Occurrence of Cabassous tatouay ( Cingulata , Dasypodidae ) in Rio Grande do Sul and its potential distribution in southern Brazil

Cabossous tatouay Desmarest, 1804 is considered a rare species in southern South America, and Rio Grande do Sul State, Brazil, records of the species are scarce and inaccurate. This study reports 40 localities for C. tatouay, and provides a map of the species’ potential distribution using ecological niche modeling (ENM). The ENM indicated that in this region C. tatouay is associated with open grasslands, including the areas of “Pampas” and the open fields in the highlands of the Atlantic Forest. This study contributes to the information about the greater naked-tailed armadillo in southern Brazil, and provides data key to its future conservation.

That Cabassous tatouay, greater naked-tailed armadillo, is a poorly known species has been reported by Wetzel (1982), Redford (1985), Ubaid et al. (2010), Gonzales & Lanfranco (2010) and Hayssen (2014).The southernmost records of the species are from the Uruguayan provinces of Maldonado and Lavalleja (Coitiño et al., 2013).From here the range extends northwards to the state of Pará, in northern Brazil.In between the species has been recorded from the southern, southeastern, and central regions of Brazil (Hayssen, 2014).It is also found in southern Paraguay (Hayssen, 2014) and northeastern Argentina, where it appears restricted to Misiones province (Abba et al., 2012).
In southern Brazil, the species appears in the regional listings of both Ihering (1892) and Silva (1994), but in both cases without any information on location or habitat.Oliveira & Vilella (2003) mentioned the occurrence of C. totouay in the west and southwest regions of the state, and Kasper et al. (2007) indicated the occurrence of this species in the central region of RS, where it may be in population decline.The only accurate and confirmed record is from an archaeological site located at east of State, associated to pioneer formations under fluvial influence of the Rio Grande basin.An assessment of potential geographic ranges of armadillos in Brazil (Anacleto et al., 2006) highlighted the small number of C. tatouay records, and suggested that Brazilian Pampa was not a favored habitat.Indeed, C. tatouay does seem relatively uncommon in this habitat, except in Uruguay, the southern limit of its distribution, where some 10 records are available from the Pampas (Coitiño et al., 2013).
There are few ecological data from the northernmost part of the range of C. tatouay.Medri et al. (2011) pointed to the occurrence of the species in forested areas, and its absence from areas of intensive agriculture or severely degraded localities.According to Carter & Encarnação (1983), C. tatouay changes its burrow every day, and does not use the same shelter twice.Medri et al. (2011) reported that the species has a home range of some 409 hectares, feeds exclusively on ants, is predominantly crepuscular or nocturnal, and has only a single offspring at any one time.
The greater naked-tailed armadillo is considered of Least Concern at the global level, with only minor conservation action being proposed (Gonzales & Abba, 2014).However, all available information on the species in Brazil has been derived from a handful of wild-based observations, and these are insufficient to assess the status of the species in the country (Chiarello et al., 2008).Cabassous tatouay is considered vulnerable to extinction in Uruguay, and is a national priority for conservation (Gonzales & Lanfranco, 2010).With regard to Rio Grande do Sul, there is no precise information on the population status or even the basic inventories to enable analysis of the current distribution of the species (Oliveira & Vilella, 2003).
Thus, the present study aims to contribute to the knowledge of C. tatouay in RS by reporting new areas of occurrence, expanding the known distribution of the species, and providing a map of its potential distribution using ecological niche modeling.

MATERIAL AND METHODS
Cabassous tatouay distributional data.All records of C. tatouay were obtained between 2000 and 2012 through monitoring and inventories of mammals in different regions of the Rio Grande do Sul State, plus records of opportunity.The main methods employed were: camera trap records, searching for tracks and signs, nocturnal censuses, ad libitum observations, and monitoring mammalian roadkill on regional highways.The results were part of a general program of mammal surveys in the state and did not specifically target C. tatouay and covered a range of regional environments.Thus, any reported habitat selection is not the result of selectivity in survey design.
Specimens found dead and with good conditions were deposited in the scientific collections of the Museu de Ciências Naturais, Universidade Luterana do Brasil, Canoas, RS.For specific determination of C. tatouay, we followed Eisenberg & Redford (1999), Achaval et al. (2007), Medri et al. (2011), and Gonzales & Lanfranco (2010) in using the following as diagnostic external characteristics: body large and robust (6-12 kg), shell with 10-13 flexible dorsal bands and tail lacking osteoderms; head large, cephalic shield with symmetrical plates; ears large, widely-separated, and with granular external surface; snout short and wide; front and back legs both with five digits, and large nails, especially on the third digit; dental formula varying between 7 to 10 upper, and 8 to 9 lower teeth.Sampling sites were widely dispersed throughout RS state, and a variety of habitat types including forests, open fields in the highlands of the Atlantic Forest, and formations of Brazilian Pampas.All records of C. tatouay were georeferenced.
Ecological niche modeling.A potential distribution map was produced using the Maxent algorithm version 3.2.1 (Phillips et al., 2006)  The quality of the model prediction was evaluated using ROC (Receiver Operating Characteristics) that relates the sensitivity and specificity parameters of the model (Phillips et al., 2006).The calculation of the area under the ROC curve, also known as Area Under the Curve (AUC), provides a single measure of the model performance.The AUC ranges from 0 to 1, where values close to 1 indicate high performance, while those less than 0.5 indicate poor performance (Peterson et al., 2008).To evaluate the sensitivity, we tracked the number of test points present in the area predicted by the model (Elith et al., 2006).To identify the variables that most influenced the distribution of C. tatouay, we ran a Jackknife test using Maxent (Phillips et al., 2006).This test measures the predictive effects of each variable in the model when verifying the quality of the model only with the variable in test and the omitted variable in test.
Seven predictor variables were used (Tab.I).These variables were obtained from the Modeling Group for Biodiversity Studies, of the Brazilian Institute for Space Research, AMBDATA/INPE (http://www.dpi.inpe.br/Ambdata/referencias.php).All environmental information was organized in grids in the ASCII-raster format, using the geodetic coordinate projection system "Lat Long," Datum WGS-84, with a spatial resolution of 30 arc-seconds or approximately 1 km 2 .

RESULTS
Field observations registered 40 occurrences of C. tatouay from 30 municipalities within RS.Twenty-seven sites were located in the Pampa biome, eleven in the Atlantic Forest, and two in the areas of transition from Pampa biome to Atlantic Forest (IBGE, 2004) (Tab.II).A total of twelve specimen records were obtained by identifying tracks, ten individuals were directly observed, nine carcasses were obtained from hunters, five carcasses were found as highway roadkill, and four individuals were recorded with camera traps.
For ecological niche modeling, 37 spatially unique records were used.Annual average temperature was the variable that most influenced the model distribution of C. tatouay, followed by precipitation in driest month (Tab.II).
The operating characteristic (OC) curve evaluates the performance of the model.OC analysis gave a value of 0.99, which is regarded as excellent, and indicates that the results were not random.As evaluated by the test points, model sensitivity was 100%, and all the test points positioned in the potential distribution area (Fig. 1).

DISCUSSION
Ecological niche modeling results (Fig. 2) showed the areas most ecological and geographic suitable for C. tatouay were associated with the two main mountainous areas of Rio Grande do Sul State, and the valley between them.In our model the areas with highest scores were Serra Geral in northeastern (with altitudes of 700 -1400 m), "Serra do Sudeste" in southeast, and the "Planície Central", a relatively flat area, between these two montainous regions (Fig. 2).Gallery forests appear to be strongly linked with C. tatouay presence, especially when associated with areas o rock-rich natural grasslands.However, even disturbed areas with farming and the extensive cultivation of soya, rice, or exotic trees were used by the greater naked-tailed armadillo.In the generated ecological niche model, the most important variables for determining C. tatouay presence were average annual temperature (responsible for 52.3% of the variance), and precipitation in driest month (28.3%).These variables were similar to other models for the species, like presented by Abba et al. (2012) and Coitiño et al. (2013).
Because of the heterogeneity of sampling methods, it is impossible to fully quantify the expended sampling effort.However, it is possible to we are confident that the absence of records in north-northwest portion of the state reflects, if not absence, low densities of this species.
There were also no records of C. tatouay in pioneer formations under marine influence in the east.However, it is interesting to note that 67.5% of our records came from Pampas region, a biome that the species is not considered to prefer (Anacleto et al., 2006).We present here more records of C. tatouay for Pampas region than Anacleto et al. (2006) for all Brazil.Exactly why this is so requires further consideration, since the species seems genuinely uncommon in Pampas landscapes in neighboring Argentina and Uruguay (Abba et al., 2012;Coitiño et al., 2013).
Approximately 12.5% of the records came from roadkill.While this demonstrates that highways may be a threat to the conservation of C. tatouay.The extent of this impact cannot currently be measured, because of the lack of systematic sampling targeted at locations where roadkill was confirmed.Although the available studies on this subject are scarce and have not reported C. tatouay (Rosa & Mahus, 2004;Tumeleiro et al., 2006;Hengemühle & Cademartori, 2008), Fontana et al. (2003) listed roadkill as the direct cause of population decline of 2.5% of the species of conservation interest in RS.Indeed the frequency of roadkill may be higher than estimated here, since many animals survive the initial impact only to die concealed in vegetation a few meters from the road.Moreover, unless surveys are frequent, roadkill frequencies may be underestimated because of scavengers: at least some of which are human (on one occasion, a local habitant asked for the carcass while we were examining it).In combination, this shows the need for urgent action by the licensing agencies in Rio Grande do Sul state to implement existing  regulations on the construction of highways in manners that mitigate and minimize the potential for vehicle-related deaths in wildlife living in proximity to highways, and to do this for both existing highways and those under construction or undergoing widening or expansion.About 22% of records were from local hunters, confirming the data of Peters et al. (2011), who reported Dasypodidae to be one of the mammalian groups most impacted by hunting in southwestern RS.Although illegal, this activity is driven by the consumption of meat, armadillo hunting is not the results of nutritional need, but instead is a recreational and cultural practice.As the naked-tailed armadillo is the largest species of armadillo, found in the RS, it is preferred over smaller species such as Dasypus novemcinctus Linnaeus, 1758, Dasypus hybridus Desmarest, 1804, and Euphractus sexcinctus Linnaeus, 1758.
The records collected in the current study, are insufficient to estimate population trends for C. tatouay.However, it is possible to say that C. tatouay can currently be considered one of the rarest species of mid-size mammal in RS.Other species of armadillo are easily recorded as roadkill, and by direct observations and camera traps.When compared with the number of records obtained during the study of other Cingulata and of species from other groups of mid-sized mammals considered as endangered (such as carnivores), it is likely that C. tatouay is now under the threat of regional extinction in RS state.Therefore, it is suggested that conservation actions be directed at C. tatouay, especially in regions indicated as having high ecological and geographical suitability for the species.These actions could involve: the creation of local of field stations or nature parks that have armadillo protection as their prime focus; the implementation of wildlife-friendly construction for existing and new highways; active counterhunting surveillance; enforcement of existing legislation; and encouraging future studies that monitor C. tatouay population trends in RS (as well other areas within its natural range).

Fig. 1 .
Fig. 1.Area Under the Curve produced by the Maxent algorithm, plus sensitivity and specificity response of the ecological niche model for Cabassous tatouay Desmarest, 1804 for the State of Rio Grande do Sul, Brazil.

Fig. 2 .
Fig. 2. Model ecological niche of Cabassous tatouay Desmarest, 1804 produced by Maxent algorithm, and a visualization of its potential distribution in the state of Rio Grande do Sul, Brazil.Stars indicate new locality records for the species, circles indicate historical records to the present study within South America.
by applying the basic parameters suggested by the program and deploying randomization of training points (random seed).Overlapping or very close points were removed, and C. tatouay records were divided into two sets, one for training (75% of points to run the model) and the other for testing (25% of points to evaluate the model).The model of potential geographical distribution generated by Maxent was then imported and edited by the ArcView 3.3 program.