Wild vertebrate roadkill in the Chapada dos Veadeiros National Park, Central Brazil

.


Introduction
Roads have recently been recognized as an important cause of habitat fragmentation and wildlife decline around the world (Coffin 2007).The construction of a new road or highway across a preserved landscape has direct and indirect impacts on local wildlife (Sherewood et al. 2002).Roads may induce direct mortality to wildlife through collisions with vehicles, create barriers to animal movement and consequently isolate populations, produce edge effects, alter animal behavior and reproductive ability, disrupt water supplies to and from wetlands, or intensify the toxic contamination and pollution of habitats alongside roads (Andrews 1990, Spelleberg 1998, Lode ´2000, Trombulak & Frissel 2000).In addition, the creation or enlargement of roads may damage natural environments by creating new road cuttings and embankments that scar the landscape, increase traffic noise, and increase road lighting (Sherewood et al. 2002;Forman et al. 2003).
Because of these threats, a number of recent studies on road ecology have been conducted, especially in Europe, the United States, and Australia (Coffin 2007).These studies have emphasized the need to plan road systems and have indicated that conservation should be a major goal of road planning (Forman et al. 2003).In spite of the great influence of roads on the dynamics of wildlife populations, most studies examined their effects only on specific groups, such as amphibians (Hels & Buchwald 2001), birds (Reijnen et al. 1995), mammals (Meunier et al. 1999), or reptiles (Hartmann et al. 2011), or even individual species, such as the Iberian lynx (Felis pardina) (Ferreras et al. 1992), the eastern barred bandicoot (Perameles gunnii) (Mallick et al. 1998) or the Mediterranean snake (Hierophis viridiflavus) (Capula et al. 2014).A few studies have dealt with the impact of roads on the local vertebrate community, describing the species that are susceptible to road mortality and identifying those whose conservation status may be threatened by roads (Yanes et al. 1995, Lode ´2000, Dodd et al. 2004, Boitet & Mead, 2014).
The effects of roads may be more severe in protected areas (Coelho et al. 2008).A protected area is defined as an area of land that is especially dedicated to the protection and maintenance of biological diversity and that is designated and managed to achieve specific conservation goals.The effectiveness of a protected area depends on conservation of biodiversity not only within the park or reserve but also in its vicinity (Chape et al. 2005, West et al. 2006).In Brazil, most protected areas are crossed by roads or highways, or have transportation infrastructure at or close to their borders, such as Iguac¸u National Park (Parana ´State, 185.262,20 ha), Emas National Park (Goia ´s State, 131.800 ha), and A ´guas Emendadas Ecological Station (Distrito Federal, 10,400 ha), so they experience the previously described effects of roads ( Ca ˆndido Jr. et al. 2002, Rodrigues et al. 2002, Lima & Obara 2004, Coelho et al. 2008).Brazil's road network extends over 1,713,885 km and is expected to increase due to new government incentives (PAC II -Brasil 2010).New studies should be conducted to determine the impact of these new roads on the Brazilian fauna, particularly in protected areas.In this article we investigate roads as a cause of vertebrate mortality in the Chapada dos Veadeiros National Park, located in Central Brazil, and address the composition, seasonal patterns, and influence of environment on wildlife mortality due to roads.

The study area
The study was conducted in Central Brazil, in the state of Goia ´s (GO), on two roads that border the Chapada dos Veadeiros National Park (CVNP) (13°51'S to 14°10'S, and 47°25'W to 47°4 2'W).The main road, GO 239, runs between Alto Paraı ´so and Colinas and comprises 72 km (22 km of paved and 50 km of unpaved road).It provides the main access to the official gate of the CVNP, located in Sa ˜o Jorge village (Figure 1).Additional data was obtained from highway BR 010, which runs between Alto Paraı ´so and Teresina de Goia ´s and includes 64 km of paved road.The CVPN covers approximately 65,512 ha of relatively undisturbed Cerrado.It is located in a mountainous region, the Planalto Central Goiano (Felfili et al. 2007), which has altitudes ranging between approximately 620 m and 1,700 m and includes Central Brazil's highest peak (Pouso Alto, 1,784 m).The climate is type Aw in the Ko ¨ppen classification, receiving annually 1500-1750 mm of a highly predictable and strongly seasonal precipitation, restricted almost entirely to October-March (Nimer 1989).Long-term climatic data from the Chapada dos Veadeiros region are summarized in Figure 2. Average temperatures vary between 20°C and 26°C (Silva et al. 2001).The vegetation is characterized by a predominance of gallery forest formations at low elevations and Cerrado with montane savannas at high elevations (Felfili et al. 2007).

Impact of roads
Roads as a cause of vertebrate mortality were investigated through surveys of animals killed by vehicles and by recording dead vertebrates observed on both roads on non-consecutive days, between November 2006 to April 2008.Road surveys were undertaken once every three days.The surveys were conducted by driving at 25-50 km h -1 along GO 239 and at 60-80 km h -1 along BR 010.When a dead animal was spotted, we stopped the car for closer inspection, identified the animal to the lowest taxonomic level possible, recorded the location with GPS, and removed the carcass from the road.Carcasses in good condition were preserved and added to the vertebrate collections of Universidade de Brası ´lia.
The frequency of roadkill was calculated by dividing the number of individuals killed by the number of kilometers covered.The average frequencies of animals and species killed were compared between the dry season (April to September) and the rainy season (October to March).A paired t-test was used to evaluate the difference between dry and rainy seasons.Multiple regression was used to evaluate whether an association existed among roadkill, precipitation, and the structure of nearby vegetation.

Estimates
We constructed a species accumulation curve for all roadkilled species and for each taxonomic class found on the road, using the individual-based rarefaction method (with the nonparametric Mao Tau estimator and with 95% confidence intervals) (Gotelli & Colwell 2001).The function of richness (Mao Tau) was computed as the accumulation function of individuals of each species.Species richness estimators were computed using nonparametric incidence-based estimators (Bootstrap, Chao 2, ICE, Jacknife 1 and 2) and abundancebased estimators (ACE and Chao 1) (Colwell & Coddington 1994, Colwell 2011).Both species accumulation curve and species richness estimators were performed to evaluate the sampling effort, allowing a better comparison among the groups of vertebrates found and pointing out which groups are underestimated through the method.Species rarefaction and richness estimators were performed with Estimates 8.2.0 software.

Correlation of landscape with roadkill
The attributes of landscapes near roads were assessed at 1 km intervals along the entire extent of the roads, both in the preserve and in the adjacent environment, using a 1:25000 topographic map.Landscape categories were assigned according to the proportions of environmental variables contained within 50 m on both sides of the the road shoulders on each kilometer.The categories included cerrado, grassland, forest, and pasture.In addition, the presence or absence of watercourses and the presence of paved or unpaved roads surfaces were recorded at the same intervals.
To examine the relationship between road-killed vertebrates and the road landscape, we performed a Canonical Correspondence Analysis (CCA; Ter Braak 1986), associating variation in one matrix (vertebrate abundance as the dependent variable) with variation in another matrix (landscape characteristics as the independent variable).In this analysis, we investigated whether a connection existed between specific habitat characteristics and the abundance of particular vertebrate groups.CCA was performed with CANOCO 4.5 (Ter Braak & Smilauer 1998), using the following options: focus scaling set on symmetric, biplot scaling type, downweighting of rare species, Monte Carlo test with 1,000 permutations of the reduced model, and unrestricted permutations.
Vertebrate mortality was significantly higher during the wet season than during the dry season (paired t-test, t ¼ 4.935, df ¼ 285, P o 0.000) and mortality differed between the two seasons for all taxonomic groups: amphibians (paired t-test, t ¼ À 2.576, df ¼ 23, P ¼ 0.017), birds (paired t-test, t ¼ À 3.184, df ¼ 125, P ¼ 0.002), mammals (paired t-test, t ¼ À 2.854, df ¼ 45, P ¼ 0.007) and reptiles (paired t-test, t ¼ À 2.141, df ¼ 89, In addition, a stepwise multiple regression analysis showed that there was no association between vertebrate roadkill and precipitation (r ¼ 0.489; F ¼ 3.138; P ¼ 0.107).Species rarefaction curves did not reach stability (Figure 3).The species richness estimators produced estimates greater than the actual recorded species richness for all taxonomic groups (Table 3).The major estimates with the major standard deviations were recorded with Chao 1 and 2 and the minor estimates were recorded with Bootstrap estimator (Table 3).
A summary of road and landscape attributes appears in Table 4. On the basis of 9,999 permutations of a Monte Carlo test and the first canonical axis, we found a significant relationship In addition, all canonical axes were significant (trace ¼ 0.205, F 4,35 ¼ 4.444, P o 0.001).Amphibians were associated with nearby forest and paved roads, birds were associated with nearby pasture, reptiles were associated with nearby grassland, and mammals were associated with unpaved road (Figure 4).

Discussion
The high number of road-killed vertebrates in CVNP indicates that the roads surrounding the park may significantly affect the wild vertebrates populations.The species richness found in our study (138 species) was greater than that reported by other long-term studies of Brazilian roads or motorways.Coelho et al. (2008) recorded 92 species on two roads in the northern coastal plain of Rio Grande do Sul; Rodrigues et al. (2002) found 100 species on four roads that delimit A ´guas Emendadas Ecological Station, in Distrito Federal;and Prada (2004) found 83 species on six roads in northeastern Sa ˜o Paulo. Fisher et al. (2003) recorded 140 species of vertebrates on BR 262, in the Pantanal wetlands of Mato Grosso do Sul, over the course of an 8-year-long study.The road mortality around CVNP is also high in comparison with other areas worldwide, such as Australia (Taylor & Goldingay 2004, n ¼ 53), Europe (Lode ´2000, n ¼ 97), and North America (Ashley and Robinson 1996, n ¼ 100).However, the overall number of 0.096 road-killed vertebrates per kilometer was similar to that of other studies in Brazil, ranging from 0.078 animals km -1 on BR-383, in Rondo ˆnia (Turci & Bernarde 2009), to 0.138 on BR-307, in Acre (Pinheiro & Turci 2013).
Despite the high number of vertebrates found during our surveys, the results are still an underestimate of total roadkill.The survey method of the present study (searches by car) allows the record of many roadkills, but due to the car speed it may overlook small animals, such as small amphibians and birds (Coelho et al. 2008).Besides, by not searching the roads constantly (more than twice a day, for example), some animals are not recorded because they quickly disappear.For instance, scavenging predators could have removed numerous small dead animals from the roads (Antworth et al. 2005).In addition, some animals may have been thrown away from the road by the collision or may have been only wounded by the collision and died far from the road (Slater 2002).Nevertheless, the method of search by car permits the inspection of a great road extension in a short amount of time and it is comparable to similar surveys.Rarefaction curves for all groups do not reach stability, a common trend in roadkill studies (Santana 2009, Pinheiro & Turci 2013).Estimates also show that a high number of species for all groups are affected by the roads in CVNP.For example, at least 47 species of snake inhabit the park (Franc¸a & Braz 2013), of which we found only 37 on the Table 3. Richness estimators of road-killed vertebrates on Chapada dos Veadeiros roads.

Estimators
All (143) Amphibians ( 12) Birds ( 63) Mammals ( 23) Reptiles ( 45) roads.Some species that are common in the Cerrado and that have been found in previous studies are the brown-banded water snake (Helicops angulatus), the southern crested caracara (Caracara plancus), the American black vulture (Coragyps atratus), the greater grison (Galictis vittata), and the ninebanded armadillo (Dasypus novemcinctus).These species will probably appear in future surveys of roads around the CVNP.Among the four vertebrate classes found as roadkill, the impact on amphibians is most likely to have been underestimated, due to their small size, their thin skin, and the slow locomotion of many amphibian species.The carcasses do not last as long on roads as those of other vertebrates because scavengers eat them rapidly or they are run over many times and are rapidly obliterated by vehicles (Hels & Buchwald 2001).Also, due to their activity patterns, seasonal reproduction, population structure, and selected habitats, amphibians are commonly more vulnerable to being hit by vehicles than other species (Hels & Buchwald 2001).Most data regarding road-killed amphibians are for large toads and frogs of the Bufonidae and Leptodactylidae families, the carcasses of which remain longer on the roads (Fahrig et al. 1995).The most commonly found amphibian in our surveys, the Schneider's toad (Rhinella schneideri), is a very common species in the Brazilian Cerrado, and individuals belonging to the genus Rhinella are often found as roadkill in different regions, such as Amazonia and the Atlantic Forest (Silva et al. 2007, Turci & Bernarde 2009).
Birds are also likely underestimated, particularly small passerines.Usually birds are hit while they are in flight and are thrown off the roads by the impact (Erritzoe et al. 2003).Many hawks and buzzards are hit while scavenging roadkill (Antworth et al. 2005).In addition, the type of habitat near roads can attract birds to the roads and increase their risk of becoming roadkill.For example, many small birds move and forage in the vegetation along the roads (Or"owski 2005).In natural landscapes crossed by roads, the roadside vegetation serves as a corridor and ecotone used by insects and birds, facilitating movement and feeding along the border of an area, but birds that spend time in this area are at higher risk of being hit by vehicles (Or"owski 2008, Rosa & Badger 2012).Most road-killed birds in the CVNP were species that use road borders frequently, such as the grassland sparrow (Ammodramus humeralis) and the blue-black grassquit (Volatinia jacarina).Although some authors have described roadkill as having low influence on the population dynamics of birds (Reijnen et al. 1995), these impacts can be proportionally greater in threatened species, such as certain grassland birds in the Cerrado, like the black-masked finch (Coryphaspiza melanotis).This species should receive special attention and mitigation programs, because it is nationally threatened species in Brazil (Brazilian National list of endangered species Portaria MMA no 444/2014).
Mammals are usually affected by roads when they have wide home ranges and terrestrial habits, which make them use roads as corridors for movement (Smith-Patten & Patten 2008).Also, many mammals are nocturnal and/or are scavengers.At night, they are targets for vehicles because they can become blind and immobilized on roads when cars' headlights shine on them (Barthelmess & Brooks 2010).Both the hoary fox (Lycalopex vetulus) and the crab-eating fox (Cerdocyon thous) are commonly hit by cars for this reason.These omnivorous, nocturnal and generalist foxes have ample distributions in South America and abundant populations (Wozencraft 2005) and are frequently found on various roads (Vieira 1996, Silveira 1999, Coelho 2003, Prada 2004).Another small mammal that was frequently found on CVNP roads was the yellow-toothed cavy (Galea flavidens).Despite its high abundance in the CVPN, this species is not common, as it is endemic to Brazil and restricted to the montane savanna habitat (Bonvincino et al. 2005, Weksler & Bonvicino 2008).Finally, other mammals that require special conservation attention are the giant anteater (Myrmecophaga tridactyla), the cougar (Puma concolor), and the maned wolf (Chrysocyon brachyurus).These are Brazilian threatened species and their populations are low in Central Brazil (Brazilian National list of endangered species Portaria MMA no 444/2014).
Snakes are the reptiles that are most likely to be hit on roads, due to their characteristic long body morphology and their habit of frequently using roads as thermoregulation sites (Bernardino & Dalrymple 1992, Bonnet et al. 1999).Most reptiles found on CVPN roads were snakes or long-bodied lizards, such as Ophiodes aff.striatus.Terrestrial and sit-and-wait genera, such as Bothrops, Crotalus, and Oxyrhopus, and slow-moving reptiles, such as Polychrus acutirostris and the worm lizards of the genus Amphisbaena, are commonly hit by vehicles.No reptile species are listed as threatened in Central Brazil.However, species with life histories characterized by low reproductive rates and low adult mortality, such as the Vanderhaege's toad-headed turtle (Phrynops vanderhaegei), and huge snakes and vipers, are more vulnerable to the demographic consequences of road mortality (Forman et al. 2003, Shepard 2009).The species richness of roadkill and the roadkill rates are influenced by seasonality as well as by type of nearby habitat (Trombulak & Frissell 2000, Rosa & Bager 2012).There are two well-defined seasons in Central Brazil, the wet or rainy season, characterized by high precipitation, and the dry season, during which almost no rain falls (Nimer 1989).Vertebrate mortality on CVNP roads was significantly higher during the wet season, as reported by previous studies in Central Brazil (Rodrigues et al. 2002).Studies show that more Brazilian mammals (Ca ´ceres et al. 2012) and anurans (Coelho et al. 2012) are killed on roads during the rainy season.The Cerrado of Central Brazil shows high seasonality and the activity of vertebrates becomes more intense after the first rainfalls (Oliveira & Marquis 2002).Even for plants, the rainy season is a more active time; although fructification of anemochorous and autochorous species occurs during the dry season, the fructification of zoochorous species is dispersed throughout the rainy season (Batalha & Mantovani 2000).Also, many animal species of Cerrado show dispersion and migratory movements and reproductive activities during the rainy season, when roads become barriers to their intense locomotion.
Amphibians were associated with patches of paved road with nearby forest habitat.In the Cerrado, gallery forests are always associated with streams and rivers, suitable environments for amphibians (Colli et al. 2002).These animals are at increased risk of being hit on high-speed paved roads and are killed mainly during the reproductive season.In contrast, there were more road-killed birds along roads next to pasture habitats.Despite their preference for forest habitats, birds that live in rural landscapes are usually associated with roads that are bordered by native vegetation.These roads can be attractive habitat and potentially a population sink, and offer a higher abundance of insects than other areas, attracting birds (Or"owski 2008).
In contrast with amphibians and birds, most of the reptiles and small mammals of the Cerrado are associated with open areas, and the majority of snakes found in the CVNP are found in grasslands or open savanna environments (Franc¸a & Braz 2013).Roadkilled mammals were strongly associated with unpaved road.In spite of the fact that mammals inhabit mainly forest habitats in the Cerrado (Johnson et al. 1999), most species have wide home ranges and may traverse the entire Cerrado.Compared with paved roads, unpaved roads appear to be more suitable and less exposed environments that enable the movements of CVNP mammals.
Due to the importance of the CVNP as a protected area that harbors and maintains the biodiversity of the Cerrado, it is important to remain monitoring the impact of the park's roads on wild vertebrates.Further studies may use recent indices and analyses to measure the impact of roads on the fauna of the CVNP and to establish priorities for mitigation programs (Coelho et al. 2008, Bager & Rosa 2011).Some mitigation actions are urgently needed, such as highway fencing in combination with safe crossing opportunities for wildlife (Clevenger et al. 2001).

Figure 1 .
Figure 1.Study area showing the Central Brazil with Chapada dos Veadeiros National Park and the roads GO 239 and BR 010.

Figure 2 .
Figure 2. Monthly data of precipitation mean and road mortality rates of vertebrate classes of Chapada dos Veadeiros National Park during 2006 -2008.Bars indicate vertebrate groups: Amphibians: Grey bars with the black board; Birds: White bars; Mammals: Black bars; Reptile: Grey bars.

Figure 3 .
Figure 3. Individual-based rarefaction curves of roadkilled amphibians, birds, reptiles and mammals of Chapada dos Veadeiros National Park.

Table 1 .
Vertebrate species killed during dry and rainy seasons on highways GO-239 and BR-010 in Chapada dos Veadeiros National Park, Central Brazil.

Table 2 .
Road-kill frequencies and incidence-intervals of vertebrates on highways GO-239 and BR-010 in Central Brazil.

Table 4 .
Means for landscape characteristics of 35 quadrats of roads in the Chapada dos Veadeiros National Park.Plot of Canonical Correspondence Analysis comparing matrices of structural habitat characteristics with roadkill vertebrates sampling data.The plot shows the position of each vertebrate class among arrays on first two canonical axes.Lengths of environmental vectors indicate significance strength and points of arrows represent centroids of impact of environmental variables on each vertebrate class' distributions among arrays.