Domestic Dog Invasion in an Agroforestry Mosaic in Southern Bahia,Brazil

Abstract

Portuguese

Although the value of agroforests for biodiversity conservation has been frequently highlighted, little is known about the susceptibility of this production system to biological invasions. Drawing on a camera-trap dataset obtained in 39 sites in an agroforestry mosaic in southern Bahia, Brazil, we investigated whether the conversion of native forests into agroforests and management intensification in agroforests favor the invasion by the most common carnivore worldwide, the domestic dog. We also examined whether domestic dog invasion is more associated with human activity in agroforests than in native forests. While the number of invading dogs was higher in agroforests than in native forests (11 compared to 7 dogs per site), management intensification in agroforests led to a higher mean number of visits per dog. In both habitats (not only agroforests) visits by domestic dogs tended to be concentrated on times of the day (around midday) and days of the week (Monday to Saturday) when there is greater human activity. Despite being permeable to native species, agroforests may act as sink or trap areas given their higher susceptibility to invasion, potentially limiting the value of this production system for biodiversity conservation. Moreover, local management intensification, which has been expanding worldwide, increases the intensity of such invasions, further decreasing the value of agroforests. The value of agroforestry mosaics for conservation thus depends on the management of invasive species and at least in the case of dogs, one of the most common and widely distributed invasive species, this management should focus on the habits and behavior of humans.

Keywords

alien species Atlantic forest land-sharing strategies wildlife-friendly agriculture

Introduction

Reconciling biodiversity conservation with farming production is presently one of the great challenges of mankind [1,2]. This reconciliation has been discussed from two distinct perspectives: (1) the land-sparing approach, based on agriculture intensification, so that larger areas of native vegetation can be set aside for conservation [3,4] and (2) the land-sharing approach, based on wildlife-friendly agricultural systems [5,6]. One of the main production systems associated with the land-sharing approach is agroforestry, a system of extensive management where agricultural products are cultivated in association with native or exotic trees [7]. Nowadays, extensive areas of native forests in the tropics have been converted into cacao and coffee agroforests [8,9]. Several studies have suggested that these agroforestry mosaics may favor the reconciliation of agricultural production and biodiversity conservation [6,10,11], but worldwide management intensification of agroforests may limit such benefits. [11,12]. To increase short term income levels, agroforests shaded by native trees have been converted into plantations shaded by few species of exotic trees or unshaded monocultures [11,13].

Agroforests shaded by native trees harbor a significant component of native faunas [14 –16], especially in comparison to more intensively-managed areas. However, given that disturbance facilitates colonization by invasive species [17,18], agroforests, especially those more intensively-managed, should be more susceptible to invasion than native forest remnants. This greater susceptibility may limit the conservation value of agroforestry systems. Biological invasions have been growing worldwide [19,20], and are today the second greatest threat to biodiversity [21,22], affecting different levels of ecological organization, from individual behavior [23] to ecosystem processes [24]. It has already been shown that native generalist species tend to increase and forest specialist species decrease in agroforests compared to native forests [25,26]. Although scarce, evidence in the literature indeed indicates that these systems are more susceptible to invasions than native forests [27 –29].

The domestic dog - the most abundant carnivore in the world [30], introduced wherever man has settled [31] - can be considered an invasive species by all three perspectives on biological invasions [32]: biogeographical, ecological-evolutionary, and nature conservation. The first is based on the breakdown of geographic barriers to dispersal, which in the case of dogs was indeed only possible after their association with humans [33]. The second focuses on novel interspecific interactions resulting from the spread or even the increase in the dominance of a species in a community. Because domestic dogs explore natural environments and interact with native fauna [30,34,35], they also fit the ecological-evolutionary perspective of invasive species. Finally, the nature conservation perspective specifically requires a negative impact of the invasive species on native species, and this is well-established for domestic dogs, which impact native species through lethal and non-lethal effects of predation [36,37], competition [38,39], and disease transmission [40,41].

Indeed, it has been proposed that, although considered a domesticated species, dogs are better understood as a subsidized predator, which is maintained at much higher population density than native predators by its association with humans [37]. This is because in vast areas where it now occurs, especially rural landscapes, dogs are free-ranging, occupying home ranges that include areas of native vegetation and preying upon wildlife [42].Moreover, because dogs in these rural landscapes usually follow men at work [43], frequency of occurrence in different habitats may be influenced by human activities. In agroforests, the presence of domestic dogs may thus be facilitated by human activities, which are more frequent in more intensively-managed systems.

However, although recent studies have shown that domestic dogs frequently use native vegetation in the tropics [e.g. 44–46], little is known about the presence of dogs in agroforestry mosaics. In a previous study conducted in one of the most important agroforestry regions in Brazil, we found that the number of records of domestic dogs better explains the decline of native mammals in agroforests than either the reduction of native forest cover in the surroundings or the local vegetation simplification [47]. Despite this evidence that dogs pose a threat to wildlife in agroforests, we still do not know which factors drive the distribution and abundance of dogs in agroforestry mosaics. We address this knowledge gap by investigating three interrelated hypotheses: (1) the conversion of native forests into agroforests and (2) management intensification in agroforests favor the invasion by domestic dogs, and (3) the invasion by these exotic animals is more associated with human activity in agroforests than in native forests. We draw on a standardized sampling protocol with camera-traps in 30 cacao agroforests and nine native forests throughout a 64,000-ha agroforestry mosaic. With this dataset, we examined whether the number of domestic dogs and the mean number of visits per dog (1) are higher in agroforests than in native forests and (2) increase in agroforests as management intensifies, and (3) whether days and times of visits by dogs are more concentrated in the periods of greater human activity in agroforests than in native forests.

Methods

Study area and sampling sites

The study was carried out in a 64,000-ha agroforestry mosaic in the Una and Arataca municipalities, in the cacao-growing region of southern Bahia, Brazil [48]. The climate is hot and humid, with no well-defined dry season [49]; average annual precipitation varies between 1,200 and 1,800 mm, and average annual temperature is 24.5°C [50]. The region was originally covered by Atlantic Forest, classified as “Southern Bahian Moist Forest” [51]. Nowadays 50% of the landscape is covered by native forests and 15% by cacao agroforests (Fig. 1). The studied agroforests, locally known as cabrucas, are cacao plantations where native trees from the original forest are left for shade; the understory is composed predominantly of cacao trees, and the herbaceous vegetation is absent or periodically removed [52]. The studied native forests are well-conserved old-growth forests, but may have suffered different levels of selective logging in the past.

Fig. 1.

Map of Brazil outlining the State of Bahia (a), map of Bahia showing the location of the study agroforestry mosaic (b), and land use in the study agroforestry mosaic, indicating the location of sampling sites (c).

Despite harboring a simplified mammal fauna, from which large felids (jaguar - Panthera onca, and puma - Puma concolor) and ungulates (tapir - Tapirus terrestris, and peccaries - Tayassu pecari and Pecari tajacu) are rare or locally extinct, the studied mosaic of native forests and agroforests is an important area for species of conservation concern, especially small primates (golden-headed lion tamarin - Leontopithecus chrysomelas, and Wied's marmoset - Callithrix kuhlii) [16].

We sampled 39 sites throughout the agroforestry mosaic, 30 in agroforests and nine in native forest remnants adjacent to nine of these agroforests (Fig. 1). The 30 agroforest sites were chosen to capture the variation in management intensification. Criteria for site selection were: (1) shaded cacao agroforests where exotic trees did not exceed 50 percent of the canopy layer (cabrucas), but which varied widely in management intensification associated with tree density and composition in the canopy layer, cacao density, and the clearance of herbaceous vegetation (these aspects were verified in the field, Appendix 1); (2) a minimum distance of 1.5 km between sites; and (3) sites that could be accessed (i.e., avoiding remote areas too far from any roads) [47]. For nine of these 30 agroforests sites, we sampled also an adjacent native forest in a paired design. Sites within a agroforest-native forest pair were between 200 and 450 m from each other, but a minimum distance of 100 m from the edge between native forest and agroforest. The nine agroforests with paired native forests were a minimum distance of 2.5 km apart; the sampled native forests were old-growth well-conserved forests, and the size of native forest and agroforest patches varied among the nine pairs.

Data collection

A camera-trap was installed in each of the 39 sampling sites, placed at 20–30 cm from the ground, baited with a small piece of sardine and bananas to increase the detection of dogs and ensure the precision of abundance estimates, while not affecting habitat use [53]. In all 39 sites, traps were active over four sampling sessions of ~ 30 days each, which occurred (for all sites) from July to October 2007 and 2008, and from January to April 2008 and 2009. Due to trap malfunctions, the sampling effort varied from 94 to 127 camera-days among the nine sites in native forest remnants, and from 90 to 126 camera-days among the 30 agroforests, totaling, respectively, 1,021 and 3,325 camera-days in each habitat and 4,346 camera-days in the agroforestry mosaic. Recorded dogs were identified from characteristics such as fur color, sex, size, breed, scars, and markings to obtain the number of individuals per sampling site, and the days and times of visits (Fig. 2). From a total of 716 photos of domestic dogs, only 23 (3.2%) did not allow individual identification.

Fig. 2.

Domestic dogs recorded in camera traps in cacao agroforests in southern Bahia, Brazil.

During camera-trap data collection, in all 30 agroforest sites we measured 14 vegetation structure variables to quantify management intensification (Appendix 1). Measurements were taken in four plots of 10 × 25 m, every 20 m, in an area of 25 × 100 m around the camera trap of each site.

Data analysis

Agroforests versus adjacent native forests

All the analyses comparing agroforests and native forests considered only the nine agroforest-native forest pairs. We used abundance models [54] modified from occupancy models [55], to investigate the effect of habitat type on the number of invading dogs. These models contain two parameters: abundance (λ) and probability of detection (r, the probability that an individual is detected), estimated from the capture history of individuals among the four sampling sessions [54]. The set of candidate models included a constant model (both parameters, r and λ, constant), and three models with r as a function of sampling effort and λ either constant, as a function of habitat type or as a function of the nine agroforest-native forest pairs.

We also used generalized linear mixed-effects models (GLMM) to investigate the effect of habitat type on the mean number of visits among individuals recorded at each site (modeled as a variable with normal distribution using log as the link function). The set of candidate models included a constant model (no fixed factors), and models with either sampling effort, habitat type, or both sampling effort and habitat type as fixed factors. In all models the nine agroforest-native forest pairs were considered as a random factor. The number of visits was calculated for each individual at each site considering all records except the sequential ones (i.e., those in intervals shorter than 1 hour) and averaged among all individuals recorded at a site.

To examine whether visits by dogs are more concentrated at times of greater human activity in agroforests than in native forests, we used the following tests: the Watson test of homogeneity between two circular samples [56] to test if the time of visits differs between the two habitats (agroforests and native forests), the Rayleigh test [57] to determine whether visits in each habitat are uniformly distributed throughout the day and, if not, the Von Mises estimate [56] to estimate the time when visits are concentrated. Finally, to test whether the days of visits by dogs are more associated with working days in agroforests than in native forests, we used Chi-square tests, considering two classes, working days (Monday to Saturday) and rest days (Sunday) and calculating the expected number of visits from the proportion of weekdays in each of these two classes. To guarantee independence, in both analyses of time of the day and day of the week we excluded sequential records of the same individual and used the median time or day of record for each individual.

Agroforests with varying management intensification

We ran a PCA in a correlation matrix to rank the 30 agroforest sites according to the 14 measured vegetation variables, and used the scores of the sites on the first two axes of this analysis (PCA1 and PCA2) as variables representing management intensification (see Results). To investigate the influence of management intensification on the number of invading dogs, we used abundance models modified from occupancy models [54]. The candidate set included a constant model (with abundance λ and detectability r constant), and four models with r as a function of sampling effort and λ either constant, as a function of only PCA1, only PCA2 or both PCA1 and PCA2.

We also used generalized linear models (GLM) to investigate the influence of management intensification on the mean number of visits among individuals recorded at each site (calculated as previously described and modeled as a variable with normal distribution using log as the link function). The set of candidate GLM models included a constant model, and seven models with only sampling effort, only PCA1, only PCA2, both PCA1 and 2, PCA1 and sampling effort, PCA2 and sampling effort, and PCA1, PCA2 and sampling effort as independent variables.

In all abundance models, given the small size of sampling sites in relation to the home range of dogs, we interpreted the parameter abundance as the number of individuals that use the sampling sites [58]. In all model selections, we compared candidate models using the Akaike Information Criterion (AICc) corrected for small sample size [59]. Models that present a difference in their AICc values relative to the first-ranked model (ΔAICc) smaller or equal to 2 were considered equally plausible. All the analyses were performed in Program R, using the packages “bbmle”, “car”, “CircStats”, “chron”, “MASS”, and “Unmarked” [60].

Results

Agroforests versus adjacent native forests

In the nine agroforest-native forest pairs, we observed 80 dogs, with only three of them recorded in both habitats. Of this total, 65 dogs were recorded in agroforests and only 18 in adjacent native forests. Indeed, only the model with abundance as a function of habitat type and detectability as a function of sampling effort was selected (Appendix 2), indicating that the number of invading dogs is higher in agroforests than in native forest remnants (Figura 3a). The estimated number of invading dogs was 11 per site and 99 in the set of nine sites in agroforests, and seven per site and 63 in the set of nine sites in native forests. In contrast, the mean number of visits among individuals recorded at each site did not differ consistently between habitats (Figura 3b), with only the constant model being selected (Appendix 3).

Fig. 3.

Number of individuals of domestic dogs (a) and mean number of visits among individuals (b) recorded in nine pairs of agroforests and adjacent native forests. Values of adjacent sampling sites are interconnected. Squares in (a) represent estimates obtained by the selected abundance model (Appendix 2).

In both habitats visits by dogs were not homogeneously distributed during the day (Rayleigh test: for native forests p<0.01; for agroforests p<0.01), concentrating at 13:14 h in native forests and 12:43 in agroforests (Fig. 4), and the time of visits did not differ between habitats (Watson homogeneity test: p=0.15). Similarly, in both habitats (and not only in the agroforests as predicted) visits by dogs tend to be concentrated on working days (marginally significant result of Chi-square tests in both cases; native forests ϧ²=0.70; p=0.08; agroforests ϧ²= 0.67; p=0.08) (Appendix 4).

Fig. 4.

Time of visits by domestic dogs to nine pairs of agroforests and adjacent native forests. Arrow: time at which visits concentrate; triangles: number of visits per 2-hour interval.

Agroforests with varying management intensification

The first axis of the PCA explained 27.3% of the variation in vegetation structure among the 30 agroforest sites, and represented a gradient of increasing management intensification associated with a decrease in tree shading and an increase in the frequency of clearance of the herbaceous vegetation and in the density of cacao shrubs and banana stems (Appendix 5). The second axis of the PCA explained 20.9% of the variation in vegetation structure and seems less directly related to management intensification, and more associated with the age of agroforests. Higher values of this axis represent mainly agroforests with smaller native trees, but with larger cacao shrubs (Appendix 5).

In total, we observed 145 dogs in the set of 30 agroforest sites, with none of them recorded in more than one agroforest. Only the model with detectability as a function of sampling effort and abundance constant was selected (Appendix 6). The estimated average number of dogs that use each agroforest was 12 individuals, with 360 dogs estimated to use the set of all 30 agroforests. It is noteworthy, however, that although not among the selected models, the model with abundance as a function of management intensification (PCA1) is ranked second (Appendix 6), indicating that management intensification has a positive, although weak, effect on the number of invading dogs (Figura 5a). In contrast, the two models that were selected to describe the mean number of visits among individuals recorded at each site contained management intensification (PCA1) as an independent variable (Appendix 7), indicating that the number of visits per dog increases in more intensively-managed agroforests (Figura 5b).

Fig. 5.

Number of individuals of domestic dogs (a) and mean number of visits among individuals (b) as a function of management intensification (PCA1, Appendix 5) throughout 30 agroforests. Empty dots: observed numbers; solid line and filled dots: estimates from models with management intensification as a co-variable (Appendices 6 and 7); black bars and black dashed lines: standard deviations of these estimates.

Discussion

The observed and estimated numbers of domestic dogs suggest that the intensity of invasion in agroforestry mosaics is high, representing a potential threat to the fauna of these landscapes. Not only dogs have been shown to impact native species through lethal and non-lethal effects of predation [36,37], competition [38,39], and disease transmission [40,41], but also we found in a previous study that the number of records of dogs was the main driver of the decline of native mammals in the studied agroforestry mosaic [47].

More importantly, our results corroborate the hypothesis of this study that the conversion of native forests into agroforests, as well as management intensification in agroforests, favor the invasion by domestic dogs. On the other hand, despite evidence that visits by dogs are associated with times and days of greater human activity, this association is not stronger in agroforests than in native forests, as we expected. Because only three of the 80 dogs recorded in the nine agroforest-native forest pairs were recorded in both habitats, this association of the presence of dogs with human activity is not just an artifact of the proximity of the sampled agroforest-native forest sites. Rather, despite the smaller number of dogs in native forests than in agroforests, the presence of dogs in native forests seems to be associated with human activity, although not necessarily with the activity of workers in the adjacent agroforests. Below we discuss each of these main results and their implications.

As expected, the number of invading dogs was higher in agroforests than in native forests. In contrast, the mean number of visits per individual did not vary between the two habitat types. As in other studies [44,61], most individuals were recorded only once, both in native forests and in agroforests. Thus, the higher number of records of dogs in agroforests than in native forests in the study region [16] results mainly from the entrance of a larger number of dogs, rather than from more frequent visits by the same individuals. The higher the number of individuals, the larger the probability of disease transmission to wild animals, adding to the expected negative effects of competition and predation by dogs [34,35,62,63].

Given that the number of people entering managed systems such as agroforests is much higher than in unmanaged native forests, the higher number of dogs in this agricultural system may be simply linked to the behavior of dogs following their owners to work [43]. This hypothesis is supported by our finding that there is an association between dog presence and the time of the day and the days of the week with greater human activity. Beyond the association with workers, however, the higher number of dogs in agroforests than in native forests may also be related to environmental factors. Compared to native forests, agroforests present a simplified vegetation structure, which may facilitate the access and movement of dogs, since these animals move preferentially through open habitats [64,65]. Also, vegetation thinning in agroforests may influence native fauna, with several native species becoming less frequent in agroforests than in native forests [16,29]. If large predators that may chase or kill domestic dogs are more susceptible to vegetation thinning than mid-sized and small vertebrates (preys), simplification of the native fauna may also contribute to the higher number of dogs in agroforests than in native forests.

By contrast, in agroforests the positive effect of management intensification was stronger on the mean number of visits per individual than on the number of invading dogs. More intensively-managed plantations, associated with a higher frequency of herbaceous vegetation clearance and a higher number of cacao shrubs, require a higher frequency of workers or longer working periods, and result in more open vegetation close to the ground. Both characteristics may lead to the increase in the number of visits per dog in more intensified agroforests, either because they follow their owners who enter more frequently or stay longer in these plantations, or because their movement is facilitated in more open environments [64,65]. However, these characteristics do not necessarily affect the number of dogs that visit agroforests, which may also depend on other factors, such as the quantity of dogs raised in surrounding areas (i.e. propagule pressure [66,67]) or on the number of workers, probably more associated with plantation size and other socioeconomic factors than with management intensification per se. It is important that future studies on the distribution and impacts of dogs in agroforestry mosaics focus on these socio-economic dimensions.

In summary, our findings suggest that the higher number of invading dogs in agroforests compared to native forests is associated with the larger number of people who enter and stay in these managed systems, while the higher number of visits in more intensively-managed agroforests may be associated with a higher frequency/length of stay of workers. Although both conversion of native forests into agroforests and management intensification in agroforests affect invasion by domestic dogs, they influence distinct aspects, which may be associated with differences in the presence of people between these two habitats. In fact, in both environments (and not only in agroforests) visits by domestic dogs seem to be associated with the period of human activity, concentrating during the day, as observed in other studies [e.g.44,46], and on working days during the week. This suggests that the control of invasion rests on conscious efforts by workers and local residents to limit the movements of their dogs - by keeping them home instead of allowing or stimulating dogs to follow them into agroforests.

Implications for conservation

Our results, as well as previous studies [27 –29], indicate that, though permeable to wildlife [15,68], agroforests may act as trap or sink areas, given the higher intensity of invasion compared to native forests, which should restrict the conservation value commonly attributed to agroforests [6,10]. More importantly, our study suggests that agroforest management intensification, which is becoming more pronounced on a global scale [11,12], increases the risks of invasions, potentially limiting the value of this agricultural system for reconciling food production and biodiversity conservation. Indeed, the domestic dog record rate was one of the most important factors negatively associated with the distribution of wild mammals in agroforests of the study region [47].

Furthermore, our study suggests – both by the association of domestic dog visits with the periods of greater human activity, and by the fact that native forest conversion into agroforests led to an increase in the number of dogs while management intensification in agroforests led to an increase in the number of visits per dog - that there is a strong association between the invasion by domestic dogs and human activities. Yet, this association does not depend on the type of habitat, and apparently occurs also in unmanaged systems, such as native forests. Thus the value of agroforestry mosaics for conservation depends on the proper management of invasive species, and at least in the case of dogs, one of the most common and widely distributed invasive species, this management should focus on how workers and residents treat their dogs. In contrast to other aspects of agroforest management intensification that negatively affect wildlife (e.g. thinning of shade trees and higher frequency of herbaceous vegetation clearance [47]), increased invasion by domestic dogs can be avoided with no impairment to agroforest productivity. Although no management solution to avoid the impact of dogs on wildlife will be simple, awareness and education campaigns to discourage workers to allow their dogs to follow them into agroforests can be a step forward, and could be incorporated in action plans for increasing the conservation value of agroforestry mosaics.

Footnotes

Acknowledgments

We thank the Instituto de Estudos Socioambientais do Sul da Bahia (IESB) for providing the maps and logistic support during field work; and the European Union, the Brazilian Ministry of the Environment, and Seeds of Change for funding. EF was granted a MSc scholarship and CRC a PhD scholarship from FAPESP (2011/03113-5 and 2007/54888-1, respectively), and RP had a research fellowship from CNPq (306715/2011-2) during the development of this work.

Appendix 1. Minimum,mean (± SD) and maximum values of the 14 vegetation structure variables quantified in 30 agroforests to characterize management intensification.

Variable identification	Variable description	Minimum	Mean ± SD	Maximum
ht_cacao	Average height of cacao shrubs (m)‡	2.6	4.1 ± 0.7	5.5
ht_exot	Average height of exotic woody trees (m)†	0	12.1 ± 8.4	26.8
ht_nat	Average height of native trees (m)†	12.4	20.0 ± 5.6	38.5
ht_herb	Average herbaceous vegetation height (cm)*	22.0	69.9 ± 31.5	140.0
cacao.BA	Sum of basal area of cacao shrubs (m²)‡	0.1	0.3 ± 0.1	0.6
exot.BA	Sum of basal area of exotic woody trees (m²)†	0	0.3 ± 0.5	2.1
nat.BA	Sum of basal area of native trees (m²)†	0.2	1.4 ± 0.7	2.7
n_cacao	Number of cacao shrubs‡	18	69.9 ± 31.5	140
n_exot	Number of exotic woody trees†	0	3.4 ± 3.8	15.0
n_nat	Number of native trees†	4	10.5 ± 6.9	34
n_banana	Number of banana stems†	0	4.0 ± 4.2	15
can.con	Sum of canopy connectivity index‡	1.0	27.5 ± 24.7	119.0
cacao.con	Sum of cacao connectivity index†	24.0	87.6 ± 28.7	143.0
slash	Frequency of herbaceous vegetation clearance (times*year⁻¹)	0.5	1.3 ± 0.6	2.5

†

Variables measured in four plots of 10 × 25 m, every 20 m, in an area of 25 × 100 m. In these plots, we computed banana stems, computed and measured the height and diameter of all shade trees with at least 10 cm in diameter at breast height (DBH), classifying them as native or exotic trees, and estimated the proportion of canopy of each of them that was connected to the canopy of adjacent trees through an index that varied from 0 to 4 (corresponding to the number of tree crown quadrants connected to adjacent trees).

‡

Variables measured in four half plots (5 × 25 m) inside main plots describe above. In these plots we counted and measured the height and diameter of all cacao trees and we estimated the proportion of canopy that was connected to the adjacent cacao trees.

Variable measured in12 sub-plots of 0.5 × 0.5 m, arranged at every 5 m along one of the sides and in the center of each main plot. In these sub-plots we measured the height of the herbaceous vegetation, through the height of the tallest herbaceous plant inside each of them.

The shade trees and the cacao trees were quantified at the end of the second sampling sessions with camera-traps, and the herbaceous vegetation height, which varies temporally, was quantified three times, at the end of the second, third and fourth sampling sessions.

Appendix 2. Results from the selection of abundance models for domestic dogs in nine pairs of agroforests and adjacent native forests. Models are ordered from the most plausible (the lowest to the highest AICc value),and the selected model is highlighted (light grey). For each model,information is provided on the number of parameters (K),maximum likelihood estimation (log-Lik),Akaike Information Criterion for small samples (AICc),difference in AICc relative to the best model (ΔAICc),weight of evidence (wi),and coefficients of co-variables (+ positive;- negative). Eff= Sampling effort;Pairs= nine agroforest-native forest pairs;Habitat= agroforest versus native forest;λ= abundance;r= detectability;(.) constant parameter.

Models	K	log-Lik	AICc	ΔAIC	wi	Coefficients
Models	K	log-Lik	AICc	ΔAIC	wi	Habitat	Eff
λ(Habitat) r(Eff)	5	-36.25	87.5	0	0.700	–	+
λ(.) r(.)	3	-41.29	90.3	2.9	0.153
λ(Pairs) r(Eff)	5	-37.85	90.7	3.2	0.121		+
λ(.) r(Eff)	4	-41.01	93.1	5.6	0.026		+

Appendix 3. Results from the selection of generalized linear mixed-effects models (GLMM) for the mean number of visits among individuals recorded in nine pairs of agroforests and adjacent native forests. Models are ordered from the most plausible (the lowest to the highest AICc value),and the selected model is highlighted (light grey). For each model,information is provided on the fixed factors,number of parameters (K),maximum likelihood estimation (log-Lik),Akaike Information Criterion for small samples (AICc),difference in AICc relative to the best model (ΔAICc),weight of evidence (wi),and coefficients of fixed factors (+ positive;- negative). Eff= Sampling effort;Habitat= agroforest versus native forest. All models had the nine agroforest-native forest pairs as a random-factor.

Fixed factors	K	log-Lik	AICc	ΔAIC	wi	Coefficients
Fixed factors	K	log-Lik	AICc	ΔAIC	wi	Habitat	Eff
Constant	3	-30.79	69.3	0	0.736
Habitat	4	-30.23	71.5	2.3	0.238	–
Eff	4	-32.77	76.6	7.3	0.018		+
Habitat + Eff	5	-31.98	79	9.7	0.005	–	+

Appendix 4. Observed and expected number of visits by dogs in native forests and agroforests on working days (Monday to Saturday) and rest days (Sunday). The expected value was calculated through the proportion of days in each class (work and rest).

Native forests
	Observed		Expected
	n	(%)	n	(%)
Working days	17	(94)	6	(86)
Rest days	1	(6)	1	(14)
Total	18	(100)	7	(100)
Agroforests
	Observed	Expected
	n	(%)	n	(%)
Working days	61	(94)	6	(86)
Rest days	4	(6)	1	(14)
Total	65	(100)	7	(100)

Appendix 5. Ordination of the 30 sampled agroforests (black dots) on a biplot of the first and second axes of the Principal Component Analysis (PCA) using the 14 vegetation structure variables (see Appendix 1).

Appendix 6. Results from the selection of abundance models for domestic dogs in 30 agroforests. Models are ordered from the most plausible (the lowest to the highest value of AICc),and the selected model is highlighted (light grey). For each model,information is provided on the number of parameters (K),maximum likelihood estimation (log-Lik),Akaike Information Criterion for small samples (AICc),difference in AICc of relative to the best model (ΔAICc),weight of evidence (wi),and coefficients of co-variables (+ positive;- negative). Eff= Sampling effort;PCA1 and PCA2= first and second axis from Principal Component Analysis on agroforest management intensification (Appendix 5);λ= abundance;r= detectability;(.) constant parameter.

Models	K	log-Lik	AICc	ΔAIC	wi	Coefficients
Models	K	log-Lik	AICc	ΔAIC	wi	PCA1	PCA2	Eff
λ(.) r(Eff)	4	-208.02	425.6	0.00	0.636			+
λ(PCA1) r(Eff)	5	-207.87	428.2	2.59	0.174	+		+
λ(PCA2) r(Eff)	5	-207.99	428.4	2.84	0.153		–	+
λ(PCA1 + PCA2) r(Eff)	6	-207.84	431.3	5.68	0.037	+	–	+
λ(.) r(.)	3	-216.52	439.9	14.32	0.000

Appendix 7. Results from the selection of generalized linear models (GLM) for the mean number of visits among individual recorded in 30 agroforests. Models are ordered from the most plausible (the lowest to the highest value of AICc),and the selected models are highlighted (light grey). For each model,information is provided on the independent variables,on the number of parameters (K),maximum likelihood estimation (log-Lik),Akaike Information Criterion for small samples (AICc),difference in AICc relative to the best model (ΔAICc),weight of evidence (wi),and coefficients of independent variables (+ positive;-negative). Eff= Sampling effort;PCA1 and PCA2= first and second axis from the Principal Component Analysis on agroforest management intensification (Appendix 5).

Independent variables	K	log-Lik	AICc	ΔAIC	wi	Coefficients
Independent variables	K	log-Lik	AICc	ΔAIC	wi	PCA1	PCA2	Eff
PCA1	3	-37.84	82.6	0	0.309	+
PCA1 + EFF	4	-36.75	83.1	0.5	0.24	+		+
EFF	3	-38.94	84.8	2.2	0.109			+
PCA2	3	-38.94	84.8	2.2	0.107		–
PCA 1 + PCA 2	4	-37.75	85.1	2.5	0.09	+	–
PCA2 + EFF	4	-37.90	85.4	2.8	0.078		–	+
Constant	2	-41.23	86.9	4.3	0.035
PCA 1 + PCA 2 + EFF	5	-37.45	87.4	4.8	0.028	+	–	+

References

Sala

O.E.

Chapin

F.S.

Armesto

J.J.

Berlow

Bloomfield

Dirzo

Huber-Sanwald

Huenneke

L.F.

Jackson

R.B.

Kinzig

Leemans

Lodge

D.M.

Mooney

H.A.

Oesterheld

Poff

N.L.

Sykes

M.T.

Walker

B.H.

Walker

and Wall

D.H.

2000. Biodiversity - Global biodiversity scenarios for the year 2100. Science 287:1770–1774.

Foley

J.A.

Defries

Asner

G.P.

Barford

Bonan

Carpenter

S.R.

Chapin

F.S.

Coe

M.T.

Daily

G.C.

Gibbs

H.K.

Helkowski

J.H.

Holloway

Howard

E.A.

Kucharik

C.J.

Monfreda

Patz

J.A.

Prentice

I.C.

Ramankutty

and Snyder

P.K.

2005. Global consequences of land use. Science 309:570–574.

Balmford

Green

R.E.

and Scharlemann

J.P.W.

2005. Sparing land for nature: Exploring the potential impact of changes in agricultural yield on the area needed for crop production. Global Change Biology 11:1594–1605.

Green

R.E.

Cornell

S.J.

Scharlemann

J.P.W.

and Balmford

2005. Farming and the fate of wild nature. Science 307:550–555.

Phalan

Onial

Balmford

and Green

R.E.

2011. Reconciliating food production and biodiversity conservation: Land sharing and land sparing compared. Science 333:1289–1291.

Perfecto

and Vandermeer

2008. Biodiversity conservation in tropical agroecosystems - A new conservation paradigm. Annals of the New York Academy of Science 1134:173–200.

Somarriba

1992. Revisiting the past - An essay on agroforestry definition. Agroforestry Systems 19:233–240.

Rice

R.A.

and Greenberg

2000. Cacao cultivation and the conservation of biological diversity. AMBIO 29:167–173.

Bhagwat

S.A.

Willis

K.J.

Birks

H.J.B.

and Whittaker

R.J.

2008. Agroforestry: A refuge for tropical biodiversity? Trends in Ecology & Evolution 23:261–267.

10.

Schroth

and Harvey

C.A.

2007. Biodiversity conservation in cocoa production landscapes: An overview. Biodiversity and Conservation 16:2237–2244.

11.

Tscharntke

Clough

Bhagwat

S.A.

Buchori

Faust

Hertel

Holscher

Juhtbandt

Kessler

Perfecto

Scherber

Schroth

Veldkamp

and Wanger

T.C.

2011. Multifunctional shade-tree management in tropical agroforestry landscapes - A review. Journal of Applied Ecology 48:619–629.

12.

Juhrbandt

Duwe

Barkmann

Gerold

and Marggraf

2010. Structure and management of cocoa agroforestry systems in Central Sulawesi across an intensification gradient. In: Tropical rainforests and agroforests under global change: Ecological and socio-economic valuations. Tscharntke

Leuschner

Veldkamp

Faust

Guhardja

and Bidin

(Eds.), pp.115–140. Springer, New York.

13.

Franzen

and Borgerhoff-Mulder

2007. Ecological, economic and social perspectives on cocoa production worldwide. Biodiversity and Conservation 16:3835–3849.

14.

Schulze

C.H.

Waltert

Kessler

P.J.A.

Pitopang

Shahabuddin

Veddeler

Muhlenberg

Gradstein

S.R.

Leuschner

Steffan-Dewenter

and Tscharntke

2004. Biodiversity indicator groups of tropical land use systems: Comparing plants, birds, and insects. Ecological Applications 14:1321–1333.

15.

Cassano

C.R.

Schroth

Faria

Delabie

J.H.C.

and Bede

2009. Landscape and farm scale management to enhance biodiversity conservation in the cocoa producing region of southern Bahia, Brazil. Biodiversity and Conservation 18:577–603.

16.

Cassano

C.R.

Barlow

and Pardini

2012. Large Mammals in an Agroforestry Mosaic in the Brazilian Atlantic Forest. Biotropica 44:818–825.

17.

Newsome

A.E.

and Noble

I.R.

1986. Ecological and physiological characters of invading species. In: Ecology of Biological Invasions. Groves

R.H.

and Burdon

J.J.

(Eds.), pp.1–20. Cambridge University Press, Cambridge.

18.

Smallwood

K.S.

1994. Site invasibility by exotic birds and mammals. Biological Conservation 69:251–259.

19.

Hulme

P.E.

2009. Trade, transport and trouble: managing invasive species pathways in an era of globalization. Journal of Applied Ecology 46:10–18.

20.

McGeoch

M.A.

Butchart

S.H.M.

Spear

Marais

Kleynhans

E.J.

Symes

Chanson

and Hoffmann

2010. Global indicators of biological invasion: species numbers, biodiversity impact and policy responses. Diversity and Distributions 16:95–108.

21.

Vitousek

P.M.

D'Antonio

C.M.

Loope

L.L.

Rejmánek

and Westbrooks

1997. Introduced species: a significant component of human-caused global change. New Zealand Journal of Ecology 21:1–16.

22.

Baillie

J.E.M.

Hilton-Taylor

and Stuart

S.N.

2004. 2004 IUCN Red List of Threatened Species: a Global Species Assessment. IUCN, Gland, Switzerland and Cambridge.

23.

Creel

and Christianson

2008. Relationships between direct predation and risk effects. Trends in Ecology & Evolution 23:194–201.

24.

Raizada

Raghubanshi

A.S.

and Singh

J.S.

2008. Impact of invasive alien plant species on soil processes: a review. Proceedings of the National Academy of Sciences India, Section B, Biological Sciences 78:288–298.

25.

Pardini

Faria

Accacio

G.M.

Laps

R.R.

Mariano-Neto

Paciência

M.L.B.

Dixo

and Baumgarten

2009. The challenge of maintaining Atlantic forest biodiversity: A multitaxa conservation assessment of specialist and generalist species in an agro-forestry mosaic in southern Bahia. Biological Conservation 142:1178–1190.

26.

Waltert

Bobo

K.S.

Kaupa

Montoya

M.L.

Nsanyi

M.S.

and Fermon

2011. Assessing conservation values: Biodiversity and endemicity in tropical land use systems. Plos One 6:e16238.

27.

Richardson

D.M.

Binggeli

and Schroth

2004. Invasive agroforestry trees: Problems and solutions. In: Agroforestry and biodiversity conservation in tropical landscapes. Schroth

da Fonseca

G.A B.

Harvey

C.A.

Gascon

Vasconcelos

H.L.

and Izac

A.M.N.

(Eds.), pp.371–396. Island Press, Washington, D.C.

28.

Bos

M.M.

Tylianakis

J.M.

Steffan-Dewenter

and Tscharntke

2008. The invasive Yellow Crazy Ant and the decline of forest ant diversity in Indonesian cacao agroforests. Biological Invasions 10:1399–1409.

29.

Weist

Tscharntke

Sinaga

M.H.

Maryanto

and Clogh

2010. Effect of distance to forest and habitat characteristics on endemic versus introduced rat species in agroforest landscapes of Central Sulawesi, Indonesia. Mammalian Biology 75:567–571.

30.

Vanak

A.T.

and Gompper

M.E.

2010. Interference competition at the landscape level: the effect of free-ranging dogs on a native mesocarnivore. Journal of Applied Ecology 47:1225–1232.

31.

Pimentel

Lach

Zuniga

and Morrison

2000. Costs of nonindigenous species in the United States. BioScience 50:53–65.

32.

Heger

Saul

W.C.

Trepl

2013. What biological invasions ‘are’ is a matter of perspective. Journal for Nature Conservation 21:93–96.

33.

Prates

Prevosti

F.J.

Berón

2010. First records of prehispanic dogs in southern South America (Pampa-Patagonia, Argentina). Current Anthropology 51:273–280.

34.

Vanak

A.T.

and Gompper

M.E.

2009. Dogs Canis familiarisas carnivores: their role and function in intraguild competition. Mammal Review 39:265–283.

35.

Hughes

and Macdonald

D.W.

2013. A review of the interactions between free-roaming domestic dogs and wildlife. Biological Conservation 157:341–351.

36.

Silva-Rodriguez

E.A.

Verdugo

Aleuy

O.A.

Sanderson

J.G.

Ortega-Solis

G.R.

Osorio-Zuniga

Gonzalez-Acuna

2010. Evaluating mortality sources for the vulnerable pudu Pudu puda in Chile: implications for theconservation of a threatened deer. Oryx 44:97–103.

37.

Gompper

M.E.

Vanak

A.T.

2008. Subsidized predators, landscapes of fear and disarticulated carnivore communities. Animal Conservation 11:13–14.

38.

Vanak

A.T.

Gompper

M.E.

2010. Interference competition at the landscape level: the effect of free-ranging dogs on a native mesocarnivore. Journal of Applied Ecology 47:1225–1232.

39.

Butler

J.R.A.

du Toit

J.T.

Bingham

2002. Free-ranging domestic dogs (Canis familiaris) as predators and prey in rural Zimbabwe: threats of competition and disease to large wild carnivores. Biological Conservation 115:369–378.

40.

Woodroffe

1999. Managing disease threats to wild mammals. Animal conservation 2:185–193.

41.

Cleaveland

Appel

M.G.J.

Chalmers

W.S.K.

Chillingworth

Kaare

Dye

2000. Serological and demographic evidence for domestic dogs as a source of canine distemper virus infection for Serengeti wildlife. Veterinary Microbiology 72:217–227.

42.

Gompper

M.E.

2014. The dog-human-wildlife interface: assessing the scope of the problem. In: Free-Ranging Dogs and Wildlife Conservation. Gompper

M.E.

(Ed.), pp.9–54. Oxford University Press, Oxford.

43.

Serpell

2008. The domestic dog - Its evolution, behaviour, and interactions with people. Cambridge University Press, Cambridge.

44.

Srbek-Araujo

A.C.

and Chiarello

A.G.

2008. Domestic dogs in Atlantic forest preserves of south-eastern Brazil: a camera-trapping study on patterns of entrance and site occupancy rates. Brazilian Journal of Biology 68:771–779.

45.

Lacerda

A.C.R.

Tomas

W.M.

and Marinho-Filho

2009. Domestic dogs as an edge effect in the Brasília National Park, Brazil: interactions with native mammals. Animal Conservation 12:477–487.

46.

Paschoal

A.M.O.

Massara

R.L.

Santos

J.L.

and Chiarello

A.G.

2012. Is the domestic dog becoming an abundant species in the Atlantic forest? A study case in southeastern Brazil. Mammalia 76:67–76.

47.

Cassano

C.R.

Barlow

and Pardini

2014. Forest loss or management intensification? Identifying causes of mammal decline in cacao agroforests. Biological Conservation 169:14–22.

48.

SEI - Superintendência de Estudos Econômicos e Sociais da Bahia, 1999. Diagnóstico Ambiental, Litoral Sul da Bahia. Série de Estudos e Pesquisas 43:1–120.

49.

Köppen

1948. Climatologia. Fondo Cultura Economica, Mexico City.

50.

Mori

S.A.

Boom

B.M.

Carvalho

A.M.

and Santos

T.S.

1983. Southern Bahian Moist Forests. Botanical Review 49:1–155.

51.

Thomas

W.W.

2003. Natural vegetation types in southern Bahia. In: Corredor de Biodiversidade da Mata Atlântica do sul da Bahia. Prado

P.I.

Landau

E.C.

Moura

R.T.

Pinto

L.P.

Alger

and Fonseca

(Eds.). CD-ROM, Ilhéus, IESB/CI/CABS/UFMG/UNICAMP.

52.

Sambuichi

R.H.R.

2002. Fitossociologia e diversidade de espécies arbóreas em cabrucas (Mata Atlântica raleada sobre plantação de cacau) na região sul da Bahia, Brasil. Acta Botânica Brasilica 16:89–101.

53.

Gerber

B.D.

Karpanty

S.M.

and Kelly

M.J.

2012. Evaluating the potential biases in carnivore capture–recapture studies associated with the use of lure and varying density estimation techniques using photographic-sampling data of the Malagasy civet. Population Ecology 54:43–54.

54.

Royle

J.A.

2004. N-mixture models for estimating population size from spatially replicated counts. Biometrics 60:108–115.

55.

Mackenzie

D.I.

Nichols

J.D.

Lachman

G.B.

Droege

Royle

J.A.

and Langtimm

C.A.

2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83:2248–2255.

56.

Jammalamadaka

S.R.

and SenGupta

2001. Topics in Circular Statistics. World Scientific Press, Singapore.

57.

Wilkie

1983. Rayleigh Test for randomness of circular data. Journal of the Royal Statistical Society 32:311–312.

58.

McCarthy

M.A.

Moore

J.L.

Morris

W.K.

Parris

K.M.

Garrard

G.E.

Vesk

P.A.

Rumpff

Giljohann

K.M.

Camac

J.S.

Bau

S.S.

Friend

Harrison

and Yue

2013. The influence of abundance on detectability. Oikos 122:717–726.

59.

Burnham

K.P.

and Anderson

D.R.

2002. Model selection and multimodel inference: a practical information-theoretic approach. Springer, New York.

60.

R Development Core Team, 2011. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna. URL http://www.R-project.org/

61.

Paschoal

A.M.O.

2008. Predadores em fragmentos de Mata Atlântica: estudo de caso na RPPN Feliciano Miguel Abdala, Caratinga, MG. MSc Dissertation, Pontifícia Universidade Católica de Minas Gerais, Brasil.

62.

Young

J.K.

Olson

K.A.

Reading

R.P.

Amgalanbaatar

and Berger

2011. Is wildlife going to the dogs? Impacts of feral and free-roaming dogs on wildlife populations. BioScience 61:125–132.

63.

Martinez

Cesário

Silva

I.O.

Boere

2003. Domestic dogs in rural area of fragmented Atlantic Forest: potential threats to wild animals. Ciência Rural 43:1998–2003.

64.

May

S.A.

and Norton

T.W.

1996. Influence of fragmentation and disturbance on the potential impact of feral predators on native fauna in Australia forest ecosystems. Wildlife Research 23:387–400.

65.

Manor

and Saltz

2004. The impact of free-roaming dogs on gazelle kid/female ratio in a fragmented area. Biological Conservation 119:231–236.

66.

Lockwood

J.L.

Cassey

and Blackburn

2005. The role of propagule pressure in explaining species invasions. Trends in Ecology & Evolution 20:223–228.

67.

Gurevitch

Fox

G.A.

Wardle

G.M.

and Taub

2011. Emergent insights from the synthesis of conceptual frameworks for biological invasions. Ecology Letters 14:407–418.

68.

Tscharntke

Klein

A.M.

Kruess

Steffan-Dewenter

and Thies

2005. Landscape perspectives on agricultural intensification and biodiversity – Ecosystem service management. Ecology Letters 8:857–874.