Diversity, abundance and the impact of hunting on large mammals in two contrasting forest sites in northern amazon

A variety of Amazonian mammals serve as sources of food for its human inhabitants, but hunting can have a strong negative impact on them. Diversity, abundance, biomass, and average group size of medium-sized and large mammals are compared across two forest areas of the northern Amazon: the Viruá National Park (protected) and the Novo Paraíso settlement (a human settlement where hunting is permitted). Hunting pressure was also characterized in Novo Paraíso. A total of 33 mammal species were recorded. There were no significant differences in the sighting rates, relative abundance and biomass, and mammal group sizes between the two areas, although the totals of all these variables were higher in Viruá due to the higher abundance of Tayassu pecari, which was not recorded at Novo Paraíso. It is suggested that T. pecari may be on the verge of local extinction, as it was the most hunted species in the settlement area. Through interviews with 50 hunters, we estimate that 541 mammals of 20 species were hunted during the study year, resulting in an estimated biomass take of 8517 kg. While the hunting intensity in Novo Paraíso may be sustainable in the short term, the reported decline of hunting efficiency, combined with the extirpation of T. pecari, suggests that mammal abundance may decline there in the near future. In the study year, 849 hunts were carried out in a hunting effort of 4575 hours, with a maximum distance travelled of 5.4 km. There was an average of 4.82 consumers for each hunt, and a per capita harvest rate of 2.24 individuals/consumer year. Hunting was not only for subsistence, but also for retaliation, although some species may not be hunted due to cultural taboos. The need for quantification of harvesting rates to maintain hunting at sustainable levels is highlighted.


Diversity, abundance and the impact of hunting on large mammals in two contrasting forest sites in northern amazon
A variety of Amazonian mammals serve as sources of food for its human inhabitants, but hunting can have a strong negative impact on them. Diversity, abundance, biomass, and average group size of medium-sized and large mammals are compared across two forest areas of the northern Amazon: the Viru á National Park (protected) and the Novo Para í so settlement (a human settlement where hunting is permitted). Hunting pressure was also characterized in Novo Para í so. A total of 33 mammal species were recorded. Th ere were no signifi cant diff erences in the sighting rates, relative abundance and biomass, and mammal group sizes between the two areas, although the totals of all these variables were higher in Viru á due to the higher abundance of Tayassu pecari, which was not recorded at Novo Para í so. It is suggested that T. pecari may be on the verge of local extinction, as it was the most hunted species in the settlement area. Th rough interviews with 50 hunters, we estimate that 541 mammals of 20 species were hunted during the study year, resulting in an estimated biomass take of 8517 kg. While the hunting intensity in Novo Para í so may be sustainable in the short term, the reported decline of hunting effi ciency, combined with the extirpation of T. pecari , suggests that mammal abundance may decline there in the near future. In the study year, 849 hunts were carried out in a hunting eff ort of 4575 hours, with a maximum distance travelled of 5.4 km. Th ere was an average of 4.82 consumers for each hunt, and a per capita harvest rate of 2.24 individuals/ consumer year. Hunting was not only for subsistence, but also for retaliation, although some species may not be hunted due to cultural taboos. Th e need for quantifi cation of harvesting rates to maintain hunting at sustainable levels is highlighted.
Th e Brazilian Amazon represents over a third of all tropical forest on Earth (Peres 2000), but vast tracts of primary forest are being degraded by deforestation and fragmentation (Laurance and Peres 2006). As a result, much of the remaining primary forest is surrounded by human modifi ed habitats (Stone et al. 2009).
While secondary tropical forests can sustain a high abundance of wildlife, in many tropical forest landscapes the conservation value of habitat mosaics have often been severely compromised by the hunting practices of local people (Parry et al. 2007, Peres andPalacios 2007). Wild animals are an important part of the diet of many inhabitants of the Amazon forest, but hunting can have serious negative impacts on vertebrate populations (Peres 1990, Alvard et al. 1997). In addition, Amazonian peoples have become more sedentary over the last decades (Vickers 1991), and the increasing availability of fi rearms and more effi cient transport continues to intensify hunting pressure (Souza-Mazurek et al. 2000). In this context, large mammals are among the most hunted species (Redford and Robinson 1987, Bodmer 1995, Peres 2000. Th e animal mortality rate in an anthropogenic landscape is strongly related to the spatial distribution of the hunting eff ort (Sir é n et al. 2004). Overall, hunting areas located near villages and settlements in Amazonia are more frequent than more distant hunting areas (Vickers 1984, Alvard 1992, Peres and Nascimento 2006, Ohl-Schacherer et al. 2007, Parry et al. 2009). According to Alvard et al. (1997), about 87% of the animals eaten by a native community are hunted within a 10-km radius of the village, which seems to be the standard in the Neotropics (Vickers 1984, Alvard 1992, Ohl-Schacherer et al. 2007, Parry et al. 2009). As a result, the abundance of game species in areas surrounding humanoccupied sites becomes gradually less, and hunters have to travel further to capture target species (Ohl-Schacherer et al. 2007).
In continuous forest areas, where the persistence of populations in the sink habitat depends on their migration from high quality source-habitats (Pulliam 1988), the establishment of reserve areas may serve as refuges for prey populations (Fragoso et al. 2000, Bodmer andRobinson 2004), which is vital for the persistence of the hunted species (Peres 2001). Additionally, primary forests far from human settlements are highly eff ective in maintaining biodiversity, even if not offi cially protected (Souza-Mazurek et al. 2000, Peres and Palacios 2007. Hunters usually target larger prey (Emlen 1966, MacArthur andPianka 1966), although when the preferred species are exhausted the range of hunted species increases as hunters are forced to target less valuable prey (Alvard 1994, Jerozolimski andPeres 2003). Hunting pressure, nevertheless, may be moderated by taboos and prey preferences (McDonald 1977, Colding 1998. In this sense, the diff erent cultural aspects of each human population can aff ect wildlife on diff erent scales. It is necessary to know the species chosen or avoided (and why), hunting techniques used, number of animals harvested, and motivation or purpose for hunting in order to both determine the impact of this activity, and promote conservation and sustainable management (Trinca and Ferrari 2006).
Studies of large and medium-sized mammals in the State of Roraima are relatively recent and the focus has been mainly on the mammals of Marac á Ecological Station (Fragoso 1998, Mendes Pontes 1997) and the Waimiri-Atroari (Souza-Mazurek et al. 2000), Yanomami and Macuxi Indian reserves (Fragoso 2004); whereas studies in recently colonized settlement areas are virtually absent. Given the limited number of studies, the diversity of mammals, the impact of hunting and the types of wildlife used in these settlements of Roraima remain virtually unknown. Th e aim in this study was, therefore, to compare the diversity of medium and large-sized mammals in two areas of the State of Roraima, one protected (Viru á National Park) and one impacted by hunting (Novo Para í so settlement), and to characterize and quantify hunting pressure to determine the impact of hunting on the mammalian community.

Study area
Th is study took place in two study sites some 100 km apart, in the northernmost part of the Brazilian Amazon, where the annual mean and maximum temperatures are approx. 26 ° C and 40 ° C, respectively (Sombroek 2001). Th e average annual rainfall is 1500 mm, with the rainy season occurring between April and August. During the seven months of the dry season, rainfall is less than 100 mm (Sombroek 2001). Th e studied region in the south of the State of Roraima is formed by pristine highly heterogeneous vegetation mosaics formed by open forest formations ( ' Campina ' and ' Campinarana ' ) and closedcanopy ombrophilous forests. Although serving as a source forest to the impacted areas, most of these pristine forests are not legally protected and are totally accessible, the only factor preventing the human presence being remoteness (Fig. 1).

Novo Para í so settlement (impacted area)
Th e Novo Para í so settlement (01 ° 13 ′ 24.59 ′′ N; 60 ° 23 ′ 6.27 ′′ W) (Fig. 1) is a fi sh-bone human settlement (forest clearings that appear in satellite images in a fi shbone pattern, with forest clearance extending along secondary roads from the main road). Th e settlement was founded in 1982 by the Government of the State of Roraima in what was once pristine forests (Governo do Estado de Roraima 2005). It is primarily populated by non-indigenous settlers from the north and northeast of Brazil, who were attracted by the free land off ered by the Federal Government as part of an Amazon colonization program initiated in the late 1970s.
Located at km 500 on the BR-174 highway, the settlement has an area of 92.84 km ² (Governo do Estado de Roraima 2005) and approximately 120 families living in its urban center, the Novo Para í so village. Dozens of families, however, living on plots of land along the BR-432, BR-210 and BR-174 highways and adjacent side roads depend on this village for the purchase of supplies and access to health and education services.
Th e population has a low income, the main source of which is livestock and smallholder farming. As a result, frequent deforestation and forest fi res occur both inside and outside their plots to clear the land for pasture and planting crops. Because these plots are demarcated within the forest, it is easily accessed and hunting is facilitated (E. R. A. Melo unpubl.), which is practiced indiscriminately and without any control by the local government. Additionally, settlers are entitled to ' subsistence hunter's permits ' granted on request by the local offi ce of the Federal Police.

Viru á National Park (protected)
Located 70 km to the northeast of the Novo Para í so settlement, the Viru á National Park (headquarters: 01 ° 42 ′ 25 ′′ N; 61 ° 10 ′ 24 ′′ W) (Fig. 1) is a federal conservation area created in 1998. With a total area of 2 270.11 km ² (ICMBio 2011), the Park is located in a region of relatively fl at terrain, with altitudes between 45 and 326 m a.s.l. (Governo do Estado de Roraima 2005).
Viru á contains one of the 25 km ² RAPELD (long-term ecological research) grids of the PPBio (Biodiversity Research Program), consisting of 12 trails each 5 km in length. Six of the trails run north -south and six run east -west (grid: 01 ° 28 ′ 13.75 ′′ N; 61 ° 0 ′ 16.63 ′′ W) (Fig. 1); all of the trails are maintained by the park administration, which also patrols the area full time. Th e park and its surroundings are formed of untouched tracts of forest that were little used by former Amerindian inhabitants and virtually unknown to colonists. Th e RAPELD grid opened areas that had no human tracks and were equally unknown to local people. Th ere are no residents within 10 km of the grid, a distance much greater than that found by this study to be the eff ective distance around habitations to be impacted by hunting( Յ six km).
Both study areas Novo Para í so and Viru á are formed by the typical vegetation mosaic of ' Campina ' and ' Campinarana ' and closed-canopy ombrophilous forest, in this case, on relatively fl at land with altitudes ranging between 89 and 122 m a.s.l. (  Th e Committee for Ethics on Research -CER, CCS -UFPE approved the interviews; a written consent was read to those interviewees who were illiterate, or given to them to read and subsequently sign.
From the data collected in the interviews, the following were calculated for each hunter and for the settlement as a whole: hunting range (km) -the average maximum distance walked per hunt; hunting eff ort (h) -the sum of the average duration of all hunting; capture per unit of eff ort (CPUE) (kg h Ϫ1 ) -the biomass of hunted animals per time unit (Souza-Mazurek et al. 2000, Fleck 2004, Sir é n et al. 2004, Parry et al. 2009); and per capita harvest rate (HR) (individuals/consumer year) -the number of individuals of each species hunted divided by the number of consumers multiplied by the study period (one year) (Redford and Robinson 1987, Souza-Mazurek et al. 2000, Sir é n et al. 2004, Gavin 2007.

Data analysis
From the data collected during the surveys, the following were calculated for each species in each area: sighting rate (sightings/10 km walked) (Chiarello 1999), relative abundance (individuals/10 km walked) (Silva J ú nior and Mendes Pontes 2008, Galetti et al. 2009), and mean group size (individuals group Ϫ1 ). For these calculations, only the total number of kilometers walked that corresponded to the activity period of each species was considered. Th e species activity periods (diurnal, nocturnal or cathemeral) were obtained from the literature (Emmons and Feer 1997, Reid 1997, Reis et al. 2011. Furthermore, the relative biomass (kg/10 km walked) of each species per unit area was calculated by multiplying the body weight by their relative abundance, as in Galetti et al. (2009). Th e body weight of each species was considered as the arithmetic mean of the values reported by Eisenberg and Redford (1999), Emmons and Feer (1997), Reid (1997) and Reis et al. (2011). A Shapiro -Wilk (W) statistic showed that the variables had a nonparametric distribution. Statistical analyses were performed using Statistica 10.0 software.
To assess whether the population variables diff ered between the impacted and protected areas, the data were compared using the Wilcoxon t-test (Z) (signifi cance level 0.05) by pairing the values of each species in both areas and assigning ' 0 ' to the species absent in one area. Additionally, abundance ratios (Rd) were calculated to analyze the individual variation in the abundance of each species between the two areas, which consisted of paired comparisons of species-specifi c estimates of relative abundance between two populations exposed to diff erent levels of hunting pressure (Peres and Palacios 2007).

Sampling of medium and large-sized mammals
After a sampling eff ort of 840 km, 29 species were registered in both studied sites, in addition to four species whose occurrence was confi rmed by records outside of the surveys, providing a total of 33 species from seven orders ( Table 1).

Surveys of medium-and large-sized mammals
Th e surveys of medium-and large-sized mammals were performed using the line transect method (Burnham et al. 1980, Buckland et al. 1993. At Novo Para í so (impacted), three parallel trails some 3 km apart were opened behind and perpendicular to three inhabited settlement plots (Fig. 1). Th e trails were alphanumerically marked every 50 m, and measured 5 km (trail 1), 3.75 km (trail 2) and 4 km (trail 3). At Viru á (protected), three of those 12 PPBio grid trails were chosen for the surveys (N1, N2, N3; Fig. 1), which are parallel and contained the same types of vegetation as the Novo Para í so area. Th is prevented vegetation type from becoming a source of variation in mammal abundance.
Th e diurnal surveys were conducted between 07:00 and 17:30. An average of 10 km was walked each day at a mean speed of 1 km h Ϫ1 . Th e nocturnal surveys were conducted between 18:30 and 04:30, travelling on average 8 km each night at an average speed of 1.25 km h Ϫ1 . During the nocturnal surveys, headlamps were used to illuminate the trail and adjacent vegetation, and long-range fl ashlights were used to view the animals.
At Novo Para í so (impacted), mammal surveys were conducted from December 2009 to January 2010, totalling a sample of 100 km for each trail during the day, and 40 km at night -a sum total of 420 km walked in the area. At Viru á (protected), surveys were carried out in March and October 2011, and the sampling eff ort was also 420 km.
For each animal or group of animals seen during the survey, the following data were recorded: species, group size, time and location along the trail, and type of vegetation. To compose a list of mammal species for each area, sightings outside of the survey times were also considered, including animals observed on roads, in open areas, as well as those sighted during the opening of trails, plus carcasses, skins, and bones of animals found in the forest or shot by hunters. No footprints or burrows were considered due to the degree of uncertainty inherent to this type of record.

Hunter interviews and hunting impact
To characterize and quantify the hunting pressure exerted on the mammal community, 50 hunters residing in the Novo Para í so settlement (impacted) were interviewed during January and February 2011. Th e hunters interviewed were chosen with the aid of a local resident, who served as a mediator and witness for the interviews.
A semi-structured questionnaire was used to obtain the following information regarding the hunting activity of each respondent: hunting time (years), number of consumers, duration of each hunt (hours), frequency (hunts month Ϫ1 ), maximum distance (km), strategies used (type of weapon, type of hunting, use of dogs), reasons for hunting, number of animals hunted, the importance of hunting for their livelihood, and the possible existence of taboos or cultural factors that infl uence the activity of hunting. Th e hunters identifi ed the species using an illustrated guide based on images derived from Emmons and Feer (1997), Reid (1997) and Eisenberg and Redford (1999). Th e respondents signed a consent agreement, and had their anonymity guaranteed. At Novo Para í so (impacted), 209 records were made of 25 medium-and large-sized mammal species, and seven other species were recorded from indirect records (Table 1). Th e total sighting rate in Novo Para í so was 7.07 sightings/10 km, while the total relative abundance was 22.72 individuals/ 10 km (Table 2). Th e relative biomass at Novo Para í so was 126.08 kg/10 km (Table 2).
Th e species that showed the highest sighting rate and relative abundance in Novo Para í so (impacted) was Cebus apella , with 1.77 sightings/10 km and 8.7 individuals/10 km, respectively. Th e species that showed the lowest values for both variables were Panthera onca , Speothos venaticus , Dasypus novemcinctus and Dasypus kappleri (0.02 sightings/ 10 km and 0.02 individuals/10 km, respectively). Th e species with the highest relative biomass was Tapirus terrestris (39.4 kg/10 km), while Sciurus aestuans had the lowest biomass (0.04 kg/10 km). Saimiri sciureus had the largest mean group size (10.5 individuals/group) ( Table 2).
At Viru á (protected), 21 species of medium-and large-sized mammals were sighted over 231 records (Table 1). Th e total sighting rate at Viru á was 7.62 sightings/10 km, while the total relative abundance was 54.05 individuals/ 10 km, and the total relative biomass was 490.54 kg/10 km (Table 2).

Characterization and quantifi cation of hunting at the Novo Para í so settlement
Interviews with 50 hunters at the Novo Para í so settlement revealed that 20 species of mammals were hunted within one year (Table 4), with a total of 541 (SD Ϯ 41.54) individuals killed, and an estimated biomass of 8517 kg (SD Ϯ 739 kg) taken. Th e most-hunted species was D. novemcinctus , with 122 (SD Ϯ 5.61) individuals taken; consequently, this species also had the highest harvest rate (0.506 individuals/ consumers year). In this one-year period, there were 849 (SD Ϯ 21.24) hunts resulting in a hunting eff ort of 4575 (SD Ϯ 127) hours. Th e average duration of each hunt was 6.48 (SD Ϯ 4.08) hours, and the average distance travelled in each hunt was 3.21 km (SD Ϯ 2.96), with a maximum of 5.4 km (SD Ϯ 4.5).
Considering the biomass and hunting eff ort, hunting effi ciency (CPUE) was 1.87 kg h Ϫ1 . Th e number of hunts conducted each month averaged 1.41 (SD Ϯ 1.75). Th e total number of hunting consumers for the 50 hunters interviewed was 241 people, resulting in an average of 4.82 (SD Ϯ 2.84) consumers for each hunter and a per capita harvest rate (HR) of 2.24 individuals/consumer year.
All of the hunters said that obtaining food is the main reason for hunting (n ϭ 50), but only 6% (n ϭ 3) of them considered game meat as the main source of animal protein for their family. Th e other 94% (n ϭ 47) reported that the meat of domestic animals (cattle, pig, goat, chicken, duck, and turkey) formed a greater proportion of the family diet. Th e second most-cited reason for hunting was to protect property against predators (n ϭ 19). Such hunting targeted members of the order Carnivora, namely P. onca (n ϭ 50 individuals killed), Puma concolor (n ϭ 37), C. thous (n ϭ 71), E. barbara (n ϭ 32), and small felids ( Leopardus spp.) (n ϭ 55). Th e least-reported reason for hunting was for use of body parts (skin and fat) (n ϭ 1), allegedly due to more access to allopathic medicine (Fig. 3).

Cultural taboos and hunting preferences
Th e hunting activity at Novo Para í so was infl uenced by cultural taboos and hunting preferences, which were mentioned by all of the respondents (n ϭ 50) (Fig. 4). For some species, the infl uence of these taboos is positive, as hunters avoid hunting them because they believe that killing these animals " brings bad luck " . In this case, the species most frequently mentioned was Priodontes maximus (n ϭ 36) and, as an example, several hunters reported cases of serious illness, accident or death in their families after someone killed an individual of this species. Th e order Primates (n ϭ 6), and the species Myrmecophaga tridactyla (n ϭ 2), Choloepus didactylus , and Bradypus tridactylus (n ϭ 1) also fell into this category, although they were less frequently mentioned.
Th ere were also hunters who avoided killing certain species, because they were considered to have a bad taste, as in the case of T. terrestris (n ϭ 1), C. didactylus (n ϭ 1) , and B. tridactylus (n ϭ 1). Small mammals were also avoided (n ϭ 14), because of the low energetic return. Primates were also avoided by some hunters (n ϭ 5) due to empathy or superstition based on this group ' s " similarity to humans " . Th e hunting of pregnant females (n ϭ 48), young (n ϭ 48) or females accompanied by young (n ϭ 47) was also avoided due both to the small size of these individuals and empathy or awareness that the death of these individuals is more likely to impact negatively the species or the future availability of prey (Fig. 4).
Conversely, there were cases where cultural factors have acted negatively, with hunters targeting predators to defend domestic animals. Th e following species were cited as examples: P. onca (n ϭ 12), P. concolor (n ϭ 10), C. thous (n ϭ 12), E. barbara (n ϭ 6), and small cats ( Leopardus spp.) (n ϭ 23). Th ere were even reports of hunters who were paid by ranchers to kill jaguars ( P. onca ) that were attacking or even just patrolling around their herds, which we classifi ed as retaliatory hunts. Such actions, which could also be classifi ed as predatory since it is not for subsistence, are deeply rooted in the culture of the settler ' s families, who argue that it aims at protecting their livestock.

Diversity of medium and large-sized mammals
A total of 32 mammal species were recorded at Novo Para í so settlement (impacted), whereas 21 species were found in Viru á National Park (protected). Th us, the area richest in mammal species was that which has been subject to hunting and impacted by the fi sh-bone forest clearance and regional development. Th is result has a direct relation with Tayassu pecari , as this species was the most hunted and was not recorded in the Novo Para í so survey. Accordingly, its absence (or, more likely, very low abundance) may have favored the proliferation of more resilient species that exploit similar resources (e.g. Mazama americana , M. nemorivaga , Dasyprocta cristata ) in a likely process of density compensation Dolman 2000, Endo et al. 2010). Furthermore, the persistence of M. americana and M. nemorivaga in disturbed areas may also be linked to the fact that these species use secondary forests, natural or artifi cial clearings, agricultural fi elds, and areas of pioneer vegetation for feeding (Vogliotti 2008); whereas T. pecari depends on intact forests for survival (Keuroghlian et al. 2012).
Our data suggests that Pecari tajacu may proximately benefi t from the considerable decline of their competitor, T. pecari , since P. tajacu sightings rates at Novo Para í so settlement (impacted) were higher than in Viru á (protected). However, the complete (or, perhaps, eff ective) extirpation of the latter, has led hunters to shift their focus towards P. tajacu , already among the third most hunted species. Ultimately, this may also cause their extirpation.
Although there was no signifi cant diff erence between the relative abundances and biomasses in the study areas, the total values of these variables were signifi cantly higher in Viru á (protected). Th is result is also due to the relatively higher abundance of T. pecari in this area than at Novo Para í so (impacted), given that it is known to be extremely sensitive to hunting and often disappears quickly from hunted areas (Peres 2001, Altrichter et al. 2012.
Of the total relative biomass calculated for Viru á (protected), T. pecari represented 71.37%. Th is proportion is much higher than that recorded in other conservation units in Amazonia, such as others reported by Lopes and Ferrari (2000) (18.45%), Mendes Pontes (2004) (43.99%), Nascimento (2006) (19.4%), andEndo et al. (2010) (40.55%), and also higher than presented by Cullen Jr. et al. (2001) in the Brazilian Atlantic forest (between 20% and 34%). Th is suggests that the population of this species at Viru á is above the expected ecological equilibrium. We do not believe that it could be a methodological artifact since in another strictly protected area in the same region, white-lipped peccary biomasses are even higher and the highest in the Amazon (Mendes Pontes 2000, Mendes Pontes and Chivers 2007. Primates were another group that also showed relatively high abundance and biomass at Viru á (protected). Although species of this order are rarely hunted in the settlement area, their lower abundances and biomasses at Novo Para í so (impacted) suggests that larger species that are more sensitive to the eff ects of hunting (e.g. Ateles paniscus ) may be impacted (Peres and Dolman 2000).  Figure 4. Th e species and groups of mammals avoided by hunters at the Novo Para í so settlement, Roraima, northern Brazilian Amazon.

Hunting at the Novo Para í so settlement (impacted)
Th e mammal community in adjacent forests have probably been subjected to the eff ects of hunting since the settlement was created (some 37 years prior to the study). During the study period, the demand for hunting by the human population (HR) was 2.24 individuals/consumer year, a value below that found by Redford and Robinson (1987)  . Th is shows that the Novo Para í so settlement (impacted) has one of the lowest per capita consumption rates (HR) when compared with similar studies. Th is, coupled with the fact that the abundance found at Novo Para í so was quite similar to those at Viru á (protected), indicates that hunting has probably not caused major losses to the local mammal assemblage. At Novo Para í so, hunting has lost its importance as a source of animal protein, as evidenced by the reduction in the average monthly number of hunts and in hunting effi ciency. Th is is a result of the improved socio-economic conditions of the settlers and greater access to meat from domestic animals. Additionally, the number of preferred prey is diminishing because, according to the hunters, it has been necessary to increase the hunting eff ort to fi nd the larger species, which reduces hunting effi ciency and discourages hunters.
Th e hunting range at Novo Para í so (about 5.4 km) is one of the lowest ever recorded in hunted areas (e.g. Vickers 1984[8.5 km], Alvard 1992, Peres and Nascimento 2006[25 km], Ohl-Schacherer et al. 2007[10 km], Parry et al. 2009. Th is relatively limited spatial range of hunting may have positive or negative implications for conservation. A limited hunting range impacts only a small area, preserving more distant areas. However, species with extensive ranges using such an area as part of their range can be completely eliminated. In such an instance, the area impacted by hunting would be larger (Alvard 1994). Th is is probably the case for white-lipped peccaries.
When hunting has a greater radius of action, its eff ects are more diluted, and the risk of local extinction may be reduced. Th us, for hunting in restricted areas to be sustainable, an effi cient source-sink balance is required in order to ensure the constant restocking of hunted areas (Alvard 1994). Th is seems to be the case with Novo Para í so, as the low hunting rates and the relatively-high abundance of mammals are strong indications that the source areas surrounding the settlement are fulfi lling the role as a recolonization source for the human-impacted sink area.

The importance of Viru á National Park for mammal conservation in northern Amazonia
Th e Guyana Shield has one of the lowest mammal diversities and abundances in the Brazilian Amazon, and within this region, Roraima is one of the poorest (Sombroek 2001, Hoorn et al. 2010, Mendes Pontes et al. 2010). Th ough Roraima has an extensive protected areas network that covers almost two million hectares (Governo do Estado de Roraima 2012), none of these is located in the dense ombrophilous non-fl oodable terra fi rme forests of the southeast of the state, where forest biomass is one of the highest (Barbosa et al. 2010).
Very little is known of the biological value of these extensive forested areas surrounding the fi sh-bone human settlements (impacted), which, nevertheless, have systematically been colonised and destroyed since the 1970s. However, the extent of information lack is such that, on the offi cial World Bank/WWF/Government-defi ned Amazon Region Protected Areas (ARPA) map (MMA 2007), which illustrates the importance and priority areas of Roraima for protection of biological resources and creation of protected areas, the southeast of the state consists only of blank space.
In these highly impacted fi sh-bone human settlements most large mammals will ultimately be locally and regionally extirpated, as will the dense ombrophilous forests of the comparatively fl atter and richer soils of the region. Th is already appears to be the case for the highly sensitive T. pecari , which had the highest biomass of any of the surveyed mammals at Viru á (protected), but is almost completely absent from the forested areas surrounding the settlement. Viru á therefore plays a crucial role in the maintenance of a highly representative sample of the original mammalian fauna of the region, as well functioning as a repository for its invaluable biological diversity.

Conservation implications
Hunters reported that, in the past, T. pecari was the most hunted mammal at Novo Para í so. However, by 2010 this species has become only the fi fth most-hunted species, suggesting a reduction from its historical abundance. Th e R d for T. pecari ( -2.033) indicates a relative abundance of around 100 times greater in the protected area, suggesting that this species has been hunted beyond a sustainable level in forests around the settlement.
Due to the extreme sensitivity of T. pecari to hunting, this species is often locally or regionally extirpated in areas close to human settlements (Peres 1996, Naranjo and Bodmer 2007, Reyna-Hurtado and Tanner 2007. Peres (2001) found the local extinction of T. pecari in six areas to be a result of hunting. In the case of Novo Para í so (impacted), although we observed an individual killed by a hunter, showing that the species still remains in the area, our data suggest that local extinction will certainly occur, since a few individuals left in a remnant subherd cannot maintain a viable population (Biondo et al. 2011). What seems to be happening is that the source population has probably been reported, and the absence of T. pecari , demands the immediate implementation of sustainable alternatives of wildlife harvest, the immediate suppression of T. pecari hunting and population monitoring. depleted, which is highlighted in the diff erences in abundance between protected and hunted sites.
Urgent measures are needed to recover the population of T. pecari at Novo Para í so and reverse the process of local extinction. Th e fi rst and ideal step towards this recovery would be to suspend T. pecari hunting long enough to allow its population to recover, but this proposal tends to encounter resistance from the population, because T. pecari is one of the species most prized by consumers (Leeuwenberg 1997, Fragoso et al. 2000. A second alternative proposed by previous authors was the establishment of non-hunting refuges by dividing the space into hunting zones. Th us, the most hunted areas (those closest to the settlement -sink areas) would be free of hunting for some years, with hunting activities only being allowed in more remote (source) areas. After the recovery of T. pecari populations, hunting would again be allowed in the sink areas, but with the establishment of maximum extraction quotas per hunter, thus ensuring the sustainability of hunting and the population viability of the species in the long term. Although the source areas surrounding the settlement are fulfi lling the role as a recolonization source for the sink area, without sustainable hunting practices, the source fauna that replenishes this area near the settlements will eventually be depleted.
For a more accurate control, however, this evaluation should be done at least once annually, and individually for each species. A tool for assessing the sustainability of hunting is the CPUE, which allows the local population themselves to analyze the abundance of hunted species and indicate whether they are being over-exploited (Sir é n et al. 2004). Th is could be a viable alternative that could be implemented in the Roraima settlements through conservation projects, which both involve the local community and generate income.
In places where there is a low availability of meat from domestic animals, hunting pressure is much higher and tends to lose selectivity, leading to notable reductions in the densities of various vertebrate species (Lourival and Fonseca 1997). At Novo Para í so, however, meat from domestic animals is widely available for consumption, which has a positive eff ect on the conservation of mammals and the sustainability of hunting, as it reduces the harvest of wild animals.
Despite the present hunting levels being considered sustainable, hunters reported that the abundance of game mammals has decreased. In this sense, measures aimed at controlling the eff ects of hunting are necessary, even if they are only preventive. Th ese measures must come from environmental authorities and should be accompanied by educational actions and intense scrutiny to ensure that the existing environmental legislation is followed. Finally, it is essential that any measures occur with the involvement of local inhabitants to make clear the benefi ts that conservation will bring.

Conclusions
Relatively high mammal abundances, the existence of extensive contiguous source forests and the low hunting levels detected during this study suggest that hunting at the Novo Para í so settlement (impacted) as currently practiced is sustainable. Th e progressive decrease in hunting effi ciency