Vertebrate roadkill: identifying where, when and who dies on wildlife vehicle collisions on Brazilian Cerrado

Abstract

Roadkill is a serious threat to wildlife conservation. To efficiently mitigate roadkill it is necessary to understand where, when, how and who suffer higher roadkill rates and that was the main objective of this entire thesis. In the first two chapters we used roadkill data from BR-050 highway, an area of Cerrado biome in Brazil. The monitoring was executed weekly from April 2012 to March 2014, by car, at about 60 km/h. We found 1294 roadkills on the highway, 922 mammals, 265 birds and 107 reptiles. We identified 78 species, 22 mammals, 42 birds and 14 reptiles. The roadkill rate was 0.051 individuals/km/day. In the first chapter, we investigate how land cover and climate influence roadkill. We conclude that the proximity to water bodies and distance to urban areas are important variables in determining the roadkill probability for all groups. Agriculture, natural and silviculture areas also influence mammal roadkill. Silvicuture and pasture are important for birds. Minimum temperature, accumulated precipitation and mean temperature influence mammal roadkill. For birds, insolation, humidity and mean temperature are important variables. For reptiles, all climate variables affect roadkill probability. Reptiles had a higher roadkill rate on rainy/hot months and birds on dry/cold ones. In the second chapter, we aimed to compare two different methods to locate roadkill hostpots: Siriema software and models based on landscape. Siriema software and landscape models identified roadkill hotspots for 50% and 70% of the data, respectively. With relation to the location of roadkill hotspots, the Mammalia class does not appear to represent well its species, although the orders do. For birds, neither methods detected roadkill hotspots for a satisfactory number of groups. The location of roadkill hotspots was different from one year to the other, and similar between one-year data and two. The application of mitigating measures on 9% of the road would protect 31% of the specimens for Siriema hotspots and 22% for landscape hotspots. Siriema and landscape hotspots in general presented a low correlation. Landscape models have a great advantage; it is notxii necessary to have wildlife-vehicle collision data to identify the stretches of the highway with higher roadkill probability. In the third chapter, we consulted scientific researches to investigate what species-specific characteristics influence roadkilland to calculate a roadkill rate for the Cerrado species. We were able to estimate a roadkill rate for 51 species, 19 were not observed killed. Roadkill rate increases as population density, home range and weaning decrease and body mass increases. Scavengers and territorial animals have higher roadkill rates. Mammals that prefer forest habitat have lower roadkill rates. The predicted mammals roadkill rate was 1.35 ind./km/year and 181909 animals could die each year. In the fourth chapter, we evaluated general patterns of animal/human accidents, from 2007 to 2017, on federal Brazilian highways. For this, we used the public data from the Brazilian Federal Highway Police. The 44444 accidents comprised 68775 people, including 66.1% unharmed, 23.2% with minor injuries, 7.7% with serious injuries and 1.6% died. The accidents occurred mainly from April to August, on Sundays, with clear sky, at night, in rural zones, on straight stretches, involving just one vehicle, especially a car. People affected by accidents were mostly drivers, adults and men. Most of the animal/human accidents happened in Northeast Brazilian Coast. Only animal/human accidents costed R$ 1 400 342 766.91, amount of money enough to install fences and one wildlife crossing structure each two kilometers through 517 km of fourlane roads.