With wildlife crossing locations difficult to pinpoint, how do we know where to build crossing structures? Associate Editor, Manuela Gonzalez-Suarez explains how Bastille-Rousseau et al.’s new method and recent article, Optimizing the positioning of wildlife crossing structures using GPS telemetry help answer this question.
There are over 40 million km of roads in the world (100 times the distance from the moon to Earth), with many millions planned for the near future. In Europe a road is rarely far (<10 km in most areas). But even in more ‘pristine’ environments, roads can be prevalent. For example, the iconic Kruger National Park has nearly 2,300 km of roads, 850 of which are paved. Roads are a commodity for humans but often create problems for wildlife. Roads change the natural environment and can alter the movement and behaviour of many species. Individual animals may avoid crossing roads (or even their proximity) and, if they do cross, there is a risk of collision with vehicles, which can cause injuries and death to both animals and humans.
The most effective conservation measure is to avoid road development all together; the value of roadless areas is clear. However, this is not always possible, and arguably not always ethical, as roads have paved the way to development (pardon the pun). Moreover, in some cases, roads are already in place. So we must consider mitigation. Crossing structures – overpass bridges or underpass tunnels – are a well-established and useful mitigation measure to reduce collisions and maintain connectivity. However, building such structures is expensive, which means we need to ensure their location and configuration is carefully decided. Unfortunately, it is not always straightforward to identify the best location to build a crossing structure. Animals rarely use only one place to cross, and even if there are clearly preferred spots, it is not always easy to pinpoint their location.
Bastille-Rousseau and colleagues present a valuable new approach to tackle this problem. They develop two algorithms that prioritize crossing points using telemetry data. GPS data are becoming increasingly available for many species, so the proposed approach could take advantage of existing data and piggyback on ongoing studies designed to address other questions. To illustrate the proposed approaches Bastille-Rousseau and colleagues applied their algorithms to identify potential crossings within a section of a planned transnational commerce corridor in Northern Kenya using data from 156 collared elephants collected since 1998.
First, Bastille-Rousseau and colleagues defined 200 meter segments along the corridor and use the GPS data to identify where and how often elephants crossed each section. Both algorithms then use that information to optimize the location of a pre-defined number of crossings (predefined for example based on available funding). The first proposed approach, the crossing intensity (CI) algorithm, prioritizes overall use, and selects the most heavily used segments keeping them generally apart. After all, there is no point in making two crossing areas right next to each other. As situations may differ, the authors offer flexibility, and the user can specify the weight of the two components: intensity of use and distance between crossings. The second approach, the maximum coverage location problem (MLCP) algorithm, takes into account individual variation in crossing preferences. Often different animals use different areas to cross, so this approach seeks to maximize the number of distinct individuals benefitted by the crossings locations.
The illustrated example with elephant data shows that the two algorithms can offer distinct solutions, with solutions also varying with data quality. Bastille-Rousseau and colleagues analysed the complete, opportunistically gathered, dataset but also a systematically gathered, spatially unbiased subsample (for 40 elephants). Their results highlight the importance of considering the quality of the available data, and the importance of clearly identify the priority. CI algorithm focuses on high crossing intensity areas, whereas MLCP aims to benefit as many individuals as possible. CI may be particularly useful if the goal is to reduce collisions, and when crossings will be placed in unfenced areas. MLCP may be better when fencing is possible and when individual representation matters. For example, if crossing preferences of lead animals from different herds or groups are known and want to be protected.
Overall, these are promising methods and to facilitate their use the authors have created an R package called “wildxing”, which others can readily use. Since R is open-source code, this also opens the door for subsequent modification. For instance, I think a nice add-on would be to have a complementary estimation of crossing structure costs (often the costs of building crossings will vary across the landscape), so that instead of using a pre-defined number of crossings, the user could enter a pre-defined budget and optimize both location and quantity of crossing.
Read the full article, Optimizing the positioning of wildlife crossing structures using GPS telemetry in Journal of Applied Ecology.
Read a post from the authors about their work here.