Unravelling the drivers of maned wolf activity along an elevational gradient in the Atlantic Forest, south-eastern

Izar Aximoff, William Douglas Carvalho, David Romero, Carlos Eduardo Lustosa Esbérard, José Carlos Guerrero & Luís Miguel Rosalino

Mammalian Biology Zeitschrift für Säugetierkunde

ISSN 1616-5047 Volume 100 Number 2

Mamm Biol (2020) 100:187-201 DOI 10.1007/s42991-020-00017-x

1 23 Your article is protected by copyright and all rights are held exclusively by Deutsche Gesellschaft für Säugetierkunde. This e- offprint is for personal use only and shall not be self-archived in electronic repositories. If you wish to self-archive your article, please use the accepted manuscript version for posting on your own website. You may further deposit the accepted manuscript version in any repository, provided it is only made publicly available 12 months after official publication or later and provided acknowledgement is given to the original source of publication and a link is inserted to the published article on Springer's website. The link must be accompanied by the following text: "The final publication is available at link.springer.com”.

1 23 Author's personal copy

Mammalian Biology (2020) 100:187–201 https://doi.org/10.1007/s42991-020-00017-x

ORIGINAL ARTICLE

Unravelling the drivers of maned wolf activity along an elevational gradient in the Atlantic Forest, south‑eastern Brazil

Izar Aximof1 · William Douglas Carvalho2 · David Romero3 · Carlos Eduardo Lustosa Esbérard4 · José Carlos Guerrero3 · Luís Miguel Rosalino5,6

Received: 27 April 2019 / Accepted: 4 February 2020 / Published online: 10 March 2020 © Deutsche Gesellschaft für Säugetierkunde 2020

Abstract The maned wolf, the largest canid in South America, was originally distributed in areas with open natural vegetation in the Cerrado biome, Chaco and Pampas regions. The dynamics of its distribution are, however, in fux, with populations declining at the southern limit of its distribution, and areas of apparent range expansion in Brazil. Although the maned wolf’s overall distribution is well documented, little is known about its smallest-scale landscape use. Here we used a novel approach, char- acterising “favourable territories” for maned wolves using presence data and information on daily movement capacity. In this way, we used favourability distribution models to relate local landscape use by maned wolves to environmental drivers in the Serra da Mantiqueira, part of the core of the species distribution. Our results showed that the favourablity of territories for maned wolf activity increases with altitude, and with the proportion of coverage of upper montane vegetation refuges and of open habitats such as agricultural felds. Our results also show that the confguration of the environment with respect to topography is an important driver of the favourability of the landscape for maned wolf activity. Finally, we identifed some human-wildlife conficts in the surroundings of the protected area which could increase with increasing maned wolf populations. In conclusion, our results support the importance of maintaining the integrity of high-altitude open areas in the conservation of maned wolf habitat and provide useful data for maned wolf management at the core of its global current distribution. We highlight that this is the frst study to use fuzzy logic tools at the local scale to analyze the favourability of territories for maned wolf activity in a highly favourable region along an elevational gradient.

Keywords Chrysocyon brachyurus · Favourable territories · High-altitude grasslands · Human-wildlife conficts · Maned wolf activity

Introduction

The maned wolf (Chrysocyon brachyurus Illiger, 1815) is the Handling editor: Emmanuel Serrano. largest canid in South America, and is anatomically adapted to move in open areas (Dietz 1984, 1985; Childs-Sanford Izar Aximof, William Douglas Carvalho and David Romero contributed equally to this work.

* David Romero Facultad de Ciencias, Universidad de la República, Iguá [email protected] 4225, CP 11400, Montevideo, Uruguay 4 Laboratório de Diversidade de Morcegos, Departamento 1 Instituto de Pesquisas do Jardim Botânico do , de Biologia Animal, Instituto de Biologia, Universidade Escola Nacional de Botânica Tropical, Pós-Graduação em Federal Rural do Rio de Janeiro, CP 74507, Seropédica, Botânica, Rua Pacheco Leão 2040, Solar da Imperatriz, RJ 23890‑000, Brazil Horto, Rio De Janeiro, RJ 22460‑036, Brazil 5 CESAM and Departamento de Biologia, Universidade de 2 Programa de Pós‑graduação em Biodiversidade Tropical, Aveiro, 3810‑193 Aveiro, Portugal Universidade do Amapá, Rod. Juscelino Kubitschck S/N, Macapá, AP 68903‑419, Brazil 6 cE3c, Centre for Ecology, Evolution and Environmental Change, Faculdade de Ciências, Universidade de Lisboa, 3 Laboratorio de Desarrollo Sustentable y Gestión Ambiental Lisbon, Portugal del Territorio, Instituto de Ecología y Ciencias Ambientales,

Vol.:(0123456789)1 3 Author's personal copy

188 I. Aximof et al.

2005; Coelho et al. 2008; Massara et al. 2012). In Brazil, et al. 2011), may indicate that the Serra da Mantiqueira the species was originally distributed in areas of native open could act as one of the source cores for maned wolf popula- vegetation, reaching high population densities in the Brazil- tions. This highlights the importance of assessing the local ian savannah, i.e., Cerrado biome (Queirolo et al. 2011). drivers shaping the species home range movements in the However, due to habitat loss, roadkill, diseases and retali- Serra da Mantiqueira, to better understand its spatial role for ation for predation of domestic animals (Curi et al. 2010; the survival of this canid in a changing world. Freitas et al. 2015; Massara et al. 2015), the maned wolf is Although the maned wolf’s wide distribution is now well now considered vulnerable in Brazil (MMA (Ministério do documented (see Queirolo et al. 2011), as well as the envi- Meio Ambiente) 2014), with a projected population reduc- ronmental drivers that determine it (Coelho et al. 2018), tion of 30% expected in the next 20 years (Paula et al. 2013). there is a scarcity of information regarding the local spa- This decline is mostly linked to the current and forecasted tial behaviour of the species within its current expansion devastation of approximately 50% of the Brazilian savan- area (Queirolo et al. 2011), or regarding spatial confgura- nah due to agricultural expansion (Queirolo et al. 2011). tion along an elevational gradient, which may be an import Nevertheless, some marginal populations of the maned wolf factor to consider as gradient characteristics can determine show an inverse pattern. The replacement of Atlantic For- local adaptations in resource use (e.g., Myslajek et al. 2012; est by anthropogenic felds (mainly pastures), has led to the Carvalho et al. 2019). Species distribution models have been expansion of the maned wolf’s distribution into the Atlantic used successfully to establish the relationship between a spe- Forest biome of south-eastern Brazil, where the species was cies and its environment (Guisan and Zimmermann 2000; previously absent or rare (Queirolo et al. 2011). Numbers of Guisan et al. 2013; Romero et al. 2016; Coelho et al. 2018), maned wolf records in the Atlantic Forest have increased in and even to predict the presence of species in localities not recent years, mostly in altered pasture felds (Queirolo et al. previously known to be occupied (Real et al. 2017). In this 2011; Eckhardt 2016; Beca et al. 2017; Bereta et al. 2017; paper, we apply the favourability function at local scale to Xavier et al. 2017). However, there have been occasional improve our understanding of local landscape use by the records in the high-altitude natural grasslands of the Atlantic maned wolf in a protected territory in the core of the maned forest (Avila-Pires and Gouvea 1977; Geise et al. 2004). wolf global distribution, the Serra da Mantiqueira. Our One of the most representative areas of the Atlantic For- specifc objectives were to: (1) identify the territories most est is the Serra da Mantiqueira, a mountainous region rang- favourable for maned wolf activity in the Serra da Mantique- ing from 500 to 2798 m a.s.l. (Simas et al. 2005; Barreto ira; and (2) estimate environmental drivers that determinate et al. 2013), located in the most populated states in Brazil, this spatial confguration of favourable territories in a highly i.e., São Paulo, and Rio de Janeiro (IBGE favourable region along an elevational gradient. Finally, we (Instituto Brasileiro de Geografa e Estatística) 2016). The discuss these results in the context of maned wolf manage- Serra da Mantiqueira is considered an irreplaceable region ment at the core of its global current distribution. of high biodiversity value and, thus, a conservation priority area (Myers et al. 2000; Le Saout et al. 2013; Rodrigues and Oliveira 2006). In the state of Minas Gerais, maned wolves Materials and methods had only been recorded in Cerrado areas (e.g., Aragona and Setz 2001; Queirolo and Motta-Junior 2007), until their Study area diet was recently described in forested areas of the Serra da Mantiqueira (Rosa et al. 2015). The frst record of maned The study was carried out in two protected areas within the wolves in the Serra da Mantiqueira was in the state of Rio Serra da Mantiqueira: National Park and Serra do de Janeiro in 1954. The species was observed in native high- Papagaio State Park (hereinafter INP and SPSP, respectively; altitude grasslands at 2400 m a.s.l. in the Itatiaia National Fig. 1). The INP encompasses the counties of Itatiaia and Park (Avila-Pires and Gouvea 1977). Since that time, evi- Resende, in the state of Rio de Janeiro, and the counties dence of the presence of the species in the region has only of Itamonte and , in the state of Minas been registered more recently, and in just a few mammal sur- Gerais. The INP covers an area of 28,084 ha ranging from veys carried out in higher areas of the region, such as in the 540 to 2798 m a.s.l. The SPSP is located in the counties of high-altitude felds (Geise et al. 2004; Aximof et al. 2015). Aiuruoca, Alagoa, , Itamonte and in The Serra da Mantiqueira has been identifed as an area of the state of Minas Gerais, and covers an area of 22,917 ha high favourability for the species globally, being compara- ranging from 1000 to 2314 m a.s.l. These protected areas ble to the Cerrado Biome as the centre of the species range make up part of a large corridor of protected areas along in terms of high environmental favourability for this canid the Serra da Mantiqueira Mountains, including the Pedra (Coelho et al. 2018). Such high favourability, together with Selada State Park, Serra da Mantiqueira Environmental Pro- its location near to the species core distribution (Queirolo tection Area, the National Forest and numerous

1 3 Author's personal copy

Unravelling the drivers of maned wolf activity along an elevational gradient in the Atlantic… 189

Fig. 1 a South America high- lighting in the box the location of the Serra da Mantiqueira in Brazil; and b, the protected areas, where the study was car- ried out, along the elevational gradient of the Mantiqueira Range: INP and SPSP Serra do Papa- gaio State Park

Private Nature Reserves (Carvalho et al. 2015). The INP appropriate as a management unit for maned wolf conserva- and SPSP are covered by four diferent types of vegetation tion (Barbosa et al. 2010). Finally, to assess the potential for which vary as elevation increases: lower montane forest human-wildlife conficts, we carried out interviews in the (between 50 and 500 m a.s.l.), montane forest (from 500 rural areas located close to INP. to 1500 m a.s.l.), upper montane forest (between 1500 and 2000 m a.s.l.) and high-altitude felds (above 2000 m a.s.l.) Records of maned wolf presence (Segadas-Vianna and Dau 1965). The forests are dominated by a mixture of seasonal semi-deciduous and ombrophil- We sampled maned wolf presence from October 2010 to ous dense forest, including sites with the critically endan- September 2014 in the two protected areas (INP and SPSP). gered Brazilian pine -Kuntze, 1898 In INP, monthly censuses were carried out between October (Oliveira-Filho and Fontes 2000). The rural areas surround- 2010 and September 2012, and in SPSP, 30-day campaigns ing the protected areas are covered by remnants of secondary were carried out in each season between January 2010 forest, degraded grasslands, federal and a network of state and December 2014. In both locations, the censuses cov- highways and unpaved roads embedded in a matrix of agri- ered all types of land cover considered in the present study cultural lands. The most north-northwestern portion of the (“Appendix”, Fig. 4). The study area was sampled at least Serra da Mantiqueira (state of Minas Gerais), is considered twice (once in the dry season and once in the rainy season), to be a transition zone between the Atlantic forest and the totalling 1500 km of efort. The censuses covered an area Cerrado biomes (Ururahy et al. 1983; IBGE (Instituto Bra- with a wide elevational range, varying from 400 to 2500 m sileiro de Geografa e Estatística) 2004). The region’s climate a.s.l (“Appendix”, Fig. 4). is mesothermal (Koeppen 1948), with mean annual tempera- Both direct observations and tracks or faeces were con- ture and precipitation of 11.5 °C and 2150 mm, respectively sidered as signs of presence of maned wolves (Becker and (IEF-MG (Instituto Estadual de Florestas de Minas Gerais) Dalponte 1991; Reis et al. 2010), and used to defne the area 2008; Barreto et al. 2013). The study area encompasses ca. with maned wolf activity (54 records of presence in Fig. 2a). 535 km2 representing more or less seven times the aver- The rest of the study area was considered as lacking maned age home range-size described for this species in Brazil (ca. wolf activity, and, therefore, to be an unoccupied area. Then, 70 km2; Coelho et al. 2008). As such, the monitored area for the modelling procedure we generated a bufer with a was large enough to cover several animals’ home-ranges. We radius of 1.5 km around each of the 54 presence points (see generated a grid of 1 km × 1 km spatial resolution within the Silveira et al. 2009 for more details), which was consid- study area for the modelling procedure (Fig. 2). This resolu- ered to be the territory used by each maned wolf detected. tion allows the identifcation of areas of high-quality habitat The bufer was defned with a diameter of 1.5 km as this

1 3 Author's personal copy

190 I. Aximof et al.

Fig. 2 Maned wolf presence records and bufer of activity (shown in grey) in the Serra da Mantiqueira in Brazil; and b, the area with maned wolf activity in black and with no activity in white. INP Itatiaia National Park (INP); SPSP Serra do Papagaio State Park (SPSP) represents the expected daily movement capacity of maned cell depending on the type of environmental predictor con- wolves, based on values reported in the literature (Silveira sidered. The mean elevation per grid was obtained from the et al. 2009). All the 1 km × 1 km grid squares totally or par- Insituto Nacional de Pesquisas Espaciais (INPE 2019). So, tially contained in the bufer area were considered part of the topography was represented by the mean elevation and, the area with activity of the species (Fig. 2b). So, a total of the slope and north–south predominant orientation. The land 203 grids with maned wolf activity were the number of grids use category was defned as the type of landscape use (see from which was built the model. In this way, we used grids Table 1), being classifed according to the vegetation map for instead of geographical locations in the models, so solve a the Atlantic Forest (IBGE (Instituto Brasileiro de Geografa large part of the spatial autocorrelation derived from obser- e Estatística) 2004). The vegetation type was also defned vation spatial clustering or sampling bias (Romero et al. according to the vegetation classifcation of the Atlantic 2019). Forest (Segadas-Vianna and Dau 1965), and as all locations Finally, to register possible human-maned wolf conficts, were above 500 m a.s.l. the types were: montane forest (from a total of 70 local people, including residents and park staf, 500 to 1500 m a.s.l.), upper montane forest (between 1500 were interviewed in the surroundings and within the area of and 2000 m a.s.l.) and high-altitude felds (above 2000 m the INP. Specifcally, the interview consisted of the follow- a.s.l.). The population density in the study area according to ing questions: (1) when and where was the last time that you the Data Center in NASA’s Earth Observing System Data saw a maned wolf in the region?; (2) Do you know of any and Information System (EOSDIS) was used as a proxy for predation of domestic animals by maned wolves?; (3) Have the intensity of human activities (details in Table 1). you witnessed any maned wolf deaths as a result of retalia- tion for predation damages, or a road collision? Analysis of the favourability of the landscape for maned wolf activity Environmental characterization We used a logistic regression approach to evaluate which To characterize the environment in the study area we used environmental factors might be infuencing maned wolf diferent environmental predictors: topography, land use presence patterns (Hosmer et al. 2013). Based on the grids category, human activities and vegetation types (Table 1). with activity of the species (Fig. 2b), and the set of environ- We assigned a value or category to each 1 km × 1 km grid mental predictors generated, we applied the favourability

1 3 Author's personal copy

Unravelling the drivers of maned wolf activity along an elevational gradient in the Atlantic… 191

Table 1 Variables and factors used to model the maned wolf activity patterns. In bold the variables used in the modeling procedure by overcom- ing multicollinearity and correlation flters Abbreviation Variable/factor Sources

Topography ME Mean elevation (m) Obtained from Instituto Nacional de Pesquisas Espaciais (INPE 2019). http://www.dsr.inpe.br/topod​ata/ S Slope Calculated from elevation

ON/S Orientation N/S Calculated from slope Land use AGR​ Agriculture (%) Extracted from the Brazilian Institute of Geography and DOF/MOF Dense Ombrophilous Forest/Mixed Ombrophilous Forest (%) Statistics—IBGE (Instituto Brasileiro de Geografa e Estatística) (2004) UMDOF Upper Montane Dense Ombrophilous Forest (%) LIV Livestock (%) SF/MOF Seasonal Forest/Mixed Ombrophilous Forest (%) SSMF Seasonal Semideciduous Montane Forest (%) UMVR Upper Montane Vegetational Refuges (%) FS Forested Savannah (%) GWS Gramineous-Woody Savannah (%) SVI Secondary Vegetation Initial (%) Human activities PobDen Population density (inhabitants’ number × [km2]−1) Gridded Population of the World (GPW-v4). Socioeco- nomic Data and Applications Center (SEDAC). Hosted by CIESIN at the Columbia University. 2010. https​://sedac​ .ciesi​n.colum​bia.edu/data/colle​ction​/gpw-v4/ Vegetation bands HAF High-Altitude Fields (%) (above 2000 a.s.l.) Based on the classifcation described by Segadas-Vianna and UMF Upper Montane Forest (%) (ca. 1501 to 2000 m. a.s.l.) Dau (1965) MF Montane Forest (%) (ca. 501 to 1500 m. a.s.l.),

function (FF) to predict the location of areas, where maned conditions for the species to be present. The FF refects the wolf home range activities may occur (Real et al. 2006; Coe- degree (between 0 and 1) to which the probability values lho et al. 2018). This function consists of a multifactorial obtained in the model difer from that expected according logistic regression, with a model selection based on a for- to the species’ prevalence, where 0.5 indicates no diference ward–backward stepwise procedure and on Akaike Informa- between both probability values. Probability depends both tion Criteria or AIC (Burnham and Anderson 2002). Model on the response of the species to the predictors and on the building and selection was implemented using the fuzzySim overall prevalence of the species (Cramer 1999), whereas package (Barbosa 2016) for R (R Core Team 2017). The favourability values only refect the response of the species probability of maned wolf presence for each grid cell was to the predictors (Acevedo and Real 2012). Favourability used to calculate the favourability values (F) according to values were categorized into high (areas with favourabil- the following equation: ity values higher than 0.8), intermediate (with favourability values between 0.21 and 0.80), and low (with favourability F = P∕(1 − P) (n1∕n0) + (P∕[1 − P]) ,   values less than 0.2) favourability. This categorization is equivalent to defning a prediction with odds higher than where P is the probability value obtained from the multifac- 4:1 as favourable and lower than 1:4 as unfavourable (Real torial logistic regression, and n1 and n0 are grid numbers et al. 2006). corresponding to presences or pseudo-absences, respec- Prior to the modelling procedure, we standardized all tively. We applied the FF to identify the areas that contain predictors to avoid bias in the modelling results associated favourable conditions regardless of the presence/pseudo- with the diference in scale of the continuous variables. To absence ratio (Real et al. 2006), and, therefore, suitable avoid excessive multicollinearity in the model, we checked

1 3 Author's personal copy

192 I. Aximof et al. pair-wise variable correlations (Pearson’s test) applied the was slightly better at classifying territories as being of low “corSelect” function from fuzzySim package, and the Vari- favourability for maned wolf activity (specifcity = 0.743), ance Infation Factor or VIF (Zuur et al. 2009; Zar 2010). than at classifying territories as being favourable (sensibil- We evaluated the discrimination and classifcation capac- ity = 0.621) (Table 2). ity of the models with the modEvA R package (Barbosa 2016). Specifcally, we calculated the discrimination ability Territory with maned wolf activity of the models using the AUC (Lobo et al. 2008), and the classifcation capacity by assessing sensitivity, specifcity, We detected 54 locations with maned wolf presence: 36 con- Kappa and Correct Classifcation Rate (CCR) values (taking frmed by faeces, 11 by footprints and 7 by visual observa- the value of F = 0.5 as the classifcation threshold). Finally, tion. The maned wolf presence records defned two main we checked the relative weight of the variables in the model populations in the study area: one at the far north in the using a Wald’s test (Wald 1943) through the survey pack- SPSP, and another at the southern end in the INP, with age (Lumley 2004, 2018). Responses to interview questions smaller populations scattered in between (Fig. 2). All these were expressed as percentage of interviewees mentioning populations were located in areas with the highest concen- the focal issue. trations of high-altitude felds above 2000 m a.s.l. (“Appen- dix”, Fig. 4). These include two small populations, one in the north and another on the edge of the SPSP, that are located Results in areas, where part of the natural felds have been converted to pastures, mainly signal grass (Brachiaria decumbens) and Model assessment molasses grass (Melinis minutifora). All the regions within the Serra da Mantiqueira Moun- Correlations between predictor variables used in the model tains study area were classifed as being of at least inter- procedure were always below 0.7 (p > 0.05) (Table 1 and mediate favourability for maned wolves (F ≥ 0.2). A core “Appendix”, Table 4), and infation values were lower than regional nucleus of high favourability (F ≥ 0.8) for maned 1.5 (VIF values up to 10 are acceptable according to Mont- wolves was identifed in the south of the study area in the gomery and Peck 1992), and commonly accepted as the INP, and three small highly favourable core areas more to cut-of value to consider variables as non-collinear (Zuur et al. 2009). As such, all predictor variables were used in subsequent modelling procedures. The favourability model obtained acceptable scores according to the evaluation indices used to estimate discrimination and classifcation capacities. Discrimination capacity measured by the index AUC was 0.724 or “excellent” according to Hosmer and Lemeshow (2000). In general, the classifcation indices indi- cated that the model correctly classifed favourable territo- ries for maned wolves (CCR > 0.70). Specifcally, the model

Table 2 Evaluation of the maned wolf favourability model, showing the discrimination and classifcation indices values Evaluation indexes Maned wolf with favourability model

Discrimination AUC​ 0.723 Classifcation (thresholds of 0.5) Sensitivity 0.507 Specifcity 0.823 CCR​ 0.742 UPR 0.172 Fig. 3 Maned wolf favourability map, showing the favourable areas OPR 0.5 for maned wolves in the Itatiaia National Park (INP) and the Serra do Papagaio State Park (SPSP). Favourability values are shown in AUC ​area under the ROC (receiving operating characteristic) curve, grayscale; from 0 to 1, where 0 indicates unfavourable areas and 1 CCR ​ correct classifcation rate, UPR under prediction range, OPR the most favourable areas according to the environmental conditions over prediction range analysed. No grid obtained unfavourable values (f < 0.2)

1 3 Author's personal copy

Unravelling the drivers of maned wolf activity along an elevational gradient in the Atlantic… 193

Table 3 Predictor variables included in the maned wolf favourability there were no reported conservation conficts (death by model retaliation) within the INP. Rangers reported one adult and Variable Sign Wald one juvenile maned wolf consuming rubbish in a dump and exploring camping areas of the INP. The use of such High-altitude felds (+) 27.953 anthropogenic areas was confrmed by scat analysis, as we Slope (+) 7.072 detected plastic material in maned wolf faeces (“Appendix”, Upper montane vegetation refuges (%) (+) 5.566 Fig. 5 photo C). Agriculture (%) (+) 4.776 Dense ombrophilous forest (%) (+) 4.093

Signs in brackets show the positive or negative relationship between Discussion favourability and the variables in the models. The Wald parameter indicates the relative weight of every variable in the Maned Wolf Maned wolf activity and drivers in the Serra da model Mantiqueira

The Serra da Mantiqueira has been described as a central the north in the SPSP. Furthermore, 92% of the grid cells part of the maned wolf’s distributional range, based both classifed as being highly favourable for maned wolf pres- on high population density and on high favourability com- ence were in high-altitude felds, above 2000 m a.s.l (Fig. 3). pared with other parts of its range (Queirolo et al. 2011; Coelho et al. 2018). However, our results show that there Environmental drivers of favourability for maned is a gradient of favourability of areas for the maned wolf wolves within the Serra da Mantiqueira when examined at the local scale. Indeed, we found the areas with the most favourable Only fve of the eighteen candidate environmental predictors conditions for maned wolf activity are concentrated at the explained the activity patterns of maned wolves in the study far north of the SPSP and in the southern part of the study area: high-altitude felds; slope; high coverage of upper mon- area, on the border between SPSP and INP. tane vegetation refuges; high coverage of agriculture felds; In common with other authors (Dietz 1984; Queirolo and dense-mixed ombrophilous forest, ordered in decreasing et al. 2011; Coelho et al. 2018), we found that areas of high relative importance according to a Wald’s test (Wald 1943) altitude (above 2000 m a.s.l.) with low cover of upper mon- (Table 3). tane forest, and with open habitats such as high-altitude With respect to human activity in the study area and sur- felds or even agricultural felds, are more favourable for roundings, we observed domestic animals (horses and cattle) maned wolves. Areas of upper montane forest likely appear and evidence of grazing, even in natural felds in the SPSP. favourable as a result of the abrupt classifcation of the Beyond this, more than half the local people interviewed vegetation (Segadas-Vianna and Dau 1965), when in real- mentioned that they had seen maned wolves in areas sur- ity there is a relative transition between the diferent types rounding the INP (n = 37; 52.8%); and half of them stated of vegetation, forming a mosaic (“Appendix”, Fig. 4). For that these records have increased in the last 5 years (n = 35; example, in the SPSP area up to 2314 m a.s.l., there are large 50%), especially in rural areas of Resende county (“Appen- areas of high-altitude felds that form a matrix containing dix”, Fig. 5 photo A). Moreover, the team of environmental natural patches of upper montane forest of diferent sizes managers from the municipalities of Itatiaia and Resende (see Ribeiro et al. 2018). The model indicated a positive reported maned wolf presence in industrial areas of both relationship between the coverage of agricultural areas and cities (“Appendix”, Fig. 5 photos E–G). In terms of detecting the regions with higher favourability for maned wolf activity, human conficts in the surroundings of the study area, on the which probably refects the relative ease with which maned one hand, only eighteen percent of the interviewed local peo- wolves can move through and use these open areas, com- ple reported maned wolf attacks on small domestic animals pared with the Atlantic forest vegetation they have replaced (n = 13; 18.5%), such as chickens. On the other hand, ten (Dietz 1984; Queirolo et al. 2011; Coelho et al. 2018). In percent reported maned wolf killing as a retaliation measure fact, the agricultural areas in our study region are concen- (n = 7; 10%), and only four reported deaths as a result of road trated mainly in the north of the SPSP, as well as in the collisions (n = 4; 5.7%). surrounding areas further south, where the model indicated Informal interviews with park rangers also confrmed intermediate and high favourability. the use of anthropogenic areas by maned wolves. However,

1 3 Author's personal copy

194 I. Aximof et al.

Maned wolf activity in high‑altitude felds the two protected areas studied. Despite detecting maned in the Atlantic forest mountains wolf activity in the vicinity of urban and peri-urban envi- ronments (see photos in “Appendix”, Fig. 5), the model- Modelling showed that the high-altitude natural open felds ling procedure did not identify human infuence within of the Serra da Mantiqueira are more favourable for activity protected areas as a direct driver of maned wolf activity. of maned wolves. This is perhaps unsurprising given that Indeed, although domestic grazers such as horses and cat- these areas show similar structural characteristics to the tle were observed in the SPSP, the number of animals per grasslands of the Brazilian savannah, where the maned wolf unit area is very low and the SPSP management team has evolved (Coelho et al. 2008; Jácomo et al. 2009; Queirolo on-going programs to eliminate livestock grazing within the et al. 2011). This ecosystem is restricted to higher elevations park (Carvalho et al. 2015; Mendonça 2017). This result of the Atlantic forest in south-eastern and southern Brazil, shows that the two protected areas have been efective in and has many endemic and endangered plant and animal the conservation of this near endangered canid. On the other species (Martinelli 1996; Saford 1999; Aximof 2011; Car- hand, the surveys surrounding the Serra da Mantiqueira valho et al. 2015; Ortiz et al. 2017; Aximof et al. 2018). mountains (in the bufer zone) highlighted the existence of Although Rosa et al. (2015) detected maned wolf faeces in some human-maned wolf conficts related with incidences semi-deciduous forests in the county of Itamonte, our model of attacks on poultry, with subsequent persecution and retali- indicated it is common for species to avoid the forested atory killing of maned wolves. So, we would recommend areas, especially those with a dense understory which makes management and awareness plans in the towns near these movement more difcult for maned wolves (Childs-Sanford points. Furthermore, as we have documented (see “Appen- 2005; Coelho et al. 2008; Massara et al. 2012). dix”, Fig. 5 photo C), consumption of waste by maned The higher favourability of high-altitude grasslands for wolves has been increasing in urban areas, but also in pro- maned wolves may also be related to their feeding hab- tected areas, with unknown consequences for maned wolf its. Solanum lycocarpum (Solanaceae) is one of the main populations (Aragona and Setz 2001; Massara et al. 2012; food sources for maned wolves in the Serra da Mantiqueira Silva and Talamoni 2003). Such increases, typical of more (Rosa et al. 2015). This plant occurs in both Brazilian savan- generalist species, should be monitored to identify the tip- nahs and the Atlantic Forest, being more common in open ping point, from which the negative impact on maned wolfs habitats (Mentz and Oliveira 2004; Stehmann et al. 2014). starts to be greater than the possible advantages of exploring Moreover, Solanum is a genus that exhibits higher species easy to access and easy to use resources close to humans. richness and abundance with increasing elevation along the In addition, educational programs on the correct disposal mountain ranges of South America. The occurrence of Sola- of rubbish, with more efective enforcement (perhaps with num sp. in several regions, and mainly in higher areas in fnes), should be carried out with tourists, researchers and mountain ranges in the Atlantic Forest (Knapp 2002), may employees of both Protected Areas. also be contributing to the spread of the maned wolf into those areas (Bueno and Motta-Junior 2009). Indeed, given Maned wolf conservation in the Serra da that records of maned wolf in the high-altitude grasslands Mantiqueira mountains have been made in the region for several decades, regis- tered for the frst time in 1954 by Avila-Pires and Gouvea The Serra da Mantiqueira mountains are considered a con- (1977), we suspect that the occurrence of maned wolves in servation priority site within the Atlantic Forest, owing to these areas is probably natural and not the result of the cur- large forest remnants and threatened high-altitude natural rent colonization of recently cleared forest, as detected in grassland felds that host biodiversity-rich communities of other regions (Bueno and Motta-Junior 2009; Queirolo et al. animals and plants (Le Saout et al. 2013; Martinelli 2007; 2011). This result provides quantitative support for the idea Ribeiro et al. 2009; Rodrigues and Oliveira 2006). It should that the substitution of Atlantic Forest vegetation with open be noted that the threats that currently impact the region anthropogenic environments is one of the drivers of the local include real estate speculation, construction of hydroelectric expansion of the maned wolf. dams and opening of felds for mining (Carvalho et al. 2015; Ferreira et al. 2014). Thus, conservation measures aiming to Human‑maned wolf conficts maintain the integrity of high-altitude grasslands are crucial to ensure that maned wolf habitat requirements are met, thus Human-wildlife confict is one of the main threats identi- avoiding population decline as required in its management fed for maned wolves, especially in rural areas (Paula et al. plan (see Paula et al. 2008). The species has large home 2013), which are the dominant landscapes in the vicinity of range requirements, and as such conservation programs for

1 3 Author's personal copy

Unravelling the drivers of maned wolf activity along an elevational gradient in the Atlantic… 195 maned wolves must help to guarantee the preservation of these results provide useful data for maned wolf manage- large areas of land (Caro 2010), including the bufer zones of ment at the core of its current global distribution. protected areas, which are also very important for the local fauna (Paolino et al. 2016; Xavier et al. 2018). Acknowledgements We thank the Instituto Chico Mendes de Con- servação da Biodiversidade (ICMBio—SISBIO license 16968-1) and Instituto Estadual de Florestas de Minas Gerais (IEF—license 151/11—Extensão I), the administration of both protected areas for logistical support during this study. WDC is supported by Coordination Conclusions for the Improvement of Higher Education Personnel (CAPES) through a postdoctoral scholarship (CAPES-PNPD). LMR was fnancially sup- ported by the University of Aveiro (Department of Biology), CESAM We highlight that this is the frst study to use fuzzy logic (UID/AMB/50017), FCT/MEC through national funds, and the co- tools at the local scale to analyze the favourability of terri- funding by the FEDER, within the PT2020 Partnership Agreement and tories for maned wolf activity in a highly favourable region Compete 2020. We also thank Clarissa Alves da Rosa for important contributions as a reviewer and Karen Mustin for the English review along an elevational gradient. We applied a novel methodol- of this manuscript. Dr. David Romero was supported by a postdoctoral ogy to determine the regions more likely to be used by the grant from the Graduate Academic Commission (CAP, from Spanish maned wolf in protected areas in the core of its global dis- acronym Comisión Académica de Posgrado) of the Universidad de la tribution in Brazil. These results indicate that in the Serra da República (2018–2020). Mantiqueira Mountains (INP and SPSP protected areas) the maned wolf populations show a structure-oriented habitat Compliance with ethical standards preference, specifcally for grasslands or other open envi- Conflict of interest ronments such as agricultural felds. This may indicate that The authors declare that they have no confict of interest. maned wolf conservation actions in the Serra da Mantique- ira should be adapted to specifc characteristics of locally favourable territories. This type of territorial management Appendix approach has previously been highlighted by other authors as a suitable approach for maned wolf conservation. As such, See Figs. 4 and 5 and Table 4.

1 3 Author's personal copy

196 I. Aximof et al. altitude vegetation bands and b altitude vegetation of land covers, classes used in the model and the a Types transect. location of sampling Main land-cover 4 Fig.

1 3 Author's personal copy

Unravelling the drivers of maned wolf activity along an elevational gradient in the Atlantic… 197

Fig. 5 Maned wolf records. a Rural area in Resende county, b high-altitude grasslands in Itati- aia National Park, c faeces with plastic, d camera-trap record in INP, e next to the park ranger’s house in INP, f urban area in Itatiaia county, g urban area in Resende county, h on a dirt road in Itamonte County, inside Itatiaia National Park

1 3 Author's personal copy

198 I. Aximof et al. 1.00 PobDen 1.00 − 0.2611 UMF 0.4523 1.00 − 0.3050 MF 0.1263 1.00 − 0.4608 − 0.2394 HAF 0.1577 0.0725 0.0484 1.00 − 0.0216 S 0.2040 0.2258 0.0722 1.00 − 0.0767 − 0.0242 OriEO 0.1132 0.1618 0.0118 1.00 0.8284 − 0.0190 − 0.0536 OriNS 0.0072 0.0190 0.0301 1.00 − 0.0160 − 0.0180 − 0.0361 − 0.0142 SVI 0.2400 0.0240 0.0337 1.00 − 0.0059 − 0.1242 − 0.0497 − 0.0867 − 0.0352 GWS 0.0473 0.1169 1.00 0.1511 0.0417 0.0358 − 0.0642 − 0.0218 − 0.0202 − 0.0034 FS 0.2374 0.5851 0.0079 1.00 0.0802 0.0408 0.0316 0.0253 0.0838 − 0.3123 − 0.1125 UMVR 0.0325 0.0136 1.00 − 0.0343 − 0.0190 − 0.0180 − 0.0122 − 0.0024 − 0.0030 − 0.0054 − 0.0530 − 0.0314 SSMF 0.5127 0.9056 0.1044 1.00 0.1890 0.2614 − 0.2807 − 0.1903 − 0.0892 − 0.0151 − 0.0173 − 0.0217 − 0.0396 DOF/ MOF 0.3281 1.00 0.0194 − 0.2386 − 0.1263 − 0.1362 − 0.2084 − 0.1431 − 0.1658 − 0.0304 − 0.0106 − 0.0665 − 0.1172 − 0.2045 LIV 0.5004 0.1435 0.1499 1.00 0.0318 0.0091 − 0.1048 − 0.3030 − 0.3347 − 0.2363 − 0.2646 − 0.0504 − 0.0353 − 0.0344 − 0.0167 UMDOF 0.0796 0.0158 0.0214 1.00 0.0038 0.0162 − 0.0179 − 0.0466 − 0.0367 − 0.0172 − 0.0714 − 0.0146 − 0.0029 − 0.0033 − 0.0042 − 0.0076 SF/MOF 0.0179 0.0246 0.0001 0.0177 0.0298 1.00 0.0282 0.0202 − 0.0063 − 0.0189 − 0.0043 − 0.0321 − 0.0212 − 0.0030 − 0.0035 − 0.0044 − 0.0079 ​ AGR 0.1178 0.6318 0.0408 0.4196 0.0718 1.00 0.0135 0.3420 0.0139 0.0676 0.0067 0.1574 − 0.1425 − 0.4480 − 0.0382 − 0.4065 − 0.0807 − 0.0968 ME Pearson’s correlation test Pearson’s MOF PobDen UMF MF HAF S UMVR LIV SF/MOF DOF/ ME ​ AGR UMDOF SSMF FS SVI GWS OriNS OriEO 4 Table 1 . In italics and without 0.7 correlation used in the due to bold the not modeling procedure variables code names as in Table or multicollinearity correlation issues. In red, above values Variables

1 3 Author's personal copy

Unravelling the drivers of maned wolf activity along an elevational gradient in the Atlantic… 199

References brachyurus) in South America: spatial dynamics and environmen- tal drivers. Mammal Biol 92:54–61 Cramer JS (1999) Predictive performance of binary logit model in Acevedo P, Real R (2012) Favourability: concept distinctive character- unbalanced samples. J R Stat Soc 48:85–94 istics and potential usefulness. Naturwissenschaften 99:515–522 Curi NHA, Araújo AS, Campos FS, Lobato ZIP, Gennari SM, Marvulo Aragona M, Setz E (2001) Diet of the maned wolf, Chrysocyon MFV, Silva JCR, Talamoni SA (2010) Wild canids, domestic dogs brachyurus (Mammalia: Canidae), during wet and dry seasons at and their pathogens in Southeast Brazil: disease threats for canid Ibitipoca State Park, Brazil. J Zool 254:131–136 conservation. Biodivers Conserv 19:3513–3524 Avila-Pires FD, Gouvea E (1977) Mamíferos do Parque Nacional Dietz JM (1984) Ecology and social organization of the maned do Itatiaia. Boletim do Museu Nacional, Nova Série Zoologia wolf (Chrysocyon brachyurus). Smithsonian Institution Press, 291:1–29 Washington Aximof I (2011) O que perdemos com a passagem do fogo pelos cam- Dietz JM (1985) Chrysocyon brachyurus. Mammal Species 234:1–4 pos de altitude do Estado do Rio de Janeiro. Revista Biodiversi- Eckhardt B (2016) Análise da presença e estratégias de conservação dade Brasileira, Número Temático: Ecologia e Manejo do Fogo do lobo-guará Chrysocyon brachyurus (Illiger, 1815) no estado em Áreas Protegidas. 2:180–200 do Rio de Janeiro. Master thesis, Instituto de Pesquisas Jardim Aximof I, Cronemberger C, Pereira FA (2015) Amostragem de longa Botânico do Rio de Janeiro, Escola Nacional de Botânica Tropical duração por armadilhas fotográfcas dos mamíferos terrestres em Ferreira J, Aragão L, Barlow J, Barreto P, Berenguer E, Bustamante dois parques nacionais no estado do Rio de Janeiro. Oecol Aust M, Gardner T, Lees A, Lima A, Louzada J (2014) Brazil’s envi- 19:215–231 ronmental leadership at risk. Science 346:706–707 Aximof I, Carvalho WD, Abdalla LS, Rosa CA (2018) New upper alti- Freitas S, Oliveira A, Ciocheti G, Vieira MV, Matos D (2015) How tudinal range of anteater species in highlands of South America. landscape features infuence road-kill of three species of mammals North-West J, Zool, p e187701 in the Brazilian Savanna. Oecol Aust 18:35–45 Barbosa AM (2016) fuzzySim: applying fuzzy logic to binary similar- Geise L, Pereira L, Bossi D, Bergallo H (2004) Pattern of elevational ity indices inecology. Methods Ecol Evol 6:853–858 distribution and richness of non volant mammals in Itatiaia Barbosa AM, Real R, Vargas JM (2010) Use of coarse-resolution mod- National Park and its surroundings, in southeastern Brazil. Braz els of species’ distributions to guide local conservation inferences. J Biol 64:599–612 Conserv Biol 24:1378–1387 Guisan A, Zimmermann NE (2000) Predictive habitat distribution Barreto CG, Campos JB, Schwarzstein NT, Alves GSG, Coelho W models in ecology. Ecol Model 135:147–186 (2013) Análise da Região da Unidade de Conservação. In: Insti- Guisan A, Tingley R, Baumgartner JB, Naujokaitis-Lewis I, Sutclife tuto Chico Mendes de Conservação da Biodiversidade (ed) Plano PR, Tulloch AIT, Regan TJ, Brotons L, McDonald-Madden E, de Manejo Parque Nacional do Itatiaia, Encartes 2 e 3. Ministério Mantyka-Pringle C, Martin TG, Rhodes JR, Maggini R, Setter- do Meio Ambiente, Brasília, pp 1–117 feld SA, Elith J, Schwartz MW, Wintle BA, Broennimann BAO, Beca G, Vancine MH, Carvalho CS, Pedrosa F, Alves RSC, Buscariol Austin M, Ferrier S, Kearney MR, Possingham HP, Buckley YM D, Peres CA, Ribeiro MC, Galetti M (2017) High mammal species (2013) Predicting species distributions for conservation decisions. turnover in forest patches immersed in biofuel plantations. Biol Ecol Lett 16:1424–1435 Conserv 210:352–359 Hosmer DW, Lemeshow S (2000) Applied logistic regression, 2nd edn. Becker M, Dalponte JC (1991) Rastros de mamíferos silvestres brasilei- Wiley, New York ros: um guia de campo. Editora Universidade de Brasília, Brasília Hosmer DW, Lemeshow S, Sturdivant RX (2013) Applied logistic Bereta A, Freitas SR, Bueno C (2017) Novas ocorrências de Chryso- regression. Wiley, New York cyon brachyurus (Carnivora) no estado do Rio de Janeiro indi- IBGE (Instituto Brasileiro de Geografa e Estatística) (2004) Mapa de cando a expansão de sua distribuição geográfca. Bol Soc Bras Biomas do Brasil e o Mapa de Vegetação do Brasil. http://www. Mastozool 78:5–8 ibge.gov.br/home/presi​denci​a/notic​ias/notic​ia_visua​liza.php?id_ Bueno AA, Motta-Junior JC (2009) Feeding habits of the maned wolf, notic​ia=169. Accessed 20 Jan 2019 Chrysocyon brachyurus (Carnivora: Canidae), in southeast Brazil. IBGE (Instituto Brasileiro de Geografa e Estatística) (2016) Resolução Stud Neotrop Fauna Environ 44:67–75 nº4 publicada em 29/08/2016. Diário Ofcial da União, Seção 1, Burnham KP, Anderson DR (2002) Model selection and inference: página 47. http://www.jusbr​asil.com.br/diari​os/12411​7274/dou- a practical information-theoretic approach. Springer, New York secao​-1-30-08-2016-pg-47. Accessed 20 Jan 2019 Caro T (2010) Conservation by proxy: indicator, umbrella, keystone, IEF-MG (Instituto Estadual de Florestas de Minas Gerais) (2008) Plano fagship, and other surrogate species. Island Press, Washington de Manejo do Parque Estadual da Serra do Papagaio. http://www. Carvalho WD, Xavier BS, Esbérard CEL (2015) Primatas do Parque ief.mg.gov.br/image​s/stori​es/Plano​_de_Manej​o/serra​_papag​aio/ Estadual da Serra do Papagaio e RPPNs adjascentes, estado de encar​te%20ii.pdf. Accessed 15 Jan 2019 Minas Gerais. Neotrop Primates 22:25–31 INPE (Instituto Nacional de Pesquisas Espaciais) (2019) Topodata: Carvalho WD, Martins MA, Esbérard CEL, Palmeirim JM (2019) banco de dados geomorfométricos do Brasil. Variáveis geomor- Traits that allow bats of tropical lowland origin to conquer moun- fométricas locais. São José dos Campos. http://www.dsr.inpe.br/ tains: bat assemblages along elevational gradients in the South topod​ata/. Accessed 20 Jan 2019 American Atlantic Forest. J Biogeogr 46:316–331 Jácomo ATA, Kashivakura CK, Ferro C, Furtado MM, Astete SP, Childs-Sanford SE (2005) The captive maned wolf (Chrysocyon brach- Tôrres NM, Sollmann R, Silveira L (2009) Home range and spa- yurus): nutritional considerations with emphasis on management tial organization of maned wolves in the Brazilian grasslands. J of cystinuria. Master thesis, University of Maryland Mammal 90:150–157 Coelho CM, De Melo LFB, Sábato MAL, Magni EMV, Hirsch A, Knapp S (2002) Assessing patterns of plant endemism in neotropical Young RJ (2008) Habitat use by wild maned wolves (Chryso- uplands. Bot Rev 68:22–37 cyon brachyurus) in a transition zone environment. J Mammal Koeppen W (1948) Climatologia: con un estudio de los climas de la 89:97–104 tierra. Universidad Católica de Córdoba, Córdoba Coelho L, Romero D, Queirolo D, Guerrero JC (2018) Understanding Le Saout S, Hoffmann M, Shi Y, Hughes A, Bernard C, Brooks factors afecting the distribution of the maned wolf (Chrysocyon TM, Bertzky B, Butchart SH, Stuart SN, Badman T (2013)

1 3 Author's personal copy

200 I. Aximof et al.

Protected areas and efective biodiversity conservation. Science R Core Team (2017) R: a language and environment for statistical com- 342:803–805 puting. R Foundation for Statistical Computing. Vienna, Austria. Lobo J, Jiménez-Valverde A. & Real R (2008) AUC: a misleading https​://www.R-proje​ct.org/. Accessed 15 Nov 2018 measure of the performance of predictive distribution models. Real R, Barbosa AM, Vargas JM (2006) Obtaining environmental Glob Ecol Biogeogr 17:145–151. favourability functions from logistic regression. Environ Ecol Lumley T (2004) Analysis of complex survey samples. J Stat Softw Stat 13:237–245 9:1–19 Real R, Barbosa AM, Bull JW (2017) Species distributions, quantum Lumley T (2018) Survey: analysis of complex survey samples. R pack- theory, and the enhancement of biodiversity measures. Syst Biol age version 3.35. https://cran.r-proje​ ​ct.org/web/packa​ges/surve​y/ 66:453–462 index​.html. Accessed 15 Nov 2018 Reis NR, Peracchi AL, Rossaneis BK, Fregonezi MN (2010) Técnicas Martinelli G (1996) Campos de altitude. Editora Index, Rio de Janeiro de estudos aplicados aos mamíferos silvestres brasileiros. Techni- Martinelli G (2007) Mountain biodiversity in Brazil. Braz J Bot cal Books Editora, Rio de Janeiro 30:587–597 Ribeiro MC, Metzger JP, Martensen AC, Ponzoni FJ, Hirota MM Massara RL, Paschoal A, Hirsch A, Chiarello A (2012) Diet and habitat (2009) The Brazilian Atlantic Forest: how much is left, and how use by maned wolf outside protected areas in eastern Brazil. Trop is the remaining forest distributed? Implications for conservation. Conserv Sci 5:284–300 Biol Conserv 142:1141–1153 Massara R, Paschoal A, Chiarello A (2015) Gastrointestinal parasites Ribeiro JHC, Santana LD, Carvalho FA (2018) Composition, structure of maned wolf (Chrysocyon brachyurus, Illiger 1815) in a subur- and biodiversity of trees in tropical montane cloud forest patches ban area in southeastern Brazil. Braz J Biol 75:643–649 in Serra do Papagaio State Park, southeast Brazil. Edinb J Bot Mendonça JGF (2017) Campos de altitude do Parque Estadual da Serra 75:255–284 do Papagaio, Minas Gerais, Brasil: Composição forística, fto- Rodrigues F, Oliveira T (2006) Unidades de conservação e seu papel geografa e estrutura da vegetação. Master thesis, Universidade na conservação de carnívoros brasileiros. In: Morato GM, Rodri- Federal de Juiz de Fora gues FHG, Eizirik E, Mangini PR, Azevedo FCC, Marinho-Filho Mentz LA, Oliveira PL (2004) Solanum (Solanaceae) na região sul do J (eds) Manejo e Conservação de carnívoros tropicais. Edições Brasil. Instituto Anchietano de Pesquisas, São Leopoldo IBAMA, Brasília, pp 97–110 MMA (Ministério do Meio Ambiente) (2014) Portaria do Ministério Romero D, Olivero J, Brito JC, Real R (2016) Comparison of do Meio Ambiente nº 444, de 17 de dezembro de 2014: Espé- approaches to combine species distribution models based on dif- cies da fauna brasileira ameaçadas de extinção. http://www. ferent sets of predictors. Ecography 39:561–571 icmbio.gov.br/porta​ l/image​ s/stori​ es/docs-plano​ -de-acao/00-saiba​ ​ Romero D, Olivero J, Real R, Guerrero JC (2019) Applying fuzzy logic -mais/04-_PORTA​RIA_MMA_N%C2%BA_444_DE_17_DE_ to assess the biogeographical risk of dengue in South America. DEZ_DE_2014.pdf. Accessed 13 Mar 2019 Parasit Vectors. https​://doi.org/10.1186/s1307​1-019-3691-5 Montgomery DC, Peck EA (1992) Introduction to linear regression Rosa CA, Santos KK, Faria GMM, Puertas F, Passamani M (2015) Die- analysis. Wiley, New York tary behavior of the Maned wolf Chrysocyon brachyurus (Illiger, Myers N, Mittermeier RA, Mittermeier CG, Da Fonseca GA, Kent J 1815) and the record of predation of Brown tinamou Crypturellus (2000) Biodiversity hotspots for conservation priorities. Nature obsoletus (Temminck, 1815) (Tinamiformes, Tinamidae) at Man- 403:853–858 tiqueira Mountains, Southeastern Brazil. Bol Soc Bras Mastozool Myslajek RW, Nowak S, Jedrzejewska B (2012) Distribution, charac- 72:7–10 teristics and use of shelters by the Eurasian badger Meles meles Saford H (1999) Brazilian Páramos I. An introduction to the physical along an altitudinal gradient in the Western Carpathians, S Poland. environment and vegetation of the campos de altitude. J Biogeogr Folia Zool 61:152–160 26:693–712 Oliveira-Filho AT, Fontes MAL (2000) Patterns of foristic diferentia- Segadas-Vianna F, Dau L (1965) Ecology of the Itatiaia Range. Imp- tion among Atlantic Forests in Southeastern Brazil and the infu- rensa Nacional, Southeastern Brazil ence of climate 1. Biotropica 32:793–810 Silva JA, Talamoni SA (2003) Diet adjustments of maned wolves, Ortiz FR, Menezes FA, Abegg AD, Landroz EMJ, Feio RN (2017) Chrysocyon brachyurus (Illiger) (Mammalia, Canidae), subjected A new record for Melanophryniscus moreirae (Miranda-Ribeiro, to supplemental feeding in a private natural reserve, Southeastern 1920) (Amphibia, Anura, Bufonidae) in the state of Minas Gerais, Brazil. Rev Bras Zool 20:339–345 Brazil. Check List 13:2140 Silveira L, Furtado MM, Tôrres NM, Sollmann R, Uhl G, Jácomo Paolino RM, Versiani NF, Pasqualotto N, Rodrigues TF, Krepschi VG, ATA (2009) Maned wolf density in a Central Brazilian Chiarello AG (2016) Bufer zone use by mammals in a Cerrado grassland reserve. J Wildl Manag 73:68–71. https​://doi. protected area. Biota Neotrop 16:e20140117 org/10.2193/2008-051 Paula RC, Medici P, Morato RG (2008) Plano de Ação para Con- Simas FNB, Schaefer CEGR, Fernandes Filho EI, Chagas AC, Brandão servação do Lobo-guará: análise de viabilidade populacional e PC (2005) Chemistry, mineralogy and micropedology of highland de hábitat (PHVA). MMA/ICMBio/CENAP, Brasília soils on crystalline rocks of Serra da Mantiqueira, southeastern Paula R, Rodrigues F, Queirolo R, Lemos F, Rodrigues L (2013) Brazil. Geoderma 125:187–201 Avaliação do estado de conservação do Lobo-guará Chrysocyon Stehmann J, Mentz L, Agra M, Vignoli-Silva M, Giacomin L, Rodri- brachyurus (Illiger, 1815) no Brasil. Biodiversidade Brasileira gues I (2014) Solanaceae in lista de espécies da Flora do Brasil. 3:146–159 Jardim Botânico do Rio de Janeiro, Rio de Janeiro Queirolo D, Motta-Junior JC (2007) Prey availability and diet of maned Ururahy JCC, Collares JER, Santos MM, Barreto RAA (1983) Folhas wolf in Serra da Canastra National Park, sotheastern Brazil. Acta SF.23/24 Rio de Janeiro/Vitória; geologia, geomorfologia, pedolo- Theriol 52:391–402 gia, vegetação e uso potencial da terra. Projeto RADAMBRASIL: Queirolo D, Moreira JR, Soler L, Emmons LH, Rodrigues FH, Pautasso as regiões ftoecológicas, sua natureza e seus recursos econômicos AA, Cartes JL, Salvatori V (2011) Historical and current range of (4—Vegetação). Estudo Fitogeográfco, Rio de Janeiro the near threatened maned wolf Chrysocyon brachyurus in South Wald A (1943) Tests of statistical hypotheses concerning several America. Oryx 45:296–303 parameters with applications to problems of estimation. Trans Am Math Soc 54:426–482

1 3 Author's personal copy

Unravelling the drivers of maned wolf activity along an elevational gradient in the Atlantic… 201

Xavier MS, Lemos HM, Caccavo A, Bezerra A, Secco H, Gonçalves Zar JH (2010) Biostatistical analysis. Pearson Prentice Hall, Nova PR (2017) Noteworthy coastal records of the maned wolf, Chryso- Jersey cyon brachyurus (Illiger, 1815), in Southeastern Brazil. Bol Soc Zuur A, Feno E, Walker N, Saveliev A, Smith G (2009) Mixed efects Bras Mastozool 78:9–13 models and extensions in ecology with R. Springer, New York Xavier BS, Carvalho WD, Dias D, Tabosa LO, Santos CEL, Esbé- rard CEL (2018) Bat richness (Mammalia: Chiroptera) in an area Publisher’s Note Springer Nature remains neutral with regard to of montane Atlantic Forest in the Serra da Mantiqueira, state of jurisdictional claims in published maps and institutional afliations. Minas Gerais, southeast Brazil. Biota Neotrop 18:e20170496. https​://doi.org/10.1590/1676-0611-bn-2017-0496

1 3