This study was performed through opportunistic sampling by monitoring roads that surround three Federal Conservation Units located in Rio Grande do Norte, Brazil: the Parque Nacional (Parna) da Furna Feia, the Floresta Nacional (Flona) de Açu, and the Estação Ecológica (ESEC) do Seridó between 2018 and 2021. The Parna da Furna Feia (5°4'14.88"S, 37°32'1.51"W) is located in the northeastern semi-arid region and has a Caatinga biome and deciduous hyper-xerophilous vegetation, with predominant shrub tree formation. It has relatively high temperatures, with an annual average of about 27.5 °C and relative humidity of about 70%, and a scarce and irregular rainfall regime, with rainfall concentrated from February to March, with October and November being the hotter and drier (ICMBio 2020). The Flona de Açu (5°57'74.79"N, 36°94'7.14"W) has a Caatinga biome housing four vegetation groups: shrub, shrub-tree, grass-shrub, and herb-shrub (ICMBio 2019). The climate in the region is hot semi-arid of the BSh type, with an average annual temperature of 28.1 °C (Alvares et al. 2013; Rio Grande do Norte 2008). The ESEC do Seridó (6°34'44.63"N, 37°15'6.78"W) has a Caatinga biome with Savanna Estépica Park vegetation. Its climate is semi-arid BsW’h type, with temperatures ranging from 33 °C to 22 °C. It has strong and prolonged exposure to sunlight, low percentages of relative humidity in the dry months, high evaporation rates, and the greatest scarcity and irregularity of rainfall in the country (ICMBio 2004).

The opportunist sampling was conducted under the SISBIO (Sistema de Autorização e Informação em Biodiversidade) license number 40620 for the collection of biological material. Road monitoring was conducted in a vehicle with two observers at speeds between 40 and 60 km/h.

Given the lack of studies with this type of opportunistic sampling, a series of parameters were considered to standardize the samples collected and, above all, avoid collecting specimens that may have been run over more than once. This resulted in a four-year (2018–2021) sampling effort to reach the sample size for this study, during which many animals were discarded and only the specimens considered suitable for the study were taken to the laboratory. Therefore, to be collected, the animals had to be found in the first hours after sunrise. Given that a large part of the Caatinga fauna has nocturnal and crepuscular habits, it is assumed that most roadkills occur during the night or early hours of the day, when the animals are active and the drivers have limited visibility of the roads. Therefore, road monitoring was carried out shortly after sunrise, between 5:30 a.m. and 9 a.m., to collect the animals as quickly as possible after the vehicle collision, prevent the bodies from being preyed upon by opportunistic animals, and prevent more fragile specimens, such as birds, from being affected by the temperature of the environment, which increases as noon approaches. The animals also had to be fresh, up to 6 hours after death (before livor mortis), which can be judged by the still-wet appearance of the open lesions and the absence of bloating by gas and putrid odor. Additionally, they should not have their physical integrity severely compromised (severely crushed or flattened). After years of monitoring roads, we can state that a large proportion of roadkilled animals are found on the shoulders of the highways, apparently thrown off the road after impact with the vehicle. These specimens are generally found in good condition, in contrast to specimens with severely compromised physical integrity, which are generally found in the middle of the road. Finally, specimens that presented cadaveric alterations that could interfere with the analyses at the time of necropsy were also discarded.

After collection, the animal was photographed, tagged, and placed in a plastic bag inside a polystyrene box with ice bags until it was transferred to a freezer at the Ecology and Wildlife Conservation Laboratory (Laboratório de Ecologia e Conservação da Fauna Silvestre-Ecofauna) of the Federal Rural University of the Semi-Arid Region (Universidade Federal Rural do Semi-Árido-Ufersa). The Global Positioning System (GPS) coordinates were noted for each individual.

In the Ecofauna laboratory, the animals were weighed with a 0.1 g precision digital scale, followed by an external inspection to look for any signs of lesions, with the smallest indication recorded in an individual file. When possible, the sexing of birds and reptiles was performed through the visualization of the gonads and mammals through visualization of the external genitalia.

The analysis of bone lesions was conducted through inspection, palpation, and dissection of structures that presented some alterations to determine the type, region, and bone affected. As for the type, bone lesions were classified into four groups, according to Henry (2014): multiple (have more than one fracture line), transverse (perpendicular to the long axis of the bone), open (type I—with small tissue damage caused by a bone fragment that has penetrated the skin; and type IV—with limb amputation), and luxation (abnormal dislocation of a joint). Bone lesions were further distributed into five regions and 13 bones (or sets of bones) as follows: head (skull), chest (sternum and ribs), abdomen (pelvic girdle, which includes the ischium, ilium, and pubis), axial skeleton (cervical, thoracic, lumbar, sacral, and caudal vertebrae), and appendicular skeleton (humerus, radius/ulna, femur, and tibia/fibula).

The body condition score of the birds was determined by palpation of the pectoral muscles, and the animals were classified into five categories according to the development of the pectoral muscles and evidence of the keel bone, with 1 being very thin and 5 being overweight (Grespan and Raso 2014). As for the reptiles, the body condition score was determined at five levels: a score of 1 for very thin animals and a score of 5 for obese ones (Deming et al. 2008). For mammals, the Laflamme (1997) method was used, which categorizes the general body condition score of individuals through characteristics observed with the naked eye, with a score of 1 being a very thin animal and a score of 9 being an obese animal.

QCT was performed at the Focus Veterinary Diagnostic Clinic in Fortaleza, Brazil. Reptiles and birds were placed in the ventral decubitus and mammals in the dorsal decubitus for tomography examination. All animals were positioned on top of a QCT phantom containing calibrations of 0, 100, and 200 mg/cm³ of calcium hydroxyapatite arranged parallel to the axial skeleton of the specimens. The examination was performed using a GE Hi-Speed FXI computed tomography device and protocol with 120 kVp and auto mA at a speed of one rotation per second. Images were obtained in the craniocaudal direction through 3 mm-thick cross-sections with a filter for bone tissue.

After the CT scan and digitalization of the images, RadiAnt Dicom Viewer software (version 2021.2, Medixant) was used to quantify and qualify the possible bone lesions. In addition, axial sections of the trabecular bone of the tibia of birds, the first lumbar vertebra of reptiles, and the fifth lumbar vertebra of mammals were performed to estimate the individual values of bone radiodensity of each animal. Therefore, the attenuation value in Hounsfield units (HU) of the trabecular bone of the tibia (mean of the proximal, medial, and distal portions) of birds and the vertebral body of the first lumbar of reptiles and the fifth lumbar of mammals was calculated from the mean of the regions of interest (ROIs) analyzed. Each ROI was manually drawn from an ellipse, representing an area of 0.05±0.01 mm² for birds and reptiles and 0.3±0.1 mm² for mammals, and a bone radiodensity value in HU of the selected region was automatically generated. The ROIs of the 0 and 200 mg/cm³ calcium hydroxyapatite phantons were also measured in the same image as the bones of interest (Fig. 1) (Cheon et al. 2012; Lee et al. 2015). To convert HU into mg/cm³, the following equation was used (Park et al. 2015):

Figure 1. 

QCT images showing the acquisition of the ROIs from the 5^th lumbar vertebra of a crab-eating fox specimen.

BMD = 200HUt / (HUb-HUw),

where BMD is the bone mineral density in mg/cm³, HUt is the radiodensity of the trabecular bone in question, HUb is the radiodensity of the phantom containing 200 mg/cm³ of calcium hydroxyapatite, and HUw is the radiodensity of the phantom containing 0 mg/cm³ of calcium hydroxyapatite.

For the interclass comparison of the means of bone lesions in terms of types, regions, and bones affected, Kruskal-Wallis nonparametric analysis was performed, using the number of lesions as the dependent variable. The Mann-Whitney U test was used to compare necropsy and QCT methods for detecting bone lesions. To compare BMD means between vertebrate classes, between the number of bone lesions and body condition score, and between BMD and body condition score, one-way analyses of variance (ANOVAs) were performed. To compare all these means with sex, a nonparametric Kruskal-Wallis analysis was performed. To estimate the possible relationship between the number of bone lesions and BMD, simple linear regression was performed. Statistical software Statistica (version 10, StatSoft) was used for statistical analysis, and values of p < 0.05 were considered significant.

A total of 88 bone lesions were observed in 24 specimens (Table 2). The yellow-headed vulture, one black and white tegu, and two crab-eating foxes did not show any bone lesions or only had lesions in their internal organs. There was an average of 3.25 injuries per bird and 3.75 injuries per reptile and mammal, with no statistical difference between the means (N = 20, df = 3, p = 0.93).

There was no statistically significant difference between the means of the types of injuries by vertebrate class (p > 0.05). For the affected regions, fewer fractures were observed in the heads of birds than in those of mammals and reptiles (N = 24; df = 3; H = 5.46; p = 0.06), but the difference was only significant between birds and reptiles (p < 0.05). For the affected bones, there was a statistically significant difference between the means of the pelvic girdle (N = 24; df = 3; H = 6.57; p = 0.03) and caudal vertebrae (N = 24; df = 3; H = 6.57; p = 0.03), as only reptiles had bone lesions in these two regions. There was no statistical difference between the means of detection of bone lesions using QCT and necropsy (N = 48; df = 1; p > 0.05).

BMD was measured for all birds and reptiles but only nine of the 12 mammals due to the low quality of the images obtained in the three remaining animals. There was a statistically significant difference in BMD between vertebrate classes (N = 20; df = 3; F = 5.9; p = 0.01) (Fig. 2), being higher in reptiles (Fisher LSD test, p = 0.003) and mammals (Fisher LSD test, p = 0.02) than in birds. However, there was no statistical difference in BMD between reptiles and mammals (Fisher’s LSD test, p = 0.22). There was no statistical difference when BMD values were compared with the number of lesions (N = 20; df = 2; p = 0.15) or body condition score (N = 19; df = 3; F = 3.17; p = 0.06). There was no statistical difference between the means of bone lesions and sex (N = 12; df = 2; p = 0.31) or BMD and sex (N = 12; df = 2; p = 0.14) of all vertebrate classes.

Figure 2. 

BMD values by vertebrate class showing statistical difference between the means.

Birds

Birds undergo adaptive processes that reduce their body mass to improve flight capacity, such as the loss or fusion of some bones, thinning of cortical bones, and pneumatization of medullary cavities, making these bones more fragile (King and John 1984). As a result, bone fractures from vehicle collisions are usually comminuted or open type (Kim and Kwon 2016; Jang et al. 2019b). Kim and Kwon (2016) stated that collisions at low speeds result in single fractures, whereas collisions at high speeds (against vehicles) result in multiple fractures. However, transverse fractures were the most observed in this study. This was the only class that showed the four types of bone lesions (Fig. 3).

Figure 3. 

Bone lesions (yellow arrows) in crested caracara: a multiple fractures in the sternum, evidenced after opening the thorax during necropsy b transverse fracture in the left tibia, evidenced after dissection of the area c open fracture in the left radius; and d femoro-tibiotarsal dislocation in the left lower limb seen by QCT.

In this study, the most affected region was the appendicular skeleton, corroborating the findings of Cousins et al. (2012), who studied lesions in 146 specimens of New Zealand pigeon (Hemiphaga novaeseelandiae) victims of collision with vehicles and stationary structures through radiographs and necropsies. They determined that the largest portion of injuries caused by vehicle collisions occurred in the appendicular skeleton, whereas injuries caused by collisions with stationary objects (such as glass windows) mainly occurred in the coracoid, clavicle, and internal organs. They also reported a substantial decrease in the survival rate of birds affected by more than two fractures.

Studies on the characterization of bone fractures in birds of prey have also found most lesions in the appendicular skeleton. Punch (2001) reported a survival rate of approximately 34% in Falconiformes and 36% in Strigiformes received at a veterinary hospital in Australia, with most fractures happening in the appendicular skeleton (mainly carpal/metacarpal) after vehicle collisions. Komnenou et al. (2005) reported that fractures in the appendicular skeleton (mainly the humerus) were the most common in birds of prey admitted to the avian medicine clinic of a university in Greece, resulting in a release rate of 59%. Kelly and Bland (2006) reported the presence of one or two fractures, mainly in the appendicular skeleton (radius and ulna), in 55% of birds of prey admitted to a rehabilitation center in England, and 24% of these could be released. In a study by Yoon et al. (2008), the most frequent bone fractures observed were in the appendicular skeleton (mainly ulna) of birds of prey victims of vehicle collisions received at a university in the United States. Kim and Kwon (2016) reported a 7% survival rate in birds of prey received at veterinary hospitals and rehabilitation centers in South Korea, whose main injuries were bone fractures in the appendicular skeleton (mainly the humerus). Jang et al. (2019b) found that the most frequent fractures were those affecting the appendicular skeleton region, mainly the ulna (29.70%), radius (21.76%), and humerus (17.35%) of birds of prey received in two rehabilitation centers in South Korea.

However, as shown in the preceding paragraph, in previous studies the most fractured bones were those of the wings. This is in great contrast to the results we obtained, where the bones of the wings and hind limbs were affected by an equivalent number of bone lesions. Additionally, the most affected bone varied among previous studies, but none showed such a high fracture incidence in the tibial bone, as we observed in the present study.

Reptiles

Clemente et al. (2011) compared the anatomy and displacement speed of lizards and concluded that lizards move more slowly to avoid bone fractures. Due to the crawling posture of lizards, as opposed to the more upright posture of mammals, they cannot rotate their limbs to take longer strides and need to keep their feet on the ground longer to increase energy dispersion time and avoid bone fractures. Therefore, slow displacement increases the chances of reptiles being run over by vehicles. Additionally, the relevance of these factors increased with the size of the lizard species.

Although no studies have characterized bone lesions in lacertids, in this study, only reptiles were affected by bone lesions in the pelvic girdle and caudal vertebrae. Lizards are known for their ability to perform tail autotomy, which occurs in more fragile areas between the caudal vertebrae (Arnold 1984) and is mainly used as a strategy to escape predators (Passos et al. 2016). In this study, four of the seven black and white tegus were collected with tail autotomy, two of which had additional lesions along the tail, suggestive of being run over (Fig. 4). In addition to the long tail length of this species, which makes it an easy target for oncoming vehicles, tail autotomy may have occurred at the moment of impact or moments later in the last minutes of the animal’s life, affected by severe physiological stress.

Figure 4. 

Three different conditions observed among roadkilled black and white tegu specimens: a multiple tail fractures b tail autotomy, and c tail regeneration.

Despite the anatomical and physical size differences between the species, each individual was affected by approximately three lesions. However, although there was no statistical difference between the means of the types of lesions and affected regions, there was a higher incidence of lesions of the transverse type in the appendicular skeleton of birds and the multiple type in the head region of reptiles and mammals.

Four specimens collected in this study showed no bone lesions, only lesions in the internal organs, including one bird, one reptile, and two mammals (Fig. 5). Befitting of this, Ushine et al. (2021) did not associate collisions with bone fractures, only with injuries to internal organs, in a study with necropsies of 55 passerines and coraciiforms that were admitted to a rehabilitation center in Tokyo, Japan. Also, according to Williamson (2000), vehicle collisions do not always cause bone fractures but mainly rupture of organs and internal hemorrhage, which leads to the animal’s immediate death.

Figure 5. 

Lesions other than bone-related observed in crab-eating foxes: a abrasion on the left forelimb b liver rupture, and c exophthalmos.

Further studies including different variants that may influence the lesions suffered by roadkilled wild animals are required. The vehicle speed is one of those variants and theoretically may impose different outcomes in a collision event. Although studies on the influence of vehicle speed on the number of roadkilled animals exist, these studies do not describe the lesions suffered by the animals. For example, a survey in Tasmania, Australia, described that 50% of the roadkill detected resulted from vehicles moving faster than 80 km/h and that a reduction of ~20% in speed would result in a decrease of ~50% roadkill (Hobday and Minstrell 2008). They also related the vehicle speed with the different taxa, where the echidna (Tachyglossus spp.) was the main average high-speed victim and the eastern barred bandicoot (Parameles gunnii) was the main one on lower average speeds. Another study on the roads of Southern Ontario, Canada, stated that the best predictor factor for roadkill among those tested was the speed limit (Farmer and Brooks 2012). The roadkill rate predicted by their model for 20 km/h was 0.1, and for 50 km/h it increased to 0.75.

Schuhmann et al. (2014) described the BMD of humeri and tibiotarsi of common buzzard (Buteo buteo), white-tailed sea eagle (Haliaeetus albicilla) and barn owl (Tyto furcata) using the Archimedes principle. They found no significant difference between bones or sex, except for common buzzards’ juveniles, which displayed lower BMD than adults. Bertuccelli et al. (2021) measured the BMD of the humerus of five species of birds of prey using QCT and obtained mean values in HU, with no statistical difference between the species but higher values in the peregrine falcon (Falco peregrinus) and honey buzzard (Pernis apivorus) compared to the common buzzard, barn owl, and tawny owl (Strix aluco).

The BMD values of vertebrae trabecular bones in boa constrictors (Boa constrictor constrictor) described by Souza et al. (2018) were twice the values described for the reptiles in the present study, with a mean of 1,237.91 mg/cm³ and 689.61 mg/cm³, respectively. For green turtles (Chelonia mydas), Oliveira et al. (2012) measured a mean of 308.9 mg/cm³ in the bones of the dorsal vertebrae.

In a QCT study by Lee et al. (2015) on the determination of BMD in domestic dogs with hyperadrenocorticism, the control group presented a mean of 316.5 mg/cm³ in the trabecular bones of the fifth lumbar vertebra, almost half of the values found for the same C. thous bone described in the present study, which was 611.56 mg/cm³. For domestic cats, Cheon et al. (2012) found a mean of 257.9 mg/cm³ in the trabecular bones of the fifth lumbar vertebra.

As observed in the statistical analyses, there was no significant relationship between the means of the number of lesions, BMD, and body condition score of the individuals studied, leading us to infer that the amount of calcium in the bones and the nutritional status of the individuals may have a minimal (if any) influence on the number of injuries suffered after a vehicle collision. The anatomical and behavioral particularities are the determining factors of the amount, type, and location of injuries that affect the different classes of roadkilled vertebrates. However, it is suggested that further studies with larger sample numbers regarding the relationship between the number of lesions, BMD, and body condition score of wild species are needed.

Medium-to-large species are more likely to recover from vehicle collisions than small species, which are more likely to be crushed on impact or escape contact (Taylor 1971). However, although bone lesions located in the skull, particularly those of multiple types, are the most severe and potentially fatal, with less possibility of treatment, it is worth mentioning that skull lesions are not always deleterious. Lacitignola et al. (2021) reported a 25% survival rate in red foxes with head injuries after vehicle collisions at a wildlife rehabilitation center in Italy. Additionally, in the study by Jang et al. (2019a), who analyzed fractures of wild mammals received at a rehabilitation center in South Korea, mostly victims of collisions with vehicles, 25% underwent surgical procedures, recovered, and were able to be released in the wild.

Despite the limitations of the present study regarding the low sample size, the sometimes compromised state of the samples, and the use of data from dead specimens to live animals, it can be concluded that with more studies, larger samples, and more species, the rehabilitation and the number of wild animals released can be increased, since the professionals involved in the rehabilitation and release of these animals will have more information regarding the main injuries suffered by animal victims of vehicle collisions. Despite the frequent lack of subsidies for institutions dedicated to the rehabilitation and release of wild animals to function correctly, the efforts and resources invested in a single wild specimen make a big difference in the conservation of the species in question, particularly if it is an endangered species. Therefore, these efforts must be encouraged and mainly invested, since the conservation of species is essential for balancing ecosystems.

The authors have declared that no competing interests exist.

The activities were carried out in accordance with license number 40620 issued by the Sistema de Autorização e Informação em Biodiversidade of the Instituto Chico Mendes de Conservação da Biodiversidade (SISBIO, ISMBIO) for the collection of biological material, which was first issued on 2013/12/24 and is renewed annually.

This work was supported by the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (Capes), the Fundação de Apoio à Pesquisa do Rio Grande do Norte (FAPERN), and the Pró-Reitoria de Pesquisa e Pós-Graduação (PROPPG) of the Universidade Federal Rural do Semi-Árido (Ufersa).

Conceptualization: ITSF, JMAPA, CIPC. Data curation: ITSF, ARS, RS, SOC. Formal analysis: ITSF, JMAPA, CIPC. Funding acquisition: ITSF, FSC, JMAPA, CIPC. Investigation: ITSF. Methodology: ITSF, FSC, JMAPA, CIPC. Project administration: ITSF, CIPC. Resources: FSC, CIPC. Software: CIPC, RS. Supervision: CIPC. Validation: FSC, JMAPA, CIPC. Visualization: ITSF, JMAPA, CIPC. Writing—original draft: ITSF. Writing—review and editing: ITSF, FSC, JMAPA, CIPC.

Itainara Taili https://orcid.org/0000-0003-3960-551X

Fabiano Séllos Costa https://orcid.org/0000-0003-2763-4348

João Marcelo Azevedo de Paula Antunes https://orcid.org/0000-0003-3922-1428

Ayko Shimabukuro https://orcid.org/0000-0003-3726-4272

Raul dos Santos https://orcid.org/0000-0002-9764-6320

Sofia de Oliveira Cabral https://orcid.org/0000-0003-3645-3012

Cecilia Calabuig https://orcid.org/0000-0002-1243-1700

All of the data that support the findings of this study are available in the main text.