Danger under wheels: mammal roadkills in the threaten lowland Atlantic Forest in southeast Brazil

ABSTRACT Understanding the ecological impacts of roads on mammals requires periodic monitoring of roads, and identification of both temporal and spatial distribution of roadkills (i.e., roadkill hotspots). The main aim of the study was to identify the most roadkilled mammals and evaluate the temporal and spatial distribution of roadkills in the RJ-122, a highway that crosses the threaten lowland Atlantic Forest in the state of Rio de Janeiro, southeast Brazil. Between October 2017 and January 2020, an intense monitoring study was conducted, with the carcasses of the roadkilled mammals being collected three times a week. Overall, we recorded 295 roadkilled mammals belonging to 22 species, resulting in 11 roadkills per month for the RJ-122 highway. The black-eared opossum, Didelphis aurita (N=149, 51%) stood out as the most roadkilled mammal, followed by porcupine, Coendou spinosus (N= 24, 8%), crab-eating fox, Cerdocyon thous (N= 23, 8), nine-banded armadillo, Dasypus novemcinctus (N= 23, 8%), and white-tufted marmoset, Callithrix jacchus (N= 20, 7%). Roadkills on the RJ-122 varied throughout the year, being more frequent in the rainy season (N=180) than in the dry season (N=115), and were concentrated in two hotspots, indicating some critical points with high roadkill frequency. Hotspots were associated with areas with dense natural vegetation, which can function as forest corridors in this fragmented landscape. Based on our results, several mitigation measures are recommended for the RJ-122 highway.

increasing negative pressure on an already impacted wildlife (Laurance et al., 2009).
Roads have major contribution to the high levels of biodiversity loss worldwide, and roadkills are among the leading causes of direct death of wild vertebrates, surpassing impacts generated by hunting and mortality rates from natural causes (Van Der Ree et al., 2015).Due to the great urgency and growth of the road network worldwide, roadkills and other direct and indirect impacts of roads have received attention in many studies (Bager & Rosa, 2010;Cunha et al., 2010;Santana, 2012;Almeida et al., 2013;Vélez, 2014;Costa et al., 2022).Among the various threats to wildlife are, in addition to collisions with vehicles, the increase in the level of air and noise pollution, increase in temperature, and the emergence of urban settlements on the edge of roads (Bager & Rosa, 2010;Vélez, 2014).
Wildlife roadkills do not occur randomly (Filius et al., 2020).Several factors favor certain species to be roadkilled more often than others, such as biological (body size and diet) (Barthelmess & Brooks, 2010) and landscape characteristics, and the road itself (Bueno et al., 2013(Bueno et al., , 2015)), as well as seasonal variations in temperature and rainfall (Bueno & Almeida, 2010;Santos et al., 2011;Santana, 2012).The vehicular traffic, as well as the vehicle speed, are also important determining characteristics that lead to the collision of the vehicle with the animal (Cunha et al., 2010;Lester, 2015).For mammals, body size has a major influence on roadkills (Barthelmess & Brooks, 2010;Carvalho et al., 2015;Chen & Koprowski, 2019).Small mammals, such as rodents and marsupials, are especially vulnerable because drivers do usually not see them when they cross the road, or when they are in areas surrounding road edges (Laurance et al., 2009;Bueno & Almeida, 2010;Carvalho et al., 2015).Medium and large-sized mammals, on the other hand, require more energy to meet their needs, which results, for many species, in an increase in movement rates in search of food and reproductive partners (Bueno & Almeida, 2010).By moving larger areas, they become more vulnerable to collisions with vehicles.
Although the high rate of roadkills in Brazilian roads is widely discussed in many studies over the years (e.g., Abra et al., 2019Abra et al., , 2021)), there is still a paucity of accurate information on the spatial and temporal distribution of roadkills.Understanding the dynamics of wildlife-vehicle collisions allows us to indicate possible solutions to increase safety on the roads, reduce the impacts on humans and wildlife, reduce costs, and invest in mitigation measures aimed at conservation of biodiversity (Rytwinski et al., 2016;Abra et al., 2019).
The objective of this study was to evaluate the impact of roadkill on wild mammals on RJ-122, a highway that crosses the threaten lowland Atlantic Forest in southeast Brazil.Specifically, we (1) identified the most roadkilled mammals between October 2017 and January 2020, (2) evaluated the temporal distribution and the influence of climate seasonality (dry and rainy seasons) on roadkills, and (3) identified whether roadkills are spatially clustered forming roadkill hotspots.We expect (1) the black-eared opossum (Didelphis aurita, Wied-Neuwied, 1826) and the crab-eating fox [Cerdocyon thous, (Linnaeus, 1766)] to be the most roadkilled species, based on previous studies in Brazil (Coelho et al., 2008;Bueno & Almeida, 2010;Cunha et al., 2010;Zanzini et al., 2018); (2) high roadkill numbers in the dry season (Bueno & Almeida, 2010); and, (3) hotspots located near forested areas, because of the species habits registered in the Atlantic Forest (Paglia et al., 2012).

Study area.
We carried out the monitoring study in the RJ-122 highway, from Km 1 to Km 34, between the municipalities of Guapimirim (UTM 23K 708028/7507913) and Cachoeiras de Macacu (UTM 23K 736352/7507913), in the state of Rio de Janeiro, southeastern Brazil.The highway is immersed in a fragmented landscape of lowland Atlantic Forest, with stretches of native vegetation interspersed by rural and residential properties along the highway.The highway includes two paved lanes, with an average width of 3.5 m each, and a maximum speed of 80 km/h (DER-RJ, 2013).The study area is characterized by a hot and humid tropical climate, with a rainy season from November to April, and a less rainy season from May to October during the study period (INMET, 2021).
Sampling method.We sampled roadkilled mammals three times a week between October 2017 and January 2020.Sampling was not performed in May, October and December of 2019.Roadkilled mammals were identified and collected by car or motorcycle on estimated average speed of 40 km/h by a trained biologist of the NGO SOS Vida Silvestre.For each record, a field form was filled out with the following information: geographic location, kilometer and direction of the road, taxonomic group, and sex, if possible.Fortnightly the specimens collected were taken to the Laboratory of Ecology of the Veiga de Almeida University (RJ), identified, and prepared as study skins, skeletons, or stored whole in the spirits collection depending on the conditions of the material.Species were deposited in the National Museum of Rio de Janeiro collection.
The carcasses of the animals used in this study comply with and are in accordance with the SISBIO License # 30727-9.
Data analysis.We compiled a list of mammal species roadkilled, indicating threatening status at national (ICMBio, 2018) and global (IUCN, 2021) levels, the number of records for each species, and their roadkill rate.The roadkill rate was calculated by dividing the number of recorded individuals/ species by the length of the road (34 km) and the total number of days sampled (N = 165).The roadkill rate was correlated with the climatic season (dry and wet seasons) using the Chi-squared test, in order to assess if seasonality was one of the factors influencing the rate of roadkill.In addition, the Factorial Correspondence analysis was applied to investigate the association between the five most-roadkilled mammals and the months they were recorded.Analyses were conducted in the R environment, version 3.6.1 (R Core Team, 2019).
To evaluate the spatial distribution of mammal roadkill hotspots, we used the heatmap plug-in in QGIS version 3.10.6(QGIS Development Team, 2019), which uses kernel density estimation, to create a heat map (i.e., a density map) of roadkills.The Kernel density is estimated based on the number of collisions per kilometer of road, with larger numbers of clustered points resulting in larger values.Point density is calculated by adding the values of all overlapping kernel surfaces within any cell of determining size.The bandwidth of the kernel exhibits a strong influence on the estimation results, since it determines the search radius in which roadkills will contribute to the hotspot identification (Bíl et al., 2013).We used a search radius of 100 m to investigate collision clusters, which proved to be appropriate for the region analyzed.To model kernel shape, we used a quartic kernel shape function in the heatmap plug-in (QGIS Development Team, 2019).

RESULTS
We recorded 295 roadkilled mammals belonging to 22 species, resulting in 0.32 recorded individuals per kilometer per month and 11 roadkills per month for the RJ-122 highway (Tab.I).Of the 295 records, 286 were identified at the species level; the other nine could not be identified due to the poor state of conservation.The black-eared opossum (Didelphis aurita) stood out as the most roadkilled species (N = 149, 51%), followed by the porcupine, Coendou spinosus (F.Cuvier, 1823) (N = 24, 8%), the crab-eating fox, Cerdocyon thous (N = 23, 8%), the nine-banded armadillo, Dasypus novemcinctus Linnaeus, 1758 (N = 23, 8%), and the whitetufted marmoset, Callithrix jacchus (Linnaeus, 1758) (N = 20, 7%).These five species concentrated more than 80% of the records, and the opossum alone was responsible for approximately nine roadkills per month.Rodents and bats were the two groups with the highest number of roadkilled species, a total of five each (Tab.I).Tab.I. Roadkilled mammals along the RJ-122 highway between October 2017 and January 2020, in the state of Rio de Janeiro, Brazil.N = Number of individuals, Roadkill rate = number of individuals/km/day, and Threat status = status of threat of extinction according to IUCN (LC -Least Concern, EN -Endangered) and national list (Vu C1-Vulnerable to extinction, with population numbers declining).Roadkill rates less than 0.001 were designated as 0.
Two mammals are included in the national and global list of threatened species.The jaguarundi, Herpailurus yagouaroundi (É.Geoffroy, 1803) (N = 1), was classified as Vulnerable (VU) at the national list and as Least Concern (LC) at the global list.However, the current population trend indicates a decline in population numbers.The tapiti, Sylvilagus tapetillus Thomas, 1913 (N = 7), was classified as Vulnerable (VU) at both national and global lists.
The chi-square test revealed a seasonal pattern in roadkills at a significance level of 10% (p = 0.077; gl = 11; χ2 = 18.19), with high number of roadkills in the rainy season (from November to April; N = 180) than in the dry season (from May to October; N = 115).For the five most sampled species (D. aurita, C. spinosus, C. thous, D. novemcinctus and C. jacchus), there was a higher predominance of roadkills in the rainy season, except for Cerdocyon thous that had a higher frequency in the dry months.The months which the five most-roadkilled mammals were recorded differed among species: Callithrix jacchus roadkill numbers were higher in September, Dasypus novemcinctus in March, and Cerdocyon thous in the months of June and July (Fig. 1).No pattern was identified for the opossum, D. aurita, and the porcupine, C. spinosus.

DISCUSSION
Mammal roadkills at the lowland Atlantic Forest was particularly high (0.32 roadkills per kilometer per month), when compared to other sites in Brazil (Coelho et al., 2008;Cunha et al., 2010).However, comparisons are difficult since this information is not always available or was estimated by different indices (Zanzini et al., 2018).Most estimates are also available for the entire vertebrate clade, not for specific taxonomic groups, or as the number or record per kilometer, which is affected by the duration of the study.Coelho et al., (2008) was one of the few studies that evaluated mammals (and other groups) separately.In this study, vertebrate roadkill rates varied from 0.21 to 0.46 roadkills/km/month in two roads in the sandy and wet restinga, important remnants of Atlantic Forest in the south of Brazil (Coelho et al., 2008).The higher mammal roadkill rate was recorded in the summer (0.36 roadkills/km/month), followed by winter (0.35 roadkills/km/month), autumn (0.33 roadkills/km/month), and spring (0.30 roadkills/km/month).Overall, mammal roadkill numbers in RJ-122 is higher than the total mortality records for the vertebrate groups in other Brazilian roads (Cunha et al., 2010).
As expected, the most roadkilled species was the marsupial, Didelphis aurita, which corresponded to more than half of the total number of records.The black-eared opossum is a habitat generalist, locally abundant, and highly mobile species (Cunha & Vieira, 2002;Prevedello & Vieira, 2009;Ferreira et al., 2016).It inhabits forest fragments surrounding the highway (Vieira et al., 2009) and cross fragments frequently in search for new areas, food, or mates (Prevedello & Vieira, 2009).Didelphis aurita is probably the most roadkilled mammal in the Atlantic Forest (Bueno & Almeida, 2010), but other species of Didelphis are also frequently recorded in road surveys probably due to the high tolerance to anthropized areas (Smith & Dodd Jr, 2003;Pinowski, 2005;Coelho et al., 2008).These high roadkill numbers may affect important ecological functions performed  by these species, such as seed dispersal, regulation of prey population, and pest control (Cáceres & Lessa, 2012).
The porcupine (Coendou spinosus), the nine-banded armadillo (Dasypus novemcinctus), and the crab-eating fox (Cerdocyon thous) had similar mortality numbers, but only C. thous is frequently recorded in roadkill surveys (Coelho et al., 2008;Freitas et al., 2015;Orlandin et al., 2015;Cirino et al., 2022;Costa et al., 2022).As D. aurita, this medium-sized, insectivorous/omnivorous canid has opportunistic habits, and occupies forests, savannas, and disturbed areas such as cultivated fields, forest fragments, and areas around human settlements (Eisenberg & Redford, 1999).In the Atlantic Forest, the crab-eating fox most commonly inhabiting forested and edge areas, but can also use altered environments (Ferraz et al., 2010;Magioli et al., 2014;Monteiro-Alves et al., 2019).In a nearby highway (BR-040) in the state of Rio de Janeiro, C. thous roadkills were more likely to occur near areas with low percentage of crop fields, herbaceous vegetation, and urban areas, reinforcing the forest habits of the species in the region (Costa et al., 2022).However, the high roadkill numbers can also be a consequence of the consumed carcasses of other roadkilled animals found on the side of the road (Gumier-Costa & Sperber, 2009;Carvalho, 2015;Orlandin et al., 2015).This is an extraordinary type of food chain which leads to a roadkill cascade as animals seeking for food on the road are particularly prone to roadkill events.
Roadkills had a significant non-random distribution over the course of the year.The rainy season was the period with the highest number of roadkilled mammals (61%).The seasonal pattern of roadkills may be a consequence of the increased movement rates in the rainy season (Loretto & Vieira, 2006;Cáceres et al., 2012).Mammals search for food since there is greater availability of resources in this season, as well as an intense proliferation of insects (Gumier-Costa & Sperber, 2009).Another possibility is the increased search for mates, since there is an overlap with the breeding season of some species (Loretto & Vieira, 2006).Our result contradicts what was reported by Bueno & Almeida (2010) and Faria et al. (2022), who pointed out the dry season as the season with high roadkill numbers.In this way, it is evident that roadkills do not occur homogeneously, but so far no pattern was identified.
The two identified hotspots in the RJ-122 (Km 23 and 24) were in areas with dense natural vegetation.Despite the high degree of forest degradation, many species persist in these forest fragments which are structurally isolated by a matrix of pastures, plantations, and urban areas (Vieira et al., 2009).These areas can function as forest corridors, since some animals tend to avoid crossing open areas, as it increases the risk of exposure to potential predators and their preys (H-Saranholi et al., 2016;Costa et al., 2022).Indeed, the areas with the lowest number of roadkills in the 9,10 and 11) were open areas (presence only of undergrowth or small bushes) with fenced rural properties in the vicinity.A low roadkill rate was also recorded near two local rivers: Guapiaçu and Orindiaçu rivers.Several studies have associated the presence of rivers close to roads as a potentiator of roadkill events (Bueno et al., 2013;Freitas et al., 2015;Costa et al., 2022).We suggest that the low mortality rate near the two local rivers can be explained by the use of rivers as an alternative route for animals that move through the landscape, avoiding crossing the road, thus resulting in a lower mortality rate near riparian forests.However, more refined research is necessary to answer this question with proper characterization of the landscape.
The identification of two hotspots along the RJ-122 highway provides subsidies for the implementation of mitigation measures for the prevention of new wildlifevehicle collisions.We suggest three possible mitigation measures: (1) modifying driver's behavior (e.g., appropriate road signs and speed bumps should be part of the road where animals are frequently recorded), (2) educational and awareness-raising campaigns for drivers (e.g., installation of signs about the incidence of roadkills), and (3) preventing animal access to the highway (e.g., construction of physical barriers or facilitators of safe movement through wildlife passages and ecological corridors (Lester, 2015;Rytwinski et al., 2016).Although awareness-raising measures are widely used, their implementation must be associated with other mitigation measures, as indicated by Rytwinski et al. (2016).The use of fences to prevent animal access to roads has been constantly implemented, and their effectiveness has been proven (Collinson et al., 2019;Williams et al., 2019;Park et al., 2021).Rytwinski et al. (2016) verified a 54% reduction in accidents on roads that had fences.We believe that the presence of fences along stretches of the RJ-122 highway, for example, around rural homes, may be one of the reasons that lead to a decrease in the number of roadkills.Given this scenario, the construction of crossing structures, together with fences, appears to be the best solution to connect forest fragments, directing animals to safe routes where there is no passage of vehicles, thus reducing the number of roadkills and allowing animal dispersal (Rytwinski et al., 2016).
Here we have shown that the overall number of roadkilled mammals at RJ-122 was high, despite the short time period of monitoring.These numbers can be even higher since many animals, at the moment of impact, can be thrown off the road or can move away from it where they end up dying.Roadkills were concentrated mainly in two areas with dense natural vegetation.The most roadkilled species were in fact from forested areas, which reinforces the need to apply mitigation measures in areas which facilitate the movement of species and connect forest fragments over the landscape.Importantly, we recorded two threatened species during the study period, the jaguarundi (Herpailurus yagouaroundi) and the tapiti (Sylvilagus tapetillus), which are classified as Vulnerable, at the national list, and Endangered, at the global list, respectively.The results discussed here can contribute to a better understanding of roadkill dynamics, in addition to providing parameters that allow assessing the sustainability of highways, as well as improving and suggesting mitigation studies in several other Brazilian highways.

Fig. 2 .
Fig. 2. Frequency of roadkill mammals over every kilometer along the RJ-122 highway between October 2017 and January 2020, in state of Rio de Janeiro, Brazil.Highlights are km 23, km 24, km 28 and km 29.

Fig. 3 .
Fig. 3. Roadkill hotspots of mammal species recorded along the RJ-122 highway (from Km 1 to Km 34) between October 2017 and January 2020, in the state of Rio de Janeiro, Brazil.The largest number of roadkills is concentrated in two regions (black location icon), Km 23/24 (N = 22 and N = 24) and Km 28/29 (N = 18 and N = 22).