EFEITOS CASCATA DA CAÇA NA COMUNIDADE DE VERTEBRADOS E PLANTAS EM FLORESTAS DA AMAZÔNIA OCIDENTAL Tese apresentada ao programa de Pós-Graduação em Ecologia da Universidade Federal do Rio Grande do Norte, como parte das exigências para a obtenção do Título de Doutor em Ecologia Orientador: Prof. Dr. Carlos A. Peres NATAL/RN Julho de 2020 Universidade Federal do Rio Grande do Norte - UFRN Sistema de Bibliotecas - SISBI Catalogação de Publicação na Fonte. UFRN - Biblioteca Setorial Prof. Leopoldo Nelson - -Centro de Biociências - CB Scabin, Andressa Bárbara. Efeitos cascata da caça na comunidade de vertebrados e plantas em florestas da Amazônia Ocidental / Andressa Bárbara Scabin. - Natal, 2020. 128 f.: il. Tese (Doutorado) - Universidade Federal do Rio Grande do Norte. Centro de Biociências. Programa de Pós-graduação em Ecologia. Orientador: Prof. Dr. Carlos Augusto Peres. 1. Caça - Tese. 2. Florestas vazias - Tese. 3. Dispersão de sementes - Tese. 4. Estoque de carbono - Tese. 5. Amazônia - Tese. I. Peres, Carlos Augusto. II. Universidade Federal do Rio Grande do Norte. III. Título. RN/UF/BSCB CDU 639.1 Elaborado por KATIA REJANE DA SILVA - CRB-15/351 O Fim Que Se Aproxima Amazonas: mito grego menos antigo que os mitos da Amazônia. Os que vivem no Cosmo há milênios são perseguidos por mãos de ganância, olhos ávidos: minério, fogo, serragem, fim. Quem são vocês, incendiários desde sempre, ferozes construtores de ruínas? Os que queimam, impunes, a morada ancestral, projetam no céu mapas sombrios: manchas da floresta calcinada, cicatrizes de rios que não renascem. Qual Brasil se esconde atrás da humanidade amazônica? Que triste pátria delida, mais armada que amada: traidora de riquezas e verdades. Quando tudo for deserto, o mundo ouvirá rugidos de fantasmas. E todos vão escutar, numa agonia seca, o eco: Não existirão mundos, novos ou velhos, nem passado ou futuro. No solo de cinzas: o tempo-espaço vazio. Milton Hatoum “Nesses tempos de céus cinzas e chumbos, precisamos de árvores desesperadamente verdes” Mário Quintana Dedico esta tese a todos aqueles que têm dedicado suas vidas para tornar o mundo melhor e mais justo e cuja principal motivação seja o respeito à vida em todas as suas manifestações. AGRADECIMENTOS Primeiramente, aos meus pais Luiz e Cristina que sempre me apoiaram em todas as minhas decisões, mesmo que algumas das minhas escolhas significassem para eles mais ausências do que presenças. Posso, certamente, dizer que foi pelo apoio da minha família que pude ter a real liberdade de escolher meu caminho e ter o privilégio de atuar no que me faz feliz. Ainda, no âmbito da família, agradeço ao meu irmão Rafael e minha cunhada Nara por nossas profundas conversas, nas quais compartilhamos ideias, frustrações e expectativas; o que tem me ajudado a enfrentar esse momento sombrio pelo qual passa nosso país. E, falando em conversas estimulantes, não poderia deixar de agradecer os momentos de reflexão que tive com Bruna Arenque e Tatiana Franco, que sempre foram grandes apoiadoras dos meus projetos pessoais e com as quais idealizo a criação de uma organização que possa trazer impacto real na conservação ambiental e qualidade de vida das pessoas. São esses sonhos compartilhados que me ajudam a manter a esperança e motivação de trabalhar em prol de um futuro mais justo e equilibrado ambiental, social e economicamente. Algumas pessoas admiráveis que eu gostaria de agradecer por serem cientistas que me inspiram e que contribuíram imensamente na minha trajetória até aqui são Flávia Costa, Jochen Schöngart, Juliana Schietti, George Rebêlo, Júlio Cesar Voltolini, Mauro Galetti, Ricardo Bovendorp, Anderson Bueno, Anderson Feijó, Alexandre Fadigas, Carine Emer, Mauro Pichorim, Maíra Benchimol, Eduardo Venticinque e Andrew Abraham. Um agradecimento especial ao meu orientador Carlos Peres, por me incluir na “família Médio Juruá” e, com isso, possibilitar que eu trabalhasse com profissionais incríveis com os quais tenho aprendido muito: João Vitor Campos e Silva, Carolina Freitas, Helder Espírito-Santo, Joseph Hawes, Hugo Costa, Jaqueline Orlando, Franciany Braga e Clara Machado. Com essa maravilhosa equipe estamos trabalhando na consolidação do Instituto Juruá, para que nossas pesquisas científicas sejam aplicadas de modo a garantir a conservação ambiental e a melhoria da qualidade de vida dos habitantes das reservas de uso sustentável no Médio Juruá. O trabalho que estou apresentando aqui não poderia ter sido concretizado sem que muitas pessoas e instituições fossem envolvidas, e contribuíssem em diferentes etapas de sua execução. Dentre as organizações locais de Carauari e Itamarati que ofereceram um auxílio crucial em aspectos logísticos de coleta de dados em campo, posso citar: Operação Amazônia Nativa (OPAN), Associação dos Produtores Rurais de Carauari (ASPROC), Associação dos Moradores da RDS Uacarai (AMARU), Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio), Centro Estadual de Unidades de Conservação (CEUC), Instituto de Proteção Ambiental do Amazonas (IPAAM), Universidade Estadual do Amazonas (UEA/ Campus Carauari). E como essas organizações são constituídas por esforços pessoais, não posso deixar de agradecer a Renato Rocha, Ronnayana Silva, Edervan Vieira dos Santos, Manoel Cunha, Gilberto Oliveira e Givanildo Freitas. Outra contribuição fundamental que tive durante as expedições no Juruá foi o trabalho dos auxiliares de campo: Seu Bi e Malagueta (Roque), Tiozinho (Gumo do Facão), Manoel e Francisco (Bom Jesus), Pensamento (Bauana), Edilson e Antônio Carlos (Tabuleiro), Rosa e Benedito (Vila Ramalho), Caiana (Cachoeira), Ageu e Manoel (Pupunha), Pi (São José), Antônio e Ricardo (São Sebastião), Wilson e Cleberson (Itamarati), com destaque para Joaquim Gomes de Lima e Antônio Farias (Rezende) que estiveram mais ativamente trabalhando no projeto. Uma outra contribuição bastante valiosa foi a dos fotógrafos Scott Abraham e Marcos Suglia que fizeram lindos registros durante nossas expedições de campo pelo Médio Juruá. Para a identificação de espécies botânicas contei com o trabalho do parabotânico Paulo Assunção e da pesquisadora Lorena Maniguage Rincon. Para os procedimentos laboratoriais tive o apoio do Instituto Nacional de Pesquisas da Amazônia (INPA) e o trabalho da equipe do Laboratório de Diversidade Funcional de Plantas: Laura Martins, Natalia Medeiros Vicente e Alessandra Peixoto. As análises físicas e químicas do solo foram realizadas pelo químico Erison Gomes do Laboratório Temático de Solos e Plantas do INPA. Além de todo apoio de recursos humanos que tive no desenvolvimento dessa tese, o apoio dos financiadores foi imprescindível. Esse suporte não somente permitiu a execução das atividades de coleta de dados de campo e análises laboratoriais, mas também proporcionou experiências de grande impacto em meu crescimento profissional e pessoal. Dessa forma, agradeço ao Conselho de Aperfeiçoamento de Pessoal de Nível Superior (CAPES), pela concessão da bolsa de doutorado, à Rufford Foundation, National Geographic Society e Society for Conservation Biology pelo financiamento das expedições de coleta de dados e atividades de laboratório. Novamente registro minha gratidão à National Geographic Society, que junto à Lyda Hill Philanthropies, financiou meu período de 3 meses de intercâmbio no Reino Unido. O apoio desta organização também possibilitou minha participação na reunião anual da Associação de Biologia Tropical e Conservação (ATBC) em Antananarivo (Madagascar). Agradeço também a Universidade de Cambridge (Reino Unido) por financiar minha participação na Conferência de Estudantes em Ciências da Conservação (SCCS 2019), e por financiar um período de um mês de intercâmbio no Reino Unido e a Universidade East Anglia (Reino Unido) por me receber para o intercâmbio. Por fim, gostaria de agradecer todo apoio, carinho e bons momentos que minha família potiguar me proporcionou, o que tornou o processo do doutorado mais feliz: Nádia Zamboni (Chica), Adrian Pichi, Adriana Almeida, Giesta George, Lara Alves, Paulo Henrique Marinho, Paulo Fernandes, Marilia Teixeira, Carolina Freitas, Helder Espirito- Santo, Alexandre Santos, Aleksander Hada, Fernanda Lamin, Alessandra Salles e suas filhas ( Kaká, Loló, Vivi, Sosô e Ramira) e meu companheiro felino Odé. E, claro, não teria como melhor finalizar esse momento de gratidão senão oferecendo um agradecimento muito especial à todos os moradores do Médio Juruá que me receberam com muito respeito e carinho em suas casas e me mostraram como a felicidade pode estar presente nas coisas mais simples do dia a dia. Hoje é no trabalho com essas comunidades da Amazônia que tenho depositado toda minha esperança de conservação de nossa maior riqueza. RESUMO ............................................................................................................ 10 ABSTRACT .......................................................................................................... 11 INTRODUÇÃO GERAL ......................................................................................... 12 CAPÍTULO I ......................................................................................................... 28 HUNTING PRESSURE MODULATES THE COMPOSITION AND SIZE STRUCTURE OF TERRESTRIAL AND ARBOREAL VERTEBRATES IN AMAZONIAN FORESTS .......... 28 CAPÍTULO II ........................................................................................................ 66 CASCADING EFFECTS OF OVERHUNTING ON FUNCTIONAL TREE COMPOSITION IN AMAZONIAN FORESTS .................................................................................. 66 CAPÍTULO III ..................................................................................................... 102 CONSEQUENCES OF DEFAUNATION FOR ABOVEGROUND CARBON STOCKS IN AMAZONIAN FORESTS ..................................................................................... 102 Artigo de Divulgação Científica enviado para a Revista Ciência Hoje Crianças ......................................................................................................................... 127 RESUMO A caça representa uma das maiores ameaças aos animais silvestres em todo o mundo e têm causado declínio populacional acentuado de grandes vertebrados nas florestas tropicais. A implicação desse declínio pode ser imensa, uma vez que espécies mais intensamente caçadas estão geralmente envolvidas em processos chave da dinâmica florestal que inclui a dispersão e a predação de sementes. A defaunação promove ruptura dessas interações planta-animal essenciais para a regeneração florestal o que pode comprometer a manutenção da diversidade vegetal e de serviços ecossistêmicos. Esta tese explora os efeitos cascata da defaunação na comunidade de vertebrados e de plantas utilizando um gradiente de pressão de caça na região do Médio Juruá na Amazônia ocidental brasileira. No primeiro capítulo analisamos os efeitos diretos e indiretos da caça na comunidade de mamíferos e aves. Para isso estimamos a biomassa dos vertebrados, utilizando armadilhamento fotográfico no sub-bosque e no dossel da floresta em 30 sítios distribuídos ao longo do gradiente de caça. As florestas sobre-caçadas apresentaram mudanças na estrutura de tamanho da comunidade com o declínio na biomassa de espécies de maior porte sensíveis à caça e aumento da abundância de roedores noturnos, possivelmente relacionado a um mecanismo de compensação de densidade. O segundo capítulo aborda o efeito cascata de defaunação sobre a composição funcional da floresta. Neste capítulo, comparamos os traços funcionais de árvores e arvoretas inventariadas em 30 parcelas de 0.25 hectare (=7.5 ha) e 0.05 hectare (=1.5 ha) respectivamente, estabelecidas ao longo do gradiente de caça para testar a hipótese de diminuição da representatividade de traços funcionais de espécies de árvores dispersas por grandes vertebrados na comunidade vegetal. Nossos resultados indicaram uma modesta diminuição na abundância de espécies de arvoretas comparadas com árvores co- específicas dispersas por animais caçados e aumento na prevalência de arvoretas de espécies com dispersão abiótica. Contudo, esse efeito não refletiu no padrão funcional da comunidade para os traços contínuos de densidade de madeira, massa foliar/área (LMA) e massa da semente. No terceiro capítulo utilizamos dados dendrométricos do inventário florestal e amostragem de densidade de madeira, para estimar o estoque atual e futuro de carbono, com o objetivo de avaliar o potencial impacto da caça nesses estoques. Dentre as 30 parcelas analisadas, 22 tiveram uma previsão de perda de carbono para o futuro, sendo que a média estimada de perda foi de 2.2 MgC ha-¹. Para as duas unidades de conservação (UCs) localizadas na paisagem estudada a perda projetada total foi de 1.560 MgC. Considerando os valores monetários no mercado internacional de carbono, essa diminuição projetada no estoque de carbono das UCs foi valorada entre US$15.6 e US$120 milhões. Palavras-chave: caça, floresta vazia, defaunação, compensação de densidade, frugivoria, dispersão de sementes, traços funcionais, estoque de carbono. 10 ABSTRACT Overhunting is one of the greatest threats to wildlife worldwide and has caused a sharp decline in the abundance of large-bodied vertebrate populations in tropical forests. The implications are far-reaching since the most intensely hunted species are often involved in key ecological processes related to forest dynamics, including seed dispersal and predation. Defaunation disrupts these plant-animal interactions, which are essential for forest regeneration, and can compromise the maintenance of plant diversity and ecosystem services. This thesis explores the cascading effects of defaunation on vertebrate and plant communities using a hunting pressure gradient in the Médio Juruá region of western Brazilian Amazonia. In the first chapter, we analysed the direct and indirect effects of hunting on mammal and bird communities. For this, we estimated vertebrate biomass using camera trapping in both the understorey and forest canopy at 30 sites distributed along the hunting gradient. Overhunted forests showed changes in the size structure of the animal community with a decline in the biomass of large-bodied hunting-sensitive species and an increase in the abundance of nocturnal rodents, possibly related to a density compensation mechanism. The second chapter explores the cascading effects of defaunation on future forest composition and functionality. We compared functional traits of trees and saplings, inventoried in 30 plots of 0.25 ha (=7.5 ha) and 30 subplots of 0.05 ha (=1.5 ha) respectively, established along the hunting gradient, to test for an expected decrease in traits associated to seed dispersal by large vertebrates. Our results indicated a modest decrease in the abundance of large vertebrate dispersed saplings compared to conspecific trees, and an increase in the prevalence of saplings from abiotically dispersed species. However, this effect did not reflect the community-wide pattern for the continuous traits of wood density, leaf mass/area (LMA) and seed mas. In the third chapter, we used dendrometric data from our forest inventory and comprehensive wood density sampling to estimate current and future carbon stocks, and thereby assess the potential impact on carbon stocks in intensively hunted forests. Of the 30 plots sampled, 22 could lose forest carbon in the future and the mean projected loss was 2.2 MgC ha-¹. For two protected areas (PAs) in the study landscape, the projected loss was approximately 1,560 MgC. Considering the currently predicted monetary values in the international carbon market, the projected decrease in the PAs future carbon stock was valued from US$15.6 to US$120 million. Keywords: hunting, empty forest, defaunation, density compensation, frugivory, seed dispersal, functional traits, carbon storage. 11 INTRODUÇÃO GERAL A humanidade vem transformando o planeta em magnitude e velocidade comparável a processos naturais que, no passado, definiram eras geológicas, por isso, acredita-se que atualmente estamos vivendo a chamada Era do Antropoceno (Dirzo et al., 2014; Malhi, et al., 2014). Nesta Era, os humanos têm ameaçado a conservação das florestas tropicais em todo mundo por meio de diversas atividades econômicas, tais como: extração de madeira, mineração, agricultura, pecuária, urbanização, criação de infraestrutura, entre outras ações que têm promovido uma perda significativa de áreas de floresta (Malhi et al., 2014). Algumas atividades antrópicas, entretanto, são consideradas mais silenciosas por que não promovem mudanças na cobertura vegetal sendo, portanto, impossíveis de serem detectadas, mesmo utilizando as mais sofisticadas técnicas de sensoriamento remoto (Peres et al., 2006). Uma dessas atividades silenciosas é a caça de animais silvestres que, embora não promova perda de cobertura vegetal, pode comprometer a integridade e funcionalidade da floresta a longo prazo (Redford, 1992). A caça de animais silvestres representa uma das maiores ameaças às populações de vertebrados em todo o mundo, sendo o principal fator responsável pela diminuição acentuada de populações e extinções locais (Benítez-López et al., 2017). O efeito da caça é tão acentuado que, atualmente, nas florestas tropicais a abundância dos animais está mais correlacionada com padrões de caça e acessibilidade ao caçador do que a fatores como tipo de floresta, tamanho do habitat e status de conservação (Peres & Lake, 2003; Ohl-Schacherer et al., 2007; Peres & Palacios, 2007). Uma das características desse tipo de atividade humana é que a seleção dos alvos de caça não é aleatória, mas segue um padrão preferencial por aves e mamíferos de grande porte, que chegam a apresentar reduções populacionais superiores a 80% em áreas severamente caçadas (Benítez-López et al., 2017; Peres, 2000; Peres & Palacios, 2007). As espécies mais intensamente caçadas geralmente possuem baixas taxas reprodutivas e de crescimento, longo tempo de geração e grande longevidade (Bodmer et al., 1997), o que dificulta a resiliência das populações frente ao impacto. Além disso, são também espécies que estão envolvidas em processos-chave para a manutenção da diversidade florestal, como a polinização, dispersão e predação de sementes e herbivoria de plântulas (Dirzo & Miranda, 1991; Redford, 1992; Terborgh et al., 2008; Wright, 2003). O alerta para essa ameaça silenciosa que as florestas tropicais estão sofrendo levou 12 Redford (1992) a usar pela primeira vez o termo “florestas vazias”, para designar as florestas severamente afetadas pela caça, cuja fauna foi tão gravemente empobrecida que promoveu a redução drástica nas densidades populacionais de algumas espécies, ocasionando até mesmo extinções locais. Mesmo que não haja a extinção local das espécies, a diminuição significativa em abundância pode ocasionar a chamada extinção ecológica, na qual as populações são tão diminuídas que as interações-chave das quais fazem parte deixam de ser realizadas eficientemente (Redford & Feinsinger, 2001). A extinção ecológica torna-se ainda mais impactante quando as espécies que sofrem o declínio populacional estão envolvidas em interações estreitas de baixa redundância funcional, como por exemplo a dispersão de sementes grandes, que são dispersas apenas por poucas espécies de mamíferos e aves de maior tamanho corpóreo (Peres & Roosmalen 2002, Poulsen et al., 2002). A diminuição ou perda dos serviços promovidos pela fauna pode alterar a diversidade e a dinâmica das florestas, uma vez que, esses serviços são cruciais na regeneração florestal. Estudos conduzidos nos Neotrópicos, África e Ásia evidenciam que o recrutamento, abundância e distribuição espacial de muitas espécies de plantas são modificados em áreas intensamente caçadas onde os dispersores-chave estão ausentes (Terborgh et al., 2008; Wright, 2003; Vanthomme et al., 2010; Brodie et al., 2009). Uma tendência demonstrada em alguns desses estudos é que espécies vegetais com grandes sementes declinam drasticamente em florestas caçadas, ocorrendo, em contrapartida, um maior recrutamento de plântulas de espécies dispersas abioticamente ou por frugívoros não caçados (Brodie et al., 2009; Effiom, et al., 2013; Kurten, et al., 2015; Stevenson & Aldana, 2008; Wright et al., 2007; Nunez-Iturri et al., 2008; Terborgh, et al., 2008). Outra tendência é uma diminuição na distância de dispersão das sementes em relação à planta mãe (Poulsen et al., 2013), o que compromete a viabilidade do recrutamento das plântulas (efeito Janzen-Connell) e altera a distribuição espacial das espécies vegetais, promovendo padrões mais agregados (Bagchi et al., 2018) que, por apresentarem maior proximidade genética, podem ser mais suscetíveis a outras ações antrópicas, como a fragmentação e extração seletiva de madeira (Cordeiro & Howe, 2003). Em contrapartida, estudos que utilizaram experimentos de exclusão de médios e grandes vertebrados terrestres não têm evidenciado mudanças no recrutamento de plântulas (Brocardo et al., 2013), ou as mudanças são fracas (Theimer et al., 2011). Além 13 disso, as respostas podem ser diferenciadas entre as espécies de plantas devido a contrastes nas características de história de vida, incluindo taxas reprodutivas, longevidade e taxa de crescimento, fazendo com que o tamanho da semente não seja a única característica que condicione o efeito da defaunação no estabelecimento de plântulas (Camargo-Sanabria et al., 2014; Beckman & Muller-Landau, 2007). Contudo, os experimentos de exclusão não interrompem a dispersão primária promovida pelos vertebrados arborícolas, ou seja, nesse caso, não simulam perfeitamente uma floresta defaunada, além de, em alguns casos, os locais estudados já apresentarem uma fauna bastante depauperada por longo período de tempo (Camargo-Sanabria et al., 2014; Gardner et al , 2019) Somado ao efeito da defaunação na perda dos principais dispersores de sementes, há também o possível favorecimento de espécies de vertebrados menores, cujos principais competidores e predadores foram dizimados. Os frugívoros e granívoros de menor porte podem então aumentar a sua contribuição relativa em áreas caçadas, a chamada compensação de densidade (Galetti et al., 2015a; Poulsen et al., 2011; Wright, 2003; Peres & Dolman, 2000). Essa mudança na estrutura da comunidade de vertebrados com substituição de espécies grandes por espécies menores que no caso de roedores, atuam preferencialmente como predadores de sementes, por sua vez, também pode ter sérias implicações na dinâmica florestal, como foi evidenciado no trabalho de Galetti et al. (2015b), que mostrou um aumento de 4.9 vezes na predação de semente da palmeira Euterpes edulis por roedores em floresta defaunada de Mata Atlântica. Dessa forma, o efeito da defaunação na dispersão e predação de sementes depende não somente da perda dos grandes dispersores, mas também do aumento populacional dos predadores de sementes que costumam ser mais resilientes à caça (Galetti et al., 2015a; Peres & Palacios, 2007). Alguns vertebrados de grande porte como os queixadas (Tayassu pecari) também atuam como predadores de sementes e plântulas, além de promoverem a mortalidade das últimas por pisoteio e, portanto, o declínio populacional de grandes granívoros e herbívoros terrestres pode favorecer o recrutamento das plantas (Dirzo & Miranda, 1991, Wright et al, 2000). Sendo assim, como os animais caçados pertencem a diferentes guildas tróficas, eles atuam por vias complementares na regeneração florestal (Rosin et al., 2017). Por isso, é necessário entender os efeitos de declínio ou aumento populacional dos vertebrados em todas as interações que realizam com as espécies vegetais, uma vez que 14 esses efeitos têm papel fundamental no estabelecimento das plântulas e, como consequência na composição e estrutura das florestas. Dessa maneira, o efeito da caça sobre o recrutamento de plantas geralmente é negativo para as espécies de sementes grandes dispersas por animais cinegéticos, mas também pode ser nulo ou até mesmo positivo se as espécies que sofrem maior impacto da caça são responsáveis pela predação de sementes ou plântulas (Beckman & Muller-Landau, 2007; Wright et al., 2000, Dirzo et al., 2007; Harrison et al., 2013). A mudança na composição florestal como um efeito cascata da defaunação pode afetar a disponibilidade de recursos alimentares para a fauna e para o ser humano, promovendo prejuízos tanto ecológicos quanto econômicos (Effiom et al., 2014). Além disso, as espécies vegetais com maior probabilidade de sofrerem gargalos populacionais geralmente são as que produzem grandes sementes as quais, em sua maioria, irão gerar árvores de grande porte e de alta densidade de madeira (Hawes et al, 2020). Essas espécies, devido aos traços funcionais relacionados à um maior acúmulo de biomassa tem uma grande contribuição na estocagem de carbono em florestas tropicais (Bello et al., 2015; Osuri et al., 2016; Peres, et al. 2016). Consequentemente, um dos serviços ecossistêmico que potencialmente poderia ser prejudicado com a mudança na composição da vegetação das florestas defaunadas é o estoque de carbono (de Paula et al., 2018; Osuri et al., 2016 ; Peres et al., 2016; Bello et al., 2015; Schnitzer et al, 2014). Estudos globais utilizando modelagem que simulam a perda de espécies arbóreas dispersas por grandes vertebrados encontraram perdas médias de 2-12% de carbono estocado em florestas na África, América e Ásia (Osuri et al., 2016), enquanto que Peres e colaboradores (2016), utilizando simulações em diferentes cenários de depleção de grandes vertebrados na Amazônia, encontraram estimativas médias de 2.5 – 5.8%, porém com alguns locais com perda superior a 30%. Embora os modelos consistentemente sugiram a diminuição no estoque de carbono em florestas defaunadas, ainda são escassos estudos comparativos com estimativas de carbono estocado em florestas nas quais houve uma depleção de vertebrados de grande porte. Mesmo após quase trinta anos do primeiro uso do termo “florestas vazias” que emblematicamente iniciou esse campo da pesquisa, ainda existem muitas lacunas a serem preenchidas, ao mesmo tempo que representa um problema cada vez mais urgente de conservação (Poulsen et al., 2013). Embora muitos estudos tenham sido conduzidos em 15 florestas tropicais em todo mundo, as pesquisas no Brasil têm se concentrado principalmente na Mata Atlântica (Brocardo et al., 2013; Galetti, et al., 2006; Galetti et al., 2015 a e b) onde há um longo histórico de exploração dos recursos da fauna. Amazônia meio vazia? A floresta Amazônica, embora apresente elevadas taxas de caça e uma grande dependência de animais silvestres como fonte de proteína para as populações humanas (Peres, 2000), comparativamente com a Mata Atlântica ainda possui altas densidades de vertebrados de grande porte. Por outro lado, como apontado por Redford & Feisenger (2001), mesmo a diminuição acentuada na abundância das espécies pode comprometer a quantidade e qualidade de alguns serviços ecossistêmicos nas “florestas meio vazias”. Nesse sentido, a floresta Amazônica constitui um excelente modelo de estudo para entender o processo de esvaziamento das florestas bem como estabelecer limiares para o uso de recursos da fauna sem comprometimento dos serviços promovido pelas espécies. Algumas vantagens de investigação na floresta Amazônica incluem a possibilidade de uma abordagem em nível de paisagem considerando a diversidade de assentamentos humanos que exercem diferentes pressões de exploração da fauna, dependendo da maior ou menor dependência da proteína provinda da carne de caça. A abordagem em nível de paisagem pode representar um grande avanço nesse campo de pesquisa, visto que uma das principais falhas apontadas no delineamento da maior parte dos estudos é exatamente a falta de replicação espacial (Harrison et al., 2013). Adicionalmente, o fato de a Amazônia brasileira possuir grandes áreas relativamente intactas de floresta permite a seleção de sítios controle nos quais existe pouca interferência humana. Essa característica representa uma interessante oportunidade de aperfeiçoamento no delineamento amostral de pesquisas nesse campo de conhecimento, visto que possibilita a dissociação do efeito da caça com o efeito de outras perturbações antrópicas. Estrutura da Tese O objetivo dessa tese foi analisar os efeitos cascata de defaunação induzidos por caça na Amazônia ocidental brasileira e está dividida em três capítulos. O primeiro capítulo intitulado: A pressão de caça modula a composição e estrutura de tamanho de vertebrados terrestres e arborícolas em florestas na Amazônia (Hunting pressure modulates the composition and size structure of terrestrial and arboreal vertebrates in 16 Amazonian forests) discute os efeitos direto e indireto de um gradiente de caça na composição e estrutura de tamanho da comunidade de vertebrados terrestres e arborícolas na região do Médio Juruá. Neste artigo, buscamos entender se o aumento na intensidade de caça afeta as assembleias de mamíferos e aves de diferentes tamanhos corpóreos e que habitam diferente estratos da floresta. Além disso, analisamos quais guildas tróficas estão sendo mais impactadas pela caça e investigamos se nas áreas mais intensamente caçadas está ocorrendo aumento na abundância de roedores noturnos por um possível mecanismo de compensação de densidade relacionado ao declínio de grandes vertebrados. O segundo capítulo, intitulado: Efeitos cascata da caça na composição funcional de árvores em florestas na Amazônia (Cascading effects of overhunting on functional tree composition in Amazonian forests) aborda o efeito do gradiente de caça na composição funcional da floresta de terra firme na região do Médio Juruá. Neste capítulo testamos a hipótese de que florestas mais intensamente caçadas apresentam um declínio na abundância de indivíduos juvenis de espécies arbóreas com maior dependência de animais de grande porte para a dispersão de sementes. E, em contrapartida, esperamos um aumento na abundância de espécies dispersas por síndromes abióticas ou por frugívoros de pequeno porte. Além do modo de dispersão, analisamos se traços funcionais contínuos de densidade da madeira, massa foliar por área (LMA) e massa da semente são menos prevalentes em florestas que foram mais intensamente caçadas, relacionado a alteração na composição florestal. O terceiro capítulo da tese intitulado: Consequências da defaunação para os estoques de carbono acima do solo em florestas da Amazônia (Consequences of defaunation for aboveground carbono stocks in Amazonian forests) apresenta estimativas de carbono estocado no presente e no futuro em florestas sobre diferentes pressões de caça ao longo do gradiente de caça estudado. Para isso utilizamos dados dendrométricos coletados no inventário florestal das parcelas permanentes e densidade de madeira mensurada com amostras coletadas na área de estudo. Por meio de simulações de substituição na composição da comunidade baseado na densidade de madeira das espécies no estágio juvenil fizemos estimativas futuras do estoque de carbono. E testamos a hipótese da diminuição do carbono estocado em florestas sujeitas à elevada pressão de caça. As perdas e ganhos de carbono das projeções futuras foram então monetarizadas baseado em valores médios no mercado internacional de carbono. conhecimento científico sobre os efeitos cascata da defaunação contemporânea nas florestas tropicais. 17 O Médio rio Juruá O rio Juruá é o segundo mais longo tributário de água branca do rio Amazonas e está localizado na Amazônia Ocidental brasileira, no estado do Amazonas. Em sua porção média encontram-se duas grandes Unidades de Conservação (UCs) de uso sustentável: a Reserva Extrativista do Médio Juruá (RESEX Médio Juruá) e a Reserva de Desenvolvimento Sustentável Uacari (RDS Uacari). A RESEX Médio Juruá é uma UC federal que foi criada em 1997, tem 253.227 hectares e é legalmente ocupada por aproximadamente 2.000 pessoas vivendo em 13 comunidades ribeirinhas. A RDS Uacari é uma UC estadual e foi criada em 2005 possui 632.949 hectares e aproximadamente 1.200 pessoas vivem no interior da reserva em 32 comunidades. 18 Figura 1. Mapa da área de estudo na região do Médio rio Juruá, Amazonas, Brasil. A delimitação da Reserva Extrativista do Médio Juruá (RESEX Médio Juruá) e da Reserva de Desenvolvimento Sustentável Uacari (RDS Uacari) estão indicadas em preto. O trecho médio do rio Juruá navegado durante as expedições de campo está marcado em branco. As localizações das cidades de Carauari e Itamarati estão indicadas por triângulos preto e as comunidades ribeirinhas por triângulos brancos. Os sítios de coleta de dados estão marcados por círculos vermelhos. 19 A caça de animais silvestres no Médio Juruá A caça de animais silvestres na região do Médio Juruá é uma atividade que ocorre tanto no interior das unidades de conservação de uso sutentável quanto nas florestas do perímetro urbano de Carauari e Itamarati. Apesar de o peixe ser o principal recurso alimentar na região, a carne de caça representa uma importante fonte de proteína para as comunidades ribeirinhas principalmente no período da cheia do rio quando a pesca é mais escassa (Endo et al, 2016). Dentre as espécies de mamíferos e aves mais consumidas estão os veados ( Mazama nemorivaga e Mazama americana), catitus (Pecari tajacu), queixada (Tayassu pecari), paca (Cuniculus paca), cutias (Dasyprocta spp), mutuns (Crax ou Mitu spp), inhambus (Tinamus spp e Crypturellus spp) e em menor proporção os primatas (Allouatta spp. e Lagothrix spp) (Abrahams et al, 2017). Ainda que atualmente os primatas não sejam os principais alvos de caça nas comunidades riberinhas do médio Juruá, durante o período áureo da exploração da borracha na segunda metade do século XIX, os primatas, com destaque para o macaco barrigudo (Lagothrix spp), foram amplamente caçados o que afetou consideravelmente as populações dessa espécie (Peres, 1991). Delineamento Amostral e Coleta de Dados A presente tese foi desenvolvida no interior dessas duas UCs de uso sustentável em florestas de terra firme, nome dado as florestas amazônicas que não são cobertas pela 20 água durante o pulso de inundação. Além disso, foram incluídas no estudo florestas peri- urbanas da cidade de Carauari e Itamarati. Selecionamos 30 sítios ao longo de um gradiente de pressão de caça que inclui florestas peri-urbanas de alta pressão associadas a grande adensamento populacional humano até florestas sobre baixa pressão de caça relacionadas a baixa densidade populacional humana. Nesses sítios foram dispostas 30 unidades amostrais (Figura 1), compostas por um grid de 16 armadilhas fotográficas instaladas no sub-bosque e 4 armadilhas instaladas no dossel das florestas (~20 m de altura). No interior de cada um dos grids de armadilhas fotográficas foi estabelecida uma parcela para levantamento florístico de 100 m x 25 m (Figura 2). Em cada uma das unidades amostramos a comunidade de vertebrados terrestres e arborícolas, conduzimos levantamentos florísticos e de estrutura da vegetação, coletamos amostras de madeira e folhas para obtenção de traços funcionais da vegetação, bem como amostras de solo para obtenção da fertilidade do solo, uma importante variável abiótica que influencia as comunidades vegetais e animais (Figura 3). Figura 2. Desenho esquemático da unidade amostral para obtenção de dados. Os círculos em vermelho representam a distribuição espacial das armadilhas fotográficas de sub- bosque e os círculos em azul as câmeras de dossel. O retângulo verde no interior do grid representa a parcela de 100mX25m onde foram conduzidos os inventários florestais. 21 Figura 3. Prancha de imagens ilustrando as principais metodologias empregadas na coleta de dados. A: armadilhamento fotográfico no sub-bosque da floresta; B: armadilhamento fotográfico no dossel da floresta; C: marcação e medição de diâmetro de árvores e arvoretas; D: identificação botânica; E: coleta botânica; F: coleta de amostra de madeira; G: medição de altura das árvores; H: analise química para obtenção de macronutrientes do solo; I: método de deslocamento da água para obtenção de densidade da madeira; J: pesagem de transecções de folhas para obtenção da massa da folha por unidade de área (LMA- leaf mass/area). 22 Sendo assim, por meio de novas abordagens, o presente trabalho fornece contribuição significativa ao conhecimento científico sobre os efeitos cascata da defaunação contemporânea nas florestas tropicais. Como resultado direto do aqui exposto, esperamos que as ideias discutidas subsidiem propostas de manejo de recursos faunísticos em unidades de conservação de uso sustentável que garantam a manutenção das interações planta-animal imprescindíveis na dinâmica florestal. Referências Bibliográficas Abrahams MI, Peres CA, Costa HCM (2017) Measuring local depletion of terrestrial game vertebrates by central-place hunters in rural Amazonia. PLoS One 12:1–25. Bagchi, R., Swamy, V., Latorre Farfan, J. P., Terborgh, J., Vela, C. I. A., Pitman, N. C. A., & Sanchez, W. G. (2018). Defaunation increases the spatial clustering of lowland Western Amazonian tree communities. Journal of Ecology, 106(4), 1470–1482. Beckman, N. G., & Muller-Landau, H. C. (2007). Differential effects of hunting on pre- dispersal seed predation and primary and secondary seed removal of two neotropical tree species. Biotropica, 39(3), 328–339. Bello, C., Galetti, M., Pizo, M. A., Magnago, L. F. S., Rocha, M. F., Lima, R. A. F., Jordano, P. (2015). Defaunation affects carbon storage in tropical forests. Science Advances, 1 (11). Benítez-López, A., Alkemade, R., Schipper, A. M., Ingram, D. J., Verweij, P. A., Eikelboom, J. A. J., & Huijbregts, M. A. J. (2017). The impact of hunting on tropical mammal and bird populations. Science, 356(6334), 180–183. Bodmer, R. E., Eisenberg, J. F., & Redford, K. H. (1997). Hunting and the likelihood of extinction of Amazonian mammals. Conservation Biology, 11(2), 460–466. Brocardo, C. R., Zipparro, V. B., de Lima, R. A. F., Guevara, R., & Galetti, M. (2013). No changes in seedling recruitment when terrestrial mammals are excluded in a partially defaunated Atlantic rainforest. Biological Conservation, 163, 107–114. Brodie, J. F., Helmy, O. E., Brockelman, W. Y., & Maron, J. L. (2009). Bushmeat poaching reduces the seed dispersal and population growth rate of a mammal- dispersed tree. Ecological Applications, 19(4), 854–863. Camargo-Sanabria, A. A., Mendoza, E., Guevara, R., Martinez-Ramos, M., & Dirzo, R. (2014). Experimental defaunation of terrestrial mammalian herbivores alters tropical rainforest understorey diversity. Proceedings of the Royal Society B: Biological Sciences, 282(1800). Cordeiro, N. J., & Howe, H. F. (2003). Forest fragmentation severs mutualism between seed dispersers and an endemic African tree. Proceedings of the National Academy of Sciences, 100(24), 14052-14056. de Paula Mateus, D., Groeneveld, J., Fischer, R., Taubert, F., Martins, V. F., & Huth, A. (2018). Defaunation impacts on seed survival and its effect on the biomass of future 23 tropical forests. Oikos, 127(10), 1526–1538. Dirzo, R., Mendoza, E., & Ortíz, P. (2007). Size‐related differential seed predation in a heavily defaunated neotropical rain forest. Biotropica, 39(3), 355-362. Dirzo, R., & Miranda, A. (1991). Altered patterns of herbivory and diversity in the forest understory: a case study of the possible consequences of contemporary defaunation. Plant-animal interactions: evolutionary ecology in tropical and temperate regions. Wiley, New York, 273-287. Dirzo, R., Young, H. S., Galetti, M., Ceballos, G., Isaac, N. J. B., & Collen, B. (2014). Defaunation in the Anthropocene. Science, 401(6195), 401–406. Effiom, E. O., Nunez-Iturri, G., Smith, H. G., Ottosson, U., & Olsson, O. (2013). Bushmeat hunting changes regeneration of African rainforests. Proceedings of the Royal Society B: Biological Sciences, 280(1759). Effiom, Edu O., Birkhofer, K., Smith, H. G., & Olsson, O. (2014). Changes of community composition at multiple trophic levels due to hunting in Nigerian tropical forests. Ecography, 37(4), 367–377. Endo W, Peres CA, Haugaasen T (2016) Flood pulse dynamics affects exploitation of both aquatic and terrestrial prey by Amazonian floodplain settlements. Biological Conservation, 201:129–136. Galetti, M., Guevara, R., Neves, C. L., Rodarte, R. R., Bovendorp, R. S., Moreira, M., Hopkins III, J. B. & Yeakel, J. D. (2015a). Defaunation affect population and diet of rodents in Neotropical rainforests. Biological Conservation, 190, 2–7. Galetti, M., Bovendorp, R. S., & Guevara, R. (2015b). Defaunation of large mammals leads to an increase in seed predation in the Atlantic forests. Global Ecology and Conservation, 3, 824-830. Galetti, M., Donatti, C. I., Pires, A. S., Guimarães Jr, P. R., & Jordano, P. (2006). Seed survival and dispersal of an endemic Atlantic forest palm: the combined effects of defaunation and forest fragmentation. Botanical Journal of the Linnean Society, 151(1), 141-149. Gardner, C. J., Bicknell, J. E., Baldwin-Cantello, W., Struebig, M. J., & Davies, Z. G. (2019). Quantifying the impacts of defaunation on natural forest regeneration in a global meta-analysis. Nature Communications, 10(1), 4590. Harrison, R. D., Tan, S., Plotkin, J. B., Slik, F., Detto, M., Brenes, T., Itoh, A. & Davies, S. J. (2013). Consequences of defaunation for a tropical tree community. Ecology Letters, 16(5). Hawes, J. E., Vieira, I. C., Magnago, L. F., Berenguer, E., Ferreira, J., Aragão, L. E., Cardoso, A., Lees, A. C., Lennox, G. D., Tobias, J. A., Waldron, A. & Jos Barlow. (2020). A large‐scale assessment of plant dispersal mode and seed traits across human‐modified Amazonian forests. Journal of Ecology, 108(4), 1-13. Kurten, E.L., Wright, S.J., Carson, W.P., Palmer, T.M., 2015. Hunting alters seedling functional trait composition in a Neotropical forest. Ecology 96, 1923–1932. Malhi, Y., Gardner, T. A., Goldsmith, G. R., Silman, M. R., & Zelazowski, P. (2014). Tropical Forests in the Anthropocene. Annual Review of Environment and 24 Resources, 39(1), 125–159. Nunez-Iturri, G., Olsson, O., & Howe, H. F. (2008). Hunting reduces recruitment of primate-dispersed trees in Amazonian Peru. Biological Conservation, 141(6), 1536– 1546. Ohl-Schacherer, J., Shepard, G. H., Kaplan, H., Peres, C. A., Levi, T., & Yu, D. W. (2007). The sustainability of subsistence hunting by Matsigenka native communities in Manu National Park, Peru. Conservation Biology, 21(5), 1174–1185. Osuri, A. M., Ratnam, J., Varma, V., Alvarez-Loayza, P., Astaiza, J. H., Bradford, M., Fletcher, C., Ndoundou-Hochemba, Jansen, P.A., Kenfack, D. Marshall, A. R., Ramesh, B. R. , Rovero, F. & Sankaran, M. (2016). Contrasting effects of defaunation on aboveground carbon storage across the global tropics. Nature Communications, 1-7. Peres, C. A., & Dolman, P. M. (2000). Density compensation in neotropical primate communities: evidence from 56 hunted and nonhunted Amazonian forests of varying productivity. Oecologia, 122(2), 175–189. Peres, C. A. (2000). Effects of subsistence hunting on vertebrate community structure in Amazonian forests. Conservation Biology, 14(1), 240–253. Peres, Carlos A., Emilio, T., Schietti, J., Desmoulière, S. J. M., & Levi, T. (2016). Dispersal limitation induces long-term biomass collapse in overhunted Amazonian forests. Proceedings of the National Academy of Sciences, 113(4), 892–897. Peres, C. A., & Van Roosmalen, M. (2002). Primate frugivory in two species-rich Neotropical forests: implications for the demography of large-seeded plants in overhunted areas. Seed dispersal and frugivory: ecology, evolution and conservation. CABI Publishing, Wallingford, 407-421 Peres, C. A., & Palacios, E. (2007). Basin-wide effects of game harvest on vertebrate population densities in Amazonian forests: implications for animal-mediated seed dispersal. Biotropica, 39(3), 304–315. Peres, C. A, Barlow, J., & Laurance, W. F. (2006). Detecting anthropogenic disturbance in tropical forests. Trends in Ecology and Evolution, 21(5), 227–229. Peres, C. A, & Lake, I. R. (2003). Extent of Nontimber Resource Extraction in Tropical Forests: Accessibility to Game Vertebrates by Hunters in the Amazon Basin. Conservation Biology, 17(2), 521–535. Peres, C.A. (1991) Humboldt’s wooly monkeys decimated by hunting in Amazonia. Oryx, 25(2), 89-95. Poulsen, J. R., Clark, C. J., & Palmer, T. M. (2013). Ecological erosion of an Afrotropical forest and potential consequences for tree recruitment and forest biomass. Biological Conservation, 163, 122–130. Poulsen, A. J. R., Clark, C. J., & Bolker, B. M. (2011). Decoupling the effects of logging and hunting on an Afrotropical animal community. Ecological Applications, 21(5), 1819–1836. Poulsen, J. R., Clark, C. J., Connor, E. F., & Smith, T. B. (2002). Differential resource use by primates and hornbills: Implications for seed dispersal. Ecology, 83(1), 228– 25 240. Redford, K. H. (1992). The Empty forest. BioScience, 42(6), 412–422. Redford, K. H. & Feinsinger, P. (2001). The half-empty forest: sustainable use and the ecology of interactions. Conservation of exploited species, 6, 370p. Rosin, C., Poulsen, J. R., Swamy, V., & Granados, A. (2017). A pantropical assessment of vertebrate physical damage to forest seedlings and the effects of defaunation. Global Ecology and Conservation, 11, 188–195. Schnitzer, S. A., Van Der Heijden, G., Mascaro, J., & Carson, W. P. (2014). Lianas in gaps reduce carbon accumulation in a tropical forest. Ecology, 95(11), 3008–3017. Stevenson, P. R., & Aldana, A. M. (2008). Potential effects of ateline extinction and forest fragmentation on plant diversity and composition in the western Orinoco Basin, Colombia. International Journal of Primatology, 29(2), 365–377. Terborgh, J., Nuñez-iturri, G., Pitman, N. C. A., Cornejo, F. H., Alvarez, P., Swamy, V., Pringle, E. G. & Paine, T. E. (2008). Tree Recruitment in an Empty Forest. Ecological Society of America, 89(6), 1757–1768. Theimer, T. A. D. C., Gehring, C. A., Green, P. T., & Connell, J. H. (2011). Terrestrial vertebrates alter seedling composition and richness but not diversity in an Australian tropical rain forest TL - 92. Ecology, 92 (8), 1637–1647. Vanthomme, H., Bellé, B., & Forget, P. M. (2010). Bushmeat hunting alters recruitment of large-seeded plant species in Central Africa. Biotropica, 42(6), 672–679. Wright, S. J. (2003). The myriad consequences of hunting for vertebrates and plants in tropical forests. Perspectives in Plant Ecology, Evolution and Systematics, 6(1,2), 73–86. Wright, S. J., Hernandéz, A., & Condit, R. (2007). The bushmeat harvest alters seedling banks by favoring lianas, large seeds, and seeds dispersed by bats, birds, and wind. Biotropica, 39(3), 363–371. Wright, S. J., Zeballos, H., Dominguez, I., Gallardo, M. M., Moreno, M. C., & Ibáñez, R. (2000). Poachers alter mammal abundance, seed dispersal, and seed predation in a neotropical forest. Conservation Biology, 14(1), 227–239. 26 27 CAPÍTULO I HUNTING PRESSURE MODULATES THE COMPOSITION AND SIZE STRUCTURE OF TERRESTRIAL AND ARBOREAL VERTEBRATES IN AMAZONIAN FORESTS Andressa B. Scabina and Carlos A. Peresb,c, * a Programa de Pós-Graduação em Ecologia, Universidade Federal do Rio Grande do Norte - Natal, RN, Brazil. Orcid: 0000-0001-5377-2123 b Centre for Ecology, Evolution and Conservation, School of Environmental Sciences, University of East Anglia, Norwich, UK. Orcid: 0000-0002-1588-8765 c Departamento de Sistemática e Ecologia, Universidade Federal da Paraíba, João Pessoa, PB, Brazil. Author for correspondence: * Carlos A. Peres, e-mail: C.Peres@uea.ac.uk short title: Hunting effects on vertebrate community 28 Abstract Overhunting is a leading contemporary driver of tropical forest wildlife loss. The absence or extremely low densities of large-bodied vertebrates disrupts plant-animal mutualisms and consequently degrades key ecosystem services. Understanding patterns of defaunation is therefore crucial given that most tropical forests worldwide are now “half- empty”. Here we investigate changes in vertebrate community composition and size structure along a gradient of marked anthropogenic hunting pressure in the Médio Juruá region of western Brazilian Amazonia. Using a novel camera trapping grid design deployed both in the understorey and the forest canopy, we estimated the aggregate biomass of several functional groups of terrestrial and arboreal species at 30 sites along the hunting gradient. Generalized linear models (GLMs) identified hunting pressure as the most important driver of aggregate biomass for game, terrestrial, and arboreal species, as well as nocturnal rodents, frugivores, and granivores. Local hunting pressure affected vertebrate community structure as shown by both GLM and ordination analyses. The size structure of vertebrate fauna changed in heavily hunted areas due to population declines in large-bodied species and apparent compensatory increases in nocturnal rodents. Our study shows markedly altered vertebrate community structure even in remote but heavily settled areas of continuous primary forest. Depletion of frugivore and granivore populations, and concomitant density-compensation by seed predators, likely affect forest regeneration in persistently overhunted tropical forests. These findings contribute to a better understanding of how cascading effects induced by historical defaunation operate, informing wildlife management policy in tropical peri-urban, rural and wilderness areas. Keywords: wildlife; mammals; birds; defaunation; density compensation; camera trapping. 29 Introduction Overhunting is the leading driver of contemporary defaunation inducing decisive large-bodied vertebrate population abundance declines in tropical forests worldwide (Peres and Palacios 2007; Fa and Brown 2009; Harrison et al. 2016). Bird and mammal abundance can decline by over 50% and 80%, respectively, in heavily hunted tropical forest areas (Benítez-López et al. 2017). Hunting-induced defaunation is a cryptic threat that obscures the integrity of both forest biotas and their fabric of ecological interactions (Redford 1992; Wilkie et al. 2011). The absence or low densities of large-bodied vertebrates disrupts mutualistic plant-animal interactions and, consequently, key ecosystem services that ensure the forest capacity to maintain its long-term baseline dynamics (Wright et al. 2007a; Terborgh et al. 2008; Harrison et al. 2013). Beyond potentially severe ecological impacts, defaunation can aggravate socioeconomic imperatives sustaining local livelihoods and food security of rural peoples for whom wild meat remains a critical source of animal protein (Nielsen et al. 2018; Nunes et al. 2019). In general, overhunting alters the structure of Neotropical vertebrate community promoting a directional decline in large-bodied mammal and bird populations (Peres 2000; Jerozolimski and Peres 2003). Large-bodied vertebrates, particular large-bodied primates, usually have low reproductive rates, which render their populations less resilient to hunting pressure (Bodmer et al. 1997). On the other hand, small-bodied mammals such as rodents and small primate species can benefit from intensely hunted areas due to release from negative interactions (e.g. predation and resource competition), changes in habitat structure, or both (Peres and Dolman 2000; Galetti et al. 2015a; Young et al. 2015). Additionally, there is often a concomitant shift in community composition and diversity resulting in reduced functional diversity of small mammals (Bovendorp et al. 2019). However, the type and magnitude of these responses are context-dependent, not least because of environmental conditions that are mediated by other anthropogenic impacts such as habitat fragmentation and land-use change (Young et al. 2015). Nevertheless, understanding changes in small mammal communities in defaunated areas is crucial because they typically operate as seed predators and exert an important role in plant community dynamics and diversity in tropical forests (Paine et al. 2016). Among the trophic guilds most affected by overhunting are frugivores and selective browsers, which are primarily hunted for subsistence and trade (Peres and Palacios 2007; Abernethy et al. 2013). While large carnivores, such as leopards and 30 jaguars, are often persecuted in retaliation for livestock depredation, their populations may also decline due to hunting-induced co-depletion of prey populations (Ripple et al. 2015). Although several large-bodied arboreal frugivores provide high-quality dispersal services for large-seeded plants (Peres and Roosmalen 2009), studies comparing how either arboreal or terrestrial vertebrates respond to hunting pressure remain scarce. Arboreal mammals are more susceptible to habitat disturbance, such as forest cover loss, compared to sympatric terrestrial counterparts, with the greatest negative numerical responses exhibited by key seed dispersal agents (Whitworth et al. 2019). Until recently, large vertebrates have been surveyed mainly through direct observational methods along line-transects, but as game species often avoid humans, they may gradually become less detectable in persistently hunted areas (Fragoso et al. 2016). However, the widespread use of camera traps to survey terrestrial vertebrates, and more recently their arboreal counterparts (Whitworth et al. 2016; Bowler et al. 2017), provide an opportunity to advance our understanding of how hunting affects vertebrate assemblages in relation to forest vertical stratification. A global analysis suggests that ~50% of all tropical forest areas is already partially defaunated of large-bodied mammals and 20% of all protected areas have been affected by hunting, mainly in Africa and Asia (Benítez-López et al. 2019). Although the Neotropics has so far experienced intermediate defaunation rates (Fa et al. 2002), some biomes contain a severely depleted contemporary vertebrate fauna, including the Brazilian semiarid Caatinga (Miranda et al. 2018) and the Atlantic Forest (Bogoni et al. 2018). Large-bodied vertebrate depletion has become increasingly pervasive even in some of the most remote parts of the Amazon (Peres and Lake 2003). However, terrestrial vertebrate populations in the Amazon are more resilient to overhunting compared to their large-bodied aquatic counterparts, likely because many vast upland areas remain inaccessible to hunters, generating a positive source-sink dynamic that can rescue overharvested populations in heavily hunted areas (Antunes et al. 2016). Here, we assessed the effects of a quasi-experimental large-scale gradient of hunting pressure in the Médio Juruá region of western Brazilian Amazonia — including heavily settled peri-urban areas, low-human-density landscapes used by local semi- subsistence communities and vast areas of non-hunted primary forest — on the community structure of both terrestrial and arboreal forest vertebrates. We hypothesized that game depletion altered vertebrate community structure by reducing the aggregate 31 biomass of large-bodied mammal and bird species (>1kg) in heavily hunted areas. In contrast, we expected that the abundance of sympatric small-bodied rodents would increase in overhunted areas where large mammals had been depleted through a mechanism of density compensation, such as competition release. We also conjectured that arboreal species would be more affected by hunting than their terrestrial counterparts because (1) large arboreal animals are often more vocal and noisier as they move through the canopy, and therefore more detectable; and (2) large primates in our study landscape were historically overhunted during the height of the rubber-boom, and their populations may not have completely recovered despite lower contemporary levels of hunting pressure. As such, frugivore-granivores, particularly arboreal species, likely represent the most affected trophic guild. These hypotheses were addressed using a novel camera-trapping design including terrestrial and arboreal surveys at 30 sites distributed throughout a hunting pressure gradient in Médio Juruá river. Our sampling design addresses the limited spatial replication of most studies and defines hunting pressure as a continuous rather than a binary or rank variable. We therefore provided evidence on the effects of hunting on tropical forest vertebrate communities through a robustly replicated design and discuss the possible implications of defaunation to long-term forest regeneration. Material and Methods Study Area The study was carried out in the Médio Juruá region of western Brazilian Amazonia (Fig. 1), including two large contiguous sustainable-use protected areas and adjacent landscapes containing two urban clusters. This represents the middle-third section of the Juruá River, the second-longest white-water tributary of the Amazon River. The two protected areas include the 253,227 ha Médio Juruá Extractive Reserve (RESEX Médio Juruá, 5º33'54"S, 67º42'47"W), created in 1997 and legally occupied by ~ 2,000 people distributed across 13 villages; and the 632,949 ha Uacari Sustainable Development Reserve (RDS Uacari, 5º43'58"S, 67º46'53"W) created in 2005, where ~ 1,200 people occupy 32 villages. The nearest towns are Carauari (population ≈ 28,000 residents), located 88 fluvial km downstream of the RESEX Médio Juruá, and Itamarati (population ≈ 8,000), located 120 fluvial km upstream of the RDS Uacari (IBGE 2018). The Médio 32 Juruá region has a wet tropical climate with a mean annual temperature of 27.1°C and a mean annual rainfall of 3,679 mm, with the wettest period between November and April. Two different forest types comprise the study landscape: seasonally-flooded (várzea) forests, which account for ~20% of the study region, characterized by enriched Andean alluvial soils and lower floristic diversity, and the dominant (~80%) unflooded forest (terra firme), which exhibits higher floristic diversity and comparatively lower soil fertility (Hawes and Peres 2016). The current study was performed in unflooded forest on paleo-várzea sediments, thereafter as terra firme for simplicity, but we recognize that these forests, may diverge in their floristic macromosaics from so-called terra firme forests (Assis et al. 2015). Our sapling sites were established along areas that had experienced subsistence and commercial hunting to varying degrees but had no recent history of clear-cuts, wildfires, and timber extraction. We selected 30 sites spanning a wide gradient of hunting pressure along ~600 km non-linear (fluvial) distance, from the towns of Carauari to Itamarati (Fig. 1, Table S1). The sites selection was based on both distance to human settlements, physical accessibility and previous studies carried out in the area by the Médio Juruá Project (PMJ). At each of these 30 sites we established a standardized sampling protocol to obtain data on vertebrate abundance using terrestrial and arboreal camera traps; forest structure and composition; and other environmental variables that potentially influence vertebrate abundance, described as follows. 33 Figure 1. Map of the study area in the Médio Juruá region of Western Brazilian Amazonia. The reserve boundaries of the RESEX Médio Juruá and RDS Uacari are outlined in black. Coloured circles indicate the location of our 30 sampling grids. Circles are colour-coded according to our proxy of hunting pressure (see colour gradient). The main Juruá River channel is outlined in white. Grey background shows elevation where lighter shading represents higher terrain. The panel on the bottom right is a schematic drawing of the camera trap grid deployment within our study grids. Red and blue circles represent cameras deployed in the understorey and the canopy, respectively. Green rectangle at the center of the camera trapping grid indicates a 0.25-ha tree plot (100 m × 25 m). Terrestrial and Arboreal Camera Trapping Terrestrial and arboreal camera trapping were conducted from July 2017 to May 2019. We employed a modified camera-trapping design using a 4.5-ha terrestrial grid containing 16 camera-trap stations (4 × 4, spaced by 100 m), which were combined with 34 four arboreal camera trap-stations (hereafter, CTS) spaced by 300 m. Thus, each of our 30 grids contained 20 cameras (16 terrestrial and 4 arboreal), amounting to total of 480 CTS placed near the ground and 120 CTS placed in the canopy. This CTS deployment prioritized efficient sampling at the grid-scale ensuring a high probability of detection events within the area covered by the grid. We established this camera-trapping grid at each pre-selected site considering a minimum spacing of 1 km when grids were in the same landscape (Fig. 1) Arboreal camera traps were placed at ~15 m height in the main bifurcation of large low-angle branches of canopy trees to intercept natural canopy pathways, thereby maximizing detection probability. Terrestrial camera traps were deployed on basal tree boles at 15 cm from the ground to ensure detection of not only large-bodied mammal and bird species, but also small-bodied rodents and marsupials (see Palmeirim et al. 2019). All CTS were unbaited, and we did not necessarily select apparently favourable terrestrial camera-trap sites (e.g. game trails) as they were deployed systematically, but we avoided major obstacles in the field of view and the understorey was slightly cleared to maximize detectability. All CTS were exposed over a minimum period of 30 camera-trap-nights (CTNs; mean ± SD, 41.4 ± 23.7 nights per CTS). At each CTS, we recorded the (1) camera code, (2) geographic coordinates, and (3) date and time of deployment and removal. All photographs and videos were analysed based on species identifications. Consecutive records of the same species were defined as independent whenever they were spaced apart by intervals longer than 60 min. For validation of species identification in case of any margin of ambiguity, 3-5 records were sent to specialists of individual taxa. Closely related species that could not be identified to species level were grouped into a single morphospecies. Records of domestic animals, small passerines, bats, lizards, and insects were excluded from the analyses. We extracted all photo metadata including date and time of records using the camtrapR 1.1 R package (Niedballa et al. 2016). For data correction from cameras with programming problems, we used data obtained in the field during both camera deployment and removal. We therefore produced a database containing the total number of records per species (or morphospecies) by CTS and their respective sampling effort (hours). 35 Vertebrate abundance and biomass All CTS records within any given grid were summed and divided by the total sampling effort per grid and standardized by 100 CTNs to derive a species-specific abundance index for each of the 30 grids. This index was then multiplied by the species body mass (mean adult male and female) and mean observed group size (number of individuals in group-living species) in the study area to obtain an approximate metric of vertebrate biomass per sampling grid. Data on body mass were obtained from Wilman et al. (2014) and Peres (1993). Data on mean group size were derived from 3 years of monthly line-transect census effort along 95 transects placed throughout the same Juruá meta-landscape (each of which 3-4 km in length) carried out by Carlos Peres and collaborators (unpubl. data). Species were initially classified into either game or non-game species according to Abrahams et al. (2017) and Carlos Peres (unpubl. data), considering both commercial and subsistence hunting. All species were classed within five trophic levels based on a rank of dietary energy content (see Almeida-Rocha et al. 2017) and dietary data available in Wilman et al (2014) (Table S3). The lowest trophic level (1) thus includes species with high proportions of low-energy dietary items (i.e. foliage), whereas the highest trophic level (5) is represented by hyper-carnivores that exclusively consume vertebrates. We also distinguished all species into either terrestrial or arboreal depending on their locomotion mode and vertical stratification according to (Paglia et al. 2012). For scansorial species, which use both strata, we assigned them into the group in which they were recorded most frequently by our camera traps (Table S2). For nocturnal rodents, we summed the species-specific biomass estimates for spiny rats (Proechimys spp.) and morphospecies identified as either small (~100g) or very small (~15g) rodents. Grid scale biomass estimates were aggregated into ten functional groups that were not necessarily mutually exclusive, including (1) game species, (2) nocturnal rodents, (3) arboreal species, (4) terrestrial species, (6) browsers, (7) grazers, (8) frugivores, (9) omnivore- insectivores, and (10) carnivores. Proxy of hunting pressure We built a proxy of hunting pressure based on the intensity of human activity: geographic distance to and size of human settlements, including villages and towns. Previous studies in the same area have shown that distance from urban centres represents 36 a good proxy for the anthropogenic impact on large vertebrate abundance (Nichols et al. 2013; Abrahams et al. 2017). We measured the Euclidean distance from each camera grid centroid to all villages and the dry-season navigation (fluvial) distance to the towns using ArcGIS10.3. Human population size of each town was derived from IBGE (2018) census data, while village size was obtained from the Projeto Médio Juruá (PMJ) and the Sustainable Amazon Foundation (FAS) databases. Hunting pressure was therefore defined by the equation: 𝑛 𝑆(𝑣𝑖𝑙) 𝑆(𝑐𝑎𝑓) 𝑆(𝑖𝑡𝑎) 𝐻𝑃 =∑ + + √𝑑(𝑣𝑖𝑙) √𝑑(𝑐𝑎𝑓) √𝑑(𝑖𝑡𝑎) 𝑖 Where S represents the human population size at any village (vil) or towns (caf = Carauari; ita=Itamarati); d represents the Euclidean distance from each grid centroid to the nearest community or the dry-season navigation (fluvial) distance to the towns. Environmental variables For each of the 30 sites, we compiled data on all major environmental variables that could affect vertebrate abundance and biomass besides the hunting pressure, namely (1) the proportion of várzea forest area within a 40-km² buffer area (9.75 km wide) around our 30 camera-trapping grids. This was based on a 2018 Landsat 7 satellite image, which was classified using the spatial analyst tool in ArcGIS10.3; (2) water level, defined as the median Juruá River water level obtained over a 38-year time-series, corresponding to the Julian day mid-point of the camera trapping survey period within each grid. Water level data were obtained using daily readings, recorded from 1st January 1973 to 31st December 2010 at the nearby meteorological station of Porto Gavião, Carauari, Amazonas (ANA 2019). This variable provides a proxy of hydrological seasonality. Both, proportion of varzea forest and water level are strongly associated with animal abundance due to the seasonal movements of terrestrial vertebrates between várzea and terra firme forests (Costa et al. 2018); (3) the mid Julian day of the survey period within each grid; (4) density of live trees ≥10 cm diameter at breast height (DBH) within each permanent 0.25- ha (100 m × 25 m) tree plot that we surveyed inside our camera-trapping grids; and (5) soil cation exchange capacity (CEC), which we measured based on soil samples collected at each tree plot, which were analysed at the Soil Chemistry Laboratory of the National Institute for Amazon Research (INPA), Manaus. Soil chemistry analysis conducted here 37 included major macronutrients such as Ca, Mg, K and P measured as cmol kg–¹which were later pooled into a single index of soil fertility. Soil fertility is a strong predictor of vertebrate biomass in Amazonian forests, particularly primary consumers (Peres 2008). Data analysis We examined the effects of all covariates on the aggregate vertebrate biomass for each functional group: (1) game, (2) arboreal, (3) terrestrial, and (4) nocturnal rodent species, and (5) all five trophic guilds. In doing so, we assess the degree to which hunting pressure, water level, Julian day, proportion of várzea forest, tree density and soil fertility affects the (i) aggregate biomass and (ii) community composition of birds and mammals. First, we visually examined response variables through histograms. Variables with non-normal distributions were transformed using the bestNormalize 1.4.2 R package (Peterson 2017), which selects the best normalization data transformation. We calculated the variance inflation factors (VIFs) to test for multicollinearity in explanatory variables, where VIFs<4 indicate low multicollinearity (Zuur et al. 2010). None of our explanatory variables were strongly correlated so they were all entered into generalised linear models (GLMs). The spatial structure of residual models was tested using the Moran’s I autocorrelation index (Gittleman and Kot 1990). All analyses were conducted in R 3.5.3 (R development core team 2019). To examine the relative importance of our environmental variables on aggregate vertebrate biomass we applied a model averaging approach using the MuMIN 1.43.15 package in R (Bartón, 2016). Model averaging calculates multiple regression models from all possible combinations of variables using the dredge function and ranks models according to the Akaike`s information criteria (AIC). We considered as ‘best’ models those for which ΔAIC<2. When more than one model was selected, we built an average model using the model.avg function and determined the importance of the explanatory variables for each response variable from their frequency of occurrence in these models. Vertebrate community composition along the gradient of hunting pressure was further investigated through Principal Coordinates Analysis (PCoA) based on the Bray- Curtis dissimilarity matrix, using the pcoa function in the ape 5.3 R package (Paradis et al 2004). We first considered the (i) relative abundance, and (ii) aggregate biomass of the entire assemblage which were then subdivided into game and non-game species. Additionally, we performed a general linear model using the environmental covariates 38 and the hunting pressure as predictors of the scores obtained from Axes 1 and 2 of the PCoA based on the relative abundance and aggregate biomass estimates for each functional group. Finally, to investigate changes in the size structure of terrestrial and arboreal vertebrates along the hunting gradient, we built cumulative distribution functions (CDFs) of the pooled body mass data for all independently recorded species in each camera- trapping grid. Species body mass ranged over four orders of magnitude from ~15 g to ~150,000 g. For each CDF function, we calculated the ‘area under the curve’ (AUC), which was then used in a non-linear regression model to investigate how this assemblage- wide metric of size structure was affected by hunting pressure. Here, higher AUC values indicate greater dominance of small-to mid-sized species, whereas lower values indicate assemblages more heavily dominated by large-bodied species. For this analysis, we removed two outliers (two lightly hunted sites at Tabuleiro) because they represented sites near secondary forest areas that had been subjected to anthropogenic disturbances. Results Arboreal and terrestrial vertebrates Based on 22,005 CTNs, we recorded 10,284 independent detections of 71 vertebrate species (or species groups), including 57 mammals, 13 birds and 1 reptile (Table S2). A total of 21 taxa were recorded exclusively by arboreal camera traps (5,715 CTNs), 30 exclusively by terrestrial camera traps (16,290 CTNs), and 18 were recorded by both. Terrestrial vertebrates were represented by 39 species, of which agoutis (Dasyprocta spp.) and spiny rats (Proechimys spp) were the most abundant mammals detected, while small tinamous (Crypturellus spp.) and trumpeters (Psophia leucoptera) were the most frequently detected birds. Collectively, these species accounted for 39.6% of all terrestrial camera trapping records. Arboreal vertebrates were represented by 33 species, of which prehensile-tailed porcupines (Coendou spp.), large-headed capuchin monkeys (Sapajus macrocephalus), arboreal echimyid rodents, and moustached tamarins (Saguinus mystax) were most frequently detected, accounting for 41.1% of all arboreal records. Paca (Cuniculus paca), agouti (Dasyprocta spp.), collared peccary (Pecari tajacu) and grey brocket deer (Mazama nemorivaga) contributed with 50.3% of the aggregate terrestrial biomass, whereas large-headed capuchins (S. macrocephalus), white-fronted capuchins (Cebus unicolor) and black spider monkeys (Ateles chamek) accounted for 60.5% of the 39 aggregate canopy biomass. Considering all 71 species detected, only seven arboreal and 18 terrestrial species were habitually harvested throughout our study landscape (Table S2). Patterns of aggregate biomass GLM modelling showed that the aggregated biomass of most functional groups was significantly affected by site-specific hunting pressure (Fig. 2, Fig. S1). The strongest effect was observed for game species, which showed a steep decline in their aggregate biomass (Fig. 3) at sites within about 15 km of the urban centre of Carauari where hunting pressure was the highest. The overall biomass of both arboreal and terrestrial vertebrate species declined in heavily hunted areas. However, while arboreal vertebrate biomass was exclusively affected by hunting pressure, biomass of terrestrial species was positively affected by floodplain water level and percentage of varzea forest (Fig. 2, Table S3). Models also showed that hunting pressure had a negative effect on the aggregate biomass of frugivores and grazers (Fig. 2, Table S3). Conversely, the overall biomass of small- bodied rodents, represented mostly by spiny rats, increased in heavily hunted areas compared to non-hunted areas (Fig. 3). Timing of deployment of camera-traps and soil fertility also positively affected the biomass of omnivores (Fig. 2). 40 Fig. 2 Explanatory variables retained in the modelling average approach explaining the aggregate biomass of game species, nocturnal rodents, arboreal species, terrestrial species, browsers, grazers, frugivores, omnivores and carnivores. Predictor variables are listed on the left of each panel: Tree density: abundance of trees within 0.25-ha plots, proportion of várzea forest, CEC: soil cation exchange capacity, and Julian Day: mid Julian day of survey period. Coefficients and 95% confidence intervals of predictors are shown in the panels. Blue and red circles indicate significantly positive and negative effect sizes, respectively. Black circles indicate non-significant effects. 41 Fig. 3 Partial regression fits as a function of hunting pressure for the aggregated biomass of game species, nocturnal rodents, arboreal species, terrestrial species, grazers and frugivores. Y-axes show Yeo-Johnson transformed variables. X-axes are Lamberts W × F-transformed for game, grazers and frugivores species, sqrt-transformed for nocturnal rodents; and log-transformed for arboreal and terrestrial species. Grey shading along regression lines represents 95% confidence regions. Compositional changes Vertebrate community structure also varied strongly in relation to site-specific hunting pressure. This pattern can be observed in ordination space by the narrow scatter in species composition and aggregate biomass at heavily hunted sites (Fig. 4). The greater convergence in community structure at sites exposed to similar levels of hunting pressure was further confirmed in GLMs by the strong relationship between the first PCoA axis and hunting pressure (Table S4). Hunting pressure significantly explained the species ordination both in terms of relative abundance and aggregate biomass. Other variables were also related to numerical abundance of all vertebrates, including water level and proportion of várzea forest. 42 Fig. 4 Ordination based on Principal Coordinates Analysis (PCoA) using the Bray-Curtis dissimilarity matrix of vertebrate species assemblages in our study region in which level of hunting pressure increases from blue to red solid dots. Top and bottom panels show ordination plots for relative abundance and aggregate biomass, respectively. Left, central, and right panels represent all species combined, game species, and non-game species, respectively. The percentage of variance explained is reported (in brackets) for each PCoA axis. Assemblage size structure Mean vertebrate body mass detected across our camera-trapping grids ranged from 6.18 kg (SE = 1.66, N = 31) in our most hunted site to 23.70 kg (SE = 5.36, N = 43) in our least hunted site. As such, we expected an increase in the area underneath each CDF function (i.e. AUCs) at sites historically exposed to persistent hunting pressure due to greater numerical dominance of small-bodied species. AUC values ranged from 5.32 43 in our most hunted site to 4.897 in our least hunted site. The AUC values tended to be greater at sites exposed to increasingly heavier hunting pressure (Fig. 5). Fig. 5 Area Under the Curve (AUC) obtained using the cumulative distribution function (CDF) in terms of vertebrate body size and abundance built for each of our camera-trapping grids. Body mass ranks range from the smallest to the largest vertebrate species recorded within each grid (left panel). Blue and red curves represent two hypothetical body size CDFs, in which the AUC metric in (2; hatched area) is dominated by a larger proportion of large-bodied species compared to (1; blue dotted area). Non-linear regression of AUC values as a function of site-specific hunting pressure (right panel). Grey shading around the regression line represents the 95% confidence interval. Discussion Our study replicated across 30 terra firme sites along a marked hunting pressure gradient of western Brazilian Amazonia spanning a wide spectrum of peri-urban, rural, and wilderness areas, shows the dominant role of hunting pressure in terms of top-down control of forest vertebrate assemblages. In other words, the effects of human hunting pressure were clearly more important than those of key habitat features in explaining variation in aggregate vertebrate biomass, particularly of game species. Although the effect of hunting pressure on the aggregate biomass of the entire vertebrate assemblage was significant, this effect was most pronounced in predicting the biomass and community structure of species frequently harvested by subsistence hunters. 44 Vertebrate biomass across the hunting gradient Game species harvested for food along the middle section of the Juruá River, which were comprised mainly of lowland paca, collared peccary, howler monkey, spider monkey, red and grey brocket deer, and curassow (Abraham et al, 2017), had a much greater contribution to the aggregate biomass of the vertebrate community in non-hunted areas. This biomass contribution at the grid scale ranged from 47.8% in our least hunted site to only 6.2% in our most hunted site. This overall reduction in the biomass of large- bodied species is the first stage of a wider defaunation process described for other depleted tropical forest landscapes both across Amazonia (Peres and Palacios 2007) and elsewhere in the tropics (Benítez-López et al. 2017), and reflects the patterns of local extinctions in low-fecundity large-bodied mammals across the Neotropics (Bogoni et al. 2020). This may trigger several poorly documented trophic cascades, including density compensation of post-dispersal seed predators, such as the apparent release of terrestrial echimyid rodents as documented in this study. Urban-centric hunting pressure Large-bodied vertebrate declines were most evident in the wider neighbourhood of the main regional scale urban centre. We therefore reinforce findings that large-bodied vertebrate depletion in the Amazon is primarily driven at the landscape scale, rather than by local processes alone (Abrahams et al. 2017). In other words, although populations of low-fecundity, large-bodied species are depleted near small semi-subsistence rural settlements, the highest depletion rates were evidenced in peri-urban areas of larger towns that had already been well established as trade posts well before the heyday of the rubber- boom. Sites near the smaller town of Itamarati also exhibited lower biomass estimates compared to neighbouring surveyed sites, which is consistent with a peri-urban pattern of wildlife depletion, rather than one driven by any of the other environmental gradients investigated. In any case, the depletion effect of the smaller town on large-bodied game biomass was considerably weaker than that of the larger town, which is older and three- fold larger in terms of human population size, thereby markedly increasing exploitation pressure on both terrestrial and aquatic resources. Surprisingly, a large village ~50 km from Carauari and containing ~650 residents did not show marked changes in vertebrate community structure, compared to a non-hunted baseline. The urban-centric effect is not necessarily a result of commercial hunting pressure, as many urban dwellers in the 45 Amazon continue to subsidize their households with subsistence wild meat (Chaves et al. 2017). Density compensation by nocturnal rodents In addition to the direct effects on game species, we also detected an indirect effect of partial defaunation on the abundance of small mammals, particularly spiny rats (Proechimys spp.) which can be highly abundant in both várzea and terra firme forest areas along the Juruá River (Malcolm 2005). Elevated numbers of spiny rats in overhunted areas likely can indicate density compensation resulting from numeric depletion of large-bodied mammals, particularly grazers such as collared and white- lipped peccaries. Although small mammals have fast life histories due to high fecundity rates (Dobson and Oli 2007), and may respond rapidly to other environmental factors, our results fail to show an influence of local habitat variables and seasonality on their abundance. We therefore attribute the negative relationship between small and large- bodied mammals along ~ 600 km fluvial section of the Juruá River we surveyed to some form of competitive release, most likely of ungulates, which is consistent with similar compensatory dynamics in heavily hunted areas elsewhere in the tropics (Keesing and Young 2014; Galetti et al. 2015a) Higher population densities of small-bodied rodents in semi-defaunated areas potentially results in several ecological and human health impacts. Higher rodent densities often favour the dominance of generalist species, thereby reducing overall small mammal functional diversity (Pardini et al. 2009; Bovendorp et al. 2019). Furthermore, higher numbers of spiny rats likely disrupt the balance of seed predation interactions for several large-seeded plant species (Galetti et al. 2015b), which can cause a significant change in forest composition, given the importance of seed predators in maintaining plant diversity. In terms of human health, positive numerical responses in rodents may result in increases in the risk of infectious diseases for which rodents are competent reservoirs, such as hantavirus and leishmaniasis (Ashford 2000; Young et al. 2014; Muylaert et al. 2019). Vertebrate responses to hunting We expected that forest canopy structure could affect the biomass of arboreal vertebrates, but tree connectivity is probably a better driver of arboreal mammal abundance than tree density (Whitworth et al. 2019). Sensitivity to hunting in arboreal 46 mammals is consistent with the fact that hunting pressure was the only variable affecting their aggregate biomass. However, there were no differences in the effect size of hunting pressure in explaining the aggregate biomass of either arboreal or terrestrial vertebrates, suggesting that both are equally affected. Considering that arboreal mammals were primarily represented by primates, which are often high-quality dispersal agents of large- seeded plants (Chapman 1995; Lambert and Garber 1999), we suggest that reduced abundance of large primates may result in a recruitment bottleneck for at least some large- seeded tree and liana species (Nunez-Iturri et al. 2008; Peres et al. 2016). In addition, we strongly advocate that canopy camera trapping should be used as a complementary method to line-transect censuses. This method recorded 21 mammal species that were not detected by terrestrial camera trapping, including rarely recorded species in census such as canopy didelphids (e.g. Caluromys lanatus, Glironia venusta), and kinkajous (Potos flavus). The aggregate biomass of terrestrial species was negatively affected by hunting pressure, but water level had a positive effect on their abundance. Strictly terrestrial vertebrates are forced to move laterally away from várzea forests and into adjacent unflooded areas with the rise of floodwaters. These lateral movements are also related to staggered patterns of food resource availability across the várzea-terra firme interface, with ripe fruit peaks in terra firme forests occurring at the onset of the wet season, whereas this occurs much later during the late high-water season in várzea forests (Schongart et al. 2002; Haugaasen and Peres 2007; Hawes and Peres 2016). This likely attracts terrestrial vertebrates into neighbouring unflooded areas during the high-water season and, conversely, into seasonally-flooded forest as the water level recedes, exposing an attractive supply of fruits and seeds newly deposited on the forest floor (Haugaasen and Peres 2007). The annual flood pulse thus promotes seasonal movements in terrestrial species between várzea and terra firme forests, which can account for seasonal differences in terrestrial vertebrate abundance (Costa et al. 2018) in addition to the underlying turnover in tree community structure between these two neighbouring forest types (Hawes and Peres 2014). Considering trophic guilds, frugivores and grazers were the only functional groups whose aggregate biomass was affected by hunting pressure. These trophic guilds play a decisive role as either seed dispersers or seed predators of large-seeded plants, thereby affecting forest regeneration and dynamics (Wright et al. 2000; Dirzo et al. 2007). 47 Consequently, low biomass density of large-bodied frugivores can disrupt the balance between recruitment and mortality of large-seeded plants, thereby facilitating compositional transitions towards forests that are more heavily dominated by small- seeded, fast-growing species (Wright et al. 2007b; Peres et al. 2016) In this study, we did not record any white-lipped peccaries, a large-group-living ungulate forming herds of up to 250 individuals in other Amazonian forests (Haugaasen and Peres 2007) and could achieve 1000 individuals in the Médio Juruá region (Carlos Peres, pers. obs). This species represents an important source of protein for villagers and exerts a strong role in forest dynamics linked to seedling recruitment (Silman et al. 2003; Keuroghlian and Eaton 2009). This is likely a sampling artefact of short-term camera- trapping because herd return-times to any given area could be months, if not years, apart. A complementary method is therefore required to understand white-lipped peccary movements across vast landscapes and how they respond to offtake mortality. However, information from local dwellers indicate that white-lipped peccary herds, which cannot be overlooked, were conspicuously absent from the most hunted sites within 30 km of Carauari and Itamarati. Our findings also suggest that some harvest-sensitive species that are expected to be important targets for hunters were in fact, not heavily affected by hunting pressure in our study area. For example, large cats are often persecuted in retaliation for livestock depredation (Michalski et al. 2006), and have been heavily hunted in the past since the post-rubber boom skin trade of southwestern Amazonia (Antunes et al. 2016). However, large cats have not been severely depleted throughout our study landscape, and we recorded a jaguar in the vicinity of Carauari, suggesting that areas depleted of large vertebrates can still retain the largest apex predator in the Neotropics. Conclusion We have shown that even remote areas of relatively lowest human pressure Amazonian forests have undergone major shifts in vertebrate community structure in both arboreal and terrestrial species whenever they fall within the depletion envelope of larger human settlements. This suggests that marked effects of overhunting are concentrated in peri-urban areas outside protected areas, at least in landscapes that continue to benefit from healthy source-sink dynamics. Less pronounced effects of hunting within the two forest reserves are likely related to lower human densities and low dependence on 48 bushmeat by local dwellers for whom fish is the most important source of protein for most of the year in this region of Amazonia (Endo et al. 2016). Given the critical importance of protected areas in maintaining harvest-sensitive wildlife populations even outside their boundaries, community-based management is potentially a key strategy to reconcile wildlife conservation and food security for forest dwellers. Wildlife management strategies should be firmly grounded on applied science and considers how different patterns of defaunation may induce trophic cascade pathways that affect forest dynamics. Tried-and-tested conservation strategies that can promote hunting sustainability through community-based management remain at best embryonic (Campos-Silva et al., 2017), but this is one of the few available solutions in low- governance regions if we are to curb declines in forest wildlife and the long-term ecosystem services provision. Acknowledgments We are grateful to Associação dos Produtores Rurais de Carauari (ASPROC), Centro Estadual de Unidades de Conservação do Amazonas (CEUC/SDS/AM), Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio), Operação Amazônia Nativa (OPAN) and Projeto Médio Juruá (PMJ) for support on fieldwork logistics. Instituto Nacional de Pesquisas da Amazônia (INPA) for support on laboratory analysis. Finally, we are grateful to all local villagers of the RESEX Médio Juruá and RDS Uacari and dwellers of Carauari and Itamarati for their hospitality, friendship and trust. Funding sources This work was supported by the National Geographic Society [grant number: 265943], Rufford Foundation [grant number: 21911-1], Society for Conservation Biology [LACA Professional Development Award], and a DEFRA (Darwin Initiative for the Survival of Species) grant to CAP. ABS was granted a PhD studentship from Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) [Finance Code 001]. References Abernethy KA, Coad L, Taylor G, et al (2013) Extent and ecological consequences of hunting in Central African rainforests in the twenty-first century. Philos Trans R Soc B Biol Sci 368:. https://doi.org/10.1098/rstb.2012.0303 49 Abrahams MI, Peres CA, Costa HCM (2017) Measuring local depletion of terrestrial game vertebrates by central-place hunters in rural Amazonia. PLoS One 12:1–25. https://doi.org/10.1371/journal.pone.0186653 Almeida-Rocha JM de, Peres CA, Oliveira LC (2017) Primate responses to anthropogenic habitat disturbance: A pantropical meta-analysis. Biol Conserv 215:30–38. https://doi.org/10.1016/j.biocon.2017.08.018 ANA (2019) Agencia Nacional das Aguas. https://www.ana.gov.br/ Antunes AP, Fewster RM, Venticinque EM, et al (2016) Empty forest or empty rivers? A century of commercial hunting in Amazonia. Sci Adv 2:. https://doi.org/10.1126/sciadv.1600936 Ashford RW (2000) The leishmaniases as emerging and reemerging zoonoses. Int J Parasitol 30:1269–1281. https://doi.org/10.1016/B978-0-12-803265-7.00007-5 Assis RL, Haugaasen T, Schöngart J, et al (2015) Patterns of tree diversity and composition in Amazonian floodplain paleo-várzea forest. J Veg Sci 26:312–322. https://doi.org/10.1111/jvs.12229 Bartón, K., 2016. multi-model inference R package version, 1(6). Benítez-López A, Alkemade R, Schipper AM, et al (2017) The impact of hunting on tropical mammal and bird populations. Science (80- ) 356:180–183. https://doi.org/10.1126/science.aaj1891 Benítez-López A, Santini L, Schipper AM, et al (2019) Intact but empty forests? Patterns of hunting-induced mammal defaunation in the tropics. PLoS Biol 17:1– 18. https://doi.org/10.1371/journal.pbio.3000247 Bodmer RE, Eisenberg JF, Redford KH (1997) Hunting and the likelihood of extinction of Amazonian mammals. Conserv Biol 11:460–466. https://doi.org/10.1046/j.1523-1739.1997.96022.x Bogoni J., Ferraz K., Peres CA (2020) Extent, intensity and drivers of mammal defaunation: a continental-scale analysis across the neotropics. Nat Sci Reports 10:1–16. https://doi.org/10.1038/s41598-020-72010-w. Bogoni JA, Pires JSR, Graipel ME, et al (2018) Wish you were here: How defaunated is the Atlantic Forest biome of its medium- to large-bodied mammal fauna? PLoS One 13:1–23. https://doi.org/10.1371/journal.pone.0204515 Bovendorp RS, Brum FT, McCleery RA, et al (2019) Defaunation and fragmentation erode small mammal diversity dimensions in tropical forests. Ecography. 42:23– 35. https://doi.org/10.1111/ecog.03504 50 Bowler MT, Tobler MW, Endress BA, et al (2017) Estimating mammalian species richness and occupancy in tropical forest canopies with arboreal camera traps. Remote Sens Ecol Conserv 3:146–157. https://doi.org/10.1002/rse2.35 Campos-Silva JV, Peres CA, Antunes AP, et al (2017) Community-based population recovery of overexploited Amazonian wildlife. Perspect Ecol Conserv 15:266–270. https://doi.org/10.1016/j.pecon.2017.08.004 Chapman CA (1995) Primate seed dispersal: coevolution and conservation implications. Evol Anthropol Issues, News, Rev 4:74–82. https://doi.org/10.1002/evan.1360040303 Chaves WA, Wilkie DS, Monroe MC, Sieving KE (2017) Market access and wild meat consumption in the central Amazon, Brazil. Biol Conserv 212:240–248. https://doi.org/10.1016/j.biocon.2017.06.013 Costa HCM, Peres CA, Abrahams MI (2018) Seasonal dynamics of terrestrial vertebrate abundance between Amazonian flooded and unflooded forests. 1–22. https://doi.org/10.7717/peerj.5058 Dirzo R, Mendoza E, Ort P (2007) Size-related differential seed predation in a heavily defaunated neotropical rainforest. Biotropica. 39:355–362. https://doi.org/10.1111/j.1744-7429.2007.00274.x Dobson FS, Oli MK (2007) Fast and slow life histories of maammals. Ecoscience 14:292–297. https://doi.org/10.1139/z11-033 Endo W, Peres CA, Haugaasen T (2016) Flood pulse dynamics affects exploitation of both aquatic and terrestrial prey by Amazonian floodplain settlements. Biol Conserv 201:129–136. https://doi.org/10.1016/j.biocon.2016.07.006 Fa JE, Brown D (2009) Impacts of hunting on mammals in African tropical moist forests: A review and synthesis. Mamm Rev 39:231–264. https://doi.org/10.1111/j.1365-2907.2009.00149.x Fa JE, Peres CA, Meeuwig J (2002) Bushmeat exploitation in tropical forests: an intercontinental comparison. Conserv Biol 16:232–237. https://doi.org/10.1046/j.1523-1739.2002.00275.x Fragoso JMV, Levi T, Oliveira LFB, et al (2016) Line transect surveys underdetect terrestrial mammals: Implications for the sustainability of subsistence hunting. PLoS One 11:1–18. https://doi.org/10.1371/journal.pone.0152659 Galetti M, Bovendorp RS, Guevara R (2015a) Defaunation of large mammals leads to an increase in seed predation in the Atlantic forests. 3:824–830 51 Galetti M, Guevara R, Neves CL, et al (2015b) Defaunation affect population and diet of rodents in Neotropical rainforests. Biol Conserv 190:2–7. https://doi.org/10.1016/j.biocon.2015.04.032 Gittleman JL, Kot M (1990) Adaptation: statistics and a null model for estimating phylogenetic effects. Syst Zool 39:227–241. https://doi.org/10.2307/2992183 Harrison RD, Sreekar R, Brodie JF, et al (2016) Impacts of hunting on tropical forests in Southeast Asia. Conserv Biol 30:972–981. https://doi.org/10.1111/cobi.12785 Harrison RD, Tan S, Plotkin JB, et al (2013) Consequences of defaunation for a tropical tree community. Ecol Lett 16:687–694. https://doi.org/10.1111/ele.12102 Haugaasen T, Peres CA (2007) Vertebrate responses to fruit production in Amazonian flooded and unflooded forests. Biodivers Conserv 16:4165–4190. https://doi.org/10.1007/s10531-007-9217-z Hawes JE, Peres CA (2016) Patterns of plant phenology in Amazonian seasonally flooded and unflooded forests. Biotropica 48:465–475. https://doi.org/10.1111/btp.12315 Hawes JE, Peres CA (2014) Fruit-frugivore interactions in Amazonian seasonally flooded and unflooded forests. J Trop Ecol 30:381–399. https://doi.org/10.1017/S0266467414000261 IBGE (2018) Instituto Brasileiro de Geografia e Estatística. In: Censo 2018. https://www.ibge.gov.br/ Jerozolimski A, Peres CA (2003) Bringing home the biggest bacon: A cross-site analysis of the structure of hunter-kill profiles in Neotropical forests. Biol Conserv 111:415–425. https://doi.org/10.1016/S0006-3207(02)00310-5 Keesing F, Young TP (2014) Cascading consequences of the loss of large mammals in an African Savanna. Bioscience 64:487–495. https://doi.org/10.1093/biosci/biu059 Keuroghlian A, Eaton DP (2009) Removal of palm fruits and ecosystem engineering in palm stands by white-lipped peccaries (Tayassu pecari) and other frugivores in an isolated Atlantic Forest fragment. Biodivers Conserv 18:1733–1750. https://doi.org/10.1007/s10531-008-9554-6 Lambert JE., Garber PA (1999) Evolutionary and ecological implications of primate seed dispersal. Am J Primatol 45:9–28. https://doi.org/10.1002/(SICI)1098- 2345(1998)45:1<9::AID-AJP3>3.0.CO;2-%23 Malcolm, J. R., Patton, J. L., & da Silva MNF (2005) Along an Amazonian white water river. In: Mammalian diversification: from chromosomes to phylogeography. p 335 52 Michalski F, Boulhosa RLP, Faria A, Peres CA (2006) Human – wildlife conflicts in a fragmented Amazonian forest landscape : determinants of large felid depredation on livestock. 9:179–188. https://doi.org/10.1111/j.1469-1795.2006.00025.x Miranda EBP, Jácomo ATDA, Tôrres NM, et al (2018) What are jaguars eating in a half-empty forest? Insights from diet in an overhunted Caatinga reserve. J Mammal 99:724–731. https://doi.org/10.1093/jmammal/gyy027 Muylaert RL, Sabino-Santos G, Prist PR, et al (2019) Spatiotemporal dynamics of hantavirus cardiopulmonary syndrome transmission risk in Brazil. Viruses 11:. https://doi.org/10.3390/v11111011 Nichols E, Uriarte M, Peres CA, et al (2013) Human-induced trophic cascades along the fecal detritus pathway. PLoS One 8:. https://doi.org/10.1371/journal.pone.0075819 Niedballa J, Sollmann R, Courtiol A, Wilting A (2016) camtrapR: an R package for efficient camera trap data management. Methods Ecol Evol 7:1457–1462. https://doi.org/10.1111/2041-210X.12600 Nielsen MR, Meilby H, Smith-Hall C, et al (2018) The importance of wild meat in the global south. Ecol Econ 146:696–705. https://doi.org/10.1016/j.ecolecon.2017.12.018 Nunes AV, Peres CA, Constantino P de AL, et al (2019) Irreplaceable socioeconomic value of wild meat extraction to local food security in rural Amazonia. Biol Conserv 236:171–179. https://doi.org/10.1016/j.biocon.2019.05.010 Nunez-Iturri G, Olsson O, Howe HF (2008) Hunting reduces recruitment of primate- dispersed trees in Amazonian Peru. Biol Conserv 141:1536–1546. https://doi.org/10.1016/j.biocon.2008.03.020 Paglia AP, Fonseca GAB, Herrmann G, et al (2012) Annotated Checklist of Brazilian Mammals 2nd Edition. Conservation International.. 1-75. Occasional papers in conservation biology, v. 6, p. 1-82, 2012 Paine CET, Beck H, Terborgh J (2016) How mammalian predation contributes to tropical tree community structure. Ecology 97:3326–3336. https://doi.org/10.1002/ecy.1586 Palmeirim AF, Benchimol M, Peres CA, Vieira MV (2019) Moving forward on the sampling efficiency of neotropical small mammals : insights from pitfall and camera trapping over traditional live trapping. Mammal Res 64:445–454. https://doi.org/10.1007/s13364-019-00429-2 Paradis, E., Claude, J., & Strimmer K (2004) APE: an R package for analyses of 53 phylogenetics and evolution. Bioinformatics 20:289–290 Pardini R, Faria D, Accacio GM, et al (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. Biol Conserv 142:1178–1190. https://doi.org/10.1016/j.biocon.2009.02.010 Peres CA (2000) Effects of subsistence hunting on vertebrate community structure in Amazonian forests. Conserv Biol 14:240–253. https://doi.org/10.1046/j.1523- 1739.2000.98485.x Peres CA (1993) Notes on the primates of the Juruá River, Western Brazilian Amazonia. Folia Primatol 61:97–103. https://doi.org/10.1159/000156735 Peres CA (2008) Soil fertility and arboreal mammals biomass in tropical forest. Trop For community Ecol 349–364. Peres CA, Dolman PM (2000) Density compensation in neotropical primate communities: evidence from 56 hunted and nonhunted Amazonian forests of varying productivity. Oecologia 122:175–189. https://doi.org/10.1007/PL00008845 Peres CA, Emilio T, Schietti J, et al (2016) Dispersal limitation induces long-term biomass collapse in overhunted Amazonian forests. Proc Natl Acad Sci 113:892– 897. https://doi.org/10.1073/pnas.1516525113 Peres CA, Lake IR (2003) Extent of nontimber resource extraction in tropical forests: accessibility to game vertebrates by hunters in the amazon basin. Conserv Biol 17:521–535. https://doi.org/10.1046/j.1523-1739.2003.01413.x Peres CA, Palacios E (2007) Basin-wide effects of game harvest on vertebrate population densities in Amazonian forests: implications for animal-mediated seed dispersal. Biotropica 39:304–315. https://doi.org/10.1111/j.1744- 7429.2007.00272.x Peres CA, Roosmalen M van (2009) Primate frugivory in two species-rich neotropical forests: implications for the demography of large-seeded plants in overhunted areas. Seed dispersal frugivory Ecol Evol Conserv Third Int Symp Frugivores Seed Dispersal, São Pedro, Brazil, 6-11 August 2000 407–421. https://doi.org/10.1079/9780851995250.0407 Peterson RA (2017) bestNormalize: Normalizing Transformation Functions. V.1.4.2 R development core team (2019) R: a language and environment for statistical computing. Version 3.5.3, R Foundation. Vienna 54 Redford KH (1992) The Empty Forest. Bioscience 42:412–422. https://doi.org/10.2307/1311860 Ripple WJ, Newsome TM, Wolf C, et al (2015) Collapse of the world’s largest herbivores. Sci Adv 1:. https://doi.org/10.1126/sciadv.1400103 Schongart J, Piedade MTF, Ludwigshausen S, et al (2002) Phenology and stem - growth periodicity of tree species in Amazonian floodplain forests. J Trop Ecol 18:581– 597. https://doi.org/10.1017/S0266467402002389 Silman MR, Terborgh JW, Kiltie RA (2003) Population regulation of a dominant rain forest tree by a major seed predator. Ecology 84:431–438. https://doi.org/http://dx.doi.org/10.1890/0012- 9658(2003)084[0431:PROADR]2.0.CO;2 Terborgh J, Nuñez-iturri G, Pitman NCA, et al (2008) Tree recruitment in an empty forest. Ecol Soc Am 89:1757–1768. https://doi.org/https://doi.org/10.1890/07- 0479.1 Whitworth A, Macleod R, Beirne C, et al (2019) Human disturbance impacts on rainforest mammals are most notable in the canopy, especially for larger ‐ bodied species. 1–13. https://doi.org/10.1111/ddi.12930 Whitworth A, Braunholtz LD, Huarcaya RP, et al (2016) Out on a limb: Arboreal camera traps as an emerging methodology for inventorying elusive rainforest mammals. Trop Conserv Sci 9:675–698. https://doi.org/10.1177/194008291600900208 Wilkie DS, Bennett EL, Peres CA, Cunningham AA (2011) The empty forest revisited. Ann N Y Acad Sci 1223:120–128. https://doi.org/10.1111/j.1749- 6632.2010.05908.x Wilman H, Belmaker J, Jennifer S, et al (2014) EltonTraits 1 . 0 : Species-level foraging attributes of the world ’s birds and mammals. Ecology 95:2027. https://doi.org/10.1890/13-1917.1 Wright SJ, Hernandéz A, Condit R (2007a) The bushmeat harvest alters seedling banks by favoring lianas, large seeds, and seeds dispersed by bats, birds, and wind. Biotropica 39:363–371. https://doi.org/10.1111/j.1744-7429.2007.00289.x Wright SJ, Stoner KE, Beckman N, et al (2007b) The plight of large animals in tropical forests and the consequences for plant regeneration. Biotropica 39:289–291. https://doi.org/10.1111/j.1744-7429.2007.00293.x Wright SJ, Zeballos H, Dominguez I, et al (2000) Poachers alter mammal abundance, 55 seed dispersal, and seed predation in a neotropical forest. Conserv Biol 14:227– 239. https://doi.org/10.1046/j.1523-1739.2000.98333.x Young HS, Dirzo R, Helgen KM, et al (2014) Declines in large wildlife increase landscape-level prevalence of rodent-borne disease in Africa. Proc Natl Acad Sci U S A 111:7036–7041. https://doi.org/10.1073/pnas.1404958111 Young HS, Mccauley DJ, Dirzo R, Goheen JR (2015) Context-dependent effects of large wildlife declines on small mammal communities in central Kenya. Ecol Soc Am 25:348–360 Zuur AF, Ieno EN, Elphick CS (2010) A protocol for data exploration to avoid common statistical problems. Methods Ecol Evol 1:3–14. https://doi.org/10.1111/j.2041- 210x.2009.00001.x Supplementary Material Supplementary Information – Camera Trapping methods The camera trap models used in this study were: Bushnell Trophy-Cam, Reconyx HC500 Hyperfire and Browning Dark Ops. Camera traps were set to take 3 sequential photographs and/or one 10-second video. Camera sensors were set to high sensitivity. Three camera grids were operated simultaneously for an approximate period of 30 days, so that each grid was sampled only in one period (dry or rainy season). The selection of grids that were established at the same time depended on spatial logistical issues that would allow the removal and installation of cameras in a new area in the shortest amount of time, as we had about 120 cameras available. For the canopy cameras we had to build a protection made by PVC pipes, because the rain and sunlight damaged many cameras at the beginning of the study, but after the use of the protection there was a drop in the number of damaged cameras. 56 Supplementary Table S1. Information about our 30 sampling sites. Information includes code name; name of the community or locality where each unit is located; name of the protected area or town; geographic coordinates and the value of hunting pressure attributed to each sample unit. Sample unit Locality Area category Lat Long Hunting code Pressure P01 Riozinho Carauari town -4,77285 -66,95521 11.6 P02 Riozinho Carauari town -4,80499 -66,93954 13.8 P03 Riozinho Carauari town -4,759 -66,96839 10.8 P04 Riozinho Carauari town -4,78323 -66,96498 11.8 P05 Gumo do Facão RESEX Médio Juruá -5,05074 -67,12075 6.5 P06 Gumo do Facão RESEX Médio Juruá -5,05872 -67,12659 6.5 P07 Roque RESEX Médio Juruá -5,08408 -67,22807 4.9 P08 Roque RESEX Médio Juruá -5,09622 -67,22901 4.9 P09 Bom Jesus RDS Uacari -5,39348 -67,21051 3.5 P10 Bom Jesus RDS Uacari -5,40296 -67,20496 3.4 P11 Bauana RDS Uacari -5,44378 -67,27795 3.5 P12 Bauana RDS Uacari -5,4507 -67,2694 3.4 P13 Vila Ramalho RESEX Médio Juruá -5,57658 -67,50482 3.2 P14 Vila Ramalho RESEX Médio Juruá -5,58331 -67,50894 3.2 P15 Tabuleiro RESEX Médio Juruá -5,52478 -67,69476 3.1 P16 Tabuleiro RESEX Médio Juruá -5,51461 -67,69478 3.1 P17 Pupunha RDS Uacari -5,57097 -67,78001 2.9 P18 Pupunha RDS Uacari -5,57145 -67,79002 2.8 P19 São José RDS Uacari -5,7569 -67,88616 2.7 P20 São José RDS Uacari -5,74766 -67,90272 2.6 P21 Cachoeira RDS Uacari -5,79108 -67,877 2.7 P22 Cachoeira RDS Uacari -5,79838 -67,88075 2.7 P23 Xibauá RDS Uacari -5,88588 -67,86581 2.6 P24 Xibauá RDS Uacari -5,87709 -67,872 2.6 P25 São Sebastião Xerua -6,06008 -67,90543 2.4 P26 São Sebastião Xerua -6,05482 -67,91276 2.4 P27 Kiriru Itamarati town -6,37413 -68,23926 4.3 P28 Kiriru Itamarati town -6,38082 -68,2645 4.2 P29 Porão Itamarati -6,55647 -68,28748 3.1 P30 Porão Itamarati -6,56518 -68,28165 3.1 57 Supplementary Table S2. Species or (morphospecies) recorded in our camera trapping grids and the species trait values used to obtain the aggregate biomass and to classify species into functional groups of arboreal, terrestrial, game, non-game, small-bodied species and into the five trophic levels. BM: body mass; ST: stratum = T: terrestrial, A: arboreal; CT: camera trap= G: camera deployed on the ground, C: camera deployed in the canopy; GS: group size, Hunt: hunted =Y: yes, N: no. TL: trophic level. Source: Wilman et al (2004), Peres (1993) and Projeto Médio Jurua (PMJ, unpubl.data) ORDER / Species name English name BM (g) ST CT GS Hunt TL Mammals ARTIODACTYLA Mazama americana Red Brocket Deer 22,799 T G 1.1 Y 1 Mazama nemorivaga Grey Brocket Deer 16,633 T G 1.2 Y 2 Pecari tajacu Collared Peccary 21,266 T G 2.01 Y 2 CARNIVORA Atelocynus microtis Short Eared Dog 7,749 T G 1.2 N 4 Speothos venaticus Bush Dog 5,999 T G 1 N 5 Leopardus pardalis Ocelot 11,900 T G 1.3 Y 5 Leopardus wiedii Margay 3,249 T G 1 N 4 Panthera onca Jaguar 90,000 T G 1.4 Y 5 Puma concolor Puma 45,000 T G 1 N 5 Puma yagouaroundi Jaguarundi 6875 T G 1 N 4 Eira barbara Tayra 3,910 T G, C 1.3 N 5 Galictis vittata Greater Grison 3,200 T G 1 N 4 Nasua nasua South American Coati 3,793 T G 11.9 N 3 Potos flavus Kinkajou 3,000 A C 3 N 2 Procyon cancrivorus Crab-Eating Raccoon 6,949 T G 1 N 4 Pteronura brasiliensis Giant Otter 23,99 T G 8.5 Y 5 CINGULATA Priodontes maximus Giant Armadillo 45,359 T G 1.2 Y 4 Dasypus kappleri Greater Long Nosed Armadillo 9,500 T G 1 Y 4 Dasypus novemcinctus Small Armadillo 4,203 T G 1 Y 4 DIDELPHIMORPHIA Caluromys lanatus Brown-Eared Woolly Opossum 325 A C 1 N 2 Didelphis marsupialis Commom Opossum 1,091 T G, C 1 N 4 Didelphideae small* Opossum 25 T G, C 1 N 4 Glironia venusta Bushy-Tailed Opossum 114 A C 1 N 4 58 ORDER / Species name English name BM (g) ST CT GS Hunt TL Mammals DIDELPHIMORPHIA Metachirus nudicaudatus Four-Eyed Opossum 375 T G 1 N 4 Philander sp* Four-Eyed Opossum 325 T G, C 1 N 4 PERISSODACTYLA Tapirus terrestres Brazilian Tapir 150,000 T G 1.2 Y 1 PILOSA Cyclopes didactylus Pigmy Anteater 329 A G 1 N 4 Choloepus didactylus Southerntwo-Toed Sloth 5,160 A C 1 N 1 Myrmecophaga tridactyla Giant Anteater 22,333 T G 1.2 N 4 Tamandua tetradactyla Southern Tamandua 5,515 T G, C 1.1 N 4 PRIMATES Alouatta seniculus Howler Monkey 6,145 A C 6.2 Y 2 Aotus sp.* Night Monkey 1,060 A C 3.5 N 3 Ateles chamek Spider Monkey 8, 900 A C 11.7 Y 3 Cacajao calvus calvus White Bald-Headed Uacari 2,900 A C 30 N 3 Callicebus cupreus Red Titi Monkey 915 A C 3.5 N 2 Callicebus torquatus Collared Titi Monkey 1,050 A C 3.5 N 2 Cebuella pygmaea Pygmy Marmoset 125 A C 7 N 3 Cebus unicolor White-Fronted Capuchin 2,629 A G, C 19.8 Y 3 Lagothrix cana Woolly Monkey 8,500 A C 19.6 Y 2 Pithecia albicans Buffy Saki 2,800 A C 5.5 N 3 Pithecia monachus Monk Saki 1,537 A C 5.5 N 3 Saguinus fuscicollis Saddleback Tamarin 387 A G, C 5 N 3 Saguinus mystax Moustached Tamarin 618 A G, C 5 N 3 Saimiri boliviensis Squirrel Monkey 860 A G, C 47.5 N 3 Saimiri macrodon Ecuadorian Squirrel Monkey 860 A G, C 31 N 3 Sapajus macrocephalus Larged-Headed Capuchin 2,500 A G, C 18 Y 3 RODENTIA Coendou sp.* Brazilian Porcupine 4,399 A G, C 1 N 1 Cuniculus paca Lowland Paca 8,172 T G 1 Y 2 Dasyprocta spp.* Agouti 3,500 T G 1,2 Y 2 Echimyid rodent* Spiny Rat 300 A C 1 N 1 Myoprocta spp* Acouchy 966 T G 1 N 2 Proechimys spp* Spiny Rat 292 T G 1 N 1 Rhipidomys sp* Climbing Mouse 80 A C 1 N 2 Small rodent* Mouse 34 T G, C 1 N 2 59 ORDER / Species name English name BM (g) ST CT GS Hunt TL Mammals Very small rodent* Mouse 17 T G, C 1 N 2 Guerlinguetus ignitus Bolivian Squirrel 190 A G, C 1.2 N 2 Urosciurus spadiceus Southern Amazon Red Squirrel 800 A G, C 1.4 N 2 Birds CUCULIFORMES Neomorphus pucheranii Red-Billed Ground Cuckoo 330 T G 1 N 4 GALLIFORMES Mitu tuberosum Razor-billed Curassow 2,769 T G, C 1.6 Y 2 Odontophorus stellatus Wood Quail 335 T G 5.4 N 2 Penelope jacquacu Spix's Guan 1487 A G, C 4.9 Y 2 GRUIFORMES Psophia leucoptera Pale Winged Trumpeter 1,315 T G 5.8 Y 2 PICIFORMES Pteroglossus azara mariae Brown Mandibled Aracari 144 A C 4 N 3 Pteroglossus beauharnaesii Curl-Crested Aracari 215 A C 7.5 N 3 Pteroglossus inscriptus Lattered Aracari 125 A C 4.5 N 3 Ramphastos tucanus Red Billed Toucan 659 A C 2 Y 3 Selenidera reinwardtii Golden Collared Toucanet 161 A C 2 N 3 Tinamus guttatus White Throathed Tinamou 686 T G 1.5 Y 2 Tinamus major Large Tinamou 1026 T G 1 Y 3 TINAMIFORMES Crypturellus spp* Small tinamous 241 T G 1.2 Y 3 Tinamus guttatus White Throated Tinamou 686 T G 1.5 Y 2 Tinamus major Large Tinamou 1026 T G 1 Y 3 Reptiles TESTUDINATA Geochelone spp. red/yellow footed tortoise 4,580 T G 1.2 Y 4 For species with * we use the traits on Wilman et al (2014) following Table A2b 60 Supplementary Table S2b. Displays taxa on the survey list and respective taxa used as reference to obtain the trait data. Taxa Reference taxa for traits Didelphidae small Marmosa murina Philander sp. Philander mcilhennyi Coendou sp. Coendou prehensilis Dasyprocta spp. Dasyprocta fuliginosa Proechimys spp. Proechimys pattoni Myoprocta spp. Myoprocta pratti Rodentia small Hylaeamys perenesis Rodentia very small Oecomys bicolor Crypturellus sp. Crypturellus bartletti Aotus sp. Aotus nigrisceps 61 Supplementary Figure S1. Map of the study area in the Médio Juruá region of Western Amazonia. The reserve boundaries in the RESEX Médio Juruá and RDS Uacari are outlined in black. Coloured circles indicate the location of our sampling grids. Circles are colour-coded according to our proxy of hunting pressure (see colour gradient) and circle area is proportional to aggregate game biomass per camera grid. The main Juruá River channel is outlined in white. Grey background shows elevation where lighter shading represents higher terrain. 62 Supplementary Table S3. Akaike importance weights for model parameters predicting the aggregate biomass of several functional groups from the modelling average. Models were defined as equally ‘good’ if ΔAIC <2 of the best fitted model. The importance of explanatory variables was calculated by their frequency of occurrence in these models. Importance weights range from 0 (parameter with no explanatory weight) to 1(parameter in all ‘best’ models). Numbers in brackets are number of the models where the variable were presented. Model parameter Importance weights (n containing models) Game Species Rodents Arboreal Terrestrial Browsers Grazers Frugivores Omnivores Carnivores Hunting Pressure 1.00(6) 1.00(4) 1.00(2) 1.00(6) 0.26 (3) 1.00(5) 1.00(2) 0.21(1) 0.06(1) Water Level 0.57(3) 0.00(0) 0.00(0) 1.00(6) 0.00(0) 1.00(5) 0.00(0) 0.00(0) 0.00(0) Tree Density 0.22(2) 0.43(2) 0.00(0) 0.37(3) 0.33(4) 0.00(0) 0.00(0) 0.00(0) 0.55(4) % of várzea forest 0.53(3) 0.00(0) 0.31(1) 0.77(4) 0.10(1) 0.31(2) 0.28(1) 0.00(0) 0.50(4) CEC* 0.00(0) 0.00(0) 0.00(0) 0.58(4) 0.36(5) 0.11(1) 0.00(0) 0.22(1) 0.61(4) Julian Day 0.00(0) 0.63(2) 0.00(0) 0.00(0) 0.36(5) 0.42(2) 0.00(0) 1.00(3) 0.00(0) *CEC = soil cation exchange capacity 63 Supplementary Table S4. GLM results using PCoA Axes 1 and 2 for relative abundance and aggregate biomass of vertebrates for all species combined and separately in two groups of game and non-game species. HP=Hunting Pressure, Var: proportion of varzea forest, WL: water level, CEC: soil cation exchange capacity, Tree: tree density, Day: Julian day Relative Axis predicto t p Biomass Axis predicto t p Abundance r r PCoA1 HP -1.243 0.226 PCoA1 HP -3.071 0.005* Var 1.351 0.190 Var 1.322 1.993 WL -0.037 0.970 WL 0.645 0.5251 CEC 0.171 0.866 CEC -0.247 0.806 Tree 0.489 0.629 Tree -0.280 0.781 Abundance all Day 2.715 0.012* Biomass all Day 0.404 0.689 PCoA2 HP -3.085 0.005* PCoA2 HP 0.826 0.417 Var 0.485 0.632 Var -0.475 0.638 WL 0.656 0.518 WL -1.741 0.095 CEC -0.954 0.350 CEC 0.040 0.968 Tree -0.059 0.954 Tree -0.477 0.637 Day -0.110 0.913 Day -0.866 0.395 PCoA1 HP -4.338 0.000* PCoA1 HP -3.514 0.001* Var 2.238 0.035* Var 1.512 0.144 WL 2.356 0.027* WL 1.368 0.184 CEC -0.482 0.634 CEC -0.418 0.679 Abundance game Tree 1.194 0.244 Biomass game Tree 0.067 0.947 Day 3.695 0.001* Day 0.693 0.495 PCoA2 HP -1.810 0.083 PCoA2 HP 0.016 0.987 Var -0.419 0.679 Var -0.639 0.529 WL 0.518 0.609 WL -1.045 0.307 CEC -0.440 0.664 CEC 0.139 0.891 Tree -1.129 0.270 Tree -1.125 0.272 Day -1.635 0.115 Day -1.838 0.079 PCoA1 HP -1.382 0.180 PCoA1 HP -0.782 0.442 Var -0.584 0.565 Var -0.748 0.462 Abundance not WL 0.607 0.550 Biomass not WL -0.319 0.753 hunted CEC -0.979 0.338 hunted CEC 0.597 0.556 Tree 0.252 0.803 Tree -1.636 0.116 Day -1.705 0.102 Day 0.744 0.464 PCoA2 HP 2.735 0.012* PCoA2 HP -0.839 0.409 Var -0.630 0.534 Var 1.817 0.082 WL 0.303 0.764 WL -2.265 0.033 CEC 0.994 0.330 CEC -0.327 0.746 Tree -1.029 0.314 Tree 0.992 0.331 Day 0.064 0.949 Day -0.002 0.998 64 65 CAPÍTULO II CASCADING EFFECTS OF OVERHUNTING ON FUNCTIONAL TREE COMPOSITION IN AMAZONIAN FORESTS Andressa Bárbara Scabin¹, *, Flávia Regina Capellotto Costa² and Carlos A. Peres3,4 1Programa de Pós-Graduação em Ecologia, Universidade Federal do Rio Grande do Norte - Natal, RN, Brazil. 2Coordenação de Biodiversidade, Instituto Nacional de Pesquisas da Amazônia, Manaus - AM, Brazil. 3Centre for Ecology, Evolution and Conservation, School of Environmental Sciences, University of East Anglia, Norwich, UK. 4 Departamento de Sistemática e Ecologia, Universidade Federal da Paraíba, João Pessoa, PB, Brazil. Author for correspondence: * Andressa B. Scabin, Email:andressa_scabin@ufrn.edu.br Departamento de Ecologia, Centro de Biociências, Universidade Federal do Rio Grande do Norte campus Lagoa Nova, Natal, RN 59072-970, Brazil. short title: Hunting cascade effects on functional forest composition 66 ABSTRACT The wildlife “emptying” of tropical forests, mainly due to overhunting, represents a threat for long-term forest conservation and ecosystem services provision. This is because some hunted vertebrates play a pivotal role in forest regeneration as key seed dispersal and seed predator agents. Plant-animal interactions can be severely disrupted resulting in changes in forest composition and functionality. Here, we investigate the cascading effects of overhunting on tree functional composition of 30 forest sites (0.25- ha for trees and 0.05-ha for saplings) along a marked hunting pressure gradient in western Brazilian Amazonia. We sampled 4,784 trees, 6,132 saplings and assigned plant trait values including dispersal mode, seed mass, wood density and leaf mass per area (LMA), based on literature and data collected in the field. We performed a linear mixed model (LMM) using sapling-to-tree abundance ratio as response variable; water level, soil- nutrient availability, hunting pressure and dispersal mode as predictors, and plant species as a random effect. To analyse hunting effects on plant functional traits we applied the community-weighted mean (CWM) for wood density, seed mass and LMA as response variable in generalized linear models (GLMs). Our results provide evidence that sapling recruitment in relation to conspecific trees decline for species dispersed almost exclusively by large-bodied frugivores in heavily hunted forests, while abiotically dispersed trees show higher sapling abundances. Water and soil-nutrient availability affected positively the sapling-to-tree abundance ratio. The CWM for wood density, LMA, and seed mass were unaffected by hunting pressure for both trees and saplings. However, the differences of mean wood density between trees and saplings have been intensified in heavily hunted areas. We conclude that persistently hunted forests in our study region show early signs of functional changes, but these were insufficient to exert profound community-wide shifts in functional composition. Peri-urban forests in Juruá region should be monitored over time to examine whether these modest effects will become stronger at later stages of tree regeneration. This can ensure better predictions of the degree to which ecosystem services, such as long-term carbon storage, will decay in overhunted forests. KEYWORDS: hunting, empty forest, plant-animal interaction, dispersal mode, tree recruitment, functional traits, wood density, leaf mass, seed mass. 67 1. INTRODUCTION Tropical forests worldwide provide crucial ecosystem services for about 1.5 billion people (Vira et al., 2015). These services include climate regulation, carbon storage, and extraction of natural resources such as timber, fruits, oilseeds, plants with therapeutic properties, and bushmeat (Foley et al., 2007; Nunes et al., 2019; Pearce, 2001). However, to maintain the full flow of these services, forest ecosystems require a nearly complete complement of species and their interactions (Mooney et al., 2009). Overhunting has promoted a marked decline in populations of large-bodied vertebrates in tropical forests worldwide (Bennett et al., 2002; Ripple et al., 2016) and consequently disrupting important plant-animal interactions (Wilkie et al., 2011). In the Neotropics the main game vertebrate targets include tapir, peccaries, paca, primates, deer and large birds which play a key role in plant community structure through processes such as seed dispersal, and density-dependent seed predation and seedling browsing (Dirzo et al., 2014; Redford, 1992). Because approximately 80% of all woody plants in Neotropical forests rely on vertebrates for seed dispersal (Howe and Smallwood, 1982; van Roosmalen, 1985), disruptions to these mutualistic interactions are expected to significantly alter forest composition and its functionality (Kurten et al., 2015.; Wright, 2003). Defaunation can alter forest composition through a decline in populations of large- seeded animal-dispersed trees and the concomitant density-compensation of trees and woody lianas dispersed abiotically, or by non-game vertebrates (Effiom et al., 2013; Harrison et al., 2013; Nunez-Iturri et al., 2008; Terborgh et al., 2008). Alternatively, forest composition may change through the increased recruitment success of large-seeded animal dispersed trees if the most affected interactions are seed predation or seedling herbivory (Paine and Beck, 2007; Poulsen et al., 2013; Roldán and Simonetti, 2001; Wright et al., 2007). Selective hunting affects different trophic guilds simultaneously, whose interactions could be disrupted depending on the balance of prevalent agents that may either decline or indirectly increase due to overhunting (Gardner et al., 2019; Stoner et al., 1998; Wright et al., 2000). Consequently, predicting the directional pathways of compositional changes in overhunted tropical forests is far from straightforward (Beckman and Muller-Landau, 2007). Regardless of possible divergences in forest composition responses to full or partial large-vertebrate defaunation, the loss of seed dispersers results in greater changes 68 to forest functional composition compared to the loss of seed predators and herbivores (Kurten et al., 2015). Declines in the abundance of plant species dispersed by large-bodied frugivores, and their replacement by those dispersed abiotically or by small frugivores, are expected to boost plant traits associated with the latter in community-wide transitions in functional trait composition (Kurten et al., 2015). Plant species potentially succumbing to dispersal limitation include primarily large-seeded, heavy-wooded, and higher-statured trees characterizing the slow-growing flora with conservative biomass accumulation strategies (Chave et al., 2006). Therefore, the gradual replacement of conservative (slow- growing, heavy-wooded) species by acquisitive strategies (fast-growing, light-wooded species) can downgrade forest carbon storage to varying extents, as shown in large-scale modelling studies (Bello et al., 2015; Peres et al., 2016). However, community functional patterns are not only affected by species identity and performance, but also by climatic and edaphic variables (Fortunel et al., 2014). Despite the growing number of studies examining the cascading effects of defaunation on tree composition, identifying well-defined patterns of functional responses has proven elusive. Defaunation effects on tree regeneration are often investigated during only the early stages of seedling recruitment, assuming that changes at this stage class persist into future cohorts thus defining the composition and diversity of the tree community. However, Janzen-Connell effects in terms of high density- dependent mortality of seedlings near parent trees may not be reflected at later stages of tree ontogeny (Bagchi et al., 2018; Beck et al., 2013; Theimer et al., 2011). As such, long- distance dispersal limitation instead of promote sapling high mortality can increase the spatial clustering of species that rely on potentially overhunted dispersal agents (Bagchi et al., 2018). Furthermore, due to practical difficulties, most studies usually settle on approaches involving poor spatial replication, focusing on comparative designs that contrast only the extremes of the defaunation spectrum, thereby failing to consider forest sites under intermediate levels of hunting pressure, which represents most of the contemporary tropics (Bogoni et al., 2020; Harrison et al., 2016) Here, we investigate the cascading effects of defaunation on the functional tree composition of Amazonian forests along a marked gradient of hunting pressure. We use a robustly replicated space-for-time substitution approach over a large meta-landscape across structurally intact forest sites ranging from very low historical hunting pressure to persistently overhunted ones. We examine whether known declines in large-bodied 69 vertebrate populations driven by a century of overexploitation induces directional transitions in plant functional groups in terms of dispersal mode. Based on a comparative approach using quantitative inventories of two (st)age classes of saplings and canopy trees, we expected the abundance of saplings that are dispersed abiotically or by small- bodied frugivores to increase in heavily hunted areas at the expense of large-seeded animal-dispersed trees. Additionally, we hypothesised that directional changes in sapling composition would promote a concomitant alteration in community-wide patterns of wood density, seed size, and leaf mass per area (LMA). We expected these continuous functional traits to be lower in trees dispersed by abiotic agents or by hunting-insensitive small-bodied vertebrates and, thereby predictably altering community-wide functional composition along the hunting gradient. 2. MATERIAL AND METHODS 2.1 Study Area This study took place in the Médio Juruá region of western Brazilian Amazonia (Fig. 1), including two large contiguous sustainable-use protected areas and adjacent landscapes containing two urban clusters. This roughly represents the middle-third section of the Juruá River, the second-longest white-water tributary of the Amazon River. The two protected areas include the 253,227-ha Medio Juruá Extractive Reserve (RESEX Médio Juruá; 5º33'54"S, 67º42'47"W), created in 1997 and legally occupied by ~ 2,000 people distributed across 13 villages; and the 632,949-ha Uacari Sustainable Development Reserve (RDS Uacari; 5º43'58"S, 67º46'53"W) created in 2005, where ~ 1,200 people occupy 32 villages. The nearest towns are Carauari with a population of 28,076 habitants and located 88 fluvial km downstream from the RESEX Médio Juruá, and Itamarati with a population of 7,888, located 120 fluvial km upstream from the RDS Uacari (IBGE, 2018). The Médio Juruá region has a wet tropical climate with a mean annual temperature of 27.1°C and a mean annual rainfall of 3,679 mm, with the wettest period between November and April. Two different forest types comprise the study landscape: seasonally-flooded (várzea) forests, which account for ~20% of the study region, characterized by enriched Andean alluvial soils and lower floristic diversity, and the dominant (~80%) unflooded forest (terra firme), which exhibits higher floristic diversity and comparatively lower soil fertility (Hawes and Peres, 2016). The current study was performed in unflooded forest on paleo-várzea sediments, that we refer to here 70 as terra firme for simplicity, but we recognize that these forests, may diverge in their floristic macromosaics from so-called terra firme forests (Assis et al., 2015). Our tree plots were established along sites that had experienced subsistence and commercial hunting to varying degrees but had limited history of clear-cuts, wildfires and timber extraction. FIGURE 1: Map of the study area in the Médio Juruá region of western Brazilian Amazonia. The main Juruá River channel is outlined in white. The reserve boundaries of RESEX Médio Juruá and RDS Uacari are outlined in black. Coloured circles indicate the location of our 30 tree plots. Circles are colour-coded according to our proxy of hunting pressure (see colour gradient). The map background represents elevation, where brownish and green shades indicate low and high elevation, respectively. 2.2 Hunting Pressure We built a proxy of hunting pressure based on the geographic distance to and size of regional-scale human settlements, including all isolated households, villages and towns. Previous studies in the study area have shown that distance from urban centres 71 represents a good proxy of the anthropogenic impact on large vertebrate abundance (Abrahams et al., 2017; Nichols et al., 2013, Scabin and Peres, Chapter 1). We measured the Euclidean distance from each plot centroid to the nearest community and the dry- season navigation (fluvial) distance to the towns using ArcGIS10.3. Human population size was derived from IBGE census data (IBGE, 2018), and the Projeto Médio Juruá (PMJ) and the Sustainable Amazon Foundation (FAS) databases. 2.3 Study design A total of 30 sites were selected along the entire spectrum of hunting pressure throughout the Juruá study landscape based on the distance to urban centres and the nearest villages, which were settled by former rubber-tapper populations ranging from ~50 to ~600 inhabitants. Previous vertebrate surveys in the region have shown that large- bodied frugivores are either absent or occur at very low densities in the peri-urban (<30 km) forests of Carauari (Abrahams et al., 2017, Scabin and Peres, Chapter 1). From May to December 2017, we established at each forest site (1) two permanent 0.25-ha plots (25m × 100m), at least 1 km apart, where all trees and arborescent palms ≥ 10 cm diameter at breast height, DBH (hereafter “trees” ) were surveyed; and (2) 0.05-ha (5 m × 100 m) sapling subplots nested within the larger tree plots, where all juvenile trees and arborescent palms (1-5 cm DBH, ≥1 m height, hereafter “saplings”) were surveyed. All live individuals were measured and number-tagged, at 1.30 m height for trees and 1m for saplings. The floristic inventory was conducted in July 2018, when all species were identified in the field to the finest possible taxonomic level. Species identification was carried out by a professional parabotanist with over 30 years of field and herbarium experience at the Brazilian National Institute for Amazonian Research (INPA), Manaus, Brazil. Voucher specimens of all individuals that could not be identified in the field were collected and subsequently identified at the INPA herbarium with the assistance of another parabotanist and then deposited at the EAFM herbarium of the Instituto Federal de Educação, Ciência e Tecnologia do Amazonas (IFAM, Manaus). To automatically standardise plant nomenclature and correct for synonyms based on The Plant List (http://www.theplantlist.org) we used the Taxonstand 22.2 R package (Cayuela et al., 2012). 72 2.4 Sapling: tree abundance log-ratio We examined changes in plot-scale species-specific density (D) between trees and conspecific saplings at each site under varying levels of hunting pressure based on the logarithmic sapling-to-tree ratio (S:T) considering all stems (trees and palms) within each plot and subplot. This log-ratio was calculated for every species that had been recorded at least once as a tree, sapling, or both, following the equation: S:T = log10 [(Dsapling(i) + 0.1) / (Dtree(i) + 0.1)] S:T values >0 or <0 indicate a sapling density higher or lower than the co- occurring conspecific tree, respectively. Sapling subplots spanned the entire length of each tree plot and were therefore representative of the sapling layer underneath all trees within each plot. S:T was thus used as a proxy of plant recruitment so that positive values indicate a greater density of saplings compared to conspecific trees while negative values indicate lower density of saplings compared to conspecific trees. This approach has been used elsewhere (Hazelwood et al., 2020; Jones et al., 2019; Terborgh et al., 2008). 2.5 Plant functional traits We directly measured or assigned values or categories based on the literature for plant functional traits that are widely recognized to be critical for tree recruitment (Westoby, 1998) and that we expected to be indirectly affected by long-term seed dispersal limitation induced by overhunting. These include three continuous [wood density (WD), leaf mass per area (LMA) and seed mass] and a categorical trait (seed dispersal mode). Wood density (WD) To obtain species-specific wood density we used a 5.15 mm Haglöf increment borer to typically sample three wood cores from different individuals for each of the 250 most abundant tree species, amounting to a total of 700 wood core samples. Wood cores were extracted perpendicularly to the bark at about 1 m height from trees >10 cm DBH. To encompass the variation in heartwood-to-bark wood density, cylindrical samples were equivalent to the approximate length of the bole radius (DBH × 0.5). Each wood core was first wrapped in filter paper and stored in a box containing silica gel for initial drying to avoid fungal attack for 5 days and then stored in labelled plastic straws. We then 73 rehydrated all wood samples for 24 h at the Plant Functional Laboratory at INPA. We subsequently obtained the green volume of each core through the water displacement method, using a beaker of water placed on a digital balance (precision ≈ 0.01g) which was re-zeroed each time. All wood cores were then oven-dried at 105°C for 72 h, at which point they were dry-weighted. Wood density (WD) values were calculated by dividing the dry wood core weight by the green volume of each sample (Chave, 2002). As we used the water displacement method, we obtained the wood specific gravity (WSG) value, which is the relationship between the dry wood mass and the wood volume at saturation point in relation to the volume of water; for simplicity, this is referred to hereafter as wood density (WD). To obtain species-specific WD values we averaged the WD values across all samples of the same species. WD values were obtained for 54.8% of all tree stems surveyed. For the minority of genera (18%) lacking WD values obtained from our own sample, we used data from the Global Wood Density Database (Chave et al., 2009; Zanne et al., 2009) including only values from the South America tropical region. In total, 47.2% of all individuals were assigned to a WD value at the species level, 29.8% at the genus level, 21.1% at the family level and only 0.1% were assigned the mean WD value of each plot, due to complete lack of data available for these taxa. Leaf Mass per Area (LMA) LMA is a plant functional trait related to the energy investment into leaf production, so that higher LMA values are associated with greater energy investment into long-lived leaves, while lower values are related to low longevity leaves (Poorter, 2007). We collected 2,327 leaves belonging to 477 tree species within all subplots. The leaves were manually collected at random in the understorey according to access to terminal foliage, prioritizing leaves showing healthy condition. We collected one leaf by sapling. Each leaf was number-tagged and had a 4-cm² section cut using a shape cutter. All samples were placed in paper envelopes and oven-dried at 65°C for 48 h, and then weighed on a digital scale (precision ≈ 0.01g). The leaf section weight divided by 2 cm corresponds to LMA (expressed as g cm-2). Species-specific LMA was obtained for 56.3% of species and 63.6% of all stems surveyed. For all individuals for which species- level data were lacking, we imputed data hierarchically using the following sequential criteria: 1) mean genus-level value for each tree plot, if there were at least 5 individuals of that genus in the plot, 2) mean genus-level value across the whole database, 3) mean 74 family-level value for each plot, if there were at least 5 individuals of that family in the plot, 4) mean family-level across the whole database, and 5) mean value for each plot. Seed Dispersal Mode and Seed Size All woody tree species were assigned to categories according to their seed dispersal mode based on a comprehensive review of the published literature on the Amazonian flora (e.g. Amaral et al., 2009; Baraloto and Forget, 2007; Cornejo and Janovec, 2010; Hammond and Brown, 1995; van Roosmalen, 1985). We also compiled data on dry seed mass by species and/or genus, which was supplemented by data available online from the Kew Seed Information database (KEW, 2008) and TRY Plant trait database (www.try- db.org). Dry seed mass values were assigned to 87.5% of all saplings and trees, including 29.8% at the species level, 47.2% at the genus level, and 10.5 % at the family level. Functional groups related to seed dispersal mode were defined as: (1) abiotic (wind, water, ballistic), (2) scatter-hoarded by large rodents, (3) endozoochory by vertebrates species insensitive to hunting (hereafter gut:HIS) and (4) endozoochory by vertebrates sensitive to hunting (gut:HSS). Tree species dispersed by vertebrates hunting-sensitive received this category based mainly on consumption by large-bodied primates and tapir according to publications (e.g. Hawes and Peres, 2014) and our own knowledge (Carlos Peres, unpubl. data) about plant-animal interactions involving these species. Any ambiguity in data sources was resolved by the parabotanists’ acknowledge of the species interactions. 2.6 Environmental variables Considering that climatic and edaphic variables such as water and soil-nutrient availability can predict the functional composition of forests (Fortunel et al., 2014), we obtained data for each tree plot on soil fertility and water availability. At all 30 plots, we collected three soil samples of ~5g each using a soil auger with samples extracted near the plot centroid, at least 10 m apart. Soil samples were air-dried and then sealed in plastic bags until analysis in the Soil Chemistry Laboratory at National Institute for Amazonian Reasearch (INPA). Soil fertility was represented by the cation exchange capacity (CEC) as the sum of Ca+2, Mg+2, and K+ concentrations measured as cmol kg-¹. The vertical distance to the nearest drainage (VDND), which represents a proxy of water availability for plants, was calculated by the Height Above the Nearest Drainage (HAND) algorithm proposed by Rennó et al. (2008) based on the 30-m resolution Digital Elevation Model 75 (DEM) topography available from the Shuttle Radar Topography Mission (SRTM). The vertical distance grids generated by the HAND algorithm, and derived from SRTM data, are available from the Brazilian Space Research Institute (INPE) website (www.dpi.inpe.br). 2.7 Data analysis We first investigated the spatial structure of the floristic composition using a Mantel test, which compared the floristic composition matrix based on the Morisita-Horn dissimilarity and the Euclidean geographic distance matrix based on plot coordinates. The spatial structure of the floristic composition was tested for both trees and saplings separately. Distance matrices were obtained using the vegdist function in the vegan 2.5- 4 R package (Oksanen et al., 2008). In order to examine within-plot congruence in floristic composition between trees and saplings we performed a Procrustes analysis which uses uniform scaling and optimal rotation to compare two ordinations by minimizing their square differences (Oksanen et al., 2019). We attempted to explain Procrustes distances between co-occurring trees and saplings as a function of hunting pressure using linear regression. To determine to what degree hunting pressure affects the sapling-to-tree ratio (S:T, our response variable), after controlling for environmental variation, we used a Linear- Mixed Models (LMMs) with a Gaussian error structure. Hunting pressure was set as the fixed effect, together with the two environmental covariates (CEC and VDND) and seed dispersal mode as a categorical variable, while holding species as a random effect, due to the repetition of species across plots. In order to compare effect size of the explanatory variables we rescaled all the explanatory variables using the function rescale within the arm 1.11-1 R package (Gelman et al., 2020). After, we calculated the variance inflation factors (VIFs) to test for multicollinearity in explanatory variables, where VIFs<4 indicates low multicollinearity (Zuur et al., 2010). All variables were retained in the models as they were uncorrelated. Different models were built considering either only the simple effects of the environmental variables or their interactions and including the interaction between hunting pressure and dispersal mode. A model selection was then performed using ANOVAs (Zuur et al., 2009). To execute the models, we used the lmer function within lme4 package (Bates et al., 2019). To assess the relative importance of the fixed predictors compared to the influence of taxonomic identity, we calculated the 76 R² marginal and conditional values using the r.squaredGLMM function in the MuMin package (Bartón, 2016). To examine the continuous functional traits, we first investigated the Pearson correlation structure between Wood Density (WD), Leaf Mass per Area (LMA) and Seed Mass (SM). We then examined to what degree WD, LMA and SM differed between the seed dispersal modes using ANOVAs, followed by TukeyHSD tests for each plant functional trait. After, to investigate whether assemblage-wide distribution in continuous traits was affected by hunting pressure, we performed linear models (LMs). In doing so, we used the Community Weighted Mean (CWM) approach, which consisted of a weighted average of the trait values by species abundances at each plot (Muscarella and Uriarte, 2016). To better understand the effects of hunting on traits scaled by individual contributions to the community biomass, we also used the CWM weighted by the basal area of all individuals. For LMs we entered hunting pressure, water availability and soil fertility as fixed effects and the CWM as response variable with normal error distributions for WD, LMA and SM. In addition, we compared CWMs for WD, LMA and SM among trees and saplings using Analysis of Covariance (ANCOVA) to examine how trait- specific CWMs for different life stages were related to hunting pressure. All analyses were conducted in R 3.5.3 (R development Core Team 2019). 3. RESULTS Tree and sapling plots surveys We recorded a total of 4,784 trees and 6,132 saplings across 30 permanent tree plots (7.5 ha) and 30 subplots (1.5 ha), which represented 846 species, 268 genera and 67 families (Table S1). Considering all individuals, 99% were identified at the species level and the remaining 0.9% at the genus level, with only 8 individuals remaining unidentified. More than 60% of all trees and 80% of all saplings were assigned to seed dispersal by small and medium-sized frugivores (Table S2), whereas only 1.8% of all trees and 1.4% of all saplings were exclusively or virtually exclusively dispersed by harvest-sensitive game species, such as prehensile-tailed large-bodied primates (i.e. howler monkeys, Alouatta seniculus, woolly monkeys, Lagothrix spp., and spider monkeys Ateles chamek). Trees bearing indehiscent fleshy fruits that are most heavily consumed by these harvest- sensitive species include several Sapotaceae genera, such as Chrysophyllum, Ecclinusa, and Pouteria (Carlos Peres, unpubl. data). Abiotically dispersed trees represented ~8% of 77 both trees and saplings and trees scatter-hoarded by large rodents represented 22% of all trees and 9.5% of all saplings. There was no meaningful spatial structure in the plot-scale floristic composition for either trees (R²= -0.27, p = 0.56) or saplings (R²= -0.026, p = 0.54) so that spatial effects exerted only a negligible role in differences in plant species composition across our sites. The assemblage-wide species composition within the same plots are significantly similar between trees and saplings (Procrustes: r = 0.38, p = 0.02, Fig.S1), however the differences between trees and sapling could not be attributed to hunting pressure (R² = - 0.02, p=0.62, Fig. S1). Conspecific transitions in abundance The linear mixed model showed that S:T ratios were significantly affected by water and soil-nutrient availability, hunting pressure and dispersal mode (Table S3). However, these variables explained only 3.5% of the S:T ratios, while species identity accounted for 27.8% of the overall S:T variation. As such, the interaction between hunting pressure and seed dispersal mode led to a modest decrease in the S:T ratio. At higher VDND there was a greater abundance of saplings relative to conspecific trees. Conversely, more fertile soils were associated with lower abundance of saplings in relation to adult conspecifics, but the interaction between these two variables produced a positive effect on sapling abundance (Table S2). Hunting effects associated to seed dispersal mode The effect of hunting pressure on the S:T ratio was widely variable for different seed dispersal modes. Species bearing seeds dispersed almost exclusively by large primates and other harvest-sensitive game vertebrates declined in densities from trees to saplings, whereas sapling density increased compared to their adult conspecifics in species served by abiotic dispersal, such as wind-dispersed trees. Trees dispersed primarily by small-bodied frugivores that were not affected by hunting exhibited a modest decrease in sapling density compared to conspecific adults, whereas scatter-hoarded tree species did not show any consistent effect (Fig. 2). 78 FIGURE 2: Effects of hunting on plot-scale sapling-to-tree abundance for tree species dispersed by abiotic agents (wind, water, ballistic); harvest-insensitive vertebrate species (gut:HIS, endozoochory by small-bodied birds and mammals), harvest-sensitive vertebrate species (gut:HSS, endozoochory by large-bodied mammals such as large primates); and scatter-hoarding rodents (e.g. agoutis). Lines and shading represent a linear fit and the 95% confidence interval region. Correlations between plant functional traits Considering the three continuous functional traits across 783 plant species for which data were available, wood density was positively correlated with both seed mass (r=0.15, p>0.001) and LMA (r=0.33, p>0.001). Seed mass and LMA relationship was not statistically significant (r=0.04, p=0.25, Fig. S4). We detected differences in mean plant trait values for WD, LMA and SM, mainly for species dispersed by scatter-hoarders. Tree species dispersed by scatter-hoarders had the highest average wood densities, while other seed dispersal modes did not show differences in their mean WD values. Tree species dispersed by scatter-hoarders and harvest-sensitive game species showed the highest LMA and seed mass values (Fig. 3). 79 FIGURE 3: Mean functional trait values for wood density, leaf mass per area and seed mass in relation to seed dispersal modes including abiotic, gut-dispersal by harvest-insensitive species (gut:HIS), gut-dispersal by harvest-sensitive species (gut:HSS), and scatter-hoarding by large rodents. Horizontal lines, boxes, whiskers and solid dots represent mean values, the first and third quartiles, extremes values, and outliers, respectively. Lower case letter distinguishes groups that are significantly different from one another (P<0.05) Effects of hunting pressure on community level functional properties The linear models showed that CWMs for wood density, LMA and seed mass for both trees and saplings, based on either species abundance or basal area, were not associated with hunting pressure or other environmental variables. The only exception was the positive effect of VDND on tree wood density, indicating that trees that were likely exposed to marked seasonal hydrological stress were on average more heavy- wooded (R²=0.16, p=0.040, Fig.S3). CWM wood density did not differ between co- occurring trees and saplings along the hunting pressure gradient (ANCOVA, F3,56 = 0.92, p = 0.34), and the same can be detected for LMA (F3,56 = 0.03, p = 0.85) and SM (F3,56 =0.22, p = 0.63). However, there was a tendency for higher CWM wood density in trees 80 within plots exposed to higher hunting pressure, which was not accompanied by an increase in sapling wood density (Fig. 4). FIGURE 4: Comparison between Community Weighted Mean (CWM) wood density (WD), leaf mass per area (LMA) and seed mass (SM) for trees and saplings across the hunting pressure gradient. Lines and shading represent a linear fit and the 95% confidence interval regions. 4. DISCUSSION Our results reveal modest changes in the functional composition of tree assemblages in persistently overhunted forests in the mid-Juruá region of western Brazilian Amazonia. Forests subject to a history of heavy hunting pressure in mid-Juruá showed a generational decline in the recruitment of tree species dispersed by large-bodied frugivores, which were numerically compensated by higher abundances of abiotically dispersed trees. Furthermore, the mean values for tree wood density, LMA and seed mass were different among seed dispersal modes in that trees dispersed by either hunting- sensitive game species or large rodent scatter-hoarders show higher values for these traits. On the other hand, community-wide patterns of continuous plant functional traits were not significantly different along the hunting pressure gradient. Nevertheless, LMA and seed mass, were lower in the regeneration tree layer, they showed no differences in their 81 relationship with hunting pressure, while CWM wood density was similar between saplings and trees, except for forests that had been historically exposed to persistent hunting. Hunting effects on seed size and dispersal mode Peri-urban areas along the Juruá River, which historically experienced the longest and most intensive hunting footprint, still represent continuous tracts of otherwise relatively undisturbed forest that are not entirely depleted of large-bodied vertebrates as source-sink dynamics has been maintained with surrounding nonhunted areas (Abrahams et al., 2017). Nevertheless, even in light of these mitigating factors that potentially buffer the indirect effects of hunting, it was possible to detect non-random directional changes in the compositional profile of plant functional groups, which are consistent with previous studies in the Amazon (Nunez-Iturri et al., 2008; Terborgh et al., 2008), Central America (Kurten et al., 2015; Wright et al., 2007), South-east Asia (Harrison et al., 2013) and Central African forests (Effiom et al., 2013; Vanthomme et al., 2010). However, the overall effect size of hunting pressure, appears to be less pronounced than those studies that have compared truly empty forests and protected areas with no previous history of hunting. At heavily hunted Juruá sites, in addition to declines in large-bodied frugivore biomass, peccaries (Pecari tajacu) were recorded at lower densities (Peres et al., 1996; Scabin & Peres, Chapter 1). The low abundance of large-bodied seed predators may represent a marked compensatory effect that facilitates seedling recruitment (Silman et al., 2003; Wright et al., 2000). On the other hand, echimyid rodent populations, and the seed predation pressure they may exert, can benefit from large mammal declines through density compensation mechanisms (Galetti et al., 2015), as we have shown at the study sites (see Chapter 1). Accordingly, the overall quantitative balance between mutualistic and antagonistic interactions, such as seed dispersal and seed predation, likely define the direction and strength of defaunation effects on tree regeneration. Consequently in order to understand the effects of defaunation on the floristic composition of forests that have not been heavily depleted of their wildlife, it is important to further dissect which plant- animal interactions may have been either favoured or penalized given the residual assemblage of vertebrate frugivores. 82 Some studies reporting forest compositional changes due to defaunation argue that seed size is a good predictor of the strength of trophic cascades on tree regeneration (Dirzo et al., 2007; Wright et al., 2007). Despite the generally strong relationship between seed size and seed dispersal mode, merely due to the physical constraints of gut passage, some non-game species such as bats are also able to disperse large seeds (Melo et al., 2009). In addition, large scatter-hoarders such as agoutis (Dasyprocta spp.) are largely harvest- insensitive and may boost the survival and recruitment of large-seeded trees through active secondary dispersal (Galetti et al., 2010). Furthermore, patterns of sapling recruitment uncovered here may reflect diffuse interactions between large-seeded plants and non-game or harvest-insensitive game vertebrate species, which may confer a certain degree of dispersal redundancy (Peres and van Roosmalen, 2002). Our results therefore corroborate the notion that community-wide variation in seed size is not necessarily a good indicator of the historical effects of hunting on forest regeneration due to complexities involved in these interactions (Beckman and Muller-Landau, 2007). Community-wide functional composition Other plant functional traits that could potentially be indirectly affected by large- bodied vertebrate declines include LMA and WD (Kurten et al., 2015). We hypothesized that heavily hunted areas would have lower LMA and WD because of relatively higher recruitment of fast-growing abiotically dispersed trees, based on the classic fast-slow growth trade-off along the plant economic spectrum (Reich, 2014). Although we found a modest increase in the abundance of abiotically dispersed trees in the regeneration strata of overhunted forests, this was not translated into community-wide divergent trait values along the hunting pressure gradient, even though these traits differed significantly among seed dispersal modes. This is likely a consequence of the greater influence of the most abundant species traits in the community-weighted means, which in this case were trees species with intermediate traits values and dispersed by non-game or harvest-insensitive game species. Some heavy-wooded, large-seeded and high leaf-mass species – representing the slow woody biomass turnover – were almost exclusively dispersed by large-bodied frugivores sensitive to hunting. However, given that these species represented only ~2% of all plant species inventoried or 2.2% of the overall basal area, their trait values became very diluted in the community weighted means. 83 Although we did not detect changes in mean community-wide functional traits across the hunting pressure gradient, there was a signal of change in mean WD values, between trees and saplings in heavily hunted forests. There are an overall tendency of adult trees have higher average trait values than saplings because, even the understorey composition reflecting largely the tree layer (Wright et al., 2003), usually the understorey have great abundance of low-wood density species with a resource acquisitive strategy (Poorter and Bongers, 2006). However, unlike we found for seed mass and LMA, for wood density these differences were accentuated only in the most heavily hunted areas. Several environmental drivers such as enhanced CO2 fertilization can lead to a shift towards more acquisitive plant communities, and consequently faster community dynamics (Brienen et al., 2015; Laurance et al., 2004), but shifts observed here were associated only with sites experiencing heavy hunting pressure, which overrides the potential effects of other drivers. Lower values of wood density in juvenile trees can have a negative effect in the future carbon storage capacity of forests (Bello et al., 2015; Berenguer et al., 2018), since wood density is an important predictor of aboveground biomass accumulation (Chave et al., 2006). Time scale in results interpretation Different timescales of analysis will affect the interpretation of how dispersal limitation may affect tree phytodemographics, as shown by Hazelwood et al.(2020), who conducted a 11-year study of the tree plots surveyed by Terborgh et al. (2008) in the Peruvian Amazon. They reported that even when community composition diverged between hunted and non-hunted sites, these changes could not be attributed to seed dispersal mode alone and argued that some saplings could have been recruited prior to the onset of hunting. Estimating sapling age in tropical forests, is at best difficult because juveniles can remain under arrested growth in the poorly illuminated understorey for many years (Green et al., 2014). However, it is unlike that the (1-5 cm in diameter) we surveyed are older than 25 years old, or one-quarter the approximate historical period in which hunting pressure has been most intense in the mid-Juruá region. In addition, unlike most comparative studies lack adequate spatial replication, our study provides ample environmental heterogeneity and greatly reduce the likelihood of confounding effects, such as elevated light availability underneath canopy gaps, on the species composition of the sapling layer. 84 Even though our well replicated study represents an incremental improvement in investigating the cascading effects of hunting on tropical forest floristic mosaics, we agree that long-term time series can be irreplaceable in elucidating cause-effect relationships (Hazelwood et al, 2020). However, it is still difficult to reliably predict how hunting affects the tree community composition without a proper understanding of the associated mechanisms pathways. Therefore, to be sure that (1) the patterns found here will either become stronger or weaker through time, (2) dispersal limitation of trees species dispersed by large-bodied vertebrates is the dominant driver of community-wide juvenile recruitment; and (3) surveyed saplings were indeed recruited after the onset of intensive hunting, it is imperative that permanent tree plots and their underlying conditions should be monitored over many years. CONCLUSIONS Our findings suggest that persistently hunted forests in remote areas across the Amazon show modest signs of dispersal limitation as reflected in the transitions in functional species composition from trees to saplings, even if hunting pressure has been considerably reduced over the last decade. Monitoring forest dynamics is important to understand whether these effects will be aggravated or mitigated in the long term. This study highlights the importance of understanding which plant-animal interactions are most affected by overhunting. This can inform sustainable game management initiatives in tropical forests that ensure retention of not only a full complement of forest vertebrate species, but also their key mutualistic interactions. Additionally, differences in the wood density traits between trees and saplings may have an important implication in the carbon stock service provided by these forests. Acknowledgments We are grateful to Associação dos Produtores Rurais de Carauari (ASPROC), Centro Estadual de Unidades de Conservação do Amazonas (CEUC/SDS/AM), Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio), Operação Amazônia Nativa (OPAN) and Projeto Médio Juruá (PMJ) for support on fieldwork logistics. Instituto Nacional de Pesquisas da Amazônia (INPA) for supporting on laboratory analysis. Special thanks to Paulo Apóstolo Lima Assunção and Nancy Lorena Maninguaje Rincón for assisting with plants species identification; to the lab assistants Laura N. Martins and Natália Medeiros Vicenti. Finally, we are grateful to all local villagers of the RESEX 85 Médio Juruá and RDS Uacari and dwellers of Carauari and Itamarati for their hospitality, friendship and trust. Funding sources This work was supported by the National Geographic Society [grant number: 265943], Rufford Foundation [grant number: 21911-1], Society for Conservation Biology [LACA Professional Development Award], and a DEFRA (Darwin Initiative for the Survival of Species) grant to CAP. ABS was granted a PhD studentship from Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) [Finance Code 001] References Abrahams, M.I., Peres, C.A., Costa, H.C.M., 2017. Measuring local depletion of terrestrial game vertebrates by central-place hunters in rural Amazonia. PLoS One 12, 1–25. https://doi.org/10.1371/journal.pone.0186653 Amaral, D.D., Viera, I.C.G., Salomão, R.P., Almeida, S.S. de., Jardim, M.A.G., 2009. Checklist da flora arbórea de remanescentes florestais da região metropolitana de Belém, Pará, Brasil. Bol. do Mus. Para. Emilio Goeldi, Ciências Nat. 4, 231–289. https://repositorio.museu-goeldi.br/handle/mgoeldi/1424 Assis, R.L., Haugaasen, T., Schöngart, J., Montero, J.C., Piedade, M.T.F., Wittmann, F., 2015. Patterns of tree diversity and composition in Amazonian floodplain paleo- várzea forest. J. Veg. Sci. 26, 312–322. https://doi.org/10.1111/jvs.12229 Bagchi, R., Swamy, V., Latorre Farfan, J.P., Terborgh, J., Vela, C.I.A., Pitman, N.C.A., Sanchez, W.G., 2018. Defaunation increases the spatial clustering of lowland Western Amazonian tree communities. J. Ecol. 106, 1470–1482. https://doi.org/10.1111/1365-2745.12929 Baraloto, C., Forget, P.M., 2007. Seed size, seedling morphology, and response to deep shade and damage in neotropical rain forest trees. Am. J. Bot. 94, 901–911. https://doi.org/10.3732/ajb.94.6.901 Bartón, K., 2016. multi-model inference R package version, 1(6). Bates, D., Maechler, M., Bolker, B., Walker, S., Chistensen, R.H.B., Singman, H., Dai, 86 B., Sheipl, F., Grothendieck, G., Green, O., Fox, J., 2019. Linear mixed-effects models using “Eigen” and S4. Beck, H., Snodgrass, J.W., Thebpanya, P., 2013. Long-term exclosure of large terrestrial vertebrates: Implications of defaunation for seedling demographics in the Amazon rainforest. Biol. Conserv. 163, 115–121. https://doi.org/10.1016/j.biocon.2013.03.012 Beckman, N.G., Muller-Landau, H.C., 2007. Differential effects of hunting on pre- dispersal seed predation and primary and secondary seed removal of two neotropical tree species. Biotropica 39, 328–339. https://doi.org/10.1111/j.1744- 7429.2007.00273.x Bello, C., Galetti, M., Pizo, M.A., Magnago, L.F.S., Rocha, M.F., Lima, R.A.F., Peres, C.A., Ovaskainen, O., Jordano, P., 2015. Defaunation affects carbon storage in tropical forests. Sci. Adv. 1. https://doi.org/10.1126/sciadv.1501105 Bennett, E.L., Milner-Gulland, E.J., Bakarr, M., Eves, H.E., Robinson, J.G., Wilkie, D.S., 2002. Hunting the world’s wildlife to extinction. Oryx 36, 328–329. https://doi.org/10.1017/s0030605302000637 Berenguer, E., Gardner, T.A., Ferreira, J., Aragão, L.E.O.C., Mac Nally, R., Thomson, J.R., Vieira, I.C.G., Barlow, J., 2018. Seeing the woods through the saplings: Using wood density to assess the recovery of human-modified Amazonian forests. J. Ecol. 106, 2190–2203. https://doi.org/10.1111/1365-2745.12991 Bogoni J., Ferraz K., Peres CA (2020) Extent, intensity and drivers of mammal defaunation: a continental-scale analysis across the neotropics. Nat Sci Reports 10:1–16. https://doi.org/10.1038/s41598-020-72010-w. Brienen, R.J.W. et al 2015. Long-term decline of the Amazon carbon sink. Nature 519, 344–348. https://doi.org/10.1038/nature14283 Cayuela, L., Granzow-de la Cerda, Í., Albuquerque, F.S., Golicher, D.J., 2012. Taxonstand: An r package for species names standardisation in vegetation databases. Methods Ecol. Evol. 3, 1078–1083. https://doi.org/10.1111/j.2041- 210X.2012.00232.x 87 Chave, J., 2002. Medição da densidade da madeira em árvores tropicais- Manual De Campo. Proj. PAN-AMAZONIA, Sixth Framew. Program. Chave, J., Coomes, D., Jansen, S., Lewis, S.L., Swenson, N.G., Zanne, A.E., 2009. Towards a worldwide wood economics spectrum. Ecol. Lett. 12, 351–366. https://doi.org/10.1111/j.1461-0248.2009.01285.x Chave, J., Muller-Landau, H.C.., Baker, T., Easdale, T., Steege, H.T., Webb, C.O., 2006. Regional and phylogenetic variation of wood density across 2456 neotropical tree species. Ecol. Appl. 116, 2356–2367. https://doi.org/10.3159/1095-5674(2007)134[301:scrohd]2.0.co;2 Cornejo, F.., Janovec, J., 2010. Seeds of Amazonian plants, 64th ed. Princeton University Press. Dirzo, R., Mendoza, E., Ort, P., 2007. Size-related differential seed predation in a heavily defaunated neotropical rainforest 39, 355–362. https://doi.org/10.1111/j.1744-7429.2007.00274.x Dirzo, R., Young, H.S., Galetti, M., Ceballos, G., Isaac, N.J.B., Collen, B., 2014. Defaunation in the Antrhopocene. Science (80). 401, 401–406. https://doi.org/10.1126/science.1251817 Effiom, E.O., Nunez-Iturri, G., Smith, H.G., Ottosson, U., Olsson, O., 2013. Bushmeat hunting changes regeneration of African rainforests. Proc. R. Soc. B Biol. Sci. 280, 20130246–20130246. https://doi.org/10.1098/rspb.2013.0246 Foley, J.A., Asner, G.P., Costa, M.H., Coe, M.T., DeFries, R., Gibbs, H.K., Howard, E.A., Olson, S., Patz, J., Ramankutty, N., Snyder, P., 2007. Amazonia revealed: Forest degradation and loss of ecosystem goods and services in the Amazon Basin. Front. Ecol. Environ. 5, 25–32. https://doi.org/10.1890/1540- 9295(2007)5[25:ARFDAL]2.0.CO;2 Fortunel, C., Paine, C.E.T., Fine, P.V.A., Kraft, N.J.B., Baraloto, C., 2014. Environmental factors predict community functional composition in Amazonian forests. J. Ecol. 102, 145–155. https://doi.org/10.1111/1365-2745.12160 88 Galetti, M., Bovendorp, R.S., Guevara, R., 2015. Defaunation of large mammals leads to an increase in seed predation in the Atlantic forests 3, 824–830. Galetti, M., Donatti, C.I., Steffler, C., Genini, J., Bovendorp, R.S., Fleury, M., 2010. The role of seed mass on the caching decision by agoutis, Dasyprocta leporina (Rodentia: Agoutidae). Zoologia 27, 472–476. https://doi.org/10.1590/S1984- 46702010000300022 Gardner, C.J., Bicknell, J.E., Baldwin-Cantello, W., Struebig, M.J., Davies, Z.G., 2019. Quantifying the impacts of defaunation on natural forest regeneration in a global meta-analysis. Nat. Commun. 10, 4590. https://doi.org/10.1038/s41467-019- 12539-1 Gelman, A.., Su, Y.-S., Yajima, M., Hill, J., Pittau, M.G., Kerman, J., Zheng, T., Dorie, V., 2020. Data Analysis Using Regression and Multilevel/Hierarchical Models. Green, P.T., Harms, K.E., Connell, J.H., 2014. Nonrandom, diversifying processes are disproportionately strong in the smallest size classes of a tropical forest. Proc. Natl. Acad. Sci. U. S. A. 111, 18649–18654. https://doi.org/10.1073/pnas.1321892112 Hammond, D.S., Brown, V.K., 1995. Seed size of woody plants in relation to disturbance, dispersal, soil type in wet neotropical forests. Ecology 76, 2544–2561. https://doi.org/10.2307/2265827 Harrison, R.D., Sreekar, R., Brodie, J.F., Brook, S., Luskin, M., O’Kelly, H., Rao, M., Scheffers, B., Velho, N., 2016. Impacts of hunting on tropical forests in Southeast Asia. Conserv. Biol. 30, 972–981. https://doi.org/10.1111/cobi.12785 Harrison, R.D., Tan, S., Plotkin, J.B., Slik, F., Detto, M., Brenes, T., Itoh, A., Davies, S.J., 2013. Consequences of defaunation for a tropical tree community. Ecol. Lett. 16, 687–694. https://doi.org/10.1111/ele.12102 Hawes, J.E., Peres, C.A., 2016. Patterns of plant phenology in Amazonian seasonally flooded and unflooded forests. Biotropica 48, 465–475. https://doi.org/10.1111/btp.12315 Hawes, J.E., Peres, C.A., 2014. Ecological correlates of trophic status and frugivory in 89 neotropical primates. Oikos 123, 365–377. https://doi.org/10.1111/j.1600- 0706.2013.00745.x Hazelwood, K., Paine, C.E.T., Cornejo Valverde, F.H., Pringle, E.G., Beck, H., Terborgh, J., 2020. Changes in tree community structure in defaunated forests are not driven only by dispersal limitation. Ecol. Evol. 10, 3392–3401. https://doi.org/10.1002/ece3.6133 Howe, H.F., Smallwood, J., 1982. Ecology of seed dispersal. Annu. Rev. Ecol. Syst. 13, 201–228. https://doi.org/10.1146/annurev.es.13.110182.001221 IBGE, 2018. Instituto Brasileiro de Geografia e Estatística [WWW Document]. Censo 2018. URL https://www.ibge.gov.br/ Jones, I.L., Peres, C.A., Benchimol, M., Bunnefeld, L., Dent, D.H., 2019. Instability of insular tree communities in an Amazonian mega-dam is driven by impaired recruitment and altered species composition. J. Appl. Ecol. 56, 779–791. https://doi.org/10.1111/1365-2664.13313 KEW, R.B.G., 2008. Seed information database (SID) [WWW Document]. version 7.1. Kurten, E.L., Wright, S.J., Carson, W.P., Palmer, T.M., 2015. Hunting alters seedling functional trait composition in a Neotropical forest. Ecology 96, 1923–1932. https://doi.org/10.1890/14-1735.1 Laurance, W.F., Oliveira, A.A., Laurance, S.G., 2004. Pervasive alteration of tree communities in undisturbed Amazonian forests. Nature 428, 171–175. https://doi.org/10.1038/nature02383 Melo, F.P.L., Rodriguez-Herrera, B., Chazdon, R.L., Medellin, R.A., Ceballos, G.G., 2009. Small tent-roosting bats promote dispersal of large-seeded plants in a neotropical forest. Biotropica 41, 737–743. https://doi.org/10.1111/j.1744- 7429.2009.00528.x Mooney, H., Larigauderie, A., Cesario, M., Elmquist, T., Hoegh-Guldberg, O., Lavorel, S., Mace, G.M., Palmer, M., Scholes, R., Yahara, T., 2009. Biodiversity, climate change, and ecosystem services. Curr. Opin. Environ. Sustain. 1, 46–54. 90 https://doi.org/10.1016/j.cosust.2009.07.006 Muscarella, R., Uriarte, M., 2016. Do community-weighted mean functional traits reflect optimal strategies? Proc. R. Soc. B Biol. Sci. 283. https://doi.org/10.1098/rspb.2015.2434 Nichols, E., Uriarte, M., Peres, C.A., Louzada, J., Braga, R.F., Schiffler, G., Endo, W., Spector, S.H., 2013. Human-induced trophic cascades along the fecal detritus pathway. PLoS One 8. https://doi.org/10.1371/journal.pone.0075819 Nunes, A.V., Peres, C.A., Constantino, P. de A.L., Santos, B.A., Fischer, E., 2019. Irreplaceable socioeconomic value of wild meat extraction to local food security in rural Amazonia. Biol. Conserv. 236, 171–179. https://doi.org/10.1016/j.biocon.2019.05.010 Nunez-Iturri, G., Olsson, O., Howe, H.F., 2008. Hunting reduces recruitment of primate-dispersed trees in Amazonian Peru. Biol. Conserv. 141, 1536–1546. https://doi.org/10.1016/j.biocon.2008.03.020 Oksanen, J., Blanchet, F.G., Friendly, M., Kindt, R., Legendre, P., Mcglinn, D., Minchin, P.R., O’hara, R.B., Simpson, G.L., Solymos, P., Henry, M., Stevens, H., Szoecs, E., Maintainer, H.W., 2019. Package “vegan” Title Community Ecology Package. Community Ecol. Packag. 2, 1–297. Oksanen, J., Kindt, R., Legendre, P., O’Hara, B., Simpson, G.L., Solymos, P.M., Stevens, M.H.H., & Wagner, H., 2008. The vegan package. Community Ecol. Packag. 190. Paine, C.E.T., Beck, H., 2007. Seed predation by neotropical rain forest mammals increases diversity in seedling recruitment. Ecology 88, 3076–3087. https://doi.org/10.1890/06-1835.1 Pearce, D.W., 2001. The economic value of forest ecosystems. Ecosyst. Heal. 7, 284– 296. https://doi.org/10.1046/j.1526-0992.2001.01037.x Peres, C.A.., van Roosmalen, M., 2002. Primate frugivory in two species-rich Neotropical forests: implications for the demography of large-seeded plants in 91 overhunted areas., in: Seed Dispersal and Frugivory: Ecology, Evolution and Conservation. pp. 407–421. Peres, C.A., Ecologia, D. De, Paulo, U.D.S., Postal, C., Sp, S.P., 1996. Population status of white-lipped Tayassu pecari and collared peccaries T. tajacu in. Biol. Conserv. 3207, 115–123. https://doi.org/10.1016/0006-3207(96)00010-9 Peres, C.A., Emilio, T., Schietti, J., Desmoulière, S.J.M., Levi, T., 2016. Dispersal limitation induces long-term biomass collapse in overhunted Amazonian forests. Proc. Natl. Acad. Sci. 113, 892–897. https://doi.org/10.1073/pnas.1516525113 Poorter, L., 2007. Are species adapted to their regeneration niche, adult niche, or both ? Am. Nat. 169, 433–442. Poorter, L., Bongers, F., 2006. Leaf traits are good predictors of plant performance across 53 rain forest species. Ecology 87, 1733–1743. Poulsen, J.R., Clark, C.J., Palmer, T.M., 2013. Ecological erosion of an Afrotropical forest and potential consequences for tree recruitment and forest biomass. Biol. Conserv. 163, 122–130. https://doi.org/10.1016/j.biocon.2013.03.021 R development core team (2019) R: a language and environment for statistical computing. Version 3.5.3, R Foundation. Vienna Redford, K.H., 1992. The Empty Forest. Bioscience 42, 412–422. Reich, P.B., 2014. The world-wide “fast-slow” plant economics spectrum: A traits manifesto. J. Ecol. 102, 275–301. https://doi.org/10.1111/1365-2745.12211 Rennó, C.D., Nobre, A.D., Cuartas, L.A., Soares, J.V., Hodnett, G.M., Tomasella, J., Waterloo, M.J., 2008. HAND, a new terrain descriptor using SRTM-DEM : Mapping terra firme rainforest environments in Amazonia. Remote Sens. Environ. 112, 3469–3481. https://doi.org/10.1016/j.rse.2008.03.018 Ripple, W.J., Chapron, …. H., Zhang, L., 2016. Saving the World’s Terrestrial Megafauna. Bioscience 66, 807–812. https://doi.org/10.1093/biosci/biw092 Roldán, A.I.., Simonetti, J.A., 2001. Plant-mammal interactions in tropical Bolivian 92 forests with different hunting pressures. Conserv. Biol. 15, 617–623. https://doi.org/10.1046/j.1523-1739.2001.015003617.x Silman, M.R., Terborgh, J.W., Kiltie, R.A., 2003. Population Regulation of a Dominant Rain Forest Tree by a Major Seed Predator. Ecology 84, 431–438. https://doi.org/http://dx.doi.org/10.1890/0012- 9658(2003)084[0431:PROADR]2.0.CO;2 Stoner, K.E., Vulinec, K., Wright, S.J., Peres, C.A., 1998. Hunting and plant community dynamics in tropical forests: a synthesis and future directions. Biotropica 39, 385–392. https://doi.org/10.1111/j.1744-7429.2007.00291.x Terborgh, J., Nuñez-iturri, G., Pitman, N.C.A., Cornejo, F.H., Alvarez, P., Swamy, V., Pringle, E.G., Paine, C.E.T., Terborgh, J., Timothy, E., 2008. Tree Recruitment in an Empty Forest. Ecol. Soc. Am. 89, 1757–1768.https://doi.org/10.1890/07-0479.1 Theimer, T.C.., Gehring, C.A.., Green, P.T.., Connell, J.H., 2011. Terrestrial vertebrates alter seedling composition and richness but not diversity in an Australian tropical rain forest. Ecology 92, 1637–1647. https://doi.org/10.1890/10-2231.1 van Roosmalen, M.G., 1985. Fruits of the Guianan flora. Vanthomme, H., Bellé, B., Forget, P.M., 2010. Bushmeat hunting alters recruitment of large-seeded plant species in Central Africa. Biotropica 42, 672–679. https://doi.org/10.1111/j.1744-7429.2010.00630.x Vira, B.., Wildburger, C., & Mansourian, S., 2015. Forests, trees and landscapes for food security and nutrition: a global assessment report. Westoby, M., 1998. A leaf-height-seed ( LHS ) plant ecology strategy scheme.213–227. https://doi.org/10.1023/A:1004327224729 Wilkie, D.S., Bennett, E.L., Peres, C.A., Cunningham, A.A., 2011. The empty forest revisited. Ann. N. Y. Acad. Sci. 1223, 120–128. https://doi.org/10.1111/j.1749- 6632.2010.05908.x Wright, S.J., Hernandéz, A., Condit, R., 2007. The bushmeat harvest alters seedling banks by favoring lianas, large seeds, and seeds dispersed by bats, birds, and wind. 93 Biotropica 39, 363–371. https://doi.org/10.1111/j.1744-7429.2007.00289.x Wright, S.J., Muller-Landau, H.C., Condit, R., Hubbell, S.P., 2003. Gap-dependent recruitment, realized vital rates, and size distributions of tropical trees. Ecology 84, 3174–3185. Wright, S.J., Zeballos, H., Dominguez, I., Gallardo, M.M., Moreno, M.C., Ibáñez, R., 2000. Poachers alter mammal abundance, seed dispersal, and seed predation in a neotropical forest. Conserv. Biol. 14, 227–239. https://doi.org/10.1046/j.1523- 1739.2000.98333.x Zanne, A.E.., Coomes, D., Jansen, S., Lewis, S.L.., Swenson, N.G.., Chave, J., 2009. Towards a worldwide wood economics spectrum. Ecol. Lett. Zuur, A.F.., Ieno, E.N.., Walker, N.J.., Saveliev, A.A.., Smith, G.M.., Walker, N.J., 2009. Statistics for Biology and Health. New York. Zuur, A.F., Ieno, E.N., Elphick, C.S., 2010. A protocol for data exploration to avoid common statistical problems. Methods Ecol. Evol. 1, 3–14. https://doi.org/10.1111/j.2041-210x.2009.00001.x 94 Supplementary Information Table S1: Taxonomic composition of trees ≥10 cm DBH and saplings (1-5 cm DBH), inventoried in 30 0.25-ha permanent tree plots and 30 0.05-ha sapling subplots in Médio Juruá region. Trees Saplings Family Nº genera Nº stems Nº genera Nº stems Achariaceae 1 4 1 1 Anacardiaceae 3 28 4 19 Anisophyllaceae 1 1 1 4 Annonaceae 6 88 11 280 Apocynaceae 6 28 9 45 Araliaceae 1 9 2 19 Arecaceae 6 223 6 402 Bignoniaceae 2 13 2 5 Bixaceae 1 2 1 1 Boraginaceae 1 21 1 23 Burseraceae 5 221 5 418 Caryocaraceae 1 14 2 3 Celastraceae 1 2 1 3 Chrysobalanaceae 5 306 4 331 Clusiaceae 5 46 6 88 Combretaceae 1 9 2 3 Connaraceae 0 0 1 5 Dichapetalaceae 1 11 1 16 Ebenaceae 1 4 1 29 Elaeocarpaceae 1 61 1 107 Erytrhoxylaceae 1 1 1 15 Euphorbiaceae 17 341 14 340 Fabaceae 40 533 31 600 Goupiaceae 1 14 1 1 Humiriaceae 2 15 3 10 Hypericaceae 1 13 1 5 Icacinaceae 1 3 1 3 Ixonanthaceae 1 1 0 0 Lacistemaceae 1 2 1 5 Lamiaceae 1 8 1 3 Lauraceae 9 128 9 234 Lecythidaceae 6 714 6 201 Lepidobotryaceae 1 3 0 0 Linaceae 0 0 2 5 Malpighiaceae 2 3 1 4 Malvaceae 10 165 9 127 Melastomataceae 3 37 4 108 Meliaceae 3 83 2 258 Menispermaceae 0 0 1 1 Monimiaceae 1 3 1 16 Moraceae 10 276 9 383 Myristicaceae 3 384 4 367 Myrtaceae 5 48 5 126 Nyctaginaceae 2 46 2 95 Ochnaceae 2 8 2 34 Olacaceae 6 50 6 45 To be continued .... 95 Trees Saplings Family Nº genera Nº stems Nº genera Nº stems Peraceae 2 7 0 0 Picramniaceae 0 0 1 10 Polygoniaceae 1 3 1 1 Proteaceae 1 4 1 2 Putranjivaceae 1 13 1 2 Quiinaceae 1 2 1 12 Rhizophoraceae 2 8 2 4 Rosaceaea 1 1 0 0 Rubiaceae 12 46 15 115 Rutaceae 1 1 1 2 Salicacea 3 17 2 37 Sapindaceae 5 12 6 54 Sapotaceae 7 428 6 396 Simaroubaceae 1 13 1 15 Siparunaceae 1 24 1 187 Solanaceae 1 1 1 1 Thymelaeaceae 0 0 1 1 Ulmaceae 1 3 1 3 Urticaceae 2 114 2 90 Violaceae 3 50 4 385 Vochysiaceae 3 62 3 24 Unidentified 0 5 0 3 Total 227 4784 230 6132 Table S2. Total abundance and percentage of trees and saplings categorized by dispersal mode surveyed in 30 permanent tree plots of 0.25 hectares (DBH>10 cm) and 0.05 subplots (1-5 cm DBH) located along to a hunting gradient in Médio Juruá region, Amazonas. Dispersal Mode Abiotic Gut: HIS Gut: HSS Scatter-hoarded tree sapling tree sapling tree sapling tree sapling Abundance 415 525 3,210 4,932 87 88 1,067 584 Proportion % 8.6 8.5 67.1 80.4 1.8 1.4 22.3 9.5 96 Table S3: Results from a linear-mixed model (LMM) examining the effect of hunting pressure in sapling-to-tree abundance ratio with water and soil-nutrient availability as fixed effects and species identity as the random effect. VDND: vertical distance to the nearest drainage, CEC: soil cation exchange capacity, DM: dispersal mode, HIS: hunting insensitive species, HSS: hunting sensitive species. R²c = 0.313; R²m = 0.035 Assemblage Explanatory variables *Estimate t-value p-value attribute (Intercept) -0.32 -4.240 p<0.001 VDND 0.06 2.179 0.029 CEC -0.21 -4.391 p<0.001 Log10 Hunting 0.09 1.093 0.274 abundance ratio VDND*CEC 0.23 5.350 p<0.001 saplings: trees DM: HIS1 0.44 0.097 p<0.001 DM: HSS1 0.02 0.377 0.922 DM: scatter-hoarded1 0.04 5.824 0.706 DM: HIS1*Hunting -0.20 -2.178 0.029 DM: HSS1*Hunting -0.51 -2.349 0.018 DM: scatter-hoarded1 * Hunting - 0.12 -1.117 0.264 1 Difference for Abiotic Dispersal Mode, * estimates are standardized 97 Table S4: Results from linear models (LMs) examining the effect of hunting pressure, vertical distance to the nearest drainage (VDND) and soil-nutrient (CEC) in the community weighted means (CWM) for continuous traits of wood density (WD), leaf mass per area (LMA) and seed mass (SM) for both tree, sapling and for the two life stages together weighted by tree plot basal area. variables estimate standard T- value p-value R² error CWM WD intercept 0.620 0.002 217.798 -0.06 sapling VDND 0.005 0.005 0.980 0.336 CEC -0.001 0.006 -0.247 0.807 Hunting 0.003 0.006 0.514 0.612 CWM LMA intercept -1.718 0.003 -564.395 -0.029 sapling VDND 0.002 0.006 0.337 0.739 CEC 0.002 0.006 0.328 0.746 Hunting 0.009 0.006 1.417 0.168 CWM SM intercept 0.183 0.014 12.744 -0.02 sapling VDND 0.016 0.029 0.573 0.571 CEC 0.000 0.031 0.031 0.976 Hunting 0.041 0.030 1.355 0.187 CWM WD intercept -0.204 0.002 -77.212 0.14 tree VDND 0.011 0.005 2.162 0.040* CEC 0.002 0.005 0.412 0.683 Hunting 0.010 0.005 1.770 0.084 CWM LMA intercept 0.019 0.000 138.882 tree VDND 0.000 0.000 1.264 0.218 -0.005 CEC -0.000 0.000 -0.382 0.705 Hunting 0.000 0.000 0.947 0.353 CWM SM intercept 0.377 0.020 18.018 -0.03 tree VDND 0.060 0.427 1.418 0.168 CEC -0.014 0.045 -0.313 0.757 Hunting 0.003 0.044 0.073 0.942 CWM WD intercept 1.208 0.013 90.441 0.036 tree + sapling VDND 0.053 0.027 1.975 0.059 weighted by CEC -0.021 0.028 -0.736 0.468 Basal Area Hunting -0.002 0.028 -0.092 0.928 CWM LMA intercept -0.247 0.125 -19.65 -0.040 tree + sapling VDND 0.028 0.025 1.116 0.274 weighted by CEC -0.020 0.027 -0.770 0.448 Basal Area Hunting 0.006 0.026 -0.245 0.809 CWM SM intercept 1.840 0.015 118 -0.036 tree + sapling VDND 0.028 0.031 0.899 0.377 weighted by CEC -0.030 0.033 -0.893 0.380 Basal Area Hunting 0.014 0.033 0.443 0.662 98 Table S5: Results of pairwaise Tukey-test comparing the mean value of the community weighted means (CWM) for continuous traits of wood density (WD), leaf mass per area (LMA) and seed mass (SM) between dispersal mode of abiotic, endozoochory for hunting insensitive species (gut: HIS), endozoochory for hunting sensitive species (gut: HSS) and scatter-hoarded species. Plant Trait Dispersal mode difference upper lower p-value gut HIS: abiotic -0.030 -0.062 0.000 0.060 gut HSS: abiotic -0.021 -0.111 0.069 0.930 Wood Density scatter-hoarded: abiotic 0.026 -0.016 0.070 0.378 gut HIS: gut HSS 0.009 -0.076 0.095 0.991 scatter-hoarded: HIS 0.057 0.024 0.090 0.000 scatter-hoarded: HSS 0.048 -0.042 0.138 0.523 gut HIS: abiotic 0.001 0.000 0.002 0.011 gut HSS: abiotic 0.004 0.001 0.004 0.000 LMA scatter-hoarded: abiotic 0.002 0.001 0.004 0.000 gut HIS: gut HSS 0.003 0.000 0.005 0.014 scatter-hoarded: HIS 0.001 0.000 0.002 0.000 scatter-hoarded: HSS -0.001 -0.004 0.001 0.563 gut HIS: abiotic -0.237 -0.815 0.341 0.716 gut HSS: abiotic 0.566 -1.057 2.190 0.806 Seed Mass scatter-hoarded: abiotic 2.496 1.707 3.284 0.000 gut HIS: gut HSS 0.803 -0.738 2.345 0.536 scatter-hoarded: HIS 2.733 2.130 3.336 0.000 scatter-hoarded: HSS 1.929 0.297 3.562 0.012 Figure S1. Procrustes distances between the species composition of trees >10 cm DBH and saplings (1-5 cm diameter) based on Bray-Curtis dissimilarity values, for which lighter blue represents tree plots under increasingly heavier hunting pressure. 99 Figure S2: Correlations between continuous functional traits, including wood density (WD), LMA and Seed Mass (SM) for 783 species in our database. 100 101 CAPÍTULO III CONSEQUENCES OF DEFAUNATION FOR ABOVEGROUND CARBON STOCKS IN AMAZONIAN FORESTS Andressa Bárbara Scabin1, * Joseph E. Hawes2 and Carlos A. Peres3,4 1Programa de Pós-Graduação em Ecologia, Universidade Federal do Rio Grande do Norte - Natal, RN, Brazil. 2Faculty of Environmental Sciences and Natural Resource Management, Norwegian University of Life Sciences, Universitetstunet 3, 1433 Ås, Norway 3Centre for Ecology, Evolution and Conservation, School of Environmental Sciences, University of East Anglia, Norwich, UK. 4 Departamento de Sistemática e Ecologia, Universidade Federal da Paraíba, João Pessoa, PB, Brazil. Author for correspondence: * Andressa B. Scabin, Email:andressa_scabin@ufrn.edu.br Departamento de Ecologia, Centro de Biociências, Universidade Federal do Rio Grande do Norte campus Lagoa Nova, Natal, RN 59072-970, Brazil. Short title: Carbon stocks in hunted Amazonian forests 102 Abstract Overhunting is a cryptic threat that has emptied tropical forests worldwide. The large- bodied vertebrates decline and the resulting disruptions in plant-animal interactions may promote substantial changes in forest composition. As a result, some forest ecosystem services, such as carbon storage, can be considerably downgraded. Here, we estimate the current and future forest carbon stocks of 30 sites across a hunting gradient in western Brazilian Amazonia to test the hypothesis of reduced forest carbon density as an indirect effect of defaunation. Our aboveground biomass (AGB) forest estimates are based on tree and sapling inventories conducted in thirty 0.25-ha tree plots and thirty nested 0.05-ha sapling subplots established along the hunting pressure gradient. Estimates of future forest carbon were based on a wood density replacement scenario using the pool of species present in the sapling layer. To improve the accuracy of our AGB estimates we used wood density data obtained at the study area for the 250 most abundant tree species. Our results show that in heavily hunted areas there was a higher abundance of trees, but these trees were smaller in diameter and height. Even with smaller stature, forest sites exposed to heavier hunting pressure did not show differences in basal area and current AGB estimates. Given projections of future forest carbon stocks, 22 of our 30 tree plots lost carbon. Mean projected loss was 2.2 MgC ha-¹, which represents 1.3% of the current carbon stock, but these losses could reach 6.3% representing 9.7 Mg ha-¹ of carbon erosion. For two large protected areas within the study landscape amounting to 708, 941 ha, the projected loss was approximately 1,560 MgC. Considering the monetary values in international carbon markets, the projected decrease in future carbon stock was valued from US$ 15.6 to US$ 120 million. Keywords: defaunation, seed dispersal, wood density, aboveground biomass, carbon storage, carbon market. 103 1. INTRODUCTION Overhunting is a cryptic threat that, together with habitat loss, has gradually emptied tropical forests worldwide, even inside protected areas (Benítez-López et al. 2019). Large-bodied terrestrial and arboreal vertebrates are particularly affected, as they are both the most preferred game species for subsistence and commercial hunters and exhibit low reproductive rates, thereby recovering slowly from sustained hunting pressure (Peres & Palacios 2007). With the decline of large vertebrates, critical ecological plant- animal interactions that modulate forest dynamics are either lost or reduced, thereby degrading the provision of associated ecosystem services. Ecosystem services detrimentally affected by defaunation include pollination, seed dispersal, nutrient cycling, pest control and carbon storage (Dirzo et al. 2014). Large-bodied mammals and birds often provide most of the high-quality seed dispersal services for large-seeded plants (Jordano et al. 2007; Pires et al. 2017). Animal dispersed trees bearing large seeds are generally higher statured and have higher wood density (Hawes et al. 2020, Scabin et al Chapter 2) and therefore have greater capacity to store carbon, compared to small-seeded, lighter-wooded tree species (Bello et al. 2015; Peres et al. 2016). Consequently, chronic dispersal limitation associated with extirpations or declines of large-bodied frugivores and their mutualistic interactions can significantly downgrade forest carbon storage capacity (Poulsen et al. 2013; Osuri et al. 2016; de Paula et al. 2018). For this reason, wildlife overhunting can affect carbon stocks in tropical forests, calling for controlling illegal hunting within REDD+ (Reducing Emissions from Deforestation and Forest Degradation) land use options (Bello et al. 2015; Peres et al. 2016; Brodie 2018). Models simulating forest carbon storage across the tropics show that Afrotropical, Neotropical and Southern Asian forests, whose tree flora experiences higher proportions of biotic dispersal, can lose between 2-12% of their carbon stocks due to defaunation. In contrast, carbon losses and gains in South-East Asian and Australian tropical forests, which harbour a greater proportion of abiotically dispersed trees, are far more modest at approximately 1% (Osuri et al. 2016), although estimates for South-East Asia have been questioned by Chanthorn et al. (2019) who found losses of 2.4 to 3%. These studies nevertheless show that responses of forest carbon dynamics to defaunation may diverge 104 according to regional scale differences in floristics and the degree to which large heavy- wooded trees critically depend on animal dispersal agents (Osuri et al. 2016). Across the Neotropics, two regional scales studies have simulated forest carbon loss induced by extirpations of large-seeded trees due to partial defaunation. A comparative analysis of 31 Atlantic Forest tree plots in Brazil, showed that defaunation has the potential to significantly erode forest carbon stocks, even if only a small proportion of tree species undergo local extinctions (Bello et al. 2015). Likewise, extinction simulations of large-bodied primate and tapir dispersed trees across 2345 1-ha tree plots distributed throughout the Brazilian Amazon conservatively projected an average forest carbon loss per hectare of 5.8%, increasing to as much as 37.8% in some regions (Peres et al. 2016). Although models considering different defaunation scenarios are entirely consistent in predicting a reduction in forest carbon storage as a result of plant compositional changes, there is a lack of studies based on empirical data obtained under different hunting pressure scenarios to estimate carbon stocks in tropical forests. Detecting forest carbon loss using both spatial or temporal replication designs would require long-term studies since this involves extremely slow processes of tree recruitment, growth, and mortality which can take over a century (de Paula et al. 2018). However, predictions of future forest carbon stocks can be improved based on simulations of plant community changes in the species pool of the sapling layer that will eventually replace increasingly senescent individuals. This provides a feasible alternative to derive estimates of long-term community-wide changes in aboveground biomass (hereafter, AGB). For example, monitoring saplings to understand the effects of anthropogenic impacts on wood density proved to be an interesting approach to better understand the future forest carbon storage capacity (Berenguer et al. 2018). Additionally, AGB estimates derived from wood density data obtained in situ can ensure greater accuracy in estimates incorporating regional scale variation in wood density (Baker et al. 2004). Here, we exploited a historically consolidated gradient of game hunting pressure in western Brazilian Amazonia to both quantify contemporary aboveground carbon stocks and estimate future stocks under scenarios of plant composition change. We predict losses in projected future AGB due to an increase in small-seeded, lighter-wooded tree species in the forest sapling layer, as consequence of elevated dispersal limitation in large-seeds 105 species if populations of large-bodied frugivores are depleted. To test these predictions, we derive AGB estimates of both present and future carbon stocks, based on wood density values obtained from wood core samples collected in the field and forest inventories conducted at 30 forest sites distributed across our hunting pressure gradient. Limiting hunting activity in tropical forests could provide long-term benefits to hunters by avoiding wildlife depletion and hence loss of income derived from sales of carbon credits (Brodie 2018). Having this in mind, we also monetarily value carbon losses and gains using the international carbon market values as reference, with the aim of discussing the possible financial losses related to changes in the carbon stock, thinking in terms of the use of emission control mechanisms like REDD +. METHODS 2.1 Study Area This study took place in the Médio Juruá region of western Brazilian Amazonia (Fig. S1), including two large contiguous sustainable-use protected areas and adjacent landscapes containing two urban clusters. This roughly represents the middle-third section of the Juruá River, the second-longest white-water tributary of the Amazon River. The two protected areas include the 253,227-ha Medio Juruá Extractive Reserve ( RESEX Médio Juruá; 5º33'54"S, 67º42'47"W), created in 1997 and legally occupied by ~ 2,000 people distributed across 13 villages; and the 632,949-ha Uacari Sustainable Development Reserve (RDS Uacari; 5º43'58"S, 67º46'53"W) created in 2005, where ~ 1,200 people occupy 32 villages. The nearest towns are Carauari with a population of 28,076 habitants and located 88 fluvial km downstream from the RESEX Médio Juruá, and Itamarati with a population of 7,888, located 120 fluvial km upstream from the RDS Uacari (IBGE 2018). The Médio Juruá region has a wet tropical climate with a mean annual temperature of 27.1°C and a mean annual rainfall of 3,679 mm, with the wettest period between November and April. Two different forest types comprise the study landscape: seasonally-flooded (várzea) forests, which account for ~20% of the study region, characterized by enriched Andean alluvial soils and lower floristic diversity, and the dominant (~80%) unflooded forest (terra firme), which exhibits higher floristic diversity and comparatively lower soil fertility (Hawes & Peres 2016). The current study was performed in unflooded forest on paleo-várzea sediments, which we refer to here as terra firme for simplicity, but we recognize that these forests, may diverge in their floristic 106 macromosaics from so-called terra firme forests (Assis et al. 2015). Our tree plots were established along sites that had experienced subsistence and commercial hunting to varying degrees but had limited history of clear-cuts, wildfires and timber extraction. Using our own prior knowledge of the study area, we selected 14 subregions that were homogeneously distributed along the gradient of hunting pressure. This selection was based on both distance from the nearest urban centre, physical accessibility and previous studies carried out in the area by the Médio Juruá Project (PMJ) and Programa de Monitoramento da Biodiversidade em Unidades de Conservação Estaduais do Amazonas (PROBUC). In each one of these sites we established a 0.25-ha tree plot nested with a 0.05 sapling plot. Tree plots From May to December 2017 we established 30 permanent tree plots of 0.25 ha (100m × 25m) along the hunting gradient in the study area; within each we nested thirty 0.05-ha subplots (100m × 5m) along a 100-m central corridor that was marked every 10 m with PVC pipes. Inside each plot we surveyed all trees ≥10 cm diameter at breast height (DBH) and, in the subplots we inventoried all saplings 1 - 5cm DBH. We measured and number-tagged all live trees and saplings at 1.3 m and 1.0 m in height, respectively. Forest Inventory Floristic inventories across all 30 tree plots and 30 subplots were conducted in July 2018. All trees and saplings were identified to the finest possible taxonomic level in the field by a parabotanist with over 30 years of field expeditions and herbarium experience at the National Institute for Amazon Research (INPA), Manaus, Brazil. Voucher specimens of all individuals that could not be identified in the field were collected and subsequently identified at the INPA herbarium with the assistance of another parabotanist and then deposited at the EAFM herbarium of the Instituto Federal de Educação, Ciência e Tecnologia do Amazonas (IFAM, Manaus). To automatically standardise plant nomenclature and correct for synonyms based on The Plant List (http://www.theplantlist.org) we used the Taxonstand 22.2 R package (Cayuela et al., 2012). In each permanent plots, approximately 20 trees belonging to 10 different diameter classes (1-5 cm; 6-10 cm, 11-15 cm ...., > 50 cm DAP) were selected for height measurements including four of the largest trees within the plot, as suggested by Sullivan 107 et al. (2018). We used a clinometer, to measure the inclination angle between the observer and (i) the upper tree crown and (ii) the base of the tree, and a measuring tape to record (iii) the distance between the observer and the tree. We then applied the Pythagorean theorem to these three measures to obtain the height of each tree (Chave 2006). Wood density (WD) To obtain species-specific WD we used a 5.15 mm Haglöf increment borer to typically sample three wood cores from different individuals representing each of the 250 most abundant tree species, amounting to a total of 700 wood core samples. Wood cores were extracted perpendicularly to the bark at about 1 m height from trees >10 cm DBH. To encompass the variation in heartwood-to-bark WD, cylindrical samples were equivalent to the approximate length of the bole radius (DBH × 0.5). Each wood core was first wrapped in filter paper and, to avoid fungal attack, sealed in a box containing silica gel for initial drying over 5 days and then stored in labelled plastic straws. We then rehydrated all wood samples for 24 h at the Plant Functional Laboratory at INPA. We subsequently obtained the green volume of each core through the water displacement method, using a beaker of water placed on a digital balance (precision ≈ 0.01g) which was re-zeroed each time. All wood cores were then oven-dried at 105°C for 72 h, at which point they were dry-weighed. WD values were calculated by dividing the dry wood core weight by the green volume of each sample (Chave 2002). As we used the water displacement method, we obtained the wood specific gravity (WSG) value, which is the relationship between the dry wood mass and the wood volume at saturation point in relation to the volume of water; for simplicity, this is referred to hereafter as wood density. To obtain species-specific WD values we averaged the WD values across all samples of the same species. WD values were obtained for 54.8% of all stems surveyed. For the minority of genera (18%) lacking WD values obtained from our own sample, we used data from the Global Wood Density Database (Chave et al. 2009; Zanne et al. 2009) including only values from the South America tropical region. In total, 47.2% of all individuals were assigned to a WD value at the species level, 29.8% at the genus level, 21.1% at the family level and only 0.1% were assigned the mean WD value of each plot, due to complete lack of data available for these taxa. Hunting Pressure We built a proxy of hunting pressure based on the geographic distance to and size of regional-scale human settlements, including villages and towns. Previous studies in the 108 study area have shown that distance from urban centres represents a good proxy for the anthropogenic impact on large vertebrate abundance (Nichols et al. 2013; Abrahams et al. 2017, Scabin & Peres, Chapter 1). We measured the Euclidean distance from each plot centroid to the nearest community and the dry-season navigation (fluvial) distance to the towns using ArcGIS10.3. Human population size was derived from IBGE census data (IBGE 2018), and from the Projeto Médio Juruá (PMJ) and the Sustainable Amazon Foundation (FAS) databases. Environmental variable potentially affecting AGB At all 30 plots, we collected three soil samples of ~5 grams each using a soil auger with samples extracted near the plot centroid, at least 10 m apart. Soil samples were air- dried in sealed plastic bags until analysis in the Soil Chemistry Laboratory at INPA. Soil fertility was represented by the cation exchange capacity (CEC) as the sum of Ca+2, Mg+2, and K+ concentrations, and measured as cmol kg-¹. The vertical distance to the nearest drainage (VDND), which represents a proxy of water availability for plants, was calculated by the Height Above the Nearest Drainage (HAND) algorithm proposed by Rennó et al. (2008) based on the 30-m resolution Digital Elevation Model (DEM) topography available from the Shuttle Radar Topography Mission (SRTM). The vertical distance grids generated by the HAND algorithm are available from the Brazilian Space Research Institute (INPE) website (www.dpi.inpe.br). 2.7 Data analysis First, we investigated the spatial structure of the floristic composition performing a Mantel test, which compared the floristic composition matrix based on the Morisita-Horn dissimilarity and the Euclidean geographic distance matrix based on plot coordinates. Distance matrices were obtained using the vegdist function in the vegan 2.5-4 R package (Oksanen et al. 2008). To test whether hunting pressure has affected forest structure, we assigned a hunting pressure value to each plot and performed linear regressions with tree density, DBH, basal area, and tree height. Total basal area was calculated for each 0.25-ha plot using the equation ∑𝜋(DBH𝑖/2)², where DBHi is the DBH for each tree, and then converted into basal area per hectare (m² ha-¹). To predict tree height for all unmeasured trees, we regressed 493 tree heights measured in the field with their respective DBH, to build tree 109 heights models based on DBH using the modelHD function of the BIOMASS 2.1.1 R package (Réjou-Méchain et al. 2017) with model comparisons using the lowest Residual Standard Error (RSE). The best fit model for our data was “log2” (Fig. S2), which we then used to estimate heights for all trees based on their DBH field measurements. To examine whether plot-level tree WD responded to hunting pressure we perform a linear regression. As we expected mean plot-level wood density for saplings and trees to be differently affected by hunting pressure, we tested whether the mean WD between these stage classes increased along the hunting gradient performing a pairwise z-test, and then regressed the z-scores with hunting pressure values. Current AGB was calculated for both each tree and whole plots using the computeAGB function in the BIOMASS 2.1.1 R package (Réjou-Méchain et al. 2017), based on the allometric equation using DBH, height and wood density, as proposed by Chave et al. (2014). We used the following equation: AGB=exp(-2.024-0.896*E+0.920*ln (WD) + 2.795 * ln (DBH)-0.0461*[ln(DBH)²]) Where, WD is wood density, DBH diameter ate breast height and E is a bioclimatic parameter (see in Chave et al 2014). We used a generalised linear model with Gaussian error structure to examine whether the current plot-level AGB was associated with hunting pressure, and water and soil- nutrient availability. Explanatory variables were tested for multicollinearity through variance inflation factors (VIFs), where VIFs<4 indicates low multicollinearity (Zuur et al. 2010). All variables were retained in the model as they were uncorrelated. To estimate a next-cohort projection of tree AGB at each plot, we used the species composition of the sapling subplots, underneath the tree plot, as a starting point. We ran 1,000 random species replacement simulations using a simple lottery model based on the wood density distribution across all saplings at each subplot and then assumed a novel future tree assemblage that retains the current DBH distribution and total basal area of the present tree plot. In other words, we assume changes in wood density, rather than changes in forest structure. In this lottery model, the probability of any conspecific or heterospecific tree-by-sapling substitution is identical to the relative abundance of each sapling species and their respective WD value. We then calculated the plot-scale AGB per simulation (N = 1000 per plot), using the allometric equation and procedure described 110 above. Finally, we calculated the differences between currently estimated AGB and all simulated estimates of total AGB per plot. AGB, expressed in Mg ha-¹, was converted into carbon stocks (MgC ha-¹) using a factor of 0.5, assuming a carbon content of ~50% of AGB (Chave et al. 2005). It is also important to note that the methods applied here to predict future carbon stocks are inherently limited in scope and likely carry large margins of uncertainty. First, we simulated a full replacement scenario in stem species composition, based on the available sapling species pool at each plot, yet species turnover occurs gradually over time and depends on the complex dynamics of tree mortality and recruitment, that can only be incorporated through long-term monitoring of forest dynamics. Second, some species benefit more than others from the absence of large-seeded species and have a competitive advantage with higher chances of recruiting (de Paula et al. 2018). Finally, we used random replacements based on WD only to account for changes in carbon density while retaining forest structure unchanged, yet plot-scale basal area explains a large fraction of the variation in AGB (Baker et al. 2004). 2. RESULTS Tree and sapling surveys We recorded a total of 4,560 trees and 5,730 saplings belonging to 837 species, 261 genera and 66 families within 30 permanent tree plots (7.5 ha) and 30 sapling subplots (1.5 ha). Considering all individuals, 99% were identified at the species level and the remaining 0.9% at the genus level; only 8 individuals could not be identified. The mean tree and sapling abundance across all plots were 637.2 ± 97.6 trees/ha (mean ± SD) and 4,086 ± 1,113 saplings/ha, respectively. The density of trees in heavily hunted areas was significant high, but trees in these areas were typically smaller both in terms of DBH and height (Fig. 1C and 1D). While the overall basal area was not significantly different with a mean value of 29.5 ±4.5 m² ha-1 (Fig. 1B). Thus, tree plots located in the most hunted areas did not show significant differences in basal area. 111 Wood density and hunting pressure We did not find significant difference in the mean wood density along to the hunting gradient (Fig. 2A). However, conspecific saplings presented lower WD in persistently hunted areas, since the sapling-to-adult ratios in these plots were well below 1 (Fig. 2B). These generational differences are not only attributed to lower WD values in juveniles, but also to differences accentuated by the high wood density values for trees. Plot-scale differences in WD between trees and saplings were significantly greater in heavily hunted areas (R² = 0.18, p=0.009). Site level hunting pressure and soil fertility were marginally correlated (r = -0.33, p=0.069, Fig S2). FIGURE 1: Linear regression fits predicting (A) tree density, (B) tree basal area, (C) DBH, and (D) estimated height as a function of hunting pressure. Colour-coded circles represent the hunting pressure gradient, where red and blue represent higher and lower values. Circles and error bars represent means and 95% confidence limits for each plot. Line and shading represent a linear fit and the 95% confidence interval region. 112 FIGURE 2: Relationship between hunting pressure and (A) tree wood density and (B) proportional changes in mean wood density between trees and saplings, with negative values indicating lower mean wood density in saplings. Colour-coded circles represent the hunting pressure gradient, where red and blue represent higher and lower values. Contemporary aboveground biomass There was wide variation in AGB in non-hunted and lightly hunted sites compared to the low range of estimates in heavily hunted sites, which on average supported 300 Mg ha-¹ (Fig. 3). In any case, our models failed to detect an effect of any environmental variables on aggregate AGB, including water availability (t = 0.88, p =0.38), soil fertility (t= - 0.95, p=0.34) and hunting pressure (t = -0.28, p = 0.77). FIGURE 3: Relationship between our metric of hunting pressure and aboveground biomass (AGB) at the plot level. Colour-coded circles represent the hunting pressure gradient, where red and blue represent high and low values, respectively. 113 Future Forest Carbon Stocks Concerning the next-cohort projections of forest AGB, 22 of the 30 plots inventoried lost biomass on the basis of sapling-to-tree WD replacements (Fig. 4 and 5). These negative changes in AGB were on average of 4.4 ± 14.7 Mg ha-¹, indicating a mean carbon loss of 1.3%, but which could be as much as 6.3%. Contrary to our expectations, estimates of AGB change from trees to saplings were poorly predicted by the entire hunting pressure gradient, with some under intermediate hunting pressure showing gains of 11.5 Mg C ha-¹. Plots under low hunting pressure showed high variance in AGB change including losses of ~6.2 Mg C ha-¹, and all sites exposed to persistent hunting pressure consistently lost biomass. The spatial distribution of AGB gains and losses across the study area shows that peri-urban forests near Carauari and Itamarati are expected to lose biomass over time. On the other hand, plots located within the RESEX Médio Juruá are expected to gain biomass in the future. Some sparsely settled sites in the Uacari Reserve were also expected to lose 6 Mg C ha-¹, one of the highest losses projected (Fig. 4). FIGURE 4: Mean AGB losses and gains under future scenarios based on a lottery model of stem based WD replacement across the entire hunting pressure gradient. Circles and error bars represent means and 95% confidence limits for each plot based on 1000 simulations. The dotted line indicates no generational change in AGB from trees to saplings. 114 FIGURE 5: Average carbon gains and losses in MgC ha-1 for future carbon stocks projected for tree plots distributed along our hunting pressure gradient in the Médio Juruá region. Colour-coded circles represent carbon loss (red) or gain (green) and circle sizes are proportional to the magnitude of those changes. Pricing carbon losses We further provide a monetary valuation of future changes in forest carbon stocks extrapolated to all non-flooded (terra firme) forest areas amounting to ~80% of the two focal reserves (or a combined area of 708,94 ha). A mean estimated loss of 2.2 MgC ha-¹ would represent a total carbon erosion of 1,561,807 MgC for just these two sustainable use reserves. Assuming a very conservative value of US$ 10/ MgC ha-¹ based on the mean international carbon trade value for April/2020 from the EU ETS instrument (European Union Emission Trading System, www.carbonpricingdashboard.worldbank.org), the monetary value of our predicted carbon loss for the two protected areas would be approximately US$15.6 million. However, we also considered a reference carbon price 115 of US$77 per MgC, which is much closer to the international consensus in carbon trading to mitigate the effects of climate change (Poelhekke 2019). Applying this high-end value would result in a predicted monetary cost of US$120.094 million for the two reserves alone, based on the projected changes estimated here. 3. DISCUSSION The Amazon is immensely important in balancing the global carbon budget because it still secures the largest terrestrial carbon pool of any single tropical forest region on Earth (Malhi et al. 2008; Nogueira et al. 2017). However, this aggregate carbon stock is likely to severely decline in the foreseeable future unless it is effectively protected from deforestation and forest degradation (Fearnside & Laurance 2004; Nogueira et al. 2015). Forest cover in the Médio Juruá region remains largely undisturbed, having experienced a long history of low anthropogenic impacts compared to other Amazonian river basins. Even in the face of a very favourable conservation scenario, peri-urban forests in the Juruá valley present clear signs of overhunting of harvest-sensitive vertebrate game species (Abrahams et al. 2017; Scabin & Peres, chapter 1). These effects of hunting have been shown to reverberate on the functional tree composition of the forest, with marked changes in the prevalence of plant guilds associated with animal dispersal (Scabin, Costa & Peres, Chapter 2). In this study, we sought to understand to what degree directional changes in forest composition along a marked hunting pressure gradient in the Médio Juruá region was reflected in both present-day and future forest carbon stocks. Firstly, considering the forest structure, our results showed that although smaller statured (DBH and tree heights) and heavy-wooded trees were prevalent in heavily hunted areas, estimated AGB of present-day forest plots were not significantly different from those in lightly hunted areas. This may suggest that neither hunting nor other anthropogenic activities such as logging have apparently affected the biomass of these forests. On the other hand, wood density is a pivotal predictor of aboveground phytomass accumulation (Chave et al. 2006), and significant differences in mean wood density between saplings and trees in heavily hunted forests had a meaningful effect on the carbon storage capacity of future forest stands. Considering our projections of future forest carbon stocks, there were negative swings in AGB at 22 of our 30 forest sites. Contrary to our expectations, however, 116 projected losses in forest carbon were not restricted to heavily hunted sites. Instead, they were intermittently observed throughout the entire gradient of hunting pressure. Mean declines in AGB were modest and just over 1%, which is lower than estimates of 2.5– 5.8% of forest carbon loss based on modelling forest carbon transitions under different hunting scenarios at 2,345 1-ha plots throughout the Brazilian Amazon (Peres et al. 2016). Considering that our projections of future carbon stocks reshuffled the tree species composition and relative abundance based on what is currently available in the sapling layer, while assuming an overall unaltered forest structure, changes in AGB were attributed to differences in WD between saplings and trees. Small differences in wood density between trees and saplings are expected since the functional composition of the sapling layer reflects the species composition of reproductive trees (Wright et al. 2003), but a higher abundance of light-wooded pioneers is expected in the understorey (Poorter & Bongers 2006). However, this would not be enough to explain our findings since negative differences in mean wood density between trees and saplings increased in hunted areas. Higher wood densities in heavily hunted areas could be associated with lower soil fertility (Muller-Landau, 2004), but this is unlikely to explain the higher differences in wood density between trees and saplings, given that soil fertility can affect the growth rates of all stems within a tree plot. In any case, edaphic effects in this case run against our hunting pressure gradient, possibly predisposing environmental filters that select for heavy-wooded stems in older age classes at oligotrophic sites. In addition, carbon stock losses were estimated for 73% of the plots, but in 37% of them, there were carbon gains exceeding 10 Mg ha-1. The spatial congruence between our projected changes in forest carbon and the current distribution and size of human settlements was at best inconclusive and puzzling. This could be further masked, however, by historical effects of past hunting pressure which could not be properly reconstructed in this study. Several sites sampled here were within the hunting catchment of much larger past settlements established during and after the 20th century rubber boom but are now much smaller because of collapsed rubber subsidies and persistent rural-urban migration since the 1970s. This is the case of Cachoeira, for example, a local community currently containing only three households (~15 people). According to local dwellers, however, this was once much larger settlement (>200 inhabitants), thriving near a rubber plantation, that undoubtedly subsisted on a daily basis on fish and game protein. For the tree plot located near this site, we estimated one 117 of the highest negative swings in carbon stocks (~ 9 MgC ha-¹). We acknowledge therefore that the full history of local wildlife harvesting, going back half a century or longer, could not be adequately taken into account, but likely shaped some of the unexplained components of tree recruitment, particularly for slow-growing seedlings/saplings, thereby perhaps affecting the long-term spatial mosaic of forest carbon dynamics across our meta-landscape. In conclusion, carbon stocks projections based on the functional profile of both established trees and saplings recruiting in the understorey can provide a better forecast of the future plant species composition. Both carbon losses and gains were not exclusive to heavily hunted forest, but differences in wood density between saplings and trees were strongly related to hunting pressure. Therefore, a better understanding of to what degree frugivore population density affects plant functional traits is imperative to predict levels of sustainable game offtake to maintain a sufficient fabric of plant-animal interactions that can inhibit forest carbon erosion. Although the devil-in-the-detail of fully costing sustainable game management by local communities in Amazonian forests is yet to be worked out, it is clear that doing so will accrue much greater cross-scale benefits, well beyond the wildlife value, but also ecosystem services related. Acknowledgments We are grateful to Associação dos Produtores Rurais de Carauari (ASPROC), Centro Estadual de Unidades de Conservação do Amazonas (CEUC/SDS/AM), Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio), Operação Amazônia Nativa (OPAN) and Projeto Médio Juruá (PMJ) for support on fieldwork logistics. Instituto Nacional de Pesquisas da Amazônia (INPA) for support on laboratory analysis. Special thanks to Paulo Apóstolo Lima Assunção and Nancy Lorena Maninguaje Rincón for assisting with plants species identification; to the lab assistants Laura N. Martins and Natália Medeiros Vicenti. Finally, we are grateful to all local villagers of the RESEX Médio Juruá and RDS Uacari and dwellers of Carauari and Itamarati for their hospitality, friendship and trust. 118 Funding sources This work was supported by the National Geographic Society [grant number: 265943], Rufford Foundation [grant number: 21911-1], Society for Conservation Biology [LACA Professional Development Award], and a DEFRA (Darwin Initiative for the Survival of Species) grant to CAP. ABS was granted a PhD studentship from Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) [Finance Code 001]. References Abrahams, M.I., Peres, C.A. & Costa, H.C.M. (2017). Measuring local depletion of terrestrial game vertebrates by central-place hunters in rural Amazonia. PLoS One, 12, 1–25. Assis, R.L., Haugaasen, T., Schöngart, J., Montero, J.C., Piedade, M.T.F. & Wittmann, F. (2015). Patterns of tree diversity and composition in Amazonian floodplain paleo-várzea forest. J. Veg. Sci., 26, 312–322. Baker, T., Phillips, O., Malhi, Y., Almeida, S., Arroyo, L., Di Fiore, A., Erwin, T., Killeen, T.J., Laurence, S.G., Laurance, W.F., Lewis, S.L., Lloyd, J., Monteagudo, A., Neill, D. a., Patiño, S., Pitman, N.C., Silva, J.N.M. & Martínez, R.V. (2004). Variation in wood density determines spatial patterns in Amazonian forest biomass. Glob. Chang. Biol., 10, 545–562. Bello, C., Galetti, M., Pizo, M.A., Magnago, L.F.S., Rocha, M.F., Lima, R.A.F., Peres, C.A., Ovaskainen, O. & Jordano, P. (2015). Defaunation affects carbon storage in tropical forests. Sci. Adv., 1.. Benítez-López, A., Santini, L., Schipper, A.M., Busana, M. & Huijbregts, M.A.J. (2019). Intact but empty forests? Patterns of hunting-induced mammal defaunation in the tropics. PLOS Biol., 17, e3000247. Berenguer, E., Gardner, T.A., Ferreira, J., Aragão, L.E.O.C., Mac Nally, R., Thomson, J.R., Vieira, I.C.G. & Barlow, J. (2018). Seeing the woods through the saplings: Using wood density to assess the recovery of human-modified Amazonian forests. J. Ecol., 106, 2190–2203. Brodie, J.F. (2018). Carbon costs and bushmeat benefits of hunting in tropical forests. Ecol. Econ., 152, 22–26. Cayuela, L., Granzow-de la Cerda, Í., Albuquerque, F.S. & Golicher, D.J. (2012). Taxonstand: An r package for species names standardisation in vegetation databases. Methods Ecol. Evol., 3, 1078–1083. Chanthorn, W., Hartig, F., Brockelman, W.Y., Srisang, W., Nathalang, A. & Santon, J. (2019). Defaunation of large-bodied frugivores reduces carbon storage in a tropical forest of Southeast Asia. Sci. Rep., 9, 1–9. 119 Chave, J. (2002). Medição da densidade da madeira em árvores tropicais- Manual De Campo. Proj. PAN-AMAZONIA, Sixth Framew. Program. Chave, J. (2006). Measuring Tree Height for Tropical Forest Trees - A Field Manual. Chave, J., Andalo, C., Brown, S., Cairns, M.A., Chambers, J.Q., Eamus, D., Fölster, H., Fromard, F., Higuchi, N., Kira, T., Lescure, J.P., Nelson, B.W., Ogawa, H., Puig, H., Riéra, B. & Yamakura, T. (2005). Tree allometry and improved estimation of carbon stocks and balance in tropical forests. Oecologia, 145, 87–99. Chave, J., Coomes, D., Jansen, S., Lewis, S.L., Swenson, N.G. & Zanne, A.E. (2009). Towards a worldwide wood economics spectrum. Ecol. Lett., 12, 351–366. Chave, J., Muller-Landau, H.C.., Baker, T., Easdale, T., Steege, H.T. & Webb, C.O. (2006). Regional and phylogenetic variation of wood density across 2456 neotropical tree species. Ecol. Appl., 116, 2356–2367. Chave, J., Réjou-Méchain, M., Búrquez, A., Chidumayo, E., Colgan, M.S., Delitti, W.B.C., Duque, A., Eid, T., Fearnside, P.M., Goodman, R.C., Henry, M., Martínez-Yrízar, A., Mugasha, W.A., Muller-Landau, H.C., Mencuccini, M., Nelson, B.W., Ngomanda, A., Nogueira, E.M., Ortiz-Malavassi, E., Pélissier, R., Ploton, P., Ryan, C.M., Saldarriaga, J.G. & Vieilledent, G. (2014). Improved allometric models to estimate the aboveground biomass of tropical trees. Glob. Chang. Biol., 20, 3177–3190. Dirzo, R., Young, H.S., Galetti, M., Ceballos, G., Isaac, N.J.B. & Collen, B. (2014). Defaunation in the Antrhopocene. Science (80)., 401, 401–406. Fearnside, P. & Laurance, W.F. (2004). Tropical deforestation and greenhouse gas emissions. Ecol. Appl., 14, 982–986. Hawes, J.E. & Peres, C.A. (2016). Patterns of plant phenology in Amazonian seasonally flooded and unflooded forests. Biotropica, 48, 465–475. Hawes, J.E., Vieira, I.C.G., Magnago, L.F.S., Berenguer, E., Ferreira, J., Aragão, L.E.O.C., Cardoso, A., Lees, A.C., Lennox, G.D., Tobias, J.A., Waldron, A. & Barlow, J. (2020). A large-scale assessment of plant dispersal mode and seed traits across human-modified Amazonian forests. J. Ecol., 1–13. IBGE. (2018). Instituto Brasileiro de Geografia e Estatística [WWW Document]. Censo 2018. URL https://www.ibge.gov.br/ Jordano, P., Garcia, C., Godoy, J.A. & García-Castaño, J.L. (2007). Differential contribution of frugivores to complex seed dispersal patterns. Proc. Natl. Acad. Sci. U. S. A., 104, 3278–3282. Malhi, Y., Roberts, J.T., Betts, R.A., Killeen, T.J., Li, W. & Nobre, C.A. (2008). Climate change, deforestation, and the fate of the Amazon. Science (80-. )., 319, 169–172. Muller-Landau, H.C. (2004). lnterspecific and Inter-site Variation in Wood Specific Gravity of Tropical Trees. Biotropica, 36, 20–32. 120 Nichols, E., Uriarte, M., Peres, C.A., Louzada, J., Braga, R.F., Schiffler, G., Endo, W. & Spector, S.H. (2013). Human-Induced Trophic Cascades along the Fecal Detritus Pathway. PLoS One, 8. Nogueira, E.M., Yanai, A.M., Fonseca, F.O.R. & Fearnside, P.M. (2015). Carbon stock loss from deforestation through 2013 in Brazilian Amazonia. Glob. Chang. Biol., 21, 1271–1292. Nogueira, E.M., Yanai, A.M., de Vasconcelos, S.S., de Alencastro Graça, P.M.L. & Fearnside, P.M. (2017). Carbon stocks and losses to deforestation in protected areas in Brazilian Amazonia. Reg. Environ. Chang., 18, 261–270. Oksanen, J., Kindt, R., Legendre, P., O’Hara, B., Simpson, G.L., Solymos, P.M., Stevens, M.H.H. & & Wagner, H. (2008). The vegan package. Community Ecol. Package., 190. Osuri, A.M., Ratnam, J., Varma, V., Alvarez-Loayza, P., Astaiza, J.H., Bradford, M., Fletcher, C., Ndoundou-Hockemba, M., Jansen, P.A., Kenfack, D., Marshall, A.R., Ramesh, B.R., Rovero, F. & Sankaran, M. (2016). Contrasting effects of defaunation on aboveground carbon storage across the global tropics. Nat. Commun., 7. de Paula Mateus, D., Groeneveld, J., Fischer, R., Taubert, F., Martins, V.F. & Huth, A. (2018). Defaunation impacts on seed survival and its effect on the biomass of future tropical forests. Oikos, 127, 1526–1538. Peres, C.A., Emilio, T., Schietti, J., Desmoulière, S.J.M. & Levi, T. (2016). Dispersal limitation induces long-term biomass collapse in overhunted Amazonian forests. Proc. Natl. Acad. Sci., 113, 892–897. Peres, C.A. & Palacios, E. (2007). Basin-wide effects of game harvest on vertebrate population densities in Amazonian forests: implications for animal-mediated seed dispersal. Biotropica, 39, 304–315. Pires, M.M.., Guimaraes Jr., P.R.., Galetti, M. & Jordano, P. (2017). Pleistocene megafaunal extinctions and the functional loss of long-distance seed-dispersal services. Ecography (Cop.)., 41. Poelhekke, S. (2019). How expensive should CO2 be? Fuel for the political debate on optimal climate policy. Heliyon, 5, e02936. Poorter, L. & Bongers, F. (2006). Leaf traits are good predictors of plant performance across 53 rain forest species. Ecology, 87, 1733–1743. Poulsen, J.R., Clark, C.J. & Palmer, T.M. (2013). Ecological erosion of an Afrotropical forest and potential consequences for tree recruitment and forest biomass. Biol. Conserv., 163, 122–130. Réjou-Méchain, M., Tanguy, A., Piponiot, C., Chave, J. & Hérault, B. (2017). BIOMASS an R package for estimating above-ground biomass and its uncertainty in tropical forests. Methods Ecol. Evol., 1163–1167. 121 Rennó, C.D., Nobre, A.D., Cuartas, L.A., Soares, J.V., Hodnett, G.M., Tomasella, J. & Waterloo, M.J. (2008). HAND, a new terrain descriptor using SRTM-DEM : Mapping terra- fi rme rainforest environments in Amazonia. Remote Sens. Environ., 112, 3469–3481. Sullivan, M.J.P. et al. (2018). Field methods for sampling tree height for tropical forest biomass estimation. Methods Ecol. Evol., 2018, 1–11. Wright, S.J., Muller-Landau, H.C., Condit, R. & Hubbell, S.P. (2003). Gap-dependent recruitment, realized vital rates, and size distributions of tropical trees. Ecology, 84, 3174–3185. Zanne, A.E.., Coomes, D., Jansen, S., Lewis, S.L.., Swenson, N.G.. & Chave, J. (2009). Towards a worldwide wood economics spectrum. Ecol. Lett. Zuur, A.F., Ieno, E.N. & Elphick, C.S. (2010). A protocol for data exploration to avoid common statistical problems. Methods Ecol. Evol., 1, 3–14. 122 Supplementary information Figure S1: Map of the study area in the Médio Juruá region of western Brazilian Amazonia. The main Juruá River channel is outlined in white. The reserve boundaries of RESEX Médio Juruá and RDS Uacari are outlined in black. Coloured circles indicate the location of our 30 tree plots. Circles are colour-coded according to our proxy of hunting pressure (see colour gradient). The map background represents elevation, where brownish and green shades indicate low and high elevation, respectively. 123 Figure S2. Pearson correlation between our metric of hunting pressure on soil fertility – cation exchange capacity (CEC). Colour-coded circles represent the hunting pressure gradient, where red and blue represent higher and lower values, respectively. Figure S3. Model comparison showing the best fits for field measurements of tree DBH data to estimate tree heights. Table S1. Results from linear model analysing the effect of hunting pressure, vertical distance to the nearest drainage (VDND) and soil-nutrient (CEC) in the Aboveground Biomass (AGB) in thirty tree plot established a long to a hunting gradient in médio Juruá region variables estimate standard T- value p-value R² error intercept 4.409 0.038 114.089 -0.05 AGB VDND 0.070 0.796 0.880 0.387 CEC -0.080 0.084 -0.957 0.347 Hunting -0.024 0.083 -0.288 0.776 124 CONCLUSÃO GERAL Esta tese, trouxe para a Amazônia Brasileira a temática da síndrome das florestas vazias, que até então só havia sido estudada em nível nacional para biomas bastante defaunados como a Mata Atlântica. Apesar da Amazônia ainda ser a maior floresta tropical do mundo e estar aparentemente conservada aos “olhares” dos satélites, ameaças invisíveis podem comprometer a conservação à longo prazo de sua diversidade e dos serviços ecossistêmicos associados. No primeiro capítulo mostramos que florestas peri- urbanas que foram mais intensamente caçadas na região do Médio Juruá apresentam alterações na estrutura de tamanho da comunidade de vertebrados terrestres e arborícolas relacionado ao declínio na biomassa de vertebrados de grande porte e uma sub- compensação por roedores noturnos. Além disso, a pressão de caça parece estar promovendo um controle top-down nessas comunidades, já que os modelos que utilizamos para explicar a biomassa dos diversos grupos funcionais de vertebrados tanto terrestres quanto arborícolas indicaram como melhor preditor a pressão de caça, ao invés de variáveis ambientais relacionadas à cobertura vegetal, fertilidade do solo, proximidade da várzea e período de amostragem. No segundo capítulo, demonstramos que as florestas sobre um regime de caça mais intenso apresentaram modestas alterações direcionais na composição funcional das árvores no que se refere ao modo de dispersão, de forma que, espécies arbóreas dispersas por vertebrados de grande porte diminuíram suas abundâncias de indivíduos juvenis enquanto que espécies dispersas abioticamente aumentaram suas contribuições nessas florestas. Contudo esses efeitos não foram evidenciados no padrão funcional da comunidade para os traços de massa da semente, densidade de madeira e massa foliar por área. Embora para a densidade de madeira conseguimos reconhecer tendência diferentes no valor médio desse traço entre árvores e arvoretas ao longo do gradiente de caça estudado. Essas diferenças do traço funcional de densidade de madeira para árvores juvenis e adultas tiveram um reflexo direto nas estimativas de carbono apresentadas no terceiro capítulo. No terceiro capítulo, mostramos que áreas persistentemente caçadas apesar de possuírem uma maior abundância de árvores, estas são menores em diâmetro e altura e mesmo com os altos valores de densidade de madeira, a biomassa acima do solo estimada para os estoques atuais foi bastante similar entre todos os sítios estudados. Por outro lado, 125 as estimativas de carbono para o futuro, demonstram que pode haver uma perda de carbono em 22 dos 30 sítios estudados, sendo o balanço final entre possíveis perdas e ganhos o de uma perda média de 4.4 Mg/ha. Apesar das estimativas de perda de carbono não terem sido exclusivas das áreas mais intensamente caçadas, as evidências de que a diferença na densidade de madeira entre árvores adultas e juvenis se intensificaram nessas áreas já fornece um forte indicativo de que essas florestas poderão perder carbono no futuro. Considerando que o estudo foi realizando em duas unidades de conservação de uso sustentável e áreas urbanas adjacentes acreditamos que os resultados apresentados podem tanto alertar para os impactos da sobre-caça sobre a fauna, flora e serviços ecossistêmicos associados, quanto fornecer embasamento teórico para iniciativa de manejo de caça nas unidades de conservação. Esse estudo alerta, portanto, que os modelos de manejo da fauna devem ir além da conservação de populações viáveis das espécies cinegéticas, mas precisam garantir um tamanho populacional ideal que mantenha as interações das quais fazem parte e que são extremamente importantes na dinâmica florestal e provisão de serviços ecossistêmicos. 126 Artigo de Divulgação Científica enviado para a Revista Ciência Hoje Crianças As Florestas Vazias Imagine que você está em uma expedição na Amazônia. Você está remando em uma canoa em meio à copa das árvores de florestas inundadas conhecidas como igapó. Encontra então um lugar para encostar sua canoa e começa a caminhar pela floresta de terra firme, floresta não coberta pela água. Em sua caminhada se depara com árvores gigantescas de mais de 20 metros de altura e ouve o canto dos passarinhos e araras. De repente você escuta uma barulheira imensa e descobre que é um grupo de macacos se movimentando nas copas das árvores comendo e derrubando frutos e sementes por todo lado. Quando você menos espera, os macacos já sumiram em busca de novas árvores com frutos. Um pouco mais adiante você avista um tucano em cima de uma palmeira saboreando alguns pequenos frutos redondos de cor roxo escuro e percebe que se trata da famosa palmeira de açaí, um fruto amazônico bastante apreciado em todo mundo. Conseguiu imaginar como seria essa visita a maior floresta tropical do mundo? Agora imagina como seria essa expedição se a floresta estivesse vazia. O que aconteceria se esses animais incríveis estivessem desaparecendo? Sem dúvida a floresta se tornaria cada vez mais silenciosa e perderia muitos de seus encantos, não é mesmo? Mas, na verdade, o sumiço desses animais causaria muito mais problemas para a floresta do que você pode imaginar. Essa floresta no futuro poderia ficar muito diferente do que conhecemos hoje, sem tantas dessas árvores grandiosas que você encontrou durante a caminhada pela floresta de terra firme. Mas o que o sumiço dos animais tem a ver com aquelas grandes árvores? Tem tudo a ver! Todos os animais desempenham um importante papel para que a floresta permaneça saudável e continue mantendo sua biodiversidade e fornecendo bens importantes para nossa sobrevivência como o ar puro, água limpa, alimentos e até mesmo as chuvas! Alguns animais como os grandes macacos, anta e os tucanos são como os “jardineiros” da floresta, por que eles comem diversos frutos, se movimentam por longas distâncias e quando fazem cocô ou quando estão comendo e deixam uma semente cair no solo é como se eles estivessem ajudando a espalhar as sementes, que justamente por serem levadas à lugares mais distantes, têm maiores chances de sobreviver e crescer como uma nova árvore. Passarinhos e outros pequenos macacos também ajudam nessa tarefa, mas, por serem menores, eles não conseguem carregar e comer os mesmos frutos grandes que um macaco barrigudo ou uma anta comem. Geralmente são essas grandes sementes que, ao germinarem, irão crescer e se transformar naquelas imensas árvores com mais de 20 metros de altura. Enquanto a maioria das sementes menores, levadas por animais menores, como os passarinhos, irão gerar árvores também menores. Mas e os frutos que caem das árvores diretamente no chão da floresta também não irão germinar? Algumas podem até germinar e virar grandes árvores, mas o que acontece é que a chance disso acontecer é bem menor por vários motivos. Um dos motivos é que essas sementes podem ser devoradas pelos animais que 127 conseguem encontrar mais facilmente os frutos e sementes que estão em grande quantidade embaixo dessas árvores. Outro motivo é que mesmo que a semente escape desses bichos famintos e germine a árvore mãe cobrirá boa parte da luz com sua copa e por isso a pequena árvore terá grande dificuldade de receber a luz que precisa para crescer. E não para por aí! Algumas espécies de plantas precisam que suas sementes passem dentro do estômago de animais para que possam germinar. Ou seja, para atingirem aqueles tamanhos imensos, as sementinhas não tiveram uma vida fácil! Por isso, quando os animais levam as sementes para longe da árvore mãe, eles aumentam muito as chances de sobrevivência dessas novas árvores. Cientistas em vários lugares do mundo tem pesquisado justamente o que pode acontecer com as florestas se os grandes animais como macacos, antas e tucanos sumissem. Mas, para que os cientistas estão pesquisando como seria a floresta sem os animais se os animais ainda estão lá? É… na verdade, em muitas florestas tropicais no mundo, os animais estão de fato desaparecendo, principalmente por causa da caça e da perda de seus habitats. Assim, quando os pesquisadores estudaram algumas florestas em que esses animais não estão mais presentes, perceberam uma menor quantidade de árvores grandes dessas que dependem dos animais para carregá-las longe da árvore mãe. No Brasil, cientistas têm estudado os animais e as plantas da Mata Atlântica e da Amazônia para entender as mudanças que podem ocorrer quando essas florestas se tornam vazias. Como já é bastante conhecido por todos nós, resta muito pouco da Mata Atlântica. E as poucas áreas de Mata Atlântica que ainda existem têm bem menos animais do que costumavam ter. Mas e a Amazônia? Ela é a maior floresta tropical do mundo! Tem muitos animais e plantas que ainda nem foram descobertos pela ciência! Os animais de lá também podem estar desaparecendo? As árvores grandes podem ficar cada vez mais difíceis de serem encontradas? São essas perguntas que os pesquisadores estão tentando responder. Quer saber mais sobre as florestas vazias e sobre outros projetos do Instituto Juruá acesse nossa página na internet: www.institutojurua.org.br Andressa Bárbara Scabin Pós-Graduação em Ecologia - Universidade Federal do Rio Grande do Norte Diretora de Comunicação do Instituto Juruá 128