Introduction
The small red brocket deer Mazama bororo is one of the most threatened, elusive and poorly known Neotropical deer. It is categorized as Vulnerable on the IUCN Red List (Duarte, Reference Duarte2008). The species inhabits remnants of the Atlantic Forest (Duarte & Jorge, Reference Duarte and Jorge2003; Weber & González, Reference Weber and González2003; Duarte, Reference Duarte2008; Duarte et al., Reference Duarte, González and Maldonado2008), the second largest rainforest of the American continent, which once stretched almost continuously along the Brazilian coast, extending inland in the south and into eastern Paraguay and north-eastern Argentina (Tabarelli et al., Reference Tabarelli, Pinto, Silva, Hirota and Bedê2005). This biome may hold up to 8% of all species of flora and fauna (Silva & Casteleti, Reference Silva, Casteleti, Galindo-Leal and Câmara2003), and has a history of intensive habitat loss and fragmentation caused by human activity. More than 88% of the original forest cover has already been cleared, and the land converted to non-forest use (Ribeiro et al., Reference Ribeiro, Metzger, Martensen, Ponzoni and Hirota2009).
The small red brocket deer is difficult to observe because it inhabits dense forests. In addition, morphological similarities with sympatric brocket deer species preclude the use of traditional methodologies for detection and species identification (Duarte & Jorge, Reference Duarte and Jorge2003; Vogliotti & Duarte, Reference Vogliotti and Duarte2009, Reference Vogliotti, Duarte, Duarte and González2010). Skins and skulls from Brazilian scientific collections revealed size, coloration and pelage differences between specimens taken from within the extent of occurrence of M. bororo and specimens of red brocket deer M. americana from other regions of Brazil (Rossi, Reference Rossi2000). Although some morphological characteristics (weight, height, length of body, circumference of thorax, and length of metacarpus and metatarsus) may be potential discriminators of the species in living individuals or when entire carcasses are available (Duarte & Jorge, Reference Duarte and Jorge2003), these characteristics are difficult to evaluate in the field using camera traps or even visual encounter data.
Because of these morphological similarities, it was only through karyotype characterization and subsequent molecular confirmation (Duarte et al., Reference Duarte, González and Maldonado2008) that the small red brocket deer was identified and described as a distinct species from the red brocket deer (2n = 32 and 2n = 52, respectively; Duarte & Jorge, Reference Duarte and Jorge2003). Karyotyping has been proposed as the most reliable tool to distinguish M. bororo from other sympatric species of brocket deer (Duarte, Reference Duarte1996; Duarte & Merino, Reference Duarte, Merino and Duarte1997; Duarte & Jorge, Reference Duarte and Jorge2003; Vogliotti & Duarte, Reference Vogliotti, Duarte, Duarte and González2010). However, to obtain karyotypes using cytogenetic protocols it is necessary to capture live individuals and collect fresh samples (blood and skin), which is difficult for this elusive species (Duarte et al., Reference Duarte, Ramalho, Lima and Jorge1999). Efforts to locate and capture the deer are expensive, time consuming and infeasible. We conducted intensive field work surveys for 4 years and were able to locate and capture only five free-living M. bororo (Vogliotti & Duarte, Reference Vogliotti and Duarte2009). This difficulty prompted us to develop faster, more efficient methods to detect this rare and elusive species and determine its geographical range, using non-invasive sampling of faecal material in combination with molecular markers (polymerase chain reaction–restriction fragment length polymorphism, PCR–RFLP; González et al., Reference González, Maldonado, Ortega, Talarico, Bidegaray-Batista, Garcia and Duarte2009).
Visually locating deer faeces in dense forest is challenging (Rivero et al., Reference Rivero, Rumiz and Taber2004). To maximize the number of faecal samples detected during surveys we used scat detection dogs, which have been proven to significantly improve the detection of faeces from rare and elusive species (Smith et al., Reference Smith, Ralls, Davenport, Adams and Maldonado2001; Oliveira et al., Reference Oliveira, Norris, Ramírez, Peres, Galetti and Duarte2012), and have been used effectively to locate faecal samples from a diversity of carnivores (Smith et al., Reference Smith, Ralls, Hurt, Adams, Parker and Davenport2003, Reference Smith, Ralls, Cypher and Maldonado2005; Harrison, Reference Harrison2006; Long et al., Reference Long, Donavan, Mackay, Zielinski and Buzas2007; Reed et al., Reference Reed, Bidlack, Hurt and Getz2011; Vynne et al., Reference Vynne, Skalski, Machado, Groom, Jácomo and Marinho-Filho2011), Xenarthrans (Vynne et al., Reference Vynne, Skalski, Machado, Groom, Jácomo and Marinho-Filho2011), and North Atlantic right whales Eubalaena glacialis (Rolland et al., Reference Rolland, Hamilton, Kraus, Davenport, Gillett and Wasser2006).
It is challenging to predict the ranges of rare and elusive species precisely but this information is critical for conservation planning. One approach that has become increasingly important is species distribution modelling (Guisan et al., Reference Guisan, Tingley, Baumgartner, Naujokaitis-Lewis, Sutcliffe and Tulloch2013), which is used to derive spatially explicit predictions of environmental and ecological suitability for a particular species. This is achieved by the identification of statistical relationships between observations of species occurrence and environmental or ecological descriptors that characterize the species’ ecological niche. Species distribution modelling has the potential to play a critical role in supporting spatial conservation decision making (Addison et al., Reference Addison, Rumpff, Bau, Carey, Chee and Jarrad2013). It has proven to be particularly useful when occurrence data are limited, and the results can be used to inform planning for new surveys (Pearson, Reference Pearson2007).
To determine the geographical range of M. bororo more accurately we conducted intensive non-invasive surveys using scat detection dogs in all potential areas of occurrence, and used the occurrence data in species distribution modelling. Our results confirmed that the small red brocket deer has the smallest geographical distribution of any deer species described to date, and its distribution appears to have decreased significantly over the past 2 centuries. We therefore suggest there is an urgent need to develop conservation and restoration actions for all remnants of the Atlantic Forest, and recommend that conservation efforts should target the reduction of the edge effects of forest fragments, palm heart exploitation and poaching.
Study area
The sampling sites were selected based on the proposed distribution of M. bororo in the Serra do Mar coastal forests ecoregion (Duarte, Reference Duarte1996). We focused on sampling in the Ombrophilous Dense Forest, and we also included other associated formations, such as the Mixed Ombrophilous Forest (Araucaria moist forest), Semideciduous Seasonal Forest (Paraná–Paraíba interior forests) and some ecotones. The chosen areas were mainly government protected areas and private reserves > 1,000 ha. Additionally, we surveyed buffer areas surrounding protected areas where the scat detection dogs were forbidden from entering. These buffer areas were adjacent to the protected areas and had similar ecological characteristics. In total, we surveyed 46 sites in the states of São Paulo, Paraná, Santa Catarina, Bahia, Espírito Santo and Minas Gerais (Fig. 1; Table 1).
Sampling sites (Table 1) for deer scat across the Brazilian Atlantic Forest. Forest type is categorized according to Veloso et al. (Reference Veloso, Rangel-Filho and Lima1991), and percentage tree cover is based on Hansen et al. (Reference Hansen, De Fries, Townshend, Carroll, Dimiceli and Sohlberg2003).
Sampling localities (see numbered locations in Fig. 1), total number of faecal samples collected, and number of samples identified, by Mazama species.
* SP, São Paulo; PR, Paraná; BA, Bahia; ES, Espírito Santo; MG, Minas Gerais; RS, Rio Grande do Sul; SC, Santa Catarina
Methods
Sample collection and storage
Searches for faecal samples were conducted during June 2003–October 2012 by one or more observers and a handler with two scat detection dogs. The dogs were trained by the military police of São Paulo State narcotics detection programme and worked freely off-lead. The searches were conducted along natural trails and the borders of rivers and forests. Although the Atlantic Forest is continually affected by deforestation and exploitation (Galindo-Leal & Câmara, Reference Galindo-Leal and Câmara2005), we had to assume that no changes in habitat or deer distribution occurred over the time period of the sampling. As we aimed only to record presence data, the sampling effort was not recorded in detail (e.g. time in hours, or meters surveyed) and the duration of surveys varied from 1 day in some areas to up to 1 week in others. This sampling scheme precluded gathering true absence data.
The collected faecal samples were stored in absolute ethanol in 50 ml sterile polypropylene conical tubes. The samples were identified by location name, collection number, geographical coordinates and date. When multiple faecal pellets were found accumulated in the same location they were considered to belong to a latrine. In these cases, to avoid resampling the same individuals, we collected only representative samples from clusters of pellets found at separate locations.
DNA extraction and PCR–RFLP
Faecal DNA extractions were performed using a Qiamp DNA Stool Mini Kit (QIAGEN, Hilden, Germany) according to the manufacturer's protocol. Extractions were carried out in a dedicated room free of PCR products to prevent contamination.
PCR amplifications of a 224 bp fragment of Cytochrome b were performed in a final volume of 25 μl and contained 1 × Invitrogen buffer, 1.5 mM MgCl2, 0.1 mM of each dNTP (deoxynucleotide triphosphate), 1 pmol μl–1 of each primer (IDMAZ224L and IDMAZ224H; González et al., Reference González, Maldonado, Ortega, Talarico, Bidegaray-Batista, Garcia and Duarte2009), 1.0 U Taq DNA Polymerase (Invitrogen), and 60–100 ng μl–1 of faecal DNA. For the thermal profile we followed González et al. (Reference González, Maldonado, Ortega, Talarico, Bidegaray-Batista, Garcia and Duarte2009). We included a negative control in each set of PCRs performed, to detect any possible contamination.
The amplified 224 bp fragment of Cytochrome b contains three informative restriction sites that can discriminate between M. bororo and its sympatric brocket deer species M. americana, M. nana and M. gouazoubira (González et al., Reference González, Maldonado, Ortega, Talarico, Bidegaray-Batista, Garcia and Duarte2009). The presence of the restriction site for ECORII or BSTNI in position 160 was used to confirm species identification because it has been demonstrated to be unique to M. bororo (González et al., Reference González, Maldonado, Ortega, Talarico, Bidegaray-Batista, Garcia and Duarte2009). PCR products that failed to digest were subjected to cycle sequencing to verify species identification. PCR products were purified using a Zymoclean™ DNA Recovery Kit (Zymo Research, Irvine, USA), and sequenced using an ABI Prism 3730xl DNA Analyzer (Applied Biosystems, Foster City, USA) with the IDMAZ224L and IDMAZ224H (forward and reverse) primers described in González et al. (Reference González, Maldonado, Ortega, Talarico, Bidegaray-Batista, Garcia and Duarte2009). Sequences were analysed using MEGA 5.1 (Tamura et al., Reference Tamura, Peterson, Peterson, Stecher, Nei and Kumar2011).
Species distribution modelling procedures
We developed species distribution models using MaxEnt v. 3.3.3k (Phillips et al., Reference Phillips, Anderson and Schapire2006), the most adequate and frequently used software for species distribution modelling (Baldwin, Reference Baldwin2009). In comparison to other approaches MaxEnt performs well with small sample sizes (Hernandez et al., Reference Hernandez, Graham, Master and Albert2006).
Occurrence data were based on only 19 sample locations, following the exclusion of multiple records in the same pixel (c. 1 km2). We used 70% of the points to train the model and 30% to test the model. They were sampled by bootstrapping 10 random partitions with replacement (Pearson, Reference Pearson2007).
The environmental layers were composed of bioclimatic variables, including mean annual temperature (bio1), temperature seasonality (the standard deviation, multiplied by 100, of the monthly temperature; bio4), annual precipitation (bio12), precipitation seasonality (bio15), and precipitation of the driest month (bio14), from the WorldClim database (Hijmans et al., Reference Hijmans, Cameron, Parra, Jones and Jarvis2005), percentage tree cover in 2000–2001 (Hansen et al., Reference Hansen, De Fries, Townshend, Carroll, Dimiceli and Sohlberg2003), altitude (Valeriano, Reference Valeriano2008), and slope obtained from the AMBDATA database. We also produced raster maps with Euclidean distance from maps of the protected areas (Ministério do Meio Ambiente, 2015), forest remnants (Ribeiro et al., Reference Ribeiro, Metzger, Martensen, Ponzoni and Hirota2009) and the Ombrophilous Dense Forest (Veloso et al., Reference Veloso, Rangel-Filho and Lima1991). The environmental layers were selected based on variables that had a high likelihood of being associated with M. bororo habitat in the humid coastal Atlantic Forest distributed throughout mountain chains.
All layers were trimmed to match the limits of the Atlantic Forest (modelling scope) and rescaled to a resolution of 1 km2. All spatial analyses were conducted using ArcMap v. 10.3.1 (ESRI, Redlands, USA). We evaluated the degree of correlation between environmental variables using PAST 3.06 (Hammer et al., Reference Hammer, Harper and Ryan2001). Two bioclimatic variables (mean annual temperature and precipitation of the driest month) were removed because of their high correlation (r > |0.75|) with altitude and precipitation seasonality, respectively. We performed a principal components analysis to select the most explicative variables. The variables identified and subsequently used in the MaxEnt analysis were bio12, bio15, bio4, distance from Ombrophilous Dense Forest, altitude, and percentage tree cover. The resulting model was evaluated using the area under the receiver operating curve, omission error and a two-proportion binomial test (Fielding & Bell, Reference Fielding and Bell1997; Elith et al., Reference Elith, Graham, Anderson, Dudik, Ferrier and Guisan2006; Pearson, Reference Pearson2007; Phillips & Dudík, Reference Phillips and Dudík2008). To predict the species’ distribution we assumed only the probabilistic values over the minimum training presence logistic threshold.
Results
Species identification from scat samples
We collected 555 deer scat samples from 30 of the 46 surveyed areas. PCR reactions amplified 497 (89%) of the samples successfully. The RFLP reaction failed for PCR products from 18 samples (3.2%). For these samples we then proceeded to sequence the PCR products directly, but only 10 of the PCR products yielded sequences of sufficient quality to determine species. The PCR–RFLP reaction and sequencing indicated that 116 (23%) samples were from M. bororo. The remaining samples were from M. gouazoubira (172; 34%), M. americana (158; 31%) and M. nana (51; 10%). The small red brocket deer was positively identified in only seven localities within the states of São Paulo, Paraná and Santa Catarina (Fig. 2).
Species distribution model for Mazama bororo, with the extent of occurrence, IUCN geographical ranges, the proposed geographical range, and locations of occurrence (numbered as in Table 1) of Mazama species. A circle with two colours indicates that two species were found at the same sampling site.
Geographical range
The size of the area obtained by linking external points of positive occurrence (extent of occurrence) was 20,000 km2, encompassing the Serra do Mar and Serra de Paranapiacaba from the states of São Paulo, Paraná and Santa Catarina. This region encompasses the Ombrophilous Dense Forest, an Atlantic Forest formation that is close to the Brazilian coastline, at 8–29°S. We did not detect M. bororo samples in the states of Espírito Santo or Bahia. To define the proposed geographical range we added a buffer zone of 30 km to the extent of occurrence, except for the areas that extended to the ocean (Fig. 2). This yielded a geographical range of 41,714 km2, which is the best estimate of the actual distribution of the species to date.
Species distribution modelling
The model developed in MaxEnt (Fig. 2) was evaluated satisfactorily by the mean area under the curve (0.9782 ± SD 0.0078) external test with 10 random partitions and was significantly different from random (P < 0.01) in a two proportion binomial test. The omission error was 0.14. The model predicted the occurrence of M. bororo with the highest values of environmental suitability at Serra de Paranapiacaba, Guaraqueçaba Environmental Protection Area and Serra do Tabuleiro State Park (Fig. 2). The variables that most contributed to the species distribution model were percentage tree cover (62.4%), distance from Ombrophilous Dense Forest (18.6%) and temperature seasonality (10.9%).
Discussion
Species of the genus Mazama display a high degree of morphological similarity (Duarte et al., Reference Duarte, González and Maldonado2008), making it difficult to distinguish them by direct field observations or camera traps (Vogliotti & Duarte, Reference Vogliotti, Duarte, Duarte and González2010). Our findings confirmed the sympatry of M. bororo, M. americana, M. nana and M. gouazoubira by the identification of more than one of the species in nine of 46 localities sampled. Furthermore, we found one or more sympatric Mazama species in three of seven localities (42.8%) where M. bororo was also detected. We detected M. nana in sympatry with M. bororo in two separate localities, and with M. gouazoubira in another locality (Fig. 2). These two species (M. nana and M. gouazoubira) are more common than M. bororo in human-altered and less pristine environments (Vogliotti, Reference Vogliotti2008). This suggests that the more suitable habitats used by M. bororo are being altered even in the protected areas where our sampling took place. This change in M. bororo habitat may be the result of edge effects (Ribeiro et al., Reference Ribeiro, Metzger, Martensen, Ponzoni and Hirota2009), palm heart exploitation, poaching (Vogliotti & Duarte, Reference Vogliotti, Duarte, Duarte and Reis2012) or global climatic changes (Allen et al., Reference Allen, Macalady, Chenchouni, Bachelet, Mc Dowell and Vennetier2010). We hypothesize that M. nana may be invading areas that were formerly occupied only by M. bororo and that this invasion may be facilitated by the degradation of the Ombrophilous Mixed Forest. In addition, M. gouazoubira, which is known to be one of the brocket species with greater ecological plasticity (Pinder & Leeuwenberg, Reference Pinder, Leeuwenberg and Duarte1997; Black-Décima et al., Reference Black-Décima, Rossi, Vogliotti, Cartes, Maffei, Duarte, Duarte and González2010) and resistance to anthropogenic habitat alteration (Rodrigues et al., Reference Rodrigues, Cerveira and Duarte2014), has been extending its range to include deforested areas that were previously covered with Atlantic Forest. The role of competition in the distribution and survival of sympatric brocket species in this area needs careful evaluation to define and mitigate better the effect of habitat changes in conservation plans for the small red brocket.
Morphometric and morphological studies of skulls and skins from various Brazilian collections (Rossi, Reference Rossi2000) failed to clearly differentiate M. bororo from M. americana, although the data supported the presence of a partially differentiated population of M. americana distributed throughout the south coast of Brazil. These results were considered to be the most reasonable evidence of the historical distribution of M. bororo in the Atlantic Forest. It is important to note that the ranges of distribution established here and by Rossi (Reference Rossi2000) were based on different methods, and there are many uncertainties involved in the reconstruction of the historical distribution. However, we suspect that the extent of the decline in distribution has been dramatic, as the species would have occupied the mountain ranges of the Serra de Paranapiacaba and the Serra do Mar, from the eastern part of the state of Santa Catarina to the state of São Paulo in the south and south-east regions of Brazil. We estimate that the historical geographical range was 114,120 km2 (Fig. 2).
The original proposed extent of the geographical distribution of M. bororo was based on the localities of provenance of the captive individuals used in the karyotypic characterization that resulted in the species description. These individuals originated in the cities of Capão Bonito, Barra do Turvo, São José dos Pinhais and Paranaguá and these localities were used to set the distribution limits to the south of São Paulo and east of Paraná states (Duarte, Reference Duarte1992, Reference Duarte1996; Duarte & Jorge, Reference Duarte and Jorge2003; Vogliotti & Duarte, Reference Vogliotti and Duarte2009).
Our survey results support an extension of the geographical range of the small red brocket deer to the latitudes 23–28°S and to the longitudes 47–49°W (Fig. 2), encompassing 37,517 km2. This range is considered to be the extent of occurrence, similar to the current area of occupancy because it represents an ecological continuum constituted by protected areas. A comparison of the historical distribution of 114,120 km2 (Rossi, Reference Rossi2000) with our results suggests that there has been an 83% decline in the distribution over the last 2 centuries. Given its recent description and hypothesized geographical range, there is little historical evidence to facilitate evaluation of the effects of anthropogenic development on the original distribution of the small red brocket deer. Currently, the habitat in the species’ area of occupancy is well-preserved, with few barriers to gene flow. If protected, habitat connectivity will help to maintain adequate levels of genetic diversity in the future.
The small red brocket is the largest endemic animal of Brazil, with perhaps the smallest geographical distribution of any Neotropical deer species. It is categorized as Vulnerable on the IUCN Red List because of its small population size and ongoing decline inferred from poaching and habitat loss, and because all of the remaining individuals are from a single subpopulation (Duarte, Reference Duarte2008). It is threatened by human encroachment (70% of the Brazilian population lives within the original limits of the Atlantic Forest), poaching, and predation by dogs (Duarte, Reference Duarte2008).
Species distribution modelling indicates that the current geographical distribution reported by IUCN should be revised to encompass northern areas with high environmental suitability, as well as the southern part of the Serra do Tabuleiro State Park and Serra do Itajaí National Park, where we collected occurrence data. In addition, areas of high suitability where no sampling was carried out or no deer species were detected during sampling should be revisited to confirm presence or absence. We also suggest that future species distribution models incorporate variables based on patch size and connectivity. Species distribution modelling provides more detailed information than traditional distribution estimates as it gives a suitability index for each 1 km2 pixel instead of a continuous distribution of values. It increases our understanding of species’ needs and gives support to conservation policies such as those for the design and establishment of protected and conservation priority areas.
We have demonstrated the importance of incorporating traditional field techniques with non-invasive sampling and sophisticated molecular tools to elucidate geographical range, extent of occurrence, and areas with the highest suitability for occupancy by the elusive small red brocket deer. Such information is essential in designing strategies for species conservation, particularly in the Neotropics, where new species with limited distributions are being discovered. Conservation efforts should aim more specifically to maintain the last remnants of the Atlantic Forest and to target the reduction of the edge effects of forest fragments, poaching and palm heart exploitation. M. bororo should be considered a flagship species and used as an indicator of the quality and restoration of the habitat.
Acknowledgements
This project was funded by the Projeto de Conservação e Utilização Sustentável da Diversidade Biológica Brasileira, Fundação de Amparo a Pesquisa do Estado de São Paulo, Conselho Nacional de Desenvolvimento Científico e Tecnológico (Brazil), Agencia Nacional de Investigación y Inovación, Programa de Desarrollo de las Ciencias Básicas, and Comisión Sectorial de Investigación Científica (Uruguay). We thank Renato Cheffer and Hermógenes Aparecido Torres for help with sample collection, and Instituto Brasileiro de Meio Ambiente, Instituto Chico Mendes de Conservação da Biodiversidade, Secretaria do Meio Ambiente do Estado do Paraná, Secretaria do Meio Ambiente do Estado de Santa Catarina and EUCATEX S.A. Indústria e Comércio, who authorized the collection of samples in the Conservation Units under their authority. Lillian Parker provided comments that greatly improved this article.
Author contributions
JMBD was the project coordinator and contributed to writing the article. ACT and JEG conducted laboratory analysis. AV contributed to non-invasive sampling and data management. MLO contributed to non-invasive sampling, laboratory analysis, distribution modelling, and writing the article. JEM provided scientific support and contributed to writing the article. SG provided scientific support, conducted laboratory analysis and contributed to writing the article.
Biographical sketches
José Maurício Barbanti Duarte has worked on various aspects of the genetics, evolution, ecology and conservation of Neotropical deer species, particularly of the genus Mazama. Ângela Cristina Talarico is a biologist, and her MSc research focused on the distribution of the small red brocket deer. Alexandre Vogliotti has worked on small red brocket deer research and conservation, and is currently focused on deer ecology and conservation planning in the Atlantic Forest. José Eduardo Garcia is interested in the conservation genetics of Brazilian wildlife. Márcio Leite Oliveira’s research is focused on distribution modelling and molecular ecology of brocket deer. Jesús E. Maldonado applies non-invasive genetic techniques to study conservation genetics, behavioural ecology and evolutionary genetics of a diversity of rare mammals. Susana González has conducted research in conservation biology for over 20 years and is co-chair of the IUCN SSC Deer Specialist Group.
