• Nenhum resultado encontrado

Influence of habitat edges on spatial and temporal occurrence patterns of mesocarnivores in a Eucalyptus dominated landscape

N/A
N/A
Protected

Academic year: 2023

Share "Influence of habitat edges on spatial and temporal occurrence patterns of mesocarnivores in a Eucalyptus dominated landscape"

Copied!
91
0
0

Texto

(1)

I

UNIVERSIDADE DE LISBOA FACULDADE DE CIÊNCIAS

DEPARTAMENTO DE BIOLOGIA ANIMAL

Influence of habitat edges on spatial and temporal occurrence patterns of mesocarnivores in a

Eucalyptus dominated landscape

Rita de Almeida Videira Pereira

Mestrado em Biolgia da Conservação

Dissertação orientada por:

Professor Doutor Luís Miguel Rosalino Professora Doutora Margarida Santos-Reis

2022

(2)

II

Index

Agradecimentos ………...V Resumo ………...VI Abstract ………..IX

1. Introduction ……….……….1

2. Study Area ………...7

3. Methods ………..……….9

3.1 Sampling design and field data collection ………..………...9

3.2 Data handling and analysis ………..13

3.2.1 Camera-trap data extraction ………...13

3.2.2 Temporal patterns ………..13

3.2.3 Spatial patterns ………...14

3.2.4 Spatio-temporal analysis ……….16

4. Results ………17

4.1 Camera-trap capture success ……….………17

4.2 Temporal patterns ……….18

4.3 Spatial patterns ……….24

4.4 Spatio-temporal analysis ………...31

5. Discussion ………..33

5.1 Use of time ………33

5.2 Space use …...………...35

5.3 Spatio-temporal analysis …..……….37

6. Conclusions ………38

7. References ………..39

8. Appendices ……….48

(3)

III

List of figures and tables Figures

Figure 2.1 - Land covers and location within Portugal of the Caniceira farmstead ………...7

Figure 2.2 - Land covers and location within Portugal of the Zambujo farmstead ………8

Figure 2.3 – Example of a habitat edge in Caniceira farmstead ………...……8

Figure 2.4 – Example of a habitat edge in Zambujo farmstead………..………8

Figure 2.5 - Location of the camera-trap stations within the different land covers in the Caniceira farmstead ………..………9

Figure 2.6 - Location of the camera-trap stations within the different land covers in the Zambujo farmstead ………10

Figure 2.7 - Example of a camera-trap attached to a tree trunk (Eucalyptus globulus) ………10

Figure 2.8 - Example of a Sherman trap, set in Zambujo to sample small mammals, showing the used cotton and bait ………11

Figure 2.9 - Pitfall trap, set in Zambujo to sample terrestrial invertebrates ……….11

Figure 2.10 - Location of the Sherman trap sampling lines within the different land covers of the Caniceira farmstead ………12

Figure 2.11 - Location of the Sherman trap sampling lines within the different land covers of the Zambujo farmstead ……….……12

Figure 4.1 - Photos captured by camera-trap during our survey: Top left, red fox captured in Zambujo; top middle, common genet captured in Zambujo; top right, stone marten captured in Zambujo; bottom left, European badger captured in Caniceira and the bottom right, small mammals captured in Caniceira ………....…18

Figure 4.2 - Mesocarnivore and small mammals’ diel activity patterns, represented by a kernel density, in habitat edges (black continuous line) and interior (black dotted line). The blue and red dotted vertical lines indicate the beginning and end of the sunrise and sunset respectively and the shaded area corresponds to the activity overlap. A – Red fox, B – Common genet, C – Stone marten, D – European badger and E – Small mammals ………..19

Figure 4.3 - Mesocarnivores´ diel activity patterns and overlap, represented by a kernel density, in edge and interior habitats. The blue and red dotted vertical lines indicate the beginning and end of the sunrise and sunset respectively and the shaded area corresponds to the activity overlap between both species. The graphics A to F correspond to activity patterns in the edges and the graphics G to L in the habitat interiors .……….……… 22

Figure 4.4 - Diel activity patterns and overlap, represented by a kernel density, between mesocarnivore species (black continuous line) and small mammals (black dotted line). The blue and red dotted vertical lines indicate the beginning and end of the sunrise and sunset respectively and the shaded area corresponds to the activity overlap. The graphics A to D correspond to activity patterns in the edge habitats and the graphics E to H in the habitat interiors. ………...………...23

(4)

IV Figure 4.5 - Coefficients and respective confidence interval of 95% estimated for each variable included in the average model that best explained the red fox relative abundance variation. 95% confidence intervals represented in red indicate a negative coefficient and in blue a positive one. A – Global model;

B – Caniceira model; C – Zambujo model ………..27 Figure 4.6 - Coefficients and respective confidence interval of 95% estimated for each variable included in the average model that best explained the common genet relative abundance variation. 95%

confidence intervals represented in red indicate a negative coefficient and in blue a positive one. A – Global model; B – Caniceira model; C – Zambujo model ………...28 Figure 4.7 - Coefficients and respective confidence interval of 95% estimated for each variable included in the average model that best explained the stone marten relative abundance variation. 95% confidence intervals represented in red indicate a negative coefficient and in blue a positive one. A – Global model;

B – Caniceira model; C – Zambujo model ………...29 Figure 4.8 - Coefficients and respective confidence interval of 95% estimated for each variable included in the average model that best explained the European badger relative abundance variation. 95%

confidence intervals represented in red indicate a negative coefficient and in blue a positive one. J – Global model; K – Caniceira model; L – Zambujo model ………...………30 Figure 4.9 - Expected times-to-encounter randomly generated from a multiresponse permutation procedure for each mesocarnivore species pair for the edge (A to F) and habitat interior (G to L). The vertical line represents the median observed time-to-encounter between the two species

………....31

Tables

Table 1.1 - Hypotheses tested in this study (H) with the corresponding rationale and predictions

………...………...4 Table 4.1 – Number of independent records for each species detected during the survey period at each habitat and edge type ………..17 Table 4.2 - Species overlap coefficient (Δ1) with their respective 95% confidence intervals (CI 95%) and the Mardia-Watson-Wheeler’s test p-value (M-W p-value) …………..………...20 Table 4.3 - Mesocarnivores and small mammals’ activity overlap coefficient (Δ1) with their respective 95% confidence intervals (CI 95%) and the Mardia-Watson-Wheele’s test p-value (M-W p-value) for edges and habitat interiors. Statistical differences are highlighted in bold …...………...21 Table 4.4 - Best models for each species for the general dataset and for Caniceira and Zambujo farmsteads. For each model we present the degrees of freedom (df), Akaike’s Information Criterion for small samples (AICc), Akaike weight (weight) and Likelihood-Ratio Test (LogLik) ....………24 Table 4.5 - Best models for each species for the general dataset and for Caniceira and Zambujo farmsteads. For each model we present the degrees of freedom (df), Akaike’s Information Criterion for small samples (AICc), Akaike weight (weight) and Likelihood-Ratio Test (LogLik) ....………25 Table 4.6 - Median observed time-to-encounter (days) for each species pair and the p-values of the test that indicates if the proportion of randomly generated times-to-encounter greater than the observed.

Values for each species pair and for edge and habitat interior (Statistical differences are highlighted in bold) ………...32

(5)

V

Agradecimentos

À The Navigator Company, S.A. pelo financiamento deste trabalho, integrado no projeto “Avaliação dos valores de biodiversidade em propriedades da The Navigator Company” e por todo o apoio logístico para a realização do trabalho de campo, e ao CE3C - Centro de Ecologia, Evolução e Alterações Ambientais pelo suporte prestado através do projeto estratégico UID/BIA700329/2019.

Ao Professor Luís Miguel Rosalino por me ter proporcionado esta oportunidade e pela incansável ajuda e orientação e por estar sempre disponível para responder às minha dúvidas e ajudar-me a encontrar as melhores soluções para os problemas que iam surgindo.

Á Professora Margarida Santos-Reis pela ajuda na escrita desta tese e por todos os seus conselhos que fizeram toda a diferença no trabalho final.

Ao Matias um agradecimento especial por me acompanhar no campo e me ensinar a montar uma camara como deve de ser. Também por estar sempre disponível para me ajudar e ensinar tudo sobre a parte estatística e de tratamento de dados. Sem a tua ajuda a minha aventura pelo R teria sido muito mais complicada e demorada.

À Beatriz pela ajuda no campo e em especial na captura dos pequenos mamíferos.

Ao Vasco que foi a primeira pessoa a ler a minha tese e por ter sempre uma palavra amiga e de motivação.

Por último mas não menos importante, à minha família por me ter apoiado sempre nas minhas decisões e por acreditarem em mim. Aos meus amigos por estarem sempre prontos para me distrair quando era necessário e por ouvirem os meus desabafos. Em especial, um agradecimento ao Bruno Miguel por me aturar nos momentos de dúvida e de sobrecarga.

(6)

VI

Resumo

O crescimento populacional humano leva a rápidas e drásticas mudança na estrutura da paisagem, frequentemente como consequência da desflorestação e fragmentação. A fragmentação modifica a paisagem e divide-a em parcelas de habitats mais pequenas e isoladas, criando cada vez mais zonas de transição entre a matriz paisagística e as manchas de outros habitats, designadas por orlas. As orlas de habitat podem afetar a distribuição e diversidade das espécies, a abundância local das populações e os movimentos individuais, e, consequentemente, influenciar as interações entre as espécies, induzindo efeitos em cascata no seio das comunidades. As orlas tanto podem constituir barreiras ao movimento de certas espécies, ser utilizadas como corredores durante as deslocações dos indivíduos. Podem também expor os organismos a diferentes condições abióticas e bióticas, promovendo a interação entre espécies que de outra maneira não ocorreriam. A abundância de certas espécies pode aumentar perto das orlas se estas as utilizarem como corredores para se deslocarem para habitats mais favoráveis ou se os dois habitats adjacentes providenciarem recursos diferentes. Os efeitos negativos das orlas são muitas vezes associados a espécies especialistas (no uso de determinados habitats ou presas), que tendem a evitar as orlas. Assim, e tendo em conta o cada vez mais rápido aumento destas estruturas paisagísticas, é importante e urgente compreender os seus efeitos, nomeadamente para espécies de mesocarnívoros; não só pela falta de informação que existe em relação a este grupo no que diz respeito às suas respostas às orlas de habitat, como pelos importantes papéis funcionais que desempenham nos ecossistemas.

Para abordar esta temática, este estudo foi realizado em duas herdades da Navigator Company, S.A. - Caniceira e Zambujo - cujas paisagens são dominadas por plantações de eucalipto, mas que também apresentam manchas de vegetação natural e outras plantações, que contribuem para uma paisagem bastante heterogénea e rica em orlas de habitat.

Neste estudo, comparando as orlas com o interior dos habitats, pretendemos estudar se, e como, as orlas influenciam a comunidade de mesocarnívoros a três níveis distintos: i) Temporalmente, estudando os padrões de atividade de cada espécie e o nível de sobreposição temporal entre cada par de espécies, ii) Espacialmente, recorrendo a modelos GLM para investigar se as características das orlas e outras variáveis ou fatores ambientais podem estar a promover eventuais alterações na abundância relativa das espécies, e iii) Espácio-temporalmente, avaliando o intervalo de tempo entre encontros de cada par de espécies presentes em ambas as zonas.

Para tal, 18 dispositivos de foto-armadilhagem (câmaras fotográficas) foram distribuídos por cada herdade, metade das quais instaladas em orlas de habitat, definidas como a linha ou área, identificável no campo, que separa dois habitats adjacentes (eucaliptal – linha de água, eucaliptal-montado, eucaliptal-pinhal, montado-linha de água) e a outra metade no interior do habitat matriz (eucaliptal, montado, linha de água), separadas por, pelo menos, 500 metros. As armadilhas fotográficas estiveram ativas 24 horas por dia, sem isco, entre Julho de 2020 e Janeiro de 2021, de modo a monitorizar a comunidade de mesocarnívoros, com foco nas seguintes espécies: raposa (Vulpes vulpes), fuinha (Martes foina), texugo europeu (Meles meles) e geneta (Genetta genetta). Num buffer de 200 metros em torno de cada armadilha fotográfica, foram recolhidos dados ambientais divididos em 7 categorias:

características da orla, fontes de alimento, fonte de água, orografia, estrutura do habitat, composição do cobertura do solo e perturbações antropogénicas. Para recolher informação acerca das fontes de alimento para estes predadores, como sejam os pequenos mamíferos e os invertebrados, foram colocadas, respetivamente, armadilhas Sherman com isco e armadilhas pitfall, ativas durante 3 noites na herdade da Caniceira e 4 noites na herdade do Zambujo, em cada tipo de orla e de interior de habitat.

No que diz respeito à dimensão temporal, os mesocarnívoros estudados, reconhecidamente com uma atividade maioritariamente noturna, mostram ter dois picos, um ao anoitecer e outro ao amanhecer, mais

(7)

VII pronunciados nas orlas de habitat do que no interior. Uma explicação plausível para este padrão poderá estar associada ao facto das orlas de habitat serem áreas mais abertas com menos vegetação herbácea e arbórea, logo com menos recursos e esconderijos. Assim, os mesocarnívoros utilizam sobretudo as orlas para se movimentarem entre os habitats onde repousam durante o dia e os habitats onde caçam durante a noite. Apresentam assim um pico ao anoitecer, quando vão em busca de alimento e outro pico ao anoitecer, quando voltam para os esconderijos onde passam o dia. Os pares de mesocarnívoros aparentam ter uma maior sobreposição no interior dos habitats demonstrando uma potencial agregação nestas zonas. Pensamos que tal pode acontecer pois o interior dos habitats apresentava, geralmente, uma maior percentagem de coberto arbóreo e esconderijos (i.e. buracos nos troncos das árvores e arbustos) que as orlas, que eram zonas abertas. Estas características permitem às espécies com capacidade arborícola alimentar-se e usar o mesmo espaço, e ao mesmo tempo, que as espécies de características não arborícolas, sem que se encontrem, evitando encontros agonísticos.

Espacialmente, não foram observadas diferenças significativas entre as abundâncias nas orlas e no interior dos habitats. As variáveis relacionadas com a orla e com a estrutura do habitat mostraram não ser fatores determinantes para explicar a abundância relativa dos mesocarnívoros. Em vez disso, a composição do habitat (tipo e percentagem de cobertura do solo presentes em redor das armadilhas fotográficas) é o que mais influencia a maior abundância relativa de mesocarnívoros. Isto sugere que, nos contextos paisagísticos estudados, os mesocarnívoros podem não percecionar as orlas como uma barreira espacial ou como um tipo habitat diferente, mas sim, podem estar a utilizá-las apenas como um corredor para se movimentarem entre diferentes tipos de habitat, possivelmente por se tratar de espécies generalistas, o que se reflete no efeito neutro das orlas.

A nível espácio-temporal, a comunidade de mesocarnívoros apresentou padrões de agregação tanto nas orlas como no interior dos habitats. A única exceção foi o par raposa e geneta, que mostraram evitar-se nas orlas de habitat. Acreditamos que esta segregação se deva ao facto de estas duas espécies partilharem nichos tróficos semelhantes. As orlas de habitat constituem uma unidade paisagística de pequena escala, com menos espaço disponível, sendo também uma área mais aberta pela menor cobertura arbustiva.

Assim, a probabilidade de encontros entre as duas espécies, e consequentes confrontos, é consideravelmente maior. Assim sendo, esta segregação pode ser uma estratégia para promover a coexistência entre os dois mesocarnívoros.

Este estudo contribuiu para demonstrar que os efeitos de orla nos mesocarnívoros ibéricos são específicos da espécie e do contexto paisagístico. No entanto, em geral, mostrou que os mesocarnívoros estudados se agregam, e têm uma maior sobreposição temporal, no interior dos habitats quando comparados com as orlas de habitats. No interior dos habitats também mostram picos de atividade menos pronunciados, que muitas vezes ocorreram mais tarde do que nas orlas. Também conseguimos detetar que o tipo de habitat (tipo e percentagem de cobertura do solo presentes em redor das armadilhas fotográficas) é mais importante do que as próprias orlas para explicar a variação na abundância relativa dos mesocarnívoros. Este conhecimento contribui para otimizar as medidas de gestão em ecossistemas alterados pelo ser humano, como áreas agrícolas e de plantações florestais, onde são criadas diferentes orlas, para garantir a conservação da biodiversidade, mantendo a rendibilidade económica. À medida que os habitats se tornam mais fragmentados, o número de orlas aumentam e estudos como este tornam- se mais importantes. Pelo que sabemos, este é o primeiro estudo direcionado especificamente aos efeitos de orla em mesocarnívoros na Península Ibérica. Estudos futuros beneficiariam de uma amostragem a longo prazo incluindo várias estações do ano e uma avaliação simultânea da disponibilidade e consumo dos diferentes recursos alimentares em cada estação do ano e tipo de habitat. Também é necessária uma análise mais extensa sobre a ocupação do espaço, recorrendo, por exemplo a técnicas de geoposicionamento (colares com GPS), para uma compreensão, a escala fina e detalhada, de como os

(8)

VIII mesocarnívoros utilizam as diferentes unidades de paisagem, incluindo os diferentes tipos de orlas e interiores de habitats matriz.

Palavras-chave: Carnivora; efeito de orla; contexto mediterrânico; fotoarmadilhagem; uso do espaço e do tempo.

(9)

IX

Abstract

Human population growth leads to drastic changes in landscape structure that often results in fragmentation. Fragmentation modifies the landscape and divides it into smaller habitat patches, creating habitat edges. These can affect the distribution, diversity, abundance and movement of species and, consequently, influence interspecific interactions, inducing cascading effects into all communities. This study was carried out in two farmsteads, whose landscapes are dominated by Eucalyptus plantations but which also present a great fragmentation of natural habitats with patches of different land covers, contributing to a very heterogeneous landscape and rich in habitat edges.

We placed18 camera-trapping devices in each farmstead, half on habitat edges and half in the habitat interiors to monitor the mesocarnivore community with a focus on the red fox (Vulpes vulpes), stone marten (Martes foina), European badger (Meles meles) and common genet (Genetta genetta). In a 200m around each camera, we collected environmental data divided into 7 categories: edge characteristics, food sources, water source, orography, habitat structure, land cover composition, and anthropogenic disturbances. To collect information about food sources, such as small mammals and invertebrates, Sherman and pitfall traps were placed, respectively, in each type of edge and habitat interior in a total of 6 sites in each farmstead.

By comparing edges with habitat interiors, we intend to study how or if habitat edges influence mesocarnivores at three levels: i) Temporally, studying the change in activity and overlapping patterns, ii) Spatially, using GLMs to investigate which variables may be influencing the relative abundance of species at the edges and the interior habitats, and iii) Spatio-temporally, by investigating if there are changes in the interactions between the species’ pairs on habitat edges compared to habitat interiors.

This study contributed to demonstrate that edge effects on Iberian mesocarnivores are species and landscape context-specific. However, in general, it showed that mesocarnivores aggregate and have a greater overlap of activity within interior habitats than in edge habitats. In the habitats interior mesocarnivores also showed less pronounced peaks of activity, often occurring later than at the edges of the habitats. We were also able to detect that the type of habitat is more important than the edges themselves in explaining the observed variation in the relative abundance of mesocarnivores.

Knowledge gathered with this study contributes to the design of better management measures in human- altered ecosystems, such as agricultural areas and forestry plantations, where different borders are created, to guarantee the conservation of biodiversity while maintaining economic profitability.

Keywords: Carnivora; edge effects; Mediterranean context; camera-trapping; time and space use

(10)

1 1. Introduction

Landscape structure is essential to understand biodiversity and species occurrence and interaction patterns (Moreira-Arce et al., 2016; Peles et al., 1999; Pita et al., 2009). It corresponds to the variability of properties of a system in spatial terms (Walz, 2011) and includes landscape composition (land covers), spatial arrangement and dynamics (Fahrig and Merriam, 1994). This structure influences several important ecological processes, such as organism’s movement and the flow of materials and nutrients (Murcia, 1995; Walz, 2011). Animal species with good dispersal ability depend more on landscape composition and less on landscape arrangement. But even for these species, the distance between patches of preferred habitat type can have an important influence on the species’ movement and dispersal patterns (Walz, 2011). Furthermore, while certain species prefer heterogeneous environments, with more diverse patches and resources, other, more specialist, can thrive only on their optimal habitat (Červinka et al., 2011; Santos and Santos-Reis, 2010), being highly sensitive to the spatial arrangement of such habitats. Therefore, data on broader land composition alone is usually not enough to understand species’ distribution and occurrence patterns, and we need to look into other components of the landscape structure, such as intra-habitat structure (e.g., understory vegetation that provides shelter), habitat edge contrasts (i.e. to which habitat a patch is contiguous to), land covers spatial distribution and representativeness (Moreira-Arce et al., 2016; Pita et al., 2009; Walz, 2011), among other.

The increased influence of humans in the environment led to rapid and drastic changes in the landscapes structure, often resulting in fragmentation. Fragmentation modifies landscapes, mainly by dividing the habitats into patches, reducing their size, increasing patch isolation, and creating wider and more contrasting habitat edges (Ewers and Didham, 2005; Laurance and Yensen, 1991; Regolin et al., 2017;

Šálek et al., 2010; Svobodová et al., 2011). Habitat edges can be identifiable as discontinuities or boundaries separating two or more distinct adjacent patches with different biotic and abiotic characteristics (Lidicker, 1999.; Lidicker and Peterson, 1999). They can correspond to different plant communities, land uses, successional or development stages (Ries and Sisk, 2004; Yahner, 1988).

Habitat edges can either be an obvious and well-defined boundary (sharp, high-contrast), often separated by empty corridors, or a transition zone (soft, low-contrast) where the two habitats gradually change from one type to the other (Lidicker, 1999.; Lidicker and Peterson, 1999; Yahner, 1988). Soft edges tend to originate weaker responses than sharp habitat edges (Ries et al., 2004). It can also be natural or induced (Fagan et al., 1999; Yahner, 1988). Natural if it results of different abiotic factors, such as the climate or topography, and induced if it results from anthropogenic disturbances and modification of the naturally occurring patches (e.g. by implementing agriculture lands; Yahner, 1988). This said, habitat edges can be hard to define, but easily identifiable by field observation (Fagan et al., 1999) and can occur at different scales depending on how the target species uses de available habitats and explore the landscape. Thus, some species do not perceive habitat edges as we do (Lidicker, 1999.). Hence, they influence different species in distinct ways and can even influence individuals differently based, for example, on its age or gender (Fagan et al., 1999; Lidicker, 1999). First of all, habitat edges can expose organisms to different abiotic conditions, such as different temperatures, wind intensity and solar incidence, for example, by creating microclimatic differences in these areas (Murcia, 1995). As such, edges can alter species movement patterns (Cantrell et al., 2002; Fletcher, Jr. et al., 2007; Maciel and Lutscher, 2013). While they can act as a barrier to the movement of some species (Fagan et al., 1999;

Ries et al., 2004; Yahner, 1988), other take advantage of them and use them as travel corridors to move faster and farther away (Heske, 1995; Maciel and Lutscher, 2013; Šálek et al., 2010). Thus, edges can affect species’ distribution, diversity and abundance, which in turn will influence the interaction among species and this may induce cascading effects into the entire communities (Cantrell et al., 2002; Ewers and Didham, 2005; Peralta et al., 2017; Regolin et al., 2017; Ries et al., 2004). Habitat edges can act as

(11)

2 a unique habitat and facilitate the interaction between species that would not be possible otherwise (Fletcher, Jr. et al., 2007). All these processes are commonly referred to as “edge effects”.

Habitat edges can have a positive (increase in abundance on or near edges), negative (avoidance) or neutral effects on species (Heske, 1995; Wimp et al., 2019). There are three most common mechanisms to explain the increase in species abundance near edges. The first one is called spillover or mass effect.

This occurs when individuals cross to an adjacent, not preferred or lower quality, habitat, exclusively because of its proximity, resulting on a higher abundance near edges (Ries and Sisk, 2004; Svobodová et al., 2011; Wimp et al., 2019); usually, they do not go very deep into this lower quality habitat. The second one is called ecotonal effect or complementary resource distribution. This happens when one or more resources is available in one of the patches, but not in the adjacent one; however the second patch also has resources that cannot be found in the first one (Lidicker, 1999; Svobodová et al., 2011). So, the resources present in both patches complement each other. There is not one specific resource available at the edge itself, but the proximity and easier access to both resources results in higher species abundance along or near edges (Svobodová et al., 2011; Yahner, 1988); these resources can include vegetation cover for shelter and foraging opportunities, for example. Lastly, habitat edges can act as an enhanced habitat, when it has resources that are absent or are rare in both the adjacent patches, acting as a different habitat type, resulting in an increase in abundance in these zones (Ries and Sisk, 2004).

In what regards the negative effects of habitat edges, these are usually attributed to species that are interior-inhabiting, habitat-specific or specialist predators, who tend to avoid habitat edges (Lidicker, 1999; Ries and Sisk, 2004; Svobodová et al., 2011; Wimp et al., 2019). Even though these species avoid edges and lower quality habitats they still might spillover into adjacent patches. Finally, neutral edge effects happen when the adjacent patches have similar types and amount of resources (Wimp et al., 2019). In this case species densities are expected to also be similar in the habitat interior and in habitat edges.

The Mediterranean region is a very important biodiversity hotspot located roughly around the Mediterranean sea naming a unique type of temperate climate characterized by a dry and very warm summer and low moderate temperatures in winter when the precipitation is concentrated (García-Ruiz et al., 2013). This characteristic climate has a direct influence in the land cover’s structure and composition, and before the alterations induced by humans, this region’s vegetation was well adapted to the seasonality in soil humidity and naturally-induced forest fires (García-Ruiz et al., 2013). This region is also known for being a very complex and rich landscapes with mosaic vegetation, originated by abundant local variations in soils, topography and water availability (García-Ruiz et al., 2013).

However, many centuries of human activities and settlements have altered and transformed the Mediterranean landscape, and now, only about 4% of its original vegetation remains (Geri et al., 2010;

Marull et al., 2014). In more recent years, the economical and social changes have led to a growth in urban and industrial areas, an intensification in the monocultures agriculture, big and often human- induced reoccurring fires and expansion of livestock, (García-Ruiz et al., 2013; Geri et al., 2010), leading to deforestation, habitat fragmentation, reduction in patch sizes (and consequent increase in habitat edges), isolation and to a landscape matrix of different stages of degradation and land-covers.

For these reasons, the Mediterranean landscape is described as very heterogeneous, in space but also in time, providing different resources through the year, and where a multitude of edge types occur between native and man-made landcovers (Alexandre et al., 2020; Curveira-Santos et al., 2017).

Although the Mediterranean landscape is nowadays characterized by being very fragmented, and consequently, having several habitat edges, not many studies have been implemented to understand the edge effects on wildlife in this region. We can find a few studies targeting plants, such as flowers or trees, and lichens (Belinchón et al., 2007; Brunialti et al., 2012; Concepción et al., 2012; González-

(12)

3 Moreno et al., 2013; Marull et al., 2014; Rosati et al., 2010; Solé-Senan et al., 2014; Torras et al., 2008), invertebrates (David et al., 1999; Holway and Suarez, 2006; Peyras et al., 2013) and even birds and nest predation (Herrando and Brotons, 2002; Patten and Bolger, 2003; Santos and Tellería, 1992). However, studies assessing the edge effects on mammals inhabiting the Mediterranean region are scarce and often use small mammals as models (Rodríguez-Pastor et al., 2016; Torre and Díaz, 2004). Regarding Mediterranean carnivores, there seems to be a major knowledge gap.

Carnivores play a crucial role in ecosystems functioning, communities and dynamics (Alexandre et al., 2020; Bencatel et al., 2018; Kowalski et al., 2015; Monterroso et al., 2020). They contribute to prey regulation, energy transference within ecosystems and act, direct and indirectly, as ecosystem engineers (Bencatel et al., 2018; Kowalski et al., 2015). They frequently occur at low population densities and require large areas, which usually means that they have a relatively sparce abundance across their habitats (Rosalino et al., 2005.; Rosalino and Gheler-Costa, 2011). Mesocarnivores comprehend most of the carnivores in the Mediterranean region, and they are defined as small and mid-sized species weighing 15Kg or less (Roemer et al., 2009). These species are very different from each other in terms of ecology and overall behaviour, have a capability to succeed in diverse habitats, often living in proximity to humans (Roemer et al., 2009). When large carnivores are absent, mesocarnivores can act as apex predators and, for example, influence trophic organization (Roemer et al., 2009). This said, some important ecosystem services are provided by mesocarnivores (Bencatel et al., 2018; Cancio et al., 2016;

Herrera et al., 2016; Kowalski et al., 2015). They can act as top-down regulators, as they eat prey from different trophic levels, having a cascading effect on the entire trophic network, which can shape all the trophic levels below and alter the ecosystem composition (Bencatel et al., 2018; Prugh et al., 2009).

Mesocarnivores also act directly as vehicles for seed dispersal or indirectly by consuming other seed dispersal species (Cancio et al., 2016; Herrera et al., 2016).

As previously said, the human population has induced massive alteration in the landscapes, including fragmentation, drastic changes in land cover and understory removal. As mesocarnivores live in low populational densities (Kowalski et al., 2015), are often vulnerable to these habitat changes and persecution by humans that can easily lead to local extinctions or influence mesocarnivores’ spatial structure and distribution. These habitat changes can cause differences in foraging opportunities, induce anthropic barriers or shelter degradation, impacts that are often linked to habitat edges (Alexandre et al., 2020; Cancio et al., 2016). As such, we believe that mesocarnivores can be considered good models to study habitat edge effects in the Mediterranean region.

Although the listed negative impacts of edges on carnivores occur worldwide, some mesocarnivores still show a spatial preference for habitat edges over interior habitats, specially in mosaic landscapes, homogeneous habitats or open agricultural landscapes (Červinka et al., 2013; Oehler and Litvaitis, 1996;

Šálek et al., 2010), which tend to be not very suitable habitats for most carnivores. This affinity might be due to the fact that habitat edges might be used as travel routes (Šálek et al., 2010), either because it is easier to walk along open landscapes or because edges might act as physical barriers, making them walk along them instead of going through less suitable habitats (Šálek et al., 2010). Another explanation for the carnivores’ affinity to habitat edges is related to habitat specific distribution of resources (Svobodová et al., 2011). These resources might be potential prey such as avian nests or small mammals and other small prey, who are shown to have a higher diversity and abundance, or be more accessible, along habitat edges, (Brodie et al., 2015; Šálek et al., 2010). However, it is important to state that this affinity for habitat edges might vary depending on the season, since it has been shown that carnivores show a more obvious affinity for habitat edges in the winter then in the summer, due to variation in food sources availability (Oehler and Litvaitis, 1996). Furthermore, local habitat characteristics along the

(13)

4 edges (e.g. edge contrast, resource available in contiguous habitats, etc.; Villaseñor et al., 2015) is something that we need to consider when analysing habitat edge related patterns and effects.

Aiming to fulfil the knowledge gap on the edge effect on Mediterranean mesocarnivores, the main goal of this study is to know if and how habitat edges influence the spatial and temporal occurrence patterns of mesocarnivores in a Mediterranean context. With the increase in native habitat fragmentation in the Mediterranean region due to native landscape conversion into agriculture and silviculture areas, which alter the edge structure, it is important to understand edge impacts on wildlife and, namely, mesocarnivores, due to their important functional roles in the ecosystems. To address this goal, we analysed mesocarnivores’ ecological patterns in edges and interior habitats using three different dimensions: Spatial, temporal and spatio-temporal. To identify which factors may be determining the changes in the occurrence of mesocarnivores in edge habitats in contrast to interior habitats we tested seven hypotheses as shown in Table 1.1 based on distinct ecological driver types.

Table 1.1 – Hypotheses tested in this study (H) with the corresponding rationale and predictions.

Hypotheses and Rationale Hypotheses’ Prediction H1 – Food resources

Food availability is a basic need for species survival and it can influence mesocarnivores’ space use and abundance (Díaz-Ruiz et al., 2013).

For the target mesocarnivores’ species, the invertebrates, especially the order Coleoptera, are the main food source, followed by fruits and small mammals (Bakaloudis et al., 2012; Díaz-Ruiz et al., 2013; Hipólito et al., 2016b); but common genets consume mostly small mammals, with invertebrates and fruit assuming a complementary role (Rosalino and Santos-Reis, 2002).

Higher food availability increases the mesocarnivores’ relative abundance.

H2 – Water resources The presence of water courses and riparian vegetation will induce high biodiversity of mesocarnivores that use them as water sources, corridors, refuge or easier access to prey (Rosalino and Gheler-Costa, 2011).

Places such as these are known to be especially important in Mediterranean landscapes, where hot and dry summers are a constant (Matos et al., 2009).

The relative abundance of mesocarnivores is higher closer to water courses.

H3 – Anthropogenic disturbances Anthropogenic disturbances and human presence can

change mesocarnivores’ movement patterns, occurrence, and abundance, creating avoidance behaviours in places where humans or their activities are present (Alexandre et al., 2020; Benítez-López et al., 2010).

Sites with less anthropogenic disturbances have a higher abundance of mesocarnivores.

(14)

5 H4 – Orography

Terrain steepness influence human’s accessibility to the site, making it more difficult, and thus may influence wildlife abundance (Nisi et al., 2022). South-facing slopes can receive up to six times more solar radiation than north-facing slopes origination very different climatic conditions (Auslander et al., 2003). This said, south- facing slopes are warmer and drier, often predominantly composed of short-lived annual plants (Auslander et al., 2003). North-facing slopes are mainly composed of more developed trees and shrubs layers (Auslander et al., 2003).

These differences in composition can affect mesocarnivores’ occurrence (Pita et al., 2009).

North-facing and steeper slopes have a higher mesocarnivores’ relative abundance.

H5 – Habitat structure Habitat structure and composition affect greatly

mesocarnivores’ presence and abundance (Pita et al., 2009).

Rocky formations are especially important for the common genet whose latrines are often located on rock tips (Larivière and Calzada, 2001). Rocky formations are also important for the red fox and the European badger that may use dens located in rocky crevices (Macdonald and Barrett, 1993; (Revilla et al., 2000).

Tree cover is particularly important for the common genet and the stone marten, since their arboreal ability facilitates the use of hollow trees as resting places (Carvalho et al., 2015; Macdonald and Barrett, 1993). Genets are also considered forest specialists, and some populations of badgers can prefer Mediterranean scrubland as a refuge and feeding grounds (Carvalho et al., 2015; Larivière and Calzada, 2001). Herbaceous cover can be used by some mesocarnivores as feeding grounds (e.g. European badger), especially if such habitats are located nearby more closed environments (e.g. forests) (Pita et al., 2009;

Revilla et al., 2000).

Mesocarnivores’ relative abundance is higher in habitats with more rocky formations and higher shrub and herbaceous cover.

Genets and martens have higher relative abundance in habitats with higher tree cover.

(15)

6 H6 – Land cover composition

Eucalyptus plantations are usually associated with low biodiversity and mesocarnivores avoidance, because of their lack of understory vegetation and consequently reduced availability of food and shelter (Cruz et al., 2015).

The land cover composition (i.e. the distinct types of land covers present in a region) provide different environmental conditions and opportunities for mesocarnivores, which can determine their abundance pattern. Thus, it is important to understand which land cover promotes mesocarnivore abundance.

Areas with a higher percentage of more natural habitats that provide more resources have a higher mesocarnivores’ abundance.

H7 – Edge

Often edges can function as a different habitat type that results from the mixture of the characteristics of the two adjacent habitats (Fagan et al., 1999). Furthermore, edge effects result from the interaction between these. Two juxtaposed habitats experience a flow of nutrients, energy and species across their edge, and this can lead to a change in the habitat structure and prey composition, or to interspecific interactions in the different fragments near the edges (Murcia, 1995; Peyras et al., 2013; Ries and Sisk, 2004). These changes can alter mesocarnivores’

space use and abundance (Peyras et al., 2013).

The relative abundance of mesocarnivores in edges is higher than in habitat interiors.

(16)

7 2. Study Area

The study took place in central Portugal, in two different farmsteads, Caniceira and Zambujo, managed by The Navigator Company, S.A. with a forestry production objective.

Caniceira is located in the district of Santarém, Abrantes municipality, in the north of the Ribatejo region. Ribatejo is the region in Portugal where the transition between the Mediterranean and Atlantic climate is more evident and is crossed by the Tagus river (Cancela d’Abreu, 2004b). It is characterized by a heterogeneous forested landscape, predominantly covered by cork oak woodlands (Quercus suber agro-forestry system or Montado) and where extensive forestry plantations of Eucalyptus and Pinus trees are common. Goats, cows, pigs, and sheep grazing occur often in this area (Saraiva et al., 2021).

Zambujo is placed in the district of Castelo Branco, Idanha-a-Nova municipality, in the south of the Beira Interior region. This region is delimited by the Tagus river (south) and Erges river (east), representing the border with Spain and is characterized by abrupt and rocky valleys (Tomé et al., 2013) (Cancela d’Abreu, 2004a). The landscape is heterogeneous but typically dominated by Mediterranean Quercus suber and Quercus ilex woodlands and Olea europaea¸ including the var. sylvestris (Zambujeiros), groves. Recently, other land uses have been expanding throughout the region, namely Eucalyptus monocultures (Tomé et al., 2013, Cancela d’Abreu, 2004a).

These distinct features between the two areas configures different landscape contexts, as highlighted in Figures 2.1 and 2.2. One of the most important differences between the two farmsteads is the type of habitat edge that we can find in each of them. In Caniceira, we find well defined habitat edges, often separated by a empty corridor (Figure 2.3x). On the other hand, in Zambujo have more soft edges, corresponding to a gradual chance from a habitat to the other (Figure 2.4y). Also, These two area have different managements. In Caniceira the main focus is forestry production of Eucalyptus globulus and we find a constant human presence. In Zambujo, there is some management but the main focus is wildlife conservation and it doesn’t have a constant human presence. The detailed different characteristics are highlighted in Appendix 8.1.

Figure 2.1 - Land covers and location within Portugal of the Caniceira farmstead.

(17)

8 Due to its heterogeneous landscapes (Figure 2.1 and Figure 2.2), our study areas harbour a diversified community of mesocarnivores that include: red fox (Vulpes vulpes), stone marten (Martes foina), European badger (Meles meles), Eurasian otter (Lutra lutra), Egyptian mongoose (Herpestes ichneumon), and common genet (Genetta genetta) (Bencatel et al., 2018). In the Zambujo area, there are also some records of the presence of the wildcat (Felis silvestris) (Bencatel et al., 2018).

Figure 2.2 - Land covers and location within Portugal of the Zambujo farmstead.

Figure 2.3 – Example of a habitat edge in Caniceira farmstead.

Figure 2.4 – Example of a habitat edge in Zambujo farmstead.

(18)

9 3. Methods

3.1 Sampling design and field data collection - Carnivore sampling

Fieldwork took place between July 2020 and January 2021, corresponding to a total of 180 trapping nights. We set a total of 36 camera-traps from two different camera models: 21 Cuddeback Model H- 1453, set to take photos, and 15 Browning Model BTC-7E, set for videos. The cameras were placed at a distance of at least 500 m from each other (randomly in terms of each model). We placed 18 cameras in each farmstead, half in habitat edges and the other half in interior habitats. In this study we considered edges as the line or area, identifiable in the field, that separates two adjacent habitats. In both farmsteads, three cameras were set up in each habitat interior occurring in the area - Montado, Eucalyptus plantation and waterline, and other three in the Eucalyptus – Montado and in the Eucalyptus – waterline edges.

Lastly, we placed three in the Eucalyptus – pine forest edge in Caniceira and three in the Montado- waterline edge in Zambujo (Figure 2.5 and Figure 2.6). Even though in Caniceira the Pinus pinea and Pinus pinaster have a higher occupation area than the waterlines, we chose to include the waterlines instead of the Pinus habitats so that we could have a more comparable habitat representation between both our study areas. This design allowed us to monitor not only the interior of the habitat but also the different types of habitat edges, thus allowing us to test the edge effect on mesocarnivore spatial patterns, one of this study objectives.

Figure 2.5 - Location of the camera-trap stations within the studied land covers in the Caniceira farmstead.

(19)

10

Figure 2.6 - Location of the camera-trap stations within the different land covers in the Zambujo farmstead.

Cameras were placed facing north at a height that varied between 15 cm to 1m attached to tree trunks (Figure 2.7), depending on the steepness of the terrain, angled according to the target species' sizes. When necessary, the vegetation around the cameras was cleared to prevent false triggering and provide better visibility. When possible, the cameras faced animal trails located nearby to maximize detectability. No baits were used, and the cameras were set to operate 24h per day during the entire monitoring period, take three consecutive photos per trigger event, with an interval of 30 seconds between events, and record the date and time of each photograph. We visited the camera sites every 25-30 days, to replace the batteries and the memory cards.

- Prey sampling

We collected prey data for two types of the most common mesocarnivore preys in the Mediterranean region: small mammals and invertebrates (Bakaloudis et al., 2012; Coman, 1973; Rosalino et al., 2005;

Rosalino and Santos-Reis, 2002).

Figure 2.7 - Example of a camera-trap attached to a tree trunk (Eucalyptus globulus).

(20)

11 Both preys were captured using traps that were active for three nights in Caniceira (October) and four nights in Zambujo (November), sampling effort differing due to logistic constrains linked to adverse weather conditions.

Small mammals were sampled using a live trapping approach with traps set in a linear design. In each study area we set six trap lines of 20 Sherman traps each (Figure 2.8), composed of 10 large (XLF15 Folding Live Capture, 10.2x11.4x38.1 cm; H.B. Sherman Traps) and 10 small (LFA Folding Live Capture Traps, 7.6x8.9x22.9 cm; H.B.

Sherman Traps) traps, placed interspersed, within 10 meters from each other. All traps were baited with a mixture of oat flakes and sardines in oil and a handful of cotton was provided to avoid animals suffering from hypothermia while captured (Gurnell and Flowerdew, 2006). Traps were checked at sunrise to minimize the time spent by the animal inside them. Each captured individual was identified at the species level and marked with a different haircut code, to assure identification in case of recapture, and then released nearby the capture site (Gurnell and Flowerdew, 2006).

Capture and handling procedures followed national and international standards (Gannon and Sikes, 2007), and were authorized by the Portuguese Institute for Nature Conservation and Forests (ICNF) through capture licenses 752/2020/CAPT, 753/2020/CAPT, 05/2021/CAPT, and 06/2021/CAPT.

Terrestrial invertebrates were sampled using pitfall traps.

Each pitfall trap consisted of 1 L bowl-shaped container, with about 0.5 L of water mixed with a few drops of dish soap (Figure 2.9), to reduce the surface tension of the water and facilitate invertebrates’ retention in the pit-fall bottom.

Three pitfall traps were placed near each Sherman trap lines, buried in the ground with the top of the bowl levelled with the ground surface and separated by 1 meter from each other. The pitfalls were operational for the same number of days as the Sherman traps. All captured individuals were brought to the lab and grouped in to two clusters: Order Coleoptera and other invertebrates. Such grouping is based on the evidence that the Order Coleoptera is the preferred invertebrate prey for many mesocarnivores (Bakaloudis et al., 2012; Díaz-Ruiz et al., 2013; Hipólito et al., 2016b;

Rosalino and Santos-Reis, 2002).

Small mammal and terrestrial invertebrate trap lines were set in each of the habitats monitored by camera-trapping, both in Caniceira and Zambujo (Figure 2.10 and Figure 2.11). However, due to logistical problems we were not able to place a trap line in Caniceira’s Eucalyptus – pine trees edge.

Since both habitats had a similar vegetation structure (only differing in the type of tree) and were closely located, we considered that the prey data collected inside the Eucalyptus plantation could be used as a surrogate of the patterns we could detect in Eucalyptus – pine trees edges.

Figure 2.8 - Example of an Sherman trap, set in Zambujo to sample small mammals, showing the used cotton and bait.

Figure 2.9 - Pitfall trap, set in Zambujo to sample terrestrial invertebrates.

(21)

12

Figure 2.10 - Location of the Sherman trap sampling lines within the different land covers of the Caniceira farmstead.

Figure 2.11 - Location of the Sherman trap sampling lines within the different land covers of the Zambujo farmstead.

(22)

13 3.2 Data handling and analysis

3.2.1 Camera-trap data extraction

All the collected photos and videos were separated by camera-trap station and viewed one by one to exclude photos with no animal present and identify the individuals at a species level. For small mammals, such specific identification was not possible, and all photos were included in the category

“Small mammals”. All the photos were then tagged according to the number or individuals and the detected species, using Digikam 7.1.0 version software (www.digikam.org). Two photos of the same species in the same camera were only considered as independent records if they were separated by a minimum of 30 minutes, a time lag commonly used on activity patterns studies (Azevedo et al., 2018;

Curveira-Santos et al., 2017; Vilella et al., 2020).

To extract the photos’ metadata we used the package “camtrapR” within the R software version 4.1.1 (Niedballa et al., 2016). Using the ExifTool application we built a database that included: the camera station ID, its coordinates, photo time and date, species, and number of individuals of the identified animal species (Niedballa et al., 2016).

Further analyses were focused on our target species, the red fox, stone marten, European badger and common genet, as these are the mesocarnivore species that are more common in our study areas and represent two functional groups according to their arborical capabilities: cursorial (foxes and badgers) and arboreal (genets and martens).

3.2.2 Temporal patterns

The independent records of each species were used as random samples with a continuous temporal distribution. We estimated the mesocarnivores’ and small mammals’ activity patterns non- parametrically using the probability density function by applying Kernel density estimate (Ridout and Linkie, 2009). Detection times were extracted from the photos and converted from local to solar time using the “solarR” package from the R software (Perpiñan Lamigueiro, 2012). Then the time was converted into a scale between 0 and 1, and transformed into radians, where 24 hours = 2π radians (Vilella et al., 2020). This transformation into solar time allows us to compare times/data between different study areas and seasons, accounting for time zones and daylight saving as it uses the position of the sun in the sky (Foster et al., 2013).

We used the R package “overlap” (Meredith & Ridout, 2021) to fit and plot kernel density functions to visualize the general daily activity patterns of each target mesocarnivore and small mammals (Ridout and Linkie, 2009). We also compared patterns between the different mesocarnivores and between the mesocarnivores and small mammals. Additionally, we analysed the variation in activity patterns between edge and habitat interiors. This comparison was made by plotting simultaneously two density functions (overlapping kernel densities) for three different contexts: (1) density plots between edge and habitat interior for each mesocarnivore species, (2) density plots between edge and habitat interior for each pair of mesocarnivore species and (3) density plots between edge and habitat interior for each mesocarnivore species and small mammals. We estimated the overlap coefficient (Δ), ranging between 0 (no overlap) and 1 (complete overlap) ( Meredith & Ridout, 2021; Ridout and Linkie, 2009; Vilella et al., 2020) that corresponds to the area shared by the two functions that are being compared (highlighted as a shaded area in the plots). The overlap coefficient (Δ) was calculated using the estimator Δ1 since this is the most accurate for small sample sizes (Meredith & Ridout, 2021). For each of these estimators, we also calculated the 95% confidence intervals from 999 bootstrap samples (Foster et al., 2013; Ridout

(23)

14 and Linkie, 2009). Lastly, to assess the statistical differences between the activity patterns, we used the Mardia-Watson-Wheeler’s test (W) (Tasdan and Cetin, 2014) using the R package “circular” (Lund et al., 2017), with a level of significance of 0.05.

3.2.3 - Spatial patterns

- Mesocarnivore relative abundance

To explain mesocarnivores space-use patterns and understand how certain variables influence our target species’ occurrence, we used a generalized linear models (GLM) approach. We modelled how each species’ relative abundance index (RAI) varied as a response to a set of selected explanatory variables (See Table 3.1). The relative abundance index (RAI) was calculated as the number of independent records per active camera-trap nights (i.e., the number of nights that the camera was active and recording), for each camera site and each of the target species (Jenks et al., 2011).

1. 𝑅𝐴𝐼 = 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑟𝑒𝑐𝑜𝑟𝑑𝑠 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑎𝑐𝑡𝑖𝑣𝑒 𝑐𝑎𝑚𝑒𝑟𝑎 𝑡𝑟𝑎𝑝𝑠 𝑛𝑖𝑔ℎ𝑡𝑠

- Explanatory Variables

Mesocarnivores respond to variations in the environment by shifting their spatial ecology (Curveira- Santos et al., 2017), and as such, to assess which factors influence their abundance in habitat edges we decided to select 21 explanatory variables, linked to distinct ecological processes and with documented relevance for the target species (see Appendix 8.2), These were grouped into seven categories associated with the predefined hypotheses (see Introduction): edge characteristics, food resources, water resources, orography, habitat structure, land cover composition and anthropogenic disturbances. These variables were collected in situ and from a Geographical Information System (GIS) built using the software QGIS version 3.12.2.

While placing the cameras in the field, we defined a 20 meters buffer around the camera to characterize the habitat structure and visually estimate the percentage of local land covered by rocky formations, trees (namely fruit trees), shrubs and herbs. We also estimated the average shrub height and registered the Eucalyptus growth phase (if present), type of disturbances, and the accessibility to humans on a scale of 1 to 5 (from very difficult to easily accessible). This data collection was done by the same observer to avoid individual bias and reduce potential errors. The 20 meters radius was defined to allow us to visually estimate the different variables with accuracy and have a local characterization of the environment where the camera was installed.

Landscape characteristics were collected remotely using the QGIS software version 3.12.2, in a 200 meters buffer around each camera. We measured the distance to primary and secondary water lines, the percentage of land cover (adding to a total of 100%), pathway presence, edge’s range of influence (to determine if it is a sharp or soft edge) and edge density. Slope angle and orientation was also remotely collected for the exact position of the camera. We selected a 200 meters radius because it corresponds roughly to the core area of the largest mesocarnivore species present in each area – the red fox (0.15 km2; Alexandre et al., 2020).

The variable “% land cover” was composed of information from eleven different variables linked to land cover categories: Montado, Eucalyptus plantations, pastures, other hardwoods, pine trees plantations, shrublands, plant nurseries, infrastructures, inland waters, arable crops, and pathways. Due

(24)

15 to this high number of factors associated with the land cover composition of the study areas we summarise the information by performing a principal components analysis (PCA), using the % of cover of each of those categories within a 200m buffer around each camera as input data, and the function

“princomp” in the “stats” R package (Cooke et al., 2019). We did this procedure for the cameras set in each of the farmsteads, Caniceira and Zambujo, and considering all the cameras together.

Regarding food resources, we estimated small mammal relative abundance by using an adaptation of the Pounds relative abundance index (Pounds, 1981) :

1. IiNi

(𝑇×𝑅−𝐹)− ∑ 𝐶−𝑟 × 1000

Where Ni is the number of captured individuals by the Sherman traps of the species i, T the number of Sherman traps set, R the number of daily trap inspections, F the number of closed but empty traps, C the number of captures and recaptures of other species and r the number of recaptures of the species i.

Terrestrial invertebrates’ relative abundance was estimated by dividing the number of captured individuals in the pitfall traps by the number of trapping nights, per taxa: Coleoptera and other invertebrates.

- Modelling and model selection procedures

The modelling procedure follows a two steps process. First, we built the GLMs using data retrieved from all the cameras to identify the variables that influence the overall abundance of each mesocarnivore. Then, due to the differences between our two study areas (See Table 2.1), we also built separate GLMs for each farmstead (Caniceira and Zambujo). By applying this two-step modelling procedure, we were able to determine what factors influence the abundance of our target species globally, but also in two very different landscape contexts, and understand if the edge’s effect on abundance vary between both contexts.

In both modelling procedures, we first tested, for each hypothesis, the variable’s collinearity by calculating the Variance Inflation Factors (VIF), using the “corvif” function in the “AED” R package (Zuur et al., 2009). We considered a variable collinear if VIF>3 (Zuur et al., 2009). At each estimation, we removed the variable with the highest VIF and recalculate the VIF. This process was continued until all variables showed VIF values smaller than 3.

Then all candidate variables were standardized using the function “scale” in the R package “base” and the variables “pathways presence” and “camera placement” were used as binary variables (presence/absence).

We analysed the effect of our selected variables on our target species using GLMs with a gaussian distribution. Within each hypothesis, we created models corresponding to all possible combinations of the variables included in that hypothesis. However, due to the low number of cameras installed, overall and in each farmstead (36 in total and 18 for each property), we only used simultaneously a maximum of four variables for the general data models and two for the Caniceira and Zambujo models.

All created models in each hypothesis were ordered according to their Akaike Information Criterion (AICc), corrected for small sample sizes (Burnham and Anderson, 2002) in an ascending value, and the ΔAICc (difference between the AICc of each model and the lowest AICc model) was calculated. The models with ΔAICc<2 were considered as the best models for each hypothesis (Burnham and Anderson, 2002; Symonds and Moussalli, 2011).

(25)

16 In the best model for each hypothesis, we considered a variable to be an informative parameter if the 95% confidence intervals (CI) of is coefficient did not include the zero value (Arnold, 2010). The variables in those best models considered informative were used to test the eighth hypothesis, designated

"combined hypothesis", to assess if a combination of variables from different hypotheses (and thus shaping different ecological processes), produce a better result and, consequently, explain better the abundance patters for each target species. To produce these combined models, we followed the procedure described previously.

We identified the hypothesis with the higher support from our data as the one that generated the model with the lowest AICc. If more than one model in that hypothesis presented a ΔAICc<2, we applied a model averaging procedure to estimate the variable’s influence on species’ relative abundance (Symonds and Moussalli, 2011).

Finally, we tested the goodness-of-fit for the overall best model for each species, and for Caniceira and Zambujo farmsteads by using the function “nagelkerke” in the R package “rcompanion” (Mangiafico, 2022) to calculate the Likelihood-Ratio Test. This test compares two models (in our case our best model and the null model) to determine if adding complexity to the model makes it significantly more accurate (Mangiafico, 2022). A model was considered to have a good fit if p-value < 0,05.

3.2.4 Spatio-temporal analysis

As the spatial distribution patterns of species can influence their activity pattern, we decided to combine space and time use patterns. Thus, we applied the time-to-encounter analysis, which is a multi-response permutation procedure, following Karanth et al. (2017) protocol to assess the spatio-temporal segregation between our mesocarnivores.

First, we established which species of mesocarnivores are the dominant and the subordinate in each pair.

We considered the European badger as the dominant in every pair (Macdonald et al., 2004), while red foxes assumed a dominant role in the presence of stone martens and the common genets (Pereira et al., 2012). Finally, genets were considered dominant in relation to martens, as documented, (López-Martín, 2006; Santos-Reis et al., 2005). We determined these species as dominant due to their bigger body size and potential aggressive behaviour towards the other species.

For every detection record of each species at each camera location (sampling station), we calculated the minimum time-to-encounter to each of its subordinate pair’s subsequent detections. Hence, we ended up with a data set of capture intervals between interspecific captures. With this, we generated the expected statistical distribution of times-to-encounter by doing 1000 simulations where encounter times were randomly assigned to sampling stations. Then, we compared the median observed times-to- encounter with those randomly simulated. We plotted the times-to-encounter and, lastly, tested if the proportion of randomly generated times-to-encounters were statistically greater than those observed. A large p-value reflects species aggregation, meaning a smaller observed time-to-encounter than expected, a small p-value reflects species segregation, meaning a larger observed time-to-encounter than expected (Watabe et al., 2021). In the plots, the vertical line corresponds to the median observed time-to- encounter between the two species and the graphic itself represents the randomly simulated times-to- encounter. We follow this procedure for each pair of species using detection in the edge areas and the habitat interiors.

(26)

17 4. Results

4.1 Camera-trap capture success

From the 36 cameras that we installed, the minimum number of active cameras at one point was 24 and the average number of cameras not working at a given point was 8.5 with a standard deviation of 3. This was due to mechanical problems or cameras theft. Globally, we were able to record the four target mesocarnivore species, with a total of 424 independent records (Table 4.1; Figure 4.1). The mesocarnivore with the most detections was the red fox (Table 4.1), with 161 independent records, detected mainly in the Eucalyptus plantations of Caniceira (46 records) and in the Eucalyptus – Montado edge in Zambujo (16 records). The second species with the highest number of independent records was the common genet (Table 4.1), with 106 independent records, mainly in the waterlines of Caniceira (34) and in the Montado – Waterline edge of Zambujo farmstead (N=19), followed by the stone marten (Table 4.1) with 91 independent records, 25 of which in the Eucalyptus plantation of Caniceira and 13 in the waterline in Zambujo. The mesocarnivore with the lower number of independent records was the European badger (Table 4.1) with 66 independent records, 18 of which in the waterlines of Caniceira and 14 in the Montado – Waterline edge in Zambujo.

Table 4.1 – Number of independent records for each species detected during the survey period at each habitat and edge type.

* - Unable to identify at a species level

Species

Farmstead Habitat / Edge Red fox

Stone Marten

European Badger

Common Genet

Small mamma

ls*

Sub- total

Caniceira

Eucalyptus 46 25 4 1 5 81

Waterline 17 5 18 34 8 82

Montado 7 4 1 3 18 33

Eucalyptus – Pine Trees 10 8 0 0 7 25

Eucalyptus – Waterline 19 12 6 26 5 68

Eucalyptus – Montado 11 2 0 1 100 114

Zambujo

Eucalyptus 6 2 0 3 19 30

Waterline 6 13 4 3 12 38

Montado 15 5 11 4 43 78

Eucalyptus – Waterline 5 4 4 3 46 62

Eucalyptus – Montado 16 8 4 9 1 38

Montado – Waterline 3 3 14 19 14 53

Total 161 91 66 106 278

(27)

18 Regarding small mammals (Table 4.1), these were the animals with the highest number of independent records with a total of 278, although we were unable to identify the species based on the collected photos.

The Eucalyptus – Montado edge (N=100) and the Eucalyptus – waterline edge (N=46) were the habitats with more independent records in the Caniceira and Zambujo farmsteads, respectively (Table 4.1).

4.2 Temporal patterns

The mesocarnivores showed an overall bimodal activity pattern, with activity being mostly concentrated at sunset and sunrise periods. Overall, the activity peaks were less pronounced in the habitat interiors than in the edges (Figure 4.2), the exception being the European badger that in the habitat interior showed a unimodal nocturnal activity. Small mammals, although active in the same periods, evidenced the opposite pattern, with more pronounced activity peaks in the habitat interior and an unimodal pattern (Figure 4.2).

Figure 4.1 – Photos captured by camera-trap during our survey: Top left, red fox captured in Zambujo; top middle, common genet captured in Zambujo; top right, stone marten captured in Zambujo; bottom left, European badger captured in Caniceira and the bottom right, small mammals captured in Caniceira.

Referências

Documentos relacionados

The best way to achieve this goal is to apply a solvent-free, pH-neutral, hydrophobic adhesive resin layer in a separate step, as confirmed by inferior in vitro and in vivo

Por outro lado, Francisco Pereira, em sua introdução ao texto de Zurara, faz um interessante elogio ao cronista, comparando-o com Fernão Lopes: “Gomes Eannes de Zurara evitou, o

The probability of attending school four our group of interest in this region increased by 6.5 percentage points after the expansion of the Bolsa Família program in 2007 and

Os controlos à importação de géneros alimentícios de origem não animal abrangem vários aspetos da legislação em matéria de géneros alimentícios, nomeadamente

Além da recensão bibliográfica e do levantamento das aplicações já existentes para a aprendizagem de vocabulário em inglês como lí ngua estrangeira (ILE) e PLE,

Por outro lado, avaliando o exame ecográfico em relação a parâmetros qualitativos e não em termos de valores absolutos, foi possível verificar que existe diferença entre

Não obstante, e porque a posse “usurpadora” de João Álvares Neto, fundamentada em sesmaria de 1499, e a de Pero de Góis, com base na propriedade de Maria Corte Real (e de cuja

Universidade de São Paulo – USP, Av. The overall aim is to discuss the relationship between relief forms and the biodiversity of the Pantanal. The BAP is a natural environmental