**3.2 Data collection and analysis**

Radio-telemetry data were obtained between April and September 2004. Trapping efforts were conducted from February to March 2004. Large Havahart single door box-traps (106.7 cm length x 38 cm width x 38 cm height) made from a combination of tensile wire mesh and steel were set within the study site in areas where red foxes or their tracks had been observed. Traps were baited with food for human consumption or wild meat and were checked every day. Captured foxes were anesthetized using Xylazine/Ketamine (1:10 mg/kg) and Atipamezole (1 mg per 10 mg of Xylazine; Animal Care Protocol, University of Prince Edward Island 03-043), and then radio-collared (TS-37 Telemetry Solutions; 50 g). The radio-telemetry procedure used in this study followed recommendations made by White & Garrott (1990). During the less active periods (morning and early afternoon), animal locations were estimated by one observer who

The study site selected for this study was located in Stanhope (543 ha; 46°25'E; 63°06'N), Prince Edward Island National Park (Fig. 1). Although situated within the park, Stanhope is surrounded by private houses and cottages, many of them inhabited throughout the whole year. It encompasses forest patches, marshes, shrubs, ponds, dunes, beaches, roads, agricultural fields, and human-use areas (about 0.30 human dwellings per ha). Although we do not possess quantitative data regarding fox-feeding intensity in Stanhope, anecdotal information and personal observations clearly indicate that fox-feeding is a common activity in Stanhope, with residents setting out food for foxes throughout the whole year and many

Fig. 1. Map of Prince Edward Island (Canada) in relation to Canada indicating the location of the study site (Stanhope) with a star. The scale-bar refers to Prince Edward Island only.

Radio-telemetry data were obtained between April and September 2004. Trapping efforts were conducted from February to March 2004. Large Havahart single door box-traps (106.7 cm length x 38 cm width x 38 cm height) made from a combination of tensile wire mesh and steel were set within the study site in areas where red foxes or their tracks had been observed. Traps were baited with food for human consumption or wild meat and were checked every day. Captured foxes were anesthetized using Xylazine/Ketamine (1:10 mg/kg) and Atipamezole (1 mg per 10 mg of Xylazine; Animal Care Protocol, University of Prince Edward Island 03-043), and then radio-collared (TS-37 Telemetry Solutions; 50 g). The radio-telemetry procedure used in this study followed recommendations made by White & Garrott (1990). During the less active periods (morning and early afternoon), animal locations were estimated by one observer who

**3.2 Data collection and analysis** 

tourists feeding these animals during the touristic season (i.e., summer).

recorded 2-3 azimuths within 15-min intervals over a period of 1 hr. During the active periods (late afternoon and early evening), simultaneous triangulation was performed by 2-3 observers within 15-min intervals over a period of 1 hr. In addition, we also conducted 8-hr intensive (1 location every 15 min) telemetry sessions (sequential locations) twice per month (one in the afternoon: 13:00 to 21:00 and one during the night: 21:00 p.m. to 05:00 a.m.). These telemetry sessions were subjected to weather conditions, thus sometimes they were shorter than 8 hr. Tracking was done on foot using hand-held Yagi antennas and portable receivers (R-1000; Communications Specialists, Inc.). Triangulation angles were maintained between 30° and 150° (Gese, 2001). About 50% of radio-tracking locations were taken with the animal in view of the observers. Bearings were plotted immediately using LOAS 2.1 (Ecological Software Solutions 2003) to determine the accuracy of the locations. For the purposes of this study, we have not looked for statistical independence of locations (e.g., Swihart & Slade, 1985), but rather for their biological independence, using a minimum time interval between successive locations long enough to allow any radio-collared animal to cross entirely its home range. Repeated observations on the same individuals, as is the case with radio-telemetry locations, are often assumed to give rise to constant within-group correlation structures (i.e., lack of independence in the data). Although some ecologists and wildlife managers think that correlation structures represent a problem for the analysis and interpretation of wildlife telemetry data (Hansteen et al., 1997), others suggest that much can be learned by studying the causes and consequences of correlation structures in telemetry data. If a large time interval between successive locations is possible and/or the calculation of home-range size is the main goal of the study, strict adherence to the collection of non-autocorrelated data may be necessary. However, it may be difficult to translate autocorrelated data into an independent form and still retain a sample size that is adequate for the home-range size to reach an asymptote. Thus, the goals of most home-range or habitat selection studies require the collection of data which are dependent to some degree. This was the case in this study.

Habitat variables were determined using 2000 Prince Edward Island aerial photographs (1: 17500; Prince Edward Island Department of Agriculture and Forestry) of the study site, and complemented with field observations to update any land-cover changes. ArcView GIS (version 3.3; Environmental Systems Research Institute, Inc., ESRI) was used to map animal locations obtained from radio-telemetry monitoring sessions and to assign to each location a habitat cover-type. Habitat was classified into 1 of 10 cover-types: agriculture (hay fields and pasture), forest (mature white spruce and hardwood forest), water or aquatic systems (frozen during winter; including ponds and saltwater bodies), dunes, beaches, roads (mostly paved), human-use areas (residential areas, recreational areas, parking lots), abandoned fields, marshes, and shrubs.

Habitat selection was examined by comparing use and availability of habitat types within the study area using the Neu Method (Neu et al., 1974). The Neu method is a straightforward application of the χ2 goodness-of-fit test, and is usually used to compare observed counts of animals in each habitat with the counts expected if habitats were used in proportion to their availability. The method involves the calculation of confidence intervals (Bonferroni Z-statistic) around the expected proportions to determine whether the observed proportion of usage in each habitat is significantly different from expected. The usage of a particular habitat type was defined as the ratio between animal locations in each habitat type and the total number of locations recorded in the study area. Expected

Use of Telemetry Data to Investigate

Home Range and Habitat Selection in Mammalian Carnivores 289

Fig. 2. Home range (95% fixed kernel analysis) and core areas (50% fixed kernel analysis) of F1, F3, and M2. Core areas are contained within the home range boundary. Animal locations are indicated by solid black circles. Spatial overlap of the home ranges of the radio-tracked

All three red foxes used all available habitats in Stanhope, except for the abandoned field habitat. At the home-range spatial scale, red foxes selected for agricultural fields, dunes, roads, and human-use areas, while forest, marsh, and water habitat were used less than expected based on their availability (Table 2). No significant preference was observed for beach habitat. At the core-area spatial scale, red foxes selected only for dunes and roads, while beach, forest, marsh, and water habitats were used less than expected based on their availability. No significant preference was observed for human-use, agriculture and shrub

red foxes in Stanhope is shown in the bottom panel.

habitats at this spatial scale.

usage of a habitat type was defined as the ratio of the area of the particular habitat type to the total area of the study site. The study-site area was defined at the home-range spatial scale using the smallest rectangle that included all 95% fixed-kernel home-ranges (see below for more information about the fixed-kernel home-range method), and at the corearea spatial scale using the smallest rectangle that included all the 50% fixed-kernel home ranges (sensu Kazmaier et al., 2001). This corresponds to a design-2 analysis of habitat selection according to Johnson (1980) because individuals could be identified using radiotelemetry data. ArcView GIS (version 3.3; ESRI) was used to calculate all the study site areas, as well as the availability of the different habitat types comprised within each study site.

Data on home-range size and core-areas were analyzed using the Animal Movement SA version 2.0 in ArcView (version 3.3; ESRI). The minimum number of locations required to accurately assess the home-range size of each animal was estimated by plotting cumulative home-range sizes against the number of locations (i.e. asymptotic home-range; Phillipps & Catling, 1991). The minimum convex polygon (MCP; Mohr, 1947) and the 95% fixed-kernel (Seaman & Powell, 1996) methods were used to determine home-range areas. The 100% MCP was utilized because it is the most commonly reported method in the literature (Harris et al., 1990), and therefore allows for some comparison with other studies. The 95% fixed-kernel method, while not without problems, has shown the best performance in simulation trials of home-range estimators that also included MCP. The 50% fixed-kernel method was used to estimate the size and shape of the core-areas or centers of activity within home-ranges. Fixed-kernel analyses were performed with a bandwidth calculated using least-squares cross validation (Powell, 2000; Seaman et al., 1999). The overlap area in home-ranges between two individuals was estimated using ArcView (version 3.3; ESRI).

#### **3.3 Results**

Amongst the five adult red foxes captured in Stanhope, sufficient data to calculate homerange sizes were only obtained for three individuals (Table 1). Asymptotic home-range was achieved with 140 locations for both females, and with 40 locations for the male. Using the 100% MCP method, home-ranges varied between 105.7 ha and 168.8 ha while the 95% fixedkernel method resulted in home-ranges that varied between 77.4 ha and 131.3 ha (Table 1; Fig. 2). The 50% fixed-kernel method resulted in core-area values that ranged between 7.4 ha and 13.2 ha, representing about 10% of each animal's home-range (Fig. 2). Home ranges of all three foxes overlapped to some extent, with the greatest overlap (63 ha) occurring between F3 and M2. The overlap in home ranges between F1 and M2 was 8.6 ha while between F3 and F1 was 3.2 ha.


Table 1. Radio-telemetry data collected from three red foxes captured in Stanhope, Prince Edward Island National Park (Prince Edward Island, Canada).

usage of a habitat type was defined as the ratio of the area of the particular habitat type to the total area of the study site. The study-site area was defined at the home-range spatial scale using the smallest rectangle that included all 95% fixed-kernel home-ranges (see below for more information about the fixed-kernel home-range method), and at the corearea spatial scale using the smallest rectangle that included all the 50% fixed-kernel home ranges (sensu Kazmaier et al., 2001). This corresponds to a design-2 analysis of habitat selection according to Johnson (1980) because individuals could be identified using radiotelemetry data. ArcView GIS (version 3.3; ESRI) was used to calculate all the study site areas, as well as the availability of the different habitat types comprised within each study

Data on home-range size and core-areas were analyzed using the Animal Movement SA version 2.0 in ArcView (version 3.3; ESRI). The minimum number of locations required to accurately assess the home-range size of each animal was estimated by plotting cumulative home-range sizes against the number of locations (i.e. asymptotic home-range; Phillipps & Catling, 1991). The minimum convex polygon (MCP; Mohr, 1947) and the 95% fixed-kernel (Seaman & Powell, 1996) methods were used to determine home-range areas. The 100% MCP was utilized because it is the most commonly reported method in the literature (Harris et al., 1990), and therefore allows for some comparison with other studies. The 95% fixed-kernel method, while not without problems, has shown the best performance in simulation trials of home-range estimators that also included MCP. The 50% fixed-kernel method was used to estimate the size and shape of the core-areas or centers of activity within home-ranges. Fixed-kernel analyses were performed with a bandwidth calculated using least-squares cross validation (Powell, 2000; Seaman et al., 1999). The overlap area in home-ranges between two individuals was estimated using

Amongst the five adult red foxes captured in Stanhope, sufficient data to calculate homerange sizes were only obtained for three individuals (Table 1). Asymptotic home-range was achieved with 140 locations for both females, and with 40 locations for the male. Using the 100% MCP method, home-ranges varied between 105.7 ha and 168.8 ha while the 95% fixedkernel method resulted in home-ranges that varied between 77.4 ha and 131.3 ha (Table 1; Fig. 2). The 50% fixed-kernel method resulted in core-area values that ranged between 7.4 ha and 13.2 ha, representing about 10% of each animal's home-range (Fig. 2). Home ranges of all three foxes overlapped to some extent, with the greatest overlap (63 ha) occurring between F3 and M2. The overlap in home ranges between F1 and M2 was 8.6 ha while

Telemetry locations recorded 172 368 63 100% Minimum convex polygon (ha) 121.6 168.8 105.7 95% Fixed-kernel (ha) 99.5 131.3 77.4 50% Fixed-kernel (ha) 13.2 10.5 7.4

Table 1. Radio-telemetry data collected from three red foxes captured in Stanhope, Prince

Edward Island National Park (Prince Edward Island, Canada).

**F1 (♀) F3 (♀) M2 (♂)** 

site.

ArcView (version 3.3; ESRI).

between F3 and F1 was 3.2 ha.

**3.3 Results** 

Fig. 2. Home range (95% fixed kernel analysis) and core areas (50% fixed kernel analysis) of F1, F3, and M2. Core areas are contained within the home range boundary. Animal locations are indicated by solid black circles. Spatial overlap of the home ranges of the radio-tracked red foxes in Stanhope is shown in the bottom panel.

All three red foxes used all available habitats in Stanhope, except for the abandoned field habitat. At the home-range spatial scale, red foxes selected for agricultural fields, dunes, roads, and human-use areas, while forest, marsh, and water habitat were used less than expected based on their availability (Table 2). No significant preference was observed for beach habitat. At the core-area spatial scale, red foxes selected only for dunes and roads, while beach, forest, marsh, and water habitats were used less than expected based on their availability. No significant preference was observed for human-use, agriculture and shrub habitats at this spatial scale.

Use of Telemetry Data to Investigate

inhabiting the study site.

access to humans.

Home Range and Habitat Selection in Mammalian Carnivores 291

more stable home ranges (Joshi et al., 1995). Alternatively, it is possible that home ranges were small because Stanhope supports a high density of red foxes due to high fox-feeding levels. Small home ranges would allow foxes to cover their territories in a relatively short time to maintain exclusive rights to the areas and reduce intraspecific competition (Baker et al., 1998). Although this has not been investigated yet, anecdotal information suggests that there may be a correlation between fox numbers and the overabundance of anthropogenic resources within certain sectors of Prince Edward Island National Park. Observations made by wardens from the park suggest that red-fox abundance in Stanhope has actually increased during the last years. Although it is logical to expect that deaths caused by vehicles can reduce foxes abundance in the park, they can also incite an increase in the reproductive output or productivity of the fox population. Some predator species compensate for high mortality levels resulting from exploitation (e.g., trapping, hunting, etc.) by increasing their litter size or reproductive output (e.g., van Deelen & Gosselink, 2006). Although an accidental death caused by a vehicle is not the same than exploitation, it may contribute to the adverse effects that other human activities in the area or nearby (e.g., tourism, trapping, farming) may have on the red-fox population

Red foxes are typically considered habitat generalists, and their adaptability to marginal or degraded ecological conditions has been a key factor in allowing them to survive in fragmented landscapes and human-dominated areas. However, this does not necessarily mean that they do not have preferences for certain habitat types. For instance, red foxes did not show much preference for habitats such as forests, abandoned fields, and shrubs, in which they would normally have to hunt for prey such as small mammals. In this study, we found that red foxes used both human-use areas and dune habitats during the study period which corresponds to the kit-rearing season for this species. Our analyses also indicate that in Prince Edward Island, red foxes seem to avoid forests even though this habitat is the most available habitat type. Red foxes showed a preference for dune habitats in the study site at both large (home range) and small (core area) spatial scales, providing support for anecdotal information that suggest that it is in the dunes where their dens are located. The granular and sandy composition of dune habitats likely provides suitable habitat for excavation of dens. In addition, dune habitats are protected within the national park, thus limiting their

Our analyses also showed that the value of certain habitat cover types to red foxes varies with the spatial scale at which habitat-selection is investigated. For instance, even though red foxes used human-use areas at the large spatial scale (home range), they did not significantly select for this habitat at the small spatial scale (core area). At the small spatial scale, our results indicated that foxes selected for dunes and roads. In our study site, foxes use roads to move between different habitat cover-types and to obtain food. Fox-feeding along roads is a common human activity during the kit-rearing season. Indeed, it is very common to see many red foxes of all ages sitting along the roads waiting for humans to feed them. During the kit-rearing season, although humans inhabiting cottages and houses still set out food for foxes, these animals must find "foraging" along the roads more efficient. During this season, it is not only easy to obtain food along the roads without too much effort, but roads are situated very close to the dunes where it is suspected that their dens are located. Parental care is particularly demanding in terms of food resources, so it may be more energetically efficient to stay close to the dens. Thus, our findings provide evidence suggesting that anthropogenic food resources are


Table 2. Comparisons of use and availability of various habitat types in Stanhope (Prince Edward Island National Park) based on telemetry data from three red foxes. Habitat types are shown in order of preference. Comparisons were made using the Neu Method and Bonferroni confidence intervals. Preference was determined with Bonferroni confidence intervals (*α* = 0.001) placed on use. "+" indicates used more than expected; "-" indicates used less than expected; "0" use according to availability or non-significant difference between expected and available.

#### **3.4 Discussion**

Although home-ranges and core areas of red foxes inhabiting the study site were comparable to values observed in studies conducted in other parts of North America and Europe (e.g. Adkins & Stott, 1998; Voigt & Macdonald, 1984), they were at the lower end of the spectrum for this species. There are several possible explanations for these results. It is possible that our findings reflect the habitat requirements and movement patterns of foxes during the kit-rearing season when foxes are involved in cub rearing and lactation (especially females) and need to stay close to their dens (e.g. Saunders et al., 1995). If this is true, it is possible to assume that home ranges and core areas obtained during this study are underestimations of values at other times of the year. This hypothesis was supported by the fact that one of the females (F3) was observed rearing three pups and receiving help from the male (M2). This observation would also explain the considerable overlap in home range areas found between F3 and M2. Another possible explanation for our findings is that the small home ranges observed in this study are the result of the aggregated distribution of anthropogenic food sources. It has been shown that the distribution of non-territorial, solitary carnivores tend to become more aggregated when anthropogenic resources are concentrated into a few patches, resulting in smaller and

Human-use 300 0.517 0.077 0.114 + Dune 86 0.148 0.054 0.053 + Road 64 0.110 0.048 0.033 + Agriculture 31 0.053 0.034 0.005 + Beach 41 0.071 0.039 0.100 0 Shrub 15 0.026 0.024 0.003 0 Marsh 1 0.002 0.006 0.109 0 Forest 39 0.067 0.038 0.503 - Water 3 0.005 0.011 0.060 - Abandoned Field 0 0 0 0.019 -

Dune 67 0.249 0.056 0.118 + Road 16 0.172 0.036 0.084 + Human-use 103 0.369 0.062 0.394 0 Shrub 8 0.035 0.026 0.028 0 Agriculture 11 0.012 0.007 0.008 0 Beach 18 0.089 0.048 0.161 - Forest 27 0.067 0.028 0.142 - Marsh 1 0.003 0.004 0.01 - Water 1 0.003 0.008 0.054 - Table 2. Comparisons of use and availability of various habitat types in Stanhope (Prince Edward Island National Park) based on telemetry data from three red foxes. Habitat types are shown in order of preference. Comparisons were made using the Neu Method and Bonferroni confidence intervals. Preference was determined with Bonferroni confidence intervals (*α* = 0.001) placed on use. "+" indicates used more than expected; "-" indicates used less than expected; "0" use according to availability or non-significant difference between

Although home-ranges and core areas of red foxes inhabiting the study site were comparable to values observed in studies conducted in other parts of North America and Europe (e.g. Adkins & Stott, 1998; Voigt & Macdonald, 1984), they were at the lower end of the spectrum for this species. There are several possible explanations for these results. It is possible that our findings reflect the habitat requirements and movement patterns of foxes during the kit-rearing season when foxes are involved in cub rearing and lactation (especially females) and need to stay close to their dens (e.g. Saunders et al., 1995). If this is true, it is possible to assume that home ranges and core areas obtained during this study are underestimations of values at other times of the year. This hypothesis was supported by the fact that one of the females (F3) was observed rearing three pups and receiving help from the male (M2). This observation would also explain the considerable overlap in home range areas found between F3 and M2. Another possible explanation for our findings is that the small home ranges observed in this study are the result of the aggregated distribution of anthropogenic food sources. It has been shown that the distribution of non-territorial, solitary carnivores tend to become more aggregated when anthropogenic resources are concentrated into a few patches, resulting in smaller and

**Habitat Use proportion Expected Use** 

**proportion Preference Average SD** 

**(# locations)** 

**Habitat Observations**

*Home-range Level* 

*Core-Area Level* 

expected and available.

**3.4 Discussion** 

more stable home ranges (Joshi et al., 1995). Alternatively, it is possible that home ranges were small because Stanhope supports a high density of red foxes due to high fox-feeding levels. Small home ranges would allow foxes to cover their territories in a relatively short time to maintain exclusive rights to the areas and reduce intraspecific competition (Baker et al., 1998). Although this has not been investigated yet, anecdotal information suggests that there may be a correlation between fox numbers and the overabundance of anthropogenic resources within certain sectors of Prince Edward Island National Park. Observations made by wardens from the park suggest that red-fox abundance in Stanhope has actually increased during the last years. Although it is logical to expect that deaths caused by vehicles can reduce foxes abundance in the park, they can also incite an increase in the reproductive output or productivity of the fox population. Some predator species compensate for high mortality levels resulting from exploitation (e.g., trapping, hunting, etc.) by increasing their litter size or reproductive output (e.g., van Deelen & Gosselink, 2006). Although an accidental death caused by a vehicle is not the same than exploitation, it may contribute to the adverse effects that other human activities in the area or nearby (e.g., tourism, trapping, farming) may have on the red-fox population inhabiting the study site.

Red foxes are typically considered habitat generalists, and their adaptability to marginal or degraded ecological conditions has been a key factor in allowing them to survive in fragmented landscapes and human-dominated areas. However, this does not necessarily mean that they do not have preferences for certain habitat types. For instance, red foxes did not show much preference for habitats such as forests, abandoned fields, and shrubs, in which they would normally have to hunt for prey such as small mammals. In this study, we found that red foxes used both human-use areas and dune habitats during the study period which corresponds to the kit-rearing season for this species. Our analyses also indicate that in Prince Edward Island, red foxes seem to avoid forests even though this habitat is the most available habitat type. Red foxes showed a preference for dune habitats in the study site at both large (home range) and small (core area) spatial scales, providing support for anecdotal information that suggest that it is in the dunes where their dens are located. The granular and sandy composition of dune habitats likely provides suitable habitat for excavation of dens. In addition, dune habitats are protected within the national park, thus limiting their access to humans.

Our analyses also showed that the value of certain habitat cover types to red foxes varies with the spatial scale at which habitat-selection is investigated. For instance, even though red foxes used human-use areas at the large spatial scale (home range), they did not significantly select for this habitat at the small spatial scale (core area). At the small spatial scale, our results indicated that foxes selected for dunes and roads. In our study site, foxes use roads to move between different habitat cover-types and to obtain food. Fox-feeding along roads is a common human activity during the kit-rearing season. Indeed, it is very common to see many red foxes of all ages sitting along the roads waiting for humans to feed them. During the kit-rearing season, although humans inhabiting cottages and houses still set out food for foxes, these animals must find "foraging" along the roads more efficient. During this season, it is not only easy to obtain food along the roads without too much effort, but roads are situated very close to the dunes where it is suspected that their dens are located. Parental care is particularly demanding in terms of food resources, so it may be more energetically efficient to stay close to the dens. Thus, our findings provide evidence suggesting that anthropogenic food resources are

Use of Telemetry Data to Investigate

2007b).

al., 2007a, 2007b).

**4.1 Study site** 

such as Mkhuze Game Reserve.

above the ground (Hockey et al., 2005).

Home Range and Habitat Selection in Mammalian Carnivores 293

Lucia Wetland Park) which is approximately 3,320 km2. Regardless of its small size, Mkhuze Game Reserve supports a very diverse mammalian community that includes four of the five big-game animals expected to occur in the area: leopard (*Panthera pardus*), African elephant (*Loxodonta africana*), black rhinoceros (*Diceros bicornis*), and buffalo (*Syncerus caffer*). Wild dogs were present in Mkhuze Game Reserve until the 1930s. In this game reserve, the reintroduction of wild dogs began in 2005 as part of the Priority Species Monitoring Project. In 2004, thirteen wild dogs originating from two other South African conservation areas (Marakele National Park and Madikwe Game Reserve) were placed together in two adjoining bomas in Mkhuze Game Reserve with the purpose of bonding all the animals into one pack. Boma construction was fundamental to ensure animals were exposed to electrified fencing (Mkhuze Game Reserve is surrounded by electric fencing), habituated to game vehicles, allowed to settle, become accustomed to radiocollars and other conspecifics within a new social group, and finally ensure that territorial bonds were relaxed so they remained at the release location (Hayward et al., 2007a,

Although a variety of methods are used to assess the success of a species reintroduction program, a common recommended first step in most of these methods is to demonstrate that the species is adapting well to its new habitat during the establishment phase of the reintroduction (Hayward et al., 2007a, 2007b). Researchers generally accomplish this first step by examining habitat selection and home-range patterns, as the reestablishment of species in areas where they formerly occurred is often influenced by the suitability of habitats at proposed release sites (IUCN, 1998; Wolf et al., 1998*).* Without high quality habitats that provide adequate food, water and suitable places to forage and breed, reintroduction programmes have a low chance of success (Griffith et al., 1989; Hayward et

The main objective of this study was to examine how telemetry data can be used to quantify habitat selection and home-range patterns of a wild-dog pack during the establishment phase of a reintroduction program. We also expected that a better knowledge of the home range and habitat-selection patterns of reintroduced animals will help identify what resources and habitats are essential for the survival of wild dogs in small reserves or parks

Mkhuze Game Reserve is located between 32°06'25" to 32°56'46" E and 26°51'26" to 28°29'07" S in the subtropical zone (Fig. 3). The game reserve receives about 1,200 mm of rain annually, 60% of which falls in the summer. Mkhuze Game Reserve comprises a diverse array of habitat types, including grasslands, lakes and pans, wetlands, savannahs, thickets, woodlands and forests (van Rooyen, 2004). Two types of grasslands characterize the game reserve: lebombo-wooded grasslands and floodplains. Lebombo-wooded grasslands are mostly found on sandy soils near the bordering Lebombo Mountains but can also be present on soils composed of clay. The game reserve also includes several freshwater pans that although are mostly permanent bodies of water, may also undergo seasonal changes such as regular flooding and inundation (Van Rooyen, 2004). Although thickets and savannahs occur in various parts of the game reserve, most of the area is occupied by woodlands and forests. Woodlands have a discontinuous canopy while forests, also known as closedwoodlands, have a continuous canopy that commences at an elevation of 5 m or greater

important for foxes at both small and large spatial scales. It is, thus, possible to use our findings as indirect evidence supporting the idea that fox-feeding is altering the way that red foxes use habitats in the study site, and possibly other areas of Prince Edward Island National Park where fox-feeding is common.
