of Spanish
imperial eagles. Biological Conservation, 159, 32–36.
Ferrer M., Otarola F., & García-Ruiz J.M. (2004) Density-dependent age of first reproduction
as a buffer affecting persistence of small populations. Ecological Applications, 14, 616–
624.
Ferrer M., Penteriani V., Balbontı́n J., & Pandolfi M. (2003) The proportion of immature
breeders as a reliable early warning signal of population decline: evidence from the
Spanish imperial eagle in Doñana. Biological Conservation, 114, 463–466. da Fonseca G.A., Balmford A., Bibby C., Boitani L., Corsi F., Brooks T., Gascon C., Olivieri
S., & Mittermeier R. (2000) Following Africa’s lead in setting priorities. Nature, 405,
393–4.
Gaston K.J., Blackburn T.M., & Goldewijk K.K. (2003) Habitat conversion and global avian
biodiversity loss. Proceedings of the Royal Society, 270, 1293–300.
González L.M., Margalida A., Sánchez R., & Oria J. (2006) Supplementary feeding as an
effective tool for improving breeding success in the Spanish imperial eagle (Aquila
adalberti). Biological Conservation, 129, 477–486.
Gouveia S.F., Souza-Alves J.P., Rattis L., Dobrovolski R., Jerusalinsky L., Beltrão-Mendes R.,
& Ferrari S.F. (2016) Climate and land use changes will degrade the configuration of the
landscape for titi monkeys in eastern Brazil. Global Change Biology, -, -.
Grimm V., Franka K., Jeltsch F., Brandla R., Uchmariskib J., & Wissela C. (1996) Pattern-
oriented modelling in population ecology. 183, 151–166.
Guiler E.R. (1970) Observations on the Tasmanian devil, Sarcophilus harrisii
(Marsupialia:Dasyuridae). Australian Journal of Zoology, 18, 49–62. 80
Hanski I. & Ovaskainen O. (2000) The metapopulation capacity of a fragmented landscape.
Nature, 404, 755–8.
Hanski I., Zurita G. a, Bellocq M.I., & Rybicki J. (2013) Species-fragmented area relationship.
Proceedings of the National Academy of Sciences of the United States of America, 110,
12715–20.
Hirzel A.H. & Arlettaz R. (2003) Modeling habitat suitability for complex species distributions
by environmental-distance geometric mean. Environmental Management, 32, 614–23.
IONIA (2009) Available at: http://ionia1.esrin.esa.int.
IUCN (2012) Available at: http://www.iucnredlist.org/.
IUCN (2015) Available at: http://www.iucnredlist.org.
Jenkins C.N., Pimm S.L., & Joppa L.N. (2013) Global patterns of terrestrial vertebrate diversity
and conservation. Proceedings of the National Academy of Sciences of the United States
of America, 110, E2602–10.
Jones K.B. (1983) Movement Patterns and Foraging Ecology of Gila Monsters (Heloderma
suspectum, Cope) in Northwestern Arizona. Herpetologica, 39, 247–253.
McGarigal K. & Cushman S.A. (2005) The gradient concept of landscape structure. Issues and
Perspectives in Landscape Ecology pp. 112–119. Cambridge Univ Pr, Cambridge.
McGarigal K., Cushman S.A., & Ene E. (2012) FRAGSTATS v4: Spatial Pattern Analysis
Program for Categorical and Continuous Maps. .
Mendes S.L., de Oliveira M.M., Mittermeier R.A., & Rylands A.B. (2008)
Milton K. (1984) Habitat, diet, and activity patterns of free-ranging woolly spider monkeys
(Brachyteles arachnoides E. Geoffroy 1806). International Journal of Primatology, 5,
491–514. 81
Olivier P.I., van Aarde R.J., & Lombard A.T. (2013) The use of habitat suitability models and
species-area relationships to predict extinction debts in coastal forests, South Africa.
Diversity and Distributions, 19, 1353–1365.
Ottaviani D., Lasinio G.J., & Boitani L. (2004) Two statistical methods to validate habitat
suitability models using presence-only data. Ecological Modelling, 179, 417–443.
Ovaskainen O. & Hanski I. (2003) How much does an individual habitat fragment contribute
to metapopulation dynamics and persistence? Theoretical Population Biology, 64, 481–
495.
Pemberton D. (1990) Social organization and behaviour of the Tasmanian devil, Sarcophilus
harisii. University of Tasmania,
Rondinini C., Boitani L., Rodrigues A.S.L., Brooks T.M., Pressey R.L., Visconti P., Baillie
J.E.M., Baisero D., Cabeza M., Crooks K.R., Di Marco M., Redford K.H., Andelman S.
a, Hoffmann M., Maiorano L., Stuart S.N., & Wilson K. a (2011) Reconciling global
mammal prioritization schemes into a strategy. Philosophical transactions of the Royal
Society of London. Series B, Biological sciences, 366, 2722–8.
Rueda M., Hawkins B. a., Morales-Castilla I., Vidanes R.M., Ferrero M., & Rodríguez M.Á.
(2013a) Does fragmentation increase extinction thresholds? A European-wide test with
seven forest birds. Global Ecology and Biogeography, 22, 1282–1292.
Rueda M., Hawkins B. a., Morales-Castilla I., Vidanes R.M., Ferrero M., & Rodríguez M.Á.
(2013b) Does fragmentation increase extinction thresholds? A European-wide test with
seven forest birds. Global Ecology and Biogeography, 22, 1282–1292.
Rybicki J. & Hanski I. (2013) Species-area relationships and extinctions caused by habitat loss
and fragmentation. Ecology Letters, 16 Suppl 1, 27–38. 82
Schwitzer N., Clough D., Zahner H., Kaumanns W., Kappeler P., & Schwitzer C. (2010a)
Parasite prevalence in blue-eyed black lemurs Eulemur flavifrons in differently degraded
forest fragments. Endangered Species Research, 12, 215–225.
Schwitzer N., Clough D., Zahner H., Kaumanns W., Kappeler P., & Schwitzer C. (2010b)
Parasite prevalence in blue-eyed black lemurs Eulemur flavifrons in differently degraded
forest fragments. Endangered Species Research, 12, 215–225.
Strier K.B. (1987) Ranging behavior of woolly spider monkeys, or muriquis,Brachyteles
arachnoides. International Journal of Primatology, 8, 575–591.
Survey U.S.T. (2006) Available at: http://www.landcover.org/data/srtm/.
Swihart R.K. & Verboom J. (2004) Using Ecologically Scaled Landscape Indices to assess
biodiversity consequences of land-use decisions. Conserving biodiversity in agricultural
landscapes (ed. by R.K. Swihart and J.E. Moore), pp. 336. Purdue University Press,
Tilman D., May R.M., Lehman C.L., & Nowak M.A. (1994) Habitat destruction and the
extinction debt. Nature, 371, 65–66.
Volampeno M.S.N., Masters J.C., & Downs C.T. (2011) Home range size in the blue-eyed
black lemur (Eulemur flavifrons): A comparison between dry and wet seasons.
Mammalian Biology - Zeitschrift für Säugetierkunde, 76, 157–164.
Vos C.C., Verboom J., Opdam P.F., & Ter Braak C.J. (2001) Toward ecologically scaled
landscape indices. The American Naturalist, 157, 24–41.
83
Conclusão geral
Esta tese traz contribuições para o entendimento da relação entre o hábitat e as espécies em três diferentes temas: 1) o cuidado que devemos ter ao definir o que é hábitat; 2) o aspecto avaliado e o objetivo ajudarão a definir a extensão do estudo; 3) considerações sobre as variáveis utilizadas nos modelos, tanto variáveis resposta quanto explicativas.
O cuidado que devemos ter ao definir o que é hábitat
Características intrínsecas das espécies podem ser usadas como um primeiro passo para a definição de hábitat. Nos capítulos um e dois, a característica utilizada foi a tolerância à sombra. No capítulo três foi o tipo de cobertura vegetal preferencial para cada espécie, conforme descrito pela literatura especializada. Com essa informação, é possível selecionar
áreas e mapear os diferentes usos da terra e fazer uma aproximação mais rigorosa do que cada espécie percebe como seu hábitat. Contudo, o exercício proposto poderia nos levar a uma definição de hábitat sem limites geográficos e sabemos que tal limite existe. O que nos leva à segunda contribuição trazida por essa tese: se desejamos entender a relação entre o hábitat e as espécies é preciso que o aspecto avaliado e o objetivo do estudo estejam claros a priori.
O aspecto avaliado e o objetivo ajudarão a definir a extensão do estudo
No capítulo dois, que trata de paisagens locais, mostramos como a escala de efeito
(i.e. a extensão geográfica em que o hábitat influencia os indivíduos de determinada espécie) pode ser determinada e o que pode influenciar a extensão dessa escala. No capítulo três, que trata da área de distribuição da espécie e sua conservação, propomos que aspectos avaliados localmente, como conectividade, capacidade de suporte e nível de fragmentação, sejam avaliados em toda a área de distribuição, posto que a conservação da espécie, ou do conjunto de espécies que ocupa dada área, dependerá da rede de áreas adequadas, de sua coesão e da 84
capacidade dos indivíduos de colonizar toda sua extensão de ocorrência. O estudo da relação entre as espécies e o componente do hábitat de interesse deve ser conduzido na escala em que o fenômeno a ser investigado é relevante. No capítulo dois, vimos que a relação habitat x espécie é mais forte a uma dada extensão, e que a extensão depende de vários fatores, como a mobilidade, a taxa reprodutiva dos organismos e a resposta biológica estudada. Tais resultados têm importantes implicações para o delineamento de estudos futuros e de planos de restauração.
Espécies vegetais que possuem sementes mais pesadas ou com menor taxa reprodutiva serão influenciadas pelas características da paisagem mais localmente do que espécies vegetais com sementes mais leves e com maiores taxas reprodutivas. Para a assembleia estudada, estudos que analisem a paisagem local em extensões maiores ou menores que 100m – 200m podem não capturar os efeitos das mudanças de hábitat em tais espécies. Planos de restauração devem considerar a extensão da dispersão de sementes a partir da área fonte de propágulos.
Considerações sobre as variáveis utilizadas nos modelos
A terceira contribuição dessa tese diz respeitos às variáveis utilizadas nos modelos, tanto variáveis resposta quanto explicativas. No capítulo um, mostramos que a quantidade de hábitat explicou mais a variação na riqueza das espécies do que a área e o isolamento do fragmento, o que sugere que a substituição dessas duas variáveis por uma medida simples de quantidade de hábitat pode ser suficiente para estudarmos e prevermos a variação nos padrões de riqueza de espécies em paisagens locais. Com a proposição da hipótese da quantidade de hábitat (Fahrig 2013) e posteriores testes como os que fizemos aqui, há uma quebra do paradigma sobre tamanho e isolamento dos fragmentos, que são tratados como medidas de configuração do hábitat, quando são mais próximos de medidas de composição. A extrapolação da Teoria de Biogeografia de Ilhas (MacArthur, R H Wilson, 1967) para fragmentos de hábitat feita por Levins (1970) fez com que, desde então as pesquisas sobre os efeitos da distribuição 85
do hábitat incluíssem a área e o isolamento dos fragmentos como preditores de riqueza, transpondo-se os processos que ocorrem em ambientes insulares para ambientes continentais.
A teoria serviu ainda como base para propostas de desenho de reservas naturais, baseando-se no fato de que em áreas maiores, as probabilidades de extinção das espécies que nela habitam seriam menores (Diamond, 1975). Mas qual seria a distribuição ideal dos remanescentes de hábitat para preservar mais espécies? Vários fragmentos pequenos ou um fragmento grande equivalente em área? (Diamond, 1975; Diamond et al., 1976). Críticas a essa proposta se seguiram (Simberloff & Abele, 1976; Connor & McCoy, 1979; Zimmerman & Bierregaard,
1986) e hoje sabe-se que tanto a teoria de equilíbrio entre ilhas e a relação espécie-área – o aumento do número de espécies em função da área que ocupam – não podem ser utilizadas como base ecológica para a escolha do desenho de áreas protegidas (Zimmerman &
Bierregaard, 1986). A transposição direta da teoria de equilíbrio para ambientes continentais, ignora fatores relacionados às barreiras de dispersão e efeitos de borda que são tão distintos em ambientes insulares quando comparados com ambientes continentais (Wright & Hubbell,
1983). A hipótese da quantidade de hábitat, testada no capítulo um, se continuamente sustentada por estudos conduzidos com diferentes grupos de espécies e em regiões distintas, pode trazer grandes implicações tanto acadêmicas quanto para as práticas de manejo de paisagens, principalmente no que tangem as questões sobre SLOSS (do inglês single large or several small). Se a riqueza de espécies varia em função da quantidade total de hábitat numa dada região, não importa se as áreas adequadas para a sobrevivência da espécie estão em único fragmento ou em vários e sim o quanto de hábitat existe na paisagem local. Aliado a isso, Fahrig
(2003) mostra que o efeito da fragmentação de hábitat per se, ou seja, o efeito da fragmentação descontado o efeito da perda de hábitat, são muito mais fracos que os efeitos da perda de hábitat, sendo positivo em 68% dos casos. Em uma revisão mais detalhada, a mesma autora mostra em um trabalho recente que 75% dos efeitos significativos da fragmentação per se são positivos 86
(Fahrig, in prep.). Tais efeitos positivos da fragmentação são independentes do táxon, da medida de fragmentação usada, do grupo taxonômico, da zona climática, da capacidade de movimentação das espécies ou se a análise foi limitada a espécies especialistas. A preponderância dos efeitos positivos da fragmentação e os resultados encontrados no primeiro capítulo implicam que não há justificativa para atribuirmos um menor valor de conservação para pequenos fragmentos, quando comparados a áreas equivalentes contínuas.
No capítulo dois vimos que a extensão de resposta de uma mesma espécie pode variar de acordo com a resposta biológica utilizada. Quando a resposta biológica é a abundância da espécie, a resposta se dá a extensões menores quando comparada com a ocorrência. O delineamento de estudos e áreas protegidas deve considerar que a extensão espacial em que a abundância de espécies é afetada pela quantidade de hábitat é menor do que a extensão espacial que afeta a ocorrência das espécies. Práticas de manejo de paisagens com o intuito de conservar espécies e número de indivíduos deveriam considerar dados de ocorrência ao invés de dados de abundância, o que abrangeria áreas maiores e englobaria ambos.
No capítulo três utilizamos modelos de paisagens que utilizam parâmetros do ambiente e da espécie. Investigações posteriores sobre a relação espécie-área e sua aplicação no manejo de áreas protegidas e paisagens mostraram que o poder seu preditivo aumenta consideravelmente quando informações autoecológicas são incorporadas aos modelos (Abbott,
1983; Zimmerman & Bierregaard, 1986). Mais do que simplesmente a área disponível para a sobrevivência dos indivíduos, as características intrínsecas da espécie determinarão como dar- se-á sua relação com os fatores extrínsecos. Ressaltamos a inclusão de parâmetros que expressem a capacidade de dispersão e a área necessária para a sobrevivência de unidades reprodutivas nos estudos da relação entre espécies e distribuição espacial dos remanescentes de hábitat. A inclusão desses parâmetros nas medidas de capacidade de suporte e conectividade dão significado biológico à essas medidas que, avaliadas em toda área de distribuição da 87
espécie, traçam um panorama geral da relação atual e futura (para um exemplo, ver Gouveia et al. 2016) da espécie com os remanescentes de habitat. Dependendo do componente de habitat estudado, a extensão em que o estudo é conduzido traz diferentes resultados, que implicam em diferentes tomadas de decisões. Os efeitos da fragmentação e da conectividade funcional mudam conforme a extensão da escala considerada, já a capacidade de suporte do ambiente e a quantidade total de remanescentes são menos afetados pela extensão da área de estudo e mais sensíveis à história de vida das espécies.
A compreensão da relação entre o hábitat e as espécies ainda tem muito o que avançar. É necessário que a hipótese da quantidade de hábitat seja testada em outros contextos e com dados populacionais (Fahrig, 2015). Quanto à escala de efeito do hábitat há várias lacunas
(Jackson & Fahrig, 2015) e hipóteses que precisam ser testadas (Miguet et al., 2015) para que possamos traçar um panorama geral sobre os fatores que afetam a escala de efeito. O método proposto para avaliarmos a rede de fragmentos de hábitat que compõe toda a área de distribuição da espécie pode ser aplicado para grupos maiores de organismos, ou mesmo assembleias que ocupem uma mesma área. O método pode também ser usado juntamente com modelos climáticos e de desmatamento futuro, como fizemos em (Gouveia et al., 2016) para avaliarmos como o desmatamento e o clima afetarão a relação entre o habitat e as espécies.
REFERÊNCIAS
Connor E.F. & McCoy E.D. (1979) The Statistics and Biology of the Species-Area Relationship. The American Naturalist, 113, 791–833. Diamond J.M. (1975) The island dilemma: Lessons of modern biogeographic studies for the design of natural reserves. Biological Conservation, 7, 129–146. Diamond J.M., Terborgh J., Robert F. Whitcomb, Lynch J.F., Opler P.A., Robbins C.S., Simberloff D.S., & Abele L.G. (1976) Island biogeography and conservation: strategy and limitation. Science, 193, 1027–1032. Fahrig L. (2003) Effects of habitat fragmentation on biodiversity. Annual Review of Ecology, Evolution, and Systematics, 34, 487–515. 88
Fahrig L. (2013) Rethinking patch size and isolation effects: the habitat amount hypothesis. Journal of Biogeography, 40, 1649–1663. Fahrig L. (2015) Just a hypothesis: a reply to Hanski. Journal of Biogeography, 42, 993–994. Gouveia S.F., Souza-Alves J.P., Rattis L., Dobrovolski R., Jerusalinsky L., Beltrão-Mendes R., & Ferrari S.F. (2016) Climate and land use changes will degrade the configuration of the landscape for titi monkeys in eastern Brazil. Global Change Biology, 22, 2003–2012. Jackson H.B. & Fahrig L. (2015) Are ecologists conducting research at the optimal scale? Global Ecology and Biogeography, 24, 52–63. Levins R. (1970) Extinction. Some mathematical problems in biology (ed. by M. Gesternhaber), pp. 77–107. American Mathematical Society, Providence, RI, USA. MacArthur, R H Wilson E.O. (1967) The Theory of Island Biogeography. Princeton University Press., Miguet P., Jackson H.B., Jackson N.D., Martin A.E., & Fahrig L. (2015) What determines the spatial extent of landscape effects on species? Landscape Ecology, 1–18. Simberloff D.S. & Abele L.G. (1976) Island Biogeography: Theory and Conservation Practice. Science, 191, 285–286. Wright S.J. & Hubbell S.P. (1983) Stochastic Extinction and Reserve Size: A Focal Species Approach. Oikos, 41, 466. Zimmerman B.L. & Bierregaard R.O. (1986) Relevance of the Equilibrium Theory of Island Biogeography and Species-Area Relations to Conservation with a Case from Amazonia. Journal of Biogeography, 13, 133–143.
89
90