Roadkill isn’t something glamorous or appealing to the eye; not something many people like seeing or getting a closer look. A closer look is by far the best way to describe how I looked at roadkill in my honours year. From a scientific perspective studying roadkill is a great way to start understanding of how wildlife is impacted by roads. In 2015 I worked in Addo Elephant National Park (Addo) on roadkill. More specifically, what factors led to roadkill in Addo. This project formed part of a long-term study that the Endangered Wildlife Trust’s (EWT) Wildlife and Roads Project are conducting nationwide.
How does one study roadkill?
Firstly, we used modelling in ArcGIS. Quite a finicky yet very powerful mapping program once you get the knack of it. The idea was to identify roads that pose a low, medium and high risk for roadkill using a number of predictors. Predictors were; proximity to gates, lodges, waterholes, and the road surface (tar or gravel). So in GIS the predictors were used in modelling software to generate a transect route of high medium and low-risk roads through Addo. I drove these roads (~140km) on two consecutive days once a month of six months (Figure 1). At each roadkill event, a whole array of measurements were recorded. This was to determine whether any other factors contributing to roadkill were not used in the models. Roadkill of both vertebrates and invertebrate species were recorded in my study. Invertebrates are not often something looked at when conducting roadkill surveys, but they were included in my study due to the Addo flightless dung beetle been endemic to the region.
After the six months of data collection and some great experiences in Addo, it was time to bring all the data together. Many long hours were spent trying to figure out GIS and R (another one of those powerful programs, once you’ve mastered it, this time for stats). Finally, after much hair pulling and shouting at an unresponsive laptop the answers to my questions started to come to life. It was great timing too, with not much time until due dates.
So what was the result of these stressful times? Well, something quite cool.
Having figured out GIS, the coordinates of the roadkill were used to produce hotspot maps of roadkill (Figure 2). There were two main hotspots for roadkill in Addo, one of which fell on a high-risk road and the other on a low-risk road. This showed that some predictors used to model the risk of roads were correctly predicting roadkill occurrence. While other predictors need to be addressed. This pointed us in the direction that something else was playing a part in roadkill events. The area with a hot spot on a low-risk road (western side of the map) is a road with high traffic volumes as it’s the only connection between the northern and southern section of the reserve. This is due to the eastern road being a private concession, Gorah Elephant Camp. To add o these high volumes of traffic in the region of the hot spot there are some points of interest. These being Jack’s picnic site and the Spekboom hide. These three factors all potentially contributing to higher volumes of traffic and therefore this hotspot for roadkill in the region.
Combining the results from GIS and R we found that road surface and distance to thicket were the significant contributors to roadkill within the reserve. More roadkill occurred on tarred roads, as was expected. While, as the distance between the road and the thicket decreased there was an increase in roadkill. Both of the hotspots for roadkill fall over sections of road that have <1meter between the road the thicket. This showing that this factor is playing a significant role in roadkill events. It’s believed this is a result of visual impairment with both drivers and animals fail to detect potential danger and therefore are unable to avoid collisions. Other contributing factors to roadkill in Addo were bends and hills in the road, however, these were not significant, largely due to not enough data been collected.
In the end, roadkill in Addo was largely a result of the road surface and the distance between the road and thicket vegetation. This shows that changes need to be made to the predictors we use in modelling programs to accurately predict roadkill events.
Modelling has the potential to be successfully used in predicting roadkill hotspot probability in protected areas. From gaining a better idea of contributing factors to roadkill occurrence from this study we can develop our modelling processes. This will lead to improved roadkill monitoring systems in protected areas. All in all, we would be able to use these programs to reduce roadkill events and reduce the unnecessary impact that roads have on wildlife in our protected areas.
I would like to thank SANParks and Gorah Elephant Camp for allowing us to conduct this work in the reserve. A big thanks to all the helping hands. From those that joined me in the park driving around Addo, pretty much one long game drive (tough stuff), to the identification of some pretty squashed animals and lastly to the geniuses that understand and helped with GIS and stats.