Consistent patterns of vehicle collision risk for six mammal species

Wildlife mortality risk posed by high and low traffic roads

Wildlife mortality due to collisions with vehicles (roadkill) is one of the predominant negative effects exerted by roads on many wildlife species. Reducing roadkill is therefore a major component of wildlife conservation. Roadkill is affected by various factors, including road attributes and traffic volume. It is theorized that the effect of traffic volume on roadkill probability should be unimodal. However, empirical evidence for this theory is lacking. Using a large-scale roadkill database of 18 wildlife species in Israel, encompassing 2846 km of roads over 10 years, we explored the effects of traffic volume and road attributes (e.g., road lighting, verge vegetation) on roadkill probability with a multivariate generalized linear mixed model. A unimodal effect of traffic volume was identified for the striped hyena (Hyaena hyaena), whereas 5 species demonstrated a novel quadratic U-shaped effect (e.g., golden jackal [Canis aureus]). Four species showed a negative linear effect (e.g., wild boar [Sus scrofa]). We also identified varying effects of road attributes on roadkill. For instance, road lighting and roadside trees decreased roadkill for several species, whereas bus stops and concrete guardrails led to increased roadkill. The theorized unimodal effect of traffic volume may only apply to large, agile species, and the U-shaped effect could be related to intraspecies variability in traffic avoidance behavior. In general, we found that both high-traffic and low-traffic roads can pose a high mortality risk for wildlife. It is therefore important to monitor roadkill on low-traffic roads and adapt road attributes to mitigate roadkill. Road design for effective roadkill mitigation includes reducing the use of concrete guardrails and median barriers where possible and avoiding dense bushes in verge landscaping. These measures are complemented by employing wildlife detection systems, driver warnings, and seasonal speed reduction measures on low-traffic roads identified as roadkill hotspots.

A spatiotemporal risk prediction of wildlife-vehicle collisions using machine learning for dynamic warnings

INTRODUCTION

The technology in the automotive industry is becoming increasingly safer in the age of automated driving, but the number of accidents is still high, especially in wildlife-vehicle collisions (WVCs). To better avoid these accidents, patterns involved in these accidents must be detected.

METHOD

This paper presents a spatiotemporal risk prediction of WVCs, including various road and environmental characteristics. A process of data preparation using GIS automated by Python scripts was developed to enable a spatiotemporal link of diverse features for the subsequent predictive data analysis. Different machine learning (ML) approaches were applied- random forest (RF), feedforward neural networks (FNN), and support vector machine classifier (SVM) - including automated ML to predict the risk of WVCs. Therefore, a dataset of approximately 731,000 accidents reported to the police in Bavaria over a period of 10 years (2010-2019) was used. In addition, non-accidents were randomly generated in Python over time and space for the supervised ML processes. As the actual risk probability for WVCs and non-WVCs is not entirely known, the impact of different training ratios between accidents and non-accidents was tested on the risk prediction quality (RPQ) (25%, 50%, 75%, 90% WVCs) of the double-weighted sensitivity and single-weighted specificity rate.

RESULTS

The test yielded high mean values of RPQ as an indicator for a suitable WVC prediction. Both RF (86.6%) and FNN (86.7%) were identified as suitable choices for WVC risk prediction in terms of RPQ. The SVM yielded a lower prediction quality, even though acceptable results could be achieved within a shorter runtime.

PRACTICAL APPLICATIONS

Spatial transferability was verified since the algorithm was trained on the dataset of Bavaria (excluding Upper Bavaria) and successfully tested in Upper Bavaria. WVC forecasts were also proven through training with datasets from 2010-2017 and in prediction for 2018 and 2019.

Wombat Roadkill Was Not Reduced by a Virtual Fence. Comment on Stannard et al. Can Virtual Fences Reduce Wombat Road Mortalities? Ecol. Eng. 2021, 172, 106414

Simple Summary

Wildlife roadkill is a global problem. Devices that produce bright lights and loud noises are claimed to be effective deterrents, but there is no scientific evidence that they actually prevent collisions between vehicles and wildlife. A recent trial in Australia installed a light and sound system, marketed as a ‘virtual fence’, and tested it on bare-nosed wombats (Vombatus ursinus). The study concluded that the system was minimally effective but promising. We detected a number of conceptual and procedural flaws in that study. When we re-analysed their data, we found absolutely no evidence for the effect of the virtual fence on the roadkills of wombats.

Abstract

The roadkill of wildlife is a global problem. Much has been written about deterring wildlife from roads, but, as of yet, there is no empirical support for deterrents based on visual and/or auditory signals. A recent paper entitled ‘Can virtual fences reduce wombat road mortalities?’reported the results of a roadkill mitigation trial. The authors installed a ‘virtual fence’ system produced by iPTE Traffic Solutions Ltd. (Graz, Austria) and evaluated its effectiveness for reducing roadkills of bare-nosed wombats (Vombatus ursinus) in southern Australia. The authors recorded roadkills in a simple Before-After-Control-Impact design but did not conduct any formal statistical analysis. They also measured three contextual variables (vegetation, wombat burrows, and vehicle velocity) but did not link these to the occurrence of roadkills in space and time. The authors concluded that the iPTE virtual fence system was ‘minimally effective’, yet ‘appears promising’. Our analysis of their data, using standard inferential statistics, showed no effect of the virtual fence on roadkills whatsoever. We conclude that the iPTE system was not effective for mitigating the roadkills of bare-nosed wombats.

Animal-vehicle collisions during the COVID-19 lockdown in early 2020 in the Krakow metropolitan region, Poland

The interrelations between human activity and animal populations are of increasing interest due to the emergence of the novel COVID-19 and the consequent pandemic across the world. Anthropogenic impacts of the pandemic on animals in urban-suburban environments are largely unknown. In this study, the temporal and spatial patterns of urban animal response to the COVID-19 lockdown were assessed using animal-vehicle collisions (AVC) data. We collected AVC data over two 6-month periods in 2019 and 2020 (January to June) from the largest metropolis in southern Poland, which included lockdown months. Furthermore, we used traffic data to understand the impact of lockdown on AVC in the urban area. Our analysis of 1063 AVC incidents revealed that COVID-19 related lockdown decreased AVC rates in suburban areas. However, in the urban area, even though traffic volume had significantly reduced, AVC did not decrease significantly, suggesting that lockdown did not influence the collision rates in the urban area. Our results suggest that there is a need to focus on understanding the effects of changes in traffic volume on both human behaviour and wildlife space use on the resulting impacts on AVC in the urban area.

COVID-19 restrictions provide a brief respite from the wildlife roadkill toll

The COVID-19 pandemic provides a rare opportunity to reveal the impact of reduced human activity on wildlife. I compared traffic volume and wildlife roadkill data along 18 km of highway before, during and after a 3-month period of COVID-19 restrictions with baseline data from the previous four years. Three marsupial herbivores comprised 89% of the 1820 roadkills recorded during the 4.5-year survey period: rufous-bellied pademelon Thylogale billardierii (31.5% of total), common brushtail possum Trichosurus vulpecula (29.8%) and red-necked wallaby Notamacropus rufogriseus (27.9%). During April 2020, when human activity was most restricted in the study area, traffic volume decreased by 36% (i.e. by an average 13,520 vehicle movements per day) and wildlife roadkill decreased by 48% (i.e. from 44 to 23 roadkills). However, when restrictions eased, traffic volume and wildlife roadkill returned to baseline levels indicating that the respite was brief in terms of animal welfare and of limited conservation value for these widespread and abundant species. Nevertheless, the results of this study suggest that even short periods of traffic reduction or road closures could be used as part of a management strategy for the conservation of endangered wildlife populations and re-wildling programs where roadkill is a risk factor.

An adaptive plan for prioritizing road sections for fencing to reduce animal mortality

Mortality of animals on roads is a critical threat to many wildlife populations and is poised to increase strongly because of ongoing and planned road construction. If these new roads cannot be avoided, effective mitigation measures will be necessary to stop biodiversity decline. Fencing along roads effectively reduces roadkill and is often used in combination with wildlife passages. Because fencing the entire road is not always possible due to financial constraints, high-frequency roadkill areas are often identified to inform the placement of fencing. We devised an adaptive fence-implementation plan to prioritize road sections for fencing. In this framework, areas along roads of high, moderate, and low levels of animal mortality (respectively, roadkill hotspots, warmspots, and coldspots) are identified at multiple scales (i.e., in circles of different diameters [200-2000 m] in which mortality frequency is measured). Fence deployment is based on the relationship between the amount of fencing being added to the road, starting with the strongest roadkill hotspots, and potential reduction in road mortality (displayed in mortality-reduction graphs). We applied our approach to empirical and simulated spatial patterns of wildlife-vehicle collisions. The scale used for analysis affected the number and spatial extent of roadkill hot-, warm-, and coldspots. At fine scales (e.g., 200 m), more hotspots were identified than at coarse scales (e.g., 2000 m), but combined the fine-scale hotspots covered less road and less fencing was needed to reduce road mortality. However, many short fences may be less effective in practice due to a fence-end effect (i.e., animals moving around the fence more easily), resulting in a trade-off between few long and many short fences, which we call the FLOMS (few-long-or-many-short) fences trade-off. Thresholds in the mortality-reduction graphs occurred for some roadkill patterns, but not for others. Thresholds may be useful to consider when determining road-mitigation targets. The existence of thresholds at multiple scales and the FLOMS trade-off have important implications for biodiversity conservation.

Reflections on Potential Risk Factors of Callitrichidae Run Over in an Urban Area: A Case Report of Marmoset Deaths

Interest in the well being of wild animals in their natural habitats is still lower than that of domesticated animals and captive wild animals. Urban development is one of the events that has the greatest impact on fauna, as it affects the survival and well being of wild species in many ways. This study aimed to record death by being run over of two Callithrix penicillata (black-tufted) marmosets in a fragmented environment by urbanization and to discuss how anthropic intervention modifies species behavior and influences their well being, and to suggest measures which can reduce the occurrence of incidents, thereby contributing to maintaining the fauna and their well being.

Predicting deer–vehicle collision risk across Victoria, Australia

The risk of deer–vehicle collisions (DVCs) is increasing in south-east Australia as populations of introduced deer expand rapidly. There are no investigations of the spatial and temporal patterns of DVC or predictions of where such collisions are most likely to occur. Here, we use an analytical framework to model deer distribution and vehicle movements in order to predict DVC risk across the State of Victoria. We modelled the occurrence of deer using existing occurrence records and geographic climatic variables. We estimated patterns of vehicular movements from records of average annual daily traffic and speeds. Given the low number of DVCs reported in Victoria, we used a generalised linear regression model fitted to DVCs in California, USA. The fitted model coefficients suggested high collision risk on road segments with high predicted deer occurrence, moderate traffic volume and high traffic speed. We used the California deer model to predict collision risk on Victorian roads and validated the predictions with two independent datasets of DVC records from Victoria. The California deer model performed well when comparing predictions of collision risk to the independent DVC datasets and generated plausible DVC risk predictions across the State of Victoria.