INTRODUCTION

The functionality of protected areas as conservation units is important for the maintenance of biodiversity to all ecosystems (^{Primack 1993}; ^{Parks & Harcourt 2002}). Different types of anthropic activities are directly or indirectly associated with loss of biodiversity in tropical forests, whether linked to habitat fragmentation (^{Laurance 1999}), or even activities related to hunting, farming and livestock practices (^{Cullen et al. 2000}; ^{Ferreguetti et al. 2017}). Associated with human occupation, an important factor that may affect the native species living in these natural areas is the presence of exotic species (^{Young et al. 2011}). In Brazil, we observed that the negative impacts for the conservation of biodiversity are growing faster than the generation of data to mitigate them (^{Dechoum et al. 2013}; ^{de Andrade & Iadanza 2016}).

The introduction of invasive exotic species can pose as a major cause of extinction of species, and in the case of mammals the impacts imposed by interaction with dogs is one of the most relevant factors for the loss of individuals of the native fauna (^{Young et al. 2011}; ^{Hughes & Macdonald 2013}). It has been estimated that there are currently more than 700 million dogs (Canis lupus familiaris) in the world; in some regions the number of dogs exceeds the number of humans (^{Hughes & Macdonald 2013}). Domestic dogs can be fully dependent on human assistance ("pet") or, as result of a set of conditions, can deviate to live in wild populations ("feral"). About 75% of the world’s total domestic dog population is considered "free-ranging", i.e., owners breed them free and they usually return home only for feeding (^{Vanak & Gompper 2009}; ^{Hughes & Macdonald 2013}). These animals supplement their diet through hunting directly in the environment and, even when they are well fed, they can attack wild animals.

Studies on the impact of domestic dogs on native fauna have shown that their main impact is preying on native sympatric species (^{Young et al. 2011}). Besides, they can affect the behaviour of some animals such as activity pattern (^{Zapata-Ríos & Branch 2016}). There is also risk of disease transmission and hybridization (^{Hughes & Macdonald 2013}; ^{Lessa et al. 2016}). Another worrying fact is that dogs do not present a numerical response to their prey population decline (i.e. natural control), as would be the case of native predators. Even with decrease in populations of potential native preys, dogs can still find food by themselves near human settlements. The maintenance of dog’s populations by humans, regardless of the level of natural resources, guarantees the continuity of pressure on preys and competition with predators, which can lead to the local extirpation of native animals (^{May & Norton 1996}; ^{Galetti & Sazima 2006}).

Most of the information available regarding domestic dogs in nature reserves comes from developed countries, particularly North America and Europe, where dogs are usually well managed and cared for. Conversely, in developing country reserves information about dogs are comparatively scarce, particularly in terms of their interactions with local wildlife, despite their high abundance and widespread distribution (^{Weston et al. 2014}). Therefore, little is known about the presence of these animals in most Brazilian protected areas (e.g. ^{Lessa et al. 2016}), and even less about the invasion impacts of these species. Therefore, understanding the population abundance, density and spatial distribution of a species is essential to be effective for wildlife conservation planning (^{Cove et al. 2013}), especially when we are dealing with an exotic invasive species such as the domestic dog.

We aimed to characterize domestic dog spatial distribution using occupancy modelling and estimate abundance in the Mata das Flores State Park (MFSP). For this, we tested which elements would affect positively and negatively the presence of domestic dogs in MFSP. Based on the hypothesis that domestic dogs are strongly associated with human presence (^{Vanak & Gompper 2009}; ^{Hughes & Macdonald 2013}), we expected that the presence of domestic dogs would increase with the proximity to urban infrastructure (roads) and human settlements. We also expected a higher occurrence in areas closest to forest edges.

METHODS

Study area

The study was carried out in the Mata das Flores State Park (MFSP), a protected area by law located in the municipality of Castelo, Espírito Santo, Brazil (Fig. 1). The MFSP has a total area of approximately 800 hectares of irregular outline, located adjacent to the urban area of Castelo municipality (20°35’14" to 20°38’18" S and 041°09’21" to 041°12’09" W). The park is enclosed in the Atlantic Rainforest biome, and there are two dominant vegetation types (sensu ^{IBGEIBGE [Instituto Brasileiro de Geografia e Estatistica], 2012}): (1) the main vegetation in the lowlands is composed of a dry forest classified as Semi-Deciduous Seasonal Forest from the Lowlands; and (2) the small highland area is a wet forest classified as Dense Ombrophilous Montane Forest, with some rocky outcrops (^{IBGEIBGE 2012}). The MFSP presents an elevation ranging from 150 to 450 m asl, but with a small area represented by mountains ranging from 500 to 750 m (^{IEMA 2016}).

Data survey

We collected data for a period of five months (January May 2017) by dividing into two sampling protocols: (1) interviews and active search on trails and roads in the region and (2) sand plots to detect the presence of dogs in the MFSP.

Interviews and active search

We obtained the number of dogs whose owners live in the study area through visiting all houses in each area and interviewing their residents. We visited a total of 38 houses, 18 houses inside the MFSP area and 20 houses in the surroundings. For the surroundings, we consider houses located within a maximum radius of 1 km away from the boundaries of the MFSP. To check the presence of stray dogs whose owners do not live in the area we searched in five trails and one dirt road that intercept the MFSP. Throughout the span of this study we walked 38 km within the area monthly, totaling 190 km. Trails and dirt road were walked twice by two observers for a 3-day monthly period. The observers walked during two different times in the day: in the early morning from 6:30 to 08:30, and in the end of the day period from 15:30 to 17:30 according to other studies that performed transections (^{Chiarello 2000}; ^{Ferreguetti et al. 2017}; ^{Pereira et al. 2017}) and activity patterns of domestic dogs (^{Zapata-Ríos & Branch 2016}). All dogs found were georeferenced and photographed for identification by the local residents.

Sand plots

To estimate occupancy and detection probabilities of Canis lupus familiaris, we randomly installed 20 stations consisting of plots of sand established on the ground to record the presence of the dogs (Fig. 1). The stations were set with a minimum distance of 500 m among them. Each station had four sand plots of 1 x 1 m, 50 m distant from each other. The sampling stations were simultaneously checked for three consecutive days each month. The presence of footprints in a station in each inspection day was considered a record of dog occurrence at that place, independently of the number of footprints and individuals that might have left them. During each visit, the surface of the sand was smoothed to erase the marks of the previous footprints to allow the recording of new detections.

Data analysis

To produce an estimate of domestic dog abundance in the MFSP, we quantified all the dogs of the surrounding houses through interviews and individuals recorded during the active search. To have an estimated density value, we divided the total of individuals registered only within the MFSP by the total area of the park (800 hectares). To evaluate the spatial distribution of the species in the MFSP, we used the geographic coordinates of each observation of domestic dog to construct a distribution of occurrences map including the frequency of dogs recorded using QGIS software (^{QGIS Development Team 2017}).

We used single-season, single-species occupancy models to investigate per-survey detection probability (p), site occupancy (Ψ), and the effects of three spatial covariates on these parameters (^{MacKenzie et al. 2002}; ²⁰⁰⁶). We assumed that sites were closed to changes in occupancy during the study season (i.e. five months of sampling), and that detections occurred independently at sites. We considered each sampling station as an independent site (i.e. 20 sites) and each day of sampling with independent survey (i.e. 15 occasions). Using QGIS software (^{QGIS Development Team 2017}) and satellite imagery from Google Earth (Google, Inc., Mountain View, CA) dated May 2016, we estimated three spatial covariates: distances from the human settlement, forest edge, and roads.

We conducted all modeling in program R (version 3.1.2, http://www.r-project.org, accessed 09 Aug 2017) using the package Unmarked (^{Fiske & Chandler 2011}). We compared candidate models using Akaike’s Information Criterion corrected for small sample size (AICc; ^{Burnham & Anderson 2002}). We first modeled the effects of different site covariates on detection probability, keeping probability of occupancy constant to determine important detection covariates for the occupancy models. We then used only the most supported detection covariates in a suite of models to examine the effects of different landscape characteristics on site occupancy. We standardized all three covariates for modeling.

We evaluated eight candidate models for detection probability that included effects of the site covariates as a single covariate or as an additive model using a set of covariates. Holding this set of detection covariates constant, we compared eight candidate models containing the site covariates hypothesized to affect occupancy. All models with a ∆AICc value < 2 were considered to have a high level of support (^{Burnham & Anderson 2002}).

Fig. 1.  Map of Brazil, the state of Espírito Santo, the municipality of Castelo, and the location of Mata das Flores State Park.  

RESULTS

Abundance and spatial distribution

We recorded a total of 132 individually identified domestic dogs between interviews with surrounding residents and records through active search. We recorded an average of 2 ± 1 dogs per household interviewed. Of this total, we recorded 89 individuals inside the MFSP, through 213 detections in the five months of sampling. We estimated a density of 0.11 domestic dog per hectare (i.e. 89/800 hectares).

The spatial distribution map of the domestic dogs in the MFSP, showed a higher occurrence in the portion of the park faced to the urbanized area of the Castelo municipality, totaling about 53% of the records (Fig. 2). These records were mainly near to the border line of the reserve (Fig. 2). We also obtained a considerable number of records in the areas near the road that traverses the park, with about 36% of the records. In the portions of the reserve where the road traverses it the records of presence of dogs entered towards the interior of the forest area. The remaining 11% of the records occurred in areas close to human settlements around the park.

Occupancy and detection probabilities

Domestic dogs were observed in 14 of 20 sampling stations. From the eight detection models produced (Table 1), detection was best described by two covariates (1) distance from human settlements (human), and (2) distance from roads; both had a negative effect on detectability, with domestic dogs having a higher detection rate in sites close to human settlement and roads (Figs. 3A and 3B). From the eight occupancy models produced, only one covariate measured significantly described site occupancy (Table 1). The probability of a site being occupied by domestic dogs decreased with increasing distance from human settlements (dropping from Ψ = 0.81 to 0.02 across the observed range of distances; Fig. 3C).

DISCUSSION

The estimated domestic dog density that included only individuals recorded within the MFSP area shows the highest value of density of domestic dogs (0.11 dog per hectare, total of 89 individuals recorded within the MFSP) so far estimated for an Atlantic Forest protected area when compared to other sudies available. This high density indicates in turn a potential high intensity of impact by these exotic species on the native fauna in the reserve. For example, in an area of small native forest fragments (10–200 ha) immersed in a matrix predominantly composed of cacao plantations (Theobroma cacao) in Southern Bahia, Brazil, it was estimated 14.2 dogs per site studied, which represents a density of less than 0.02 dog per hectare (^{Santos et al. 2017}). Also, in two rural regions an area of São Paulo State, Brazil, it was estimated a density of 0.05 dog/ha (^{Torres & Prado 2010}). In a reserve of Atlantic Forest in Minas Gerais, Brazil, the estimated densities of dogs were 0.015 (^{Paschoal et al. 2012}) and 0.016 dog/ha (^{Paschoal et al. 2016}), respectively. The high density of dogs we found on MFSP may be related to the fact that this reserve has not been completely expropriated and is under the effect of the many rural properties within the park and on its surroundings. Additionally, ^{Paschoal et al. (2016)} found that domestic dog density was higher in small, private protected areas with intensified rural activities and closest to human settlements. Also, other studies showed that rural areas such as the region where the MFSP is located tend to be under higher density of domestic dogs moving inside them and which is also related to the culture of owning dogs (^{Butler & Bingham 2000}; ^{Kitala et al. 2001}). In fact, owning dogs seems to be the common rule in the study area, as all the inhabited houses had at least one dog.

Table 1 Single-season occupancy and detection models for Canis lupus familiaris in the Mata das Flores State Park, Brazil, estimated by sand plots for a period of five months (January - May 2017). Covariates: distance from forest edge (edge), distance from the road (road), and distance from human settlement (human). Ψ = occupancy, p = detectability, AICcw = Akaike weight, and K= number of parameters. 

In addition to the high density of dogs, we verified that the spatial distribution of the species is moreconcentrated in the regions close to the human settlements and near the road located in the MFSP. Probably this results from the proximity of the urban area together with the effect of the road that traverses MFSP, which facilitates the access for the dogs to seek for prey in the reserve. The species had a higher detection probability in sites closest to human settlements and roads and a much higher occupancy probability closest to human settlements. This relationship was expected because the entry of the species into protected areas is strongly linked to these two covariates. Several studies suggest that the occurrence of domestic dogs in many areas are closely linked to human densities and other anthropic factors such as roads and edges (^{Brooks 1990}; ^{Butler & Bingham 2000}; ^{Kitala et al. 2001}; ^{Reithinger et al. 2003}; ^{Lacerda et al. 2009}; ^{Torres & Prado 2010}; ^{Santos et al. 2017}). However, we found no relationship between the domestic dog and the distance from the edge as suggested by previous studies.^{Lacerda et al. (2009)}, for example, found that domestic dog occurrence was higher in areas closest to Brasília National Park edges located in Federal District, Brazil. Possibly we did not find a relationship of the domestic dog with the forest edge because the MFSP presents an irregular contour composed by three small fragments. In addition, the fact that a road intercepts the reserve makes it easier for domestic dogs to enter the interior areas of the MFSP.

Fig. 2. Map with the spatial distribution of the individuals of domestic dogs recorded in the Mata das Flores State Park,Espírito Santo, Brazil. Sampling from January 2017 to May 2017. Black stars representing human settlements locations. 

The highest density, together with the positive relationship between spatial distribution of domestic dogs and the number of households in the surroundings of the park, highlights the urgency of proposing management measures of this invasive species in MFSP. Dogs represent an important treat to biodiversity conservation worldwide, mainly due to wildlife predation, but also through environmental disturbance, disease transmission, competition and hybridization (^{Doherty et al. 2017}). In addition, domestic dogs can also affect activity patterns of potential competitors (^{Zapata-Ríos & Branch 2016}). Processes driving the negative effect of dogs on wildlife in the studied area are still unknown.

Certainly, a great contribution to this scenario is the fact that domestic dogs have a strong representation in human culture in general. However, in rural areas the problem goes beyond the culture of having a dog as a pet. Domestic dogs are bred unrestricted in rural settlements, facilitating their entry into native forest areas. In fact, ^{Kitala et al. (2001)} surveyed management practices and found that many dogs spend considerable time roaming freely, because 69% of the dogs were never restricted and only 19% of the households consistently restricted their dog’s movements. A high proportion of unrestricted dogs (77%) was also registered in a rural area and was also found in a study conducted in a major Mexican city (Merida) and three rural communities (^{Ortega-Pacheco et al. 2007}).

Our results of abundance, density and spatial may be useful to assist conservation planning and management in the MFSP. Because we provide a first estimate of domestic dog abundance and showed how they are distributed in the reserve, managers and researchers of the MFSP can could help to

develop management strategies of these individuals. Once we have mapped the distribution of domestic dogs and produced an abundance estimate for the MFSP, the next step could be to verify the impact that this high density of domestic dogs has on local vertebrate species, especially mammals, which tends to be the group of fauna most impacted by this invasive species (^{Galetti & Sazima 2006}). For this, itwould be necessary to characterize the community

of mammals of the region and to evaluate potential relationships among native and invasive species. As another important step in controlling the effects of domestic dogs in the MFSP it is strongly recommended to initiate a program of castration of males and sterilization of females in the region together with the awareness of the local population. However, this action will not be effective in reducing population growth as new individuals continue to enter the population of domestic dogs located in the park. To reduce this, it is necessary also to rely on the capture of individuals considered stray-dogs or free-ranging dogs.

Fig. 3. Detection and occupancy probabilities of domestic dog in the Mata das Flores State Park, Espírito Santo, Brazil. Estimated effect of (A) distance from human settlement and (B) distance from road on the detection probability, and (C) the distance from human settlement on the occupancy probability.