Managing the timing and speed of vehicles reduces wildlife-transport collision risk
Introduction
Wildlife collisions with transport vehicles pose a serious global problem (Litvaitis and Tash, 2008) and inspire research to investigate causes and propose solutions. For example, deer-vehicle collisions on roads are common and well-studied in North America (Huijser et al., 2007, Romin and Bissonette, 1996) and Europe (Sáenz-de-Santa-María and Tellería, 2015, Seiler, 2004). In addition to concerns about animal welfare (Sainsbury et al., 1995) and conservation status of threatened species (Dwyer et al., 2016, Jones, 2000), collisions with large animals pose direct risks to the life of humans (Langley et al., 2006, Rowden et al., 2008). Moose, for instance, are one of the largest animals struck by vehicles in North America and Europe, causing significant damage and injuries (Huijser et al., 2007, Hurley et al., 2009). Vehicle collisions with deer, although smaller than moose, frequently kill humans in North America (Williams and Wells, 2005). Collision events disrupt transportation services, and are therefore also an economic burden. Management of wildlife-vehicle collisions will become increasingly important as new transportation networks are constructed and existing networks are expanded.
Information about the spatial and temporal distribution and magnitude of wildlife-vehicle collisions is useful to managers because it can more effectively help mitigate impacts (Mountrakis and Gunson, 2009). For example, identifying a collision hotspot along a transportation network for a particular species can inform the most appropriate form of mitigation (e.g. animal exclusion or change in vehicle speed – see Visintin et al., 2016). Collision data can also be used in statistical modelling which helps to predict the probability of wildlife-vehicle collisions (Gunson et al., 2011).
The majority of wildlife-vehicle collision modelling deals with roads and traffic (Popp and Boyle, 2017, van der Ree et al., 2015), yet, the problem extends to other forms of vehicular transportation such as air (van Belle et al., 2007), railway (Belant, 1995, Onoyama et al., 1998, Wells et al., 1999) and shipping (Laist et al., 2001) operations. Regardless of the mode of transport, the modelling of collisions share some common attributes (Forman et al., 2003). The movements or presence of animals are often considered in the models and may include species-specific habitat variables (Roger and Ramp, 2009) or behavioural traits (Lee et al., 2010). Vehicle presence or movements are also considered and may be grouped into a larger category of human behaviour as humans ultimately control speeds and trajectories of vehicles (Ramp and Roger, 2008).
Some drawbacks of existing road ecology studies are that the time of collision is often not known precisely, the volume of traffic varies temporally and is difficult to accommodate in analyses (especially when time of collision is not known), and collisions might be under-reported or temporally and spatially biased. These drawbacks might obscure drivers of collision risk or provide inaccurate predictions of risk upon which mitigation might be implemented. These considerations are especially important for studies that explore temporal variation in animal activity (e.g. Dussault et al., 2006, Thurfjell et al., 2015) as reporting errors in collision times may produce spurious correlations (or lack of).
Wildlife-train collisions provide a different dataset for considering factors that influence collision risk that are often difficult to obtain. Firstly, the timetabling of trains means that their movements are much more precisely documented than those of cars. Secondly, collisions with large animals are more consistently reported for trains than cars – although under-reporting is still an issue (Dorsey et al., 2015). And thirdly, the time of collision is also reported more precisely. These benefits mean that examining factors that influence wildlife-train collision risk may help uncover factors that influence wildlife-vehicle collisions more generally.
Herein, we develop a modelling framework to predict the rate of wildlife-train collisions across a large railway network. We use kangaroos and a regional passenger railway network in south-east Australia as an example to demonstrate our methods. In addition to informing railway operators in Australia of potential kangaroo collision risks, our approach can be generalised to other species (e.g. deer) and transportation modes (e.g. road vehicles, watercraft) elsewhere in the world by incorporating parameters that accommodate patterns of species activity.
Section snippets
Study area
Our study area encompassed a 1712-kilometre passenger railway network from regional Victoria, Australia (operated by V/line, a government-owned corporation) in south-east Australia (Fig. 1). Trains operate on all sections of the network between 4 a.m. and 2 a.m. (the following day), with the largest volume occurring Monday through Friday between 7 a.m. and 9 a.m., and 4 p.m. and 6 p.m. Most trains operate at speeds of 100 km h−1 or less, however, on some sections of track commuter trains
Results
The signs of model coefficients (positive or negative) indicated plausible relationships between collision likelihood and each predictor variable. All variables except train frequency demonstrated strong influence on collision risk (Table 2). Dropping the temporal animal activity term had the largest negative effect on model fit, changing the Akaike information criteria (AIC) score from 5528 to 5740. The model residuals (using all variables) demonstrated no strong patterns with respect to
Discussion
Our model demonstrates that kangaroo-train collisions are related to train speed, kangaroo exposure to moving trains, and the coincidence of periods of high train and kangaroo activity. All of these relationships are consistent with our initial expectations and also shown in related studies on railway (Gundersen and Andreassen, 1998), road (Lao et al., 2011, Roger et al., 2012, Visintin et al., 2016) and marine collisions (Conn and Silber, 2013, van der Hoop et al., 2012) with other species.
Data accessibility
The modelling dataset and code for this work are available upon request. Please email the authors to inquire. The original collision data is V/Line’s Intellectual Property (IP) and authorisation for additional use must be granted by V/Line.
Acknowledgements
We thank V/Line for their generous collaboration; in particular, Sam Parsons, who queried and provided the collision statistics used for the study. Michaela Plein provided valuable comments on the manuscript. This project was supported by a University of Melbourne International Research Scholarship and the Australian Research Council Centre of Excellence for Environmental Decisions.
References (63)
- et al.
Some stylized facts about high-speed rail: a review of HSR experiences around the world
Transp. Policy
(2009) - et al.
Spatial wildlife-vehicle collision models: a review of current work and its application to transportation mitigation projects
J. Environ. Manage.
(2011) - et al.
Flattened fauna and mitigation: traffic victims related to road, traffic, vehicle, and species characteristics
Transport. Res. Part D: Transp. Environ.
(2006) - et al.
Risk factors associated with fatal animal-vehicle collisions in the United States, 1995–2004
Wilderness Environ. Med.
(2006) - et al.
Modeling animal-vehicle collisions considering animal-vehicle interactions
Accid. Anal. Prev.
(2011) Basic principles of ROC analysis
Semin. Nucl. Med.
(1978)- et al.
Railway ecology: Underrepresented in science?
Basic Appl. Ecol.
(2017) - et al.
Road crashes involving animals in Australia
Accid. Anal. Prev.
(2008) - et al.
Factors affecting road mortality and the suitability of road verges for butterflies
Biol. Conserv.
(2013) - et al.
Underreporting of wildlife-vehicle collisions does not hinder predictive models for large ungulates
Biol. Conserv.
(2015)
Avoidance of high traffic levels results in lower risk of wild boar-vehicle accidents
Landscape Urban Plann.
Moose collisions with vehicles and trains in northeastern Minnesota
Alces
Validation and calibration of probabilistic predictions in ecology
Methods Ecol. Evol.
Vessel speed restrictions reduce risk of collision-related mortality for North Atlantic right whales
Ecosphere
Analysis of Survival Data
Analysis of Binary Data
The Ecology of Transportation: Managing Mobility for the Environment
Kangaroos
Randomized quantile residuals
J. Comput. Graph. Stat.
Temporal and spatial distribution of moose-vehicle accidents in the Laurentides Wildlife Reserve, Quebec, Canada
Wildlife Biol.
Using citizen-collected wildlife sightings to predict traffic strike hot spots for threatened species: a case study on the southern cassowary
J. Appl. Ecol.
A working guide to boosted regression trees
J. Anim. Ecol.
Spatio-temporal data types: an approach to modeling and querying moving objects in databases
GeoInformatica
Effects of roads on animal abundance: an empirical review and synthesis
Ecol. Soc.
Road Ecology: Science and Solutions
Is my species distribution model fit for purpose? Matching data and models to applications
Glob. Ecol. Biogeogr.
The risk of moose Alces alces collision: a predictive logistic model for moose-train accidents
Wildlife Biol.
The state of spatial and spatio-temporal statistical modeling
Cited by (17)
Low audibility of trains may contribute to increased collisions with wildlife
2022, Transportation Research Interdisciplinary PerspectivesCitation Excerpt :These objectives might be practically achieved in locations of high collision risk with targeted reductions in train speed, with consistent use of train horns, and potentially with trackside wildlife warning signals. Our results showed that trains are audible at a longer time before arrival at lower speeds, and speed reductions appear to be effective in reducing collisions (Visintin et al., 2018). Reduced variation in train speed may also reduce instances of surprise from low-audibility outliers.
Where wildlife and traffic collide: Roadkill rates change through time in a wildlife-tourism hotspot
2021, Global Ecology and ConservationCitation Excerpt :The increase in visitors between 1998–1999 and 2014 may therefore have contributed to higher roadkill rates due to a higher probability of a wildlife-vehicle collision (Farmer and Brooks, 2012; Taylor and Goldingay, 2010; Visintin et al., 2018). Dawn and dusk have been shown to correlate with increased mortality rates involving trains and wildlife (Visintin et al., 2018) and similar may be occurring on Phillip island with these periods coinciding with increased activity of nocturnal and crepuscular species. Therefore, the probability of a roadkill event may vary throughout the day and traffic pulses at certain times of the day may have a disproportionate effect on roadkill rates.
Spatial analysis of wildlife-train collisions on the Czech rail network
2020, Applied GeographyCitation Excerpt :Despite the differences between road and rail transport, certain findings are valid for both transport modes. These include, for example, a temporal pattern of collisions (Bartonička et al., 2018; Haikonen & Summala, 2001; Krauze-Gryz et al., 2017), as the activity of wild animals, particularly ungulates, is usually the highest between sunset and sunrise; landscape structure (Bartonička et al., 2018; Jensen et al., 2014), as forested areas are often more hazardous; environmental conditions (Andersen et al., 1991; Gundersen & Andreassen, 1998; Gundersen et al., 1998; Popp et al., 2018), such as the snow depth and low temperatures are related to the number of collisions with moose or elk; food attractants near railway tracks (Murray et al., 2017), such as spilled grain from freight trains can attract wildlife; the density of wildlife in the vicinity of railway tracks (Jasińska et al., 2019); or vehicle speed and traffic intensity (Jasińska et al., 2019; Visintin et al., 2018). Factors that determine crash hotspots are then referred to as local factors because their presence increases the probability of crashes with wildlife at particular places (see Bartonička et al., 2018; Bíl et al., 2019).
Wildlife-vehicle collisions - Influencing factors, data collection and research methods
2020, Biological ConservationCitation Excerpt :The remaining 19% (n = 120) of the literature list were also deleted as it contains diverse other publications not analysing WVCs. These publications deal, for instance, with the economic consequences (e.g., Gren and Jägerbrand, 2019) or human injuries of WVCs (e.g., Ang et al., 2019; Conover et al., 1995), or with other accident types such as train-collisions (Visintin et al., 2018; Andreassen et al., 2005) - just to name a few examples. It would go beyond the scope of the work to name all excluded topics.
Warning systems triggered by trains increase flight-initiation times of wildlife
2020, Transportation Research Part D: Transport and EnvironmentEffects of hourly precipitation and temperature on road traffic casualties in Shenzhen, China (2010–2016): A time-stratified case-crossover study
2020, Science of the Total Environment