Survey of medium-sized and large mammals in semideciduous Atlantic Forest patches near Alfenas , southern Minas Gerais , Brazil

We surveyed, with the aid of camera traps, medium-sized and large terrestrial mammals in 8 patches of semideciduous Atlantic Forest near the city of Alfenas, Minas Gerais, Brazil. We recorded species occurrence using 2 camera traps per fragment, which were simultaneously active for 2 months each. We recorded 16 species, plus a species of armadillo that we were unable to identify to genus. The rarefaction curve tended to stabilize, evidencing the efficacy of our methods. The species richness was similar or higher to other surveys carried out in remnants of seasonal semideciduous Atlantic Forest or transitional areas. However, such seemingly high richness in a severely degraded and fragmented region, including large carnivores and endangered species, hides a wide variation in richness among fragments. This calls for urgent actions to improve the landscape for these animals to ensure their survival in the long term, as some populations are probably not viable in the long run.


Introduction
Forested areas worldwide have been suffering from changes due to intense land use associated with economic development, which has increased the amount of cultivated and urban areas, as well as road networks (Costa et al. 2005).These changes usually involve severe habitat reduction and the division of remnant areas into smaller habitat patches, which are usually isolated and surrounded by an altered landscape (the matrix), a process called fragmentation (Wilcove et al. 1998).
The Atlantic Forest in Brazil is no exception to this pattern, with less than a third of its original forest cover remaining (Ribeiro et al. 2009, Resende et al. 2018).Even worse, most Atlantic Forest remnants are not pristine.They are small forest patches at distinct regeneration stages, as only 7.9% of the patches are larger than 100 ha (Ribeiro et al. 2009, Fundação SOS Mata Atlântica andINPE 2015).Furthermore, although the "interior" subregion (largely comprised of seasonal semideciduous forest) originally occupied the largest area of the Atlantic Forest, currently only 7% of the original forest in this subregion remains (the second worst scenario among the subregions) (Ribeiro et al. 2009).
Even with its reduced and fragmented area, the Atlantic Forest is one of the most biodiverse ecosystems of the world, harboring many endemic and/or threatened species (Pinto et al. 2006).
Among vertebrates, mammals are among the groups most affected by forest fragmentation and destruction of natural habitats (Chiarello 1999, Cullen et al. 2001, Galetti et al. 2009).In this context, several mammal species in the Atlantic Forest are Vulnerable (threatened with extinction) (MMA 2014).Nonetheless, many areas, including protected regions, lack species lists or have not been surveyed at all (Cerqueira 2001).The medium-sized and large mammals of Atlantic Forest are affected by fragmentation (Ceballos et al. 2005, Costa et al. 2005), as they have a low reproductive rate, have large home ranges (Cardillo et al. 2005, Ceballos et al. 2005, Costa et al. 2005, Galetti et al. 2009), and are subject to hunting (Kasper et al. 2007, Araújo et al. 2008, Brocardo et al. 2012, Brocardo and Cândido-Júnior 2012).Even so, they are still found in many small remnants of seasonal semideciduous Atlantic Forest (Eduardo and Passamani 2009, Penido and Zanini 2012, Pires and Cademartori 2012, Magioli et al. 2014, Santos et al. 2016, Melo-Dias and Passamani 2018).
Thus, our study aims to survey medium-sized and large terrestrial mammals in 8 seasonal semideciduous Atlantic Forest remnants in a highly fragmented landscape in southern Minas Gerais, Brazil.This state has been poorly studied and has a shortage of systematic survey data.It has suffered extensively from the process of fragmentation and, even nowadays, is not free of threats.Thus, gathering basic inventory data is an essential step to support conservation and management projects.

Methods
Study sites.We carried out the study in 8 sampling patches of seasonal semideciduous Atlantic Forest (Fig. 1).The sampling sites needed to have between 20 and 40% forest cover within a buffer of 2 km 2 around the fragment and have an associated matrix of pasture.All fragments were in a highly fragmented landscape in southern Minas Gerais, Brazil, in a region that is severely degraded and fragmented, dominated by pasture, and coffee, sugar cane and corn plantations (Ayer et al. 2015, Olivetti at al. 2015).All fragments were private land, and we obtained permission from landowners before placing the camera traps.

Data collection.
We employed camera traps (Bushnell 8 MP Trophy Cam Trail Camera) to record the occurrence of mammals within the fragments (Tomas and Miranda 2003).We sampled each fragment continuously for 2 months, employing 2 camera traps per fragment.Within the fragments, we placed the camera traps turned towards preexisting trails, when available, 1-3 m away from them.When there were no available trails, we placed the 2 cameras apart from each other, in central parts of the fragment, preferably in areas showing good forest quality (higher canopy and less undergrowth cover).We programmed the traps to take 3 photographs each time they were triggered, with a 1-minute interval before resetting the trigger.Sampling effort was calculated by multiplying the number of traps in each fragment by the number of fragments by 60 (number of sampling days).The sampling effort amounted to 960 camera trap days, yielding 186 images of mammals.Data analysis.To verify if the sampling period was enough to record the richness of mammals across the landscape, we generated a species accumulation curve for the 60 sampling days, with a 95% confidence interval.We computed the curve using the software ESTIMATES and employed the estimator Jacknife 1 (Colwell et al. 2004).

Mammal classification.
Species nomenclature follow Paglia et al. (2012), except for Leopardus guttulus, which follows Trigo et al. (2013).We also classified each species according to its conservation status (IUCN Red List categories) at 3 levels, based on the official Red Lists of endangered species: regionally, for the state of Minas Gerais (COPAM 2010) and Brazil (MMA 2014), and globally (IUCN 2018).We considered mammals weighing over 1 kg to be as medium-sized and large (Chiarello 2000b), included only native species, and excluded records of domestic species.

Results
We recorded 16 medium-sized and large mammal species, from 8 orders and 11 families (Table 1).We also recorded an additional species of armadillo, which we could not determine if it belonged to genus Euphractus or Cabassous.Considering all sampling sites, the rarefaction curve tends to stabilize (Fig. 2).The recorded species richness is representative of the areas, estimating approximately 95% of the species expected in the areas.Only Myrmecophaga tridactyla and Leopardus guttulus are globally threatened (IUCN 2018), while 4 species are threatened in Brazil (MMA 2014; Table 1); all the recorded species that are classified under some threat category in the Brazilian list, plus Leopardus pardalis, are threatened in the Minas Gerais state list (COPAM 2010; Table 1).

Identification.
A medium-sized opossum, grayish with a paler face with 3 conspicuous black stripes, 1 on the middle of the forehead and the other 2 above the eyes.White ears clearly distinguish it from Didelphis aurita.
Identification.A medium-sized opossum, dorsally grayish with inconspicuous facial stripes and black ears.

Identification.
A medium-sized rabbit, and the only species of the genus occurring in the study area.
Identification.The Giant Anteater is the only species of genus Myrmecophaga.Its large size, long and furry tail, the large black stripe running from its throat towards its back, bordered by a thin white stripe, and its long and pointy snout make this species clearly distinct from any other in the region.
Identification.The Collared Anteater is the only species of its genus occurring in Brazil.Its medium-sized body, short fur with 2 clear black scapular stripes and its long snout make this species easily recognizable.
Identification.The Spotted Paca is the only mediumsized rodent in the region.It has a small tail, and short, homogeneous fur that is usually brown, with a pattern of white spots arranged in longitudinal rows.
Identification.Nasua nasua is the only species of coati that occurs in the region.It is a medium-sized procyonid, with a triangular head, long and pointed snout and short ears.This species has a brownish to orangish color with a black-ringed tail.Its face is darker, with an incomplete white ring around the eyes.
Identification.Puma concolor is a large felid with short, round ears and a long tail.Its pelage is light brown to reddish with whitish ventral parts.
Identification.Eira barbara is a medium-sized mustelid with a long body and neck, short, dorsoventrally flattened, head and short, round ears.Its pelage is dark brown with the head lighter than the dorsum and a whitish spot on the throat.The tail is long, dark and has long hairs.
Identification.Mazama americana is a medium-sized cervid with short ears in relation to the size of the head.Its body has an overall reddish color, with a more grayish head and neck.The hips are robust and slightly taller than the shoulders.
Identification.Sapajus nigritus is a medium-sized primate with an overall brown to dark brown color, a pigmented and hairless area on the face, lighter areas on the cheeks and temples and a developed and black tuft on the top of the head.It also has a long, dark tail.
Identification.Dasypus novemcinctus is a mediumsized armadillo with body armor with 9 mobile, back bands.It has long ears and a tail with rings of scales.

Discussion
In the 8 sampled fragments, we found 16 medium-sized and mammal species, which belong to 8 orders and 11 families (Table 1), plus an additional armadillo species whose identification could not be confirmed (either Euphractus sexcinctus or Cabassous unicinctus).This species richness recorded is similar or higher to other surveys carried out in remnants of seasonal semideciduous Atlantic Forest or transition physiognomies (Table 3).Although some studies have reported a higher number, all of them recorded species with more than a single method, which usually included camera traps.Probably, if we had done the same, more species would have been recorded.For example, we are aware of the presence of 2 primates in the region (Callithrix aurita and Callicebus nigrifrons; Silva et al. 2015) that were not recorded by the cameras.Our study had the highest number for species recorded using camera traps alone, which shows that our methodology was efficient at sampling the species in the region.However, the total number of species recorded is somewhat misleading.Our total richness was computed for the set of 8 fragments considered together, while the majority of the aforementioned studies were conducted in single fragments.Thus, our result masks the fact that fewer species were recorded in each fragment.Besides, there is also a large variation in richness among fragments.For example, at one extreme, a fragment (not particularly large or with good forest quality) had 13 species and 2 single records of large carnivores (Puma concolor  and Chrysocyon brachyurus).At the other end of the spectrum, in another fragment we recorded only 1 species of mammal (Nasua nasua).The latter fragment was highly degraded, with an undergrowth dominated by a thorny Acacia (Fabaceae) species, and it did not have trails.The presence of the coati may be due to the highly generalist diet of this species, which may include garbage (Alves- Costa et al. 2004).Whatever the reason for this expressive variation, this result highlights the need for simultaneous surveys of several fragments in similarly fragmented areas.Otherwise, there is the risk of having a biased estimate of the fauna of a given region, underestimating it, or worse, overestimating it and giving the impression of a brighter picture than reality.Another caveat is that sampling in some fragments may not have provided the most accurate documentation of the fauna.In 2 fragments we observed 1 (P ID2) and 5 (P ID4) additional species in the bordering pastures, although there were no new species for the regional set.Thus, we suggest conducting longer sampling periods and including more locations at each site.
Regarding IUCN Red List categories, only 2 species are classified as Vulnerable on the IUCN Red List (IUCN 2018): Myrmecophaga tridactyla and Leopardus guttulus.However, in the Brazilian list of threatened mammals (MMA 2014), more species are classified as Vulnerable, including 3 carnivores.All carnivores find food and shelter within the remnants (Costa et al. 2005) and, without the fragments they could disappear from the region due to the ongoing threats.For the Minas Gerais state list (COPAM 2010), we found all species classified in the Brazilian list plus Leopardus pardalis, which is mainly threatened by roadkills, deforestation, habitat loss, and predation by invasive and exotic species (Costa et al. 2005).In addition, Leopardus guttulus is classified at a higher risk level on the state list.Overall, nearly half of the recorded species are at some threat level, which points to the importance of these unprotected fragments in the maintenance of the regional fauna.In our survey, we did not record some species that were found in similar surveys, such as Puma yagouaroundi, Lontra longicaudis, Hydrochoerus hydrochaeris (the previous 2 probably because we did not have any large bodies of water close to the studied fragments), and Mazama gouazoubira.
Overall, our results can be seen from 2 different perspectives.On the one hand, although the region is severely degraded and fragmented, it still harbors a significant fauna of medium-sized and large mammals, including large carnivores and endangered species.This calls for urgent actions to improve the situation of the landscape for these animals to ensure their survival in the long term, as some populations are probably not viable in the long run.On the other hand, such variation in the number of species among the fragments is a serious warning that the situation may be far worse than the overall list might indicate.Some fragments seem to be almost devoid of medium-sized and large, and some species are restricted to only 1 or a few fragments, which reinforces the need for specific conservation and management actions.One action could be to improve landscape connectivity, as most fragments are isolated.The establishment of ecological corridors or stepping stones is a possible alternative, although their effectiveness vary according to the characteristic of the surrounding matrices (Perault and Lomolino 2000).

Figure 1 .
Figure 1.Location of the 8 sampling sites in southern Minas Gerais, Brazil.

Figure 2 .
Figure2.Species accumulation curve for the community of medium-sized and large mammals in 8 landscapes of semideciduous Atlantic Forest remnants near Alfenas, southern Minas Gerais, Brazil, based on 60 days of sampling with camera traps.Observed richness in red, expected richness in black, and their 95% confidence intervals.To compute this curve, we considered that we had another armadillo species (Euphractus sexcinctus or Cabassous unicinctus), which was definitely not the species already identified (Dasypus novemcinctus).

Table 1 .
Coordinates and size of the studied forest fragments in southern Minas Gerais, Brazil.

Table 1 .
Identification.Cerdocyon thous is a short, robust, medium-sized canid.It has short and round ears and a long furry tail.The general color pattern is gray, but darker in the nape.It can be confused with Lycalopex vetulus, which is less common, prefers open areas in the Cerrado biome, and has longer ears, a longer snout and a slightly lighter color.

Table 1 .
Identification.Chrysocyon brachyurus is a large canid with an orangish-brown color, long and dark limbs, large ears and a long snout.These observable characters make this species clearly distinct from any other in the region.
Identification.Leopardus guttulus is a small felid.It has a yellow to brown pelage, with rows of black ocellated spots that may form open rings.

Table 2 .
List of medium-sized and large mammals recorded in each studied forest fragment in southern Minas Gerais, Brazil, classified according to IUCN Red List categories: (VU) Vulnerable, (NT) Near Threatened, (LC) Least Concern, (DD) Data Deficient.IUCN Red List categories were assigned to each species according to the IUCN (2018) Red List ( IUCN ), Brazil's Red List ( BR ) (Ministério do Meio Ambiente 2014) and the Minas Gerais Red List ( MG ) (COPAM 2010); abbreviation for these different Red List classifications are presented in subscripts after the categories.

Table 3 .
Richness of medium-sized and large-sized mammals in surveys carried out in remnants of seasonal semideciduous Atlantic Forest or transition physiognomies (between open cerrado and forested formations).The definition of what is a medium-sized and large mammal may vary from study to study.